THE CONCEPT

OF IDENTITY
This page intentionally left blank
THE CONCEPT
OF IDENTITY

Eli Hirsch

New York Oxford

OXFORD UNIVERSITY PRESS
 Oxford University Press
Oxford New York Toronto
Delhi Bombay Calcutta Madras Karachi
Petaling Jaya Singapore Hong Kong Tokyo
Nairobi Dar es Salaam Cape Town
Melbourne Auckland
and associated companies in
Berlin Ibadan

Copyright © 1982 by Eli Hirsch

First published by Oxford University Press, 1982
First issued as an Oxford University Presss paperback, 1992
Oxford University Press, Inc.
200 Madison Avenue, New York, New York 10016
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 the publisher.
Library of Congress Cataloging-in-Publication Data
Hirsch, Eli
The Concept of Identity
"Part One . . . originally published as . . . The
persistence of objects"—Verso of t.p.
Includes index
1. Identity. 2. Object (Philosophy) 3. Space
and time. I. title
BD236.H55 1982 111 81-9508
ISBN 0-19-502995-X AACR2
ISBN 0-19-507474-2 (pbk)

2456897531
Printed in the United States of America
on acid-free paper
To the memory
of Dean Kolitch
This page intentionally left blank
 Preface

My main concern in this book is identity through time, first with

respect to ordinary bodies, then underlying matter, and eventually
persons. These issues link up at various points with other aspects
of identity, such as the spatial unity of things, the unity of kinds,
and the unity of groups. One of my concerns is to understand how
our identity concept ordinarily operates in these various respects;
but I also try to understand, especially in later chapters of the
book, why this concept is so central to our thinking, and whether
we can justify seeing the world in terms of such a concept.
Part One, with a few minor differences (mainly in footnotes),
was published several years ago as a monograph on the persistence
of objects. That work, though its circulation was quite limited,
did generate some interest, and this has encouraged me to present
it again in a more accessible form. The views expressed in Part
One are augmented, and in certain respects qualified, by the treat-
ment in Part Two of various related themes of identity. Though
both parts of this book may be said to form a single extended
discussion, the chapters in Part Two can also be read as relatively
self-contained essays, which is in fact the spirit in which they were
written.
A number of people over the years have helped me with this
book. I thank my wife Pamela, for her perceptive assistance on
both matters of style and philosophy; Milton Munitz, for being a
good teacher and friend; Roderick Chisholm, for his discussions
of identity at an NEH-sponsored seminar I attended during a
period when I wrote portions of the book; and especially Saul
viii PREFACE

Kripke, whose discussions in lectures of some of my views encour-

aged me to elaborate them further.
I am also much indebted to Alan Brody, Georges Dicker, Wil-
liam James Earle, Dean Kolitch, Joseph Margolis, and Karsten
Struhl, for their invaluable advice and comments on various por-
tions of the book during its preparation. And Part Two, especially
Chapter 7, was substantially affected by Sydney Shoemaker's exten-
sive criticisms, for which I am most grateful.
I have made use, with permission, of the following previously
published material:
"Physical Identity," The Philosophical Review, 84 (1976); por-
tions of this paper are scattered throughout Part One.
The Persistence of Objects (Philosophical Monographs, Phila-
delphia, Pennsylvania, 1976); this corresponds to Part One.
"A Sense of Unity," The Journal of Philosophy, 74, no. 9 (1978);
this is Chapter 8.

New York City E. H.

September 1981
 Contents

PART ONE: THE PERSISTENCE OF OBJECTS 1

Introduction to Part One 3

1 Continuity 7
I The Simple Continuity Analysis 7
II Qualitative Continuity 10
III Spatiotemporal Continuity 15
IV Is Continuity Necessary? 22
V Is Continuity Sufficient? 25
2 Sortals 34
I The Sortal Rule 34
II The Making of a Sortal 40
III Coming into Existence and Going out of Existence 47
IV Identity, Predication, and Constitution 57
V The Compositional Criterion 64
3 The Basic Idea of Persistence 72
I A Question about Sortal-Relativity 72
II The Basic Rule 77
III Limitations of the Basic Rule 82
IV Refining the Basic Rule 90
V Unity through Time and Space 97
VI Articulation 105
4 The Persistence of Matter 113
I A Puzzle about Matter 113
II An "Ultimate" Kind of Persistence 119
III Searching for Identity Criteria 123
IV Matter and Common Sense 128
X CONTENTS

5 The Metaphysics of Persistence 138

I Do We Need Persisting Objects? 138
II A Question about Spatiotemporal Continuity 144
III Identity Schemes 149
IV "Real" and "Fictitious" Persistence 156
V Can We Justify Our Identity Scheme? 162

PART TWO: MINDS AND BODIES 175

Introduction to Part Two 177

6 Foundations of Identity 181
I Metaphysical Priorities and Epistemological Priorities 182
II Body-Stages 184
III Temporal Parts 188
IV A Question of Priorities 192
V Spatiotemporal Continuity 196
VI Analyzing Bodily Identity 201
VII Epistemological Priorities 202
7 Matter, Causality, and Stereotypes of Identity 211
I Optimal Cases 211
II Compositional and Causal Continuity 216
III Stereotypes of Identity 227
8 A Sense of Unity 236
1 Criteria of Unity 236
II Unity and Similarity 239
III Conventionalism 244
IV An "Empiricist" Explanation 249
V Focusing on Objects 255
VI Conclusion 262
9 Natural Kinds and Natural Units 264
I Kinds and Units 264
II Kinds and Similarity Classes 267
III Is the Class of Units a Kind? 270
IV Kinds and Individuation 274
V The Basis of Kinds and Units 278
10 Constraints on Self-Identity 286
I A Strange Identity Concept 287
II Metaphysical Constraints 293
III Pragmatic Constraints 297
IV Psychological Constraints 301
V The Sense of Self 307
Index 313
 PART ONE

THE PERSISTENCE
OF OBJECTS
This page intentionally left blank
 Introduction to Part One

OUR CONCEPT of a physical object's persistence through time

seems so fundamental and primitive that it requires a special
effort to appreciate what philosophers might be after when they
ask for an analysis of this concept. Traditionally the request for
such an analysis might take the form of such questions as: What
does the identity through time of a physical object consist in?
or What is it for a physical object which exists at one time to be
the same object as a physical object which exists at another time?
In more recent literature one typically finds philosophers asking
for an account of our "identity criteria" for objects. This new
terminology, at least as I intend to employ it, still expresses very
much the same traditional request for an analysis of our identity
concept, except perhaps that to talk about "identity criteria" is
to signal more clearly one's quite reasonable willingness to settle
for an analysis which may be less than airtight and which may
allow for many borderline cases.1
When we ask with regard to physical objects what their iden-
tity through time consists in, we are asking for an account of the
unity of a physical object's career. Any physical object has a
career which stretches over a period of time, a career which we
can think of as comprised of a temporal succession of momentary
stages. The successive parts, or stages, of an object's career must
1. On the meaning of "identity criteria" see Sydney Shoemaker, Self-Knowledge
and Self-Identity (Cornell University Press, Ithaca, New York, 1963), pp. 3-5,
and David Wiggins, Identity and Spatio-Temporal Continuity (Basil Blackwell,
Oxford, 1967), p. 43.

3
4 THE CONCEPT OF IDENTITY

hang together in some distinctive way; otherwise there would be

nothing to prevent us from arbitrarily combining into a single
career the early stages of one object with the later stages of a
different object. Evidently not just any succession of object-stages
corresponds to a single persisting object; some do and some do
not. So in order for object-stages to add up to a single persisting
object they must be related in some special way. What I arn seek-
ing in Part One is an analysis or definition of what that relation-
ship is.
In a sense, of course, any succession of object-stages, however
arbitrary, does add up to something: perhaps to an event, or to
a state of affairs or, if nothing else, at least to a "merely arbitrary
succession of object-stages." What is important, however, is that
not every succession adds up to a persisting object or body (I
will use these expressions interchangeably), where this funda-
mental category is to be understood as loosely comprising items
which can straightforwardly be said to occupy space and to per-
sist through time. Clearly only certain privileged successions are
accorded the special status of uniting into a single persisting
object in this sense, which gives rise to the question as to what
the unity-making relationship is in virtue of which some suc-
cessions enjoy this special status.
Our question, I want to stress, is primarily conceptual rather
than epistemological. We are not, that is, to be thinking pri-
marily of a situation in which someone has not seen an object for
some time and a question arises as to how he can know that he
has really come across the same object again. Rather we are to
be thinking primarily of a situation in which someone contin-
uously observes an object for a stretch of time, and, as I shall
often put it, traces the object's career for that period. Our ques-
tion is what criteria of identity enter into this tracing operation.
How can we analyze what it means to judge in those optimal
circumstances that it was a single persisting object that was being
followed?
It must be emphasized, furthermore, that this is a question
about our most ordinary notion of physical persistence. We want
an account of what goes into our thought about the identity
through time of tables, trees, and other objects that we ordinarily
talk about. A philosopher may of course hold that the ordinary
notion of physical persistence is not ultimately important, per-
 INTRODUCTION TO PART ONE 5

haps because ordinary physical objects are not among the "ulti-
mate constituents of reality." Whatever might be the cogency
of this sort of claim (and I shall have something to say about it
in the course of what follows), the fact remains that we certainly
do have an ordinary way of thinking about the physical world,
and it must be of some philosophical interest to provide an analy-
sis of that way of thinking.
Our question, then, is about as clear as the notion of giving an
"analysis" (or a "definition"), which means, I think, that it is
not luminously clear at all. One important difficulty with this
notion has to do with deciding when an analysis is "circular,'"
when, that is, the concepts in terms of which it is couched de-
pend, in some sense, on the concept being analyzed. This diffi-
culty may seem potentially devastating when the concept to be
analyzed is as fundamental to our overall thought as the concept
of physical persistence. But perhaps we may provisionally adopt
a fairly tolerant attitude about this. If we can provide an account
of our identity criteria which strikes us as at least not patently
circular then we may feel that we have the kind of analytic
illumination that we sought. It may turn out, of course, that
granted even a reasonable measure of tolerance our concept of
physical persistence, or some application of that concept, will
seem to resist the sort of analysis that we are seeking. In this case
we will have to say that the concept, or some application of it, is,
in some important sense, ultimate and unanalyzable. Later, in
Chapter 4, I will in fact defend the position that our concept of
the persistence of material substance is in a sense unanalyzable.
And in Chapter 5, the final chapter of this part, I will consider a
bit more forthrightly some of the metaphysical issues that may
revolve around the idea of giving an analysis of physical persist-
ence. These issues in their full generality, however, will not be
dealt with until Part Two.
The topic that I intend to focus upon in this first part is rather
severely circumscribed. I want to examine our concept of per-
sistence as it pertains to the seemingly most central and unex-
ceptionable instances of physical objects or bodies. These would
include, I assume, such things as tables and cars, mountains and
stones, trees and flowers, cats and dogs, chunks of clay and bits
of wood. But I shall have nothing to say in this part about the
identity conditions for such nonsubstantial items as events and
6 THE CONCEPT OF IDENTITY

properties, or such corporate items as groups and forests; nor

will I enter into the very special problems which seem to affect
our concept of the persistence of persons. Some of these addi-
tional issues will be discussed in Part Two.
An object's unity through time is by no means the only philo-
sophically challenging mode of object-unity. In particular one can
raise questions about an object's unity through space which
parallel in many ways questions about its unity through time.
The spatial question would have to do with our basis for treat-
ing some, but not all, aggregates of matter as unitary objects.
This question will eventually be addressed in Chapter 3. But in
order to focus properly on the immediate question, a question
essentially about identity through time, the perspective to adopt
is one in which an aggregate of matter has (on whatever basis)
already been delineated as a unitary object and our primary con-
cern is to understand what it means to trace that object's career
through time.
 I
Continuity

I. The Simple Continuity Analysis

WE WANT to understand the nature of the unity-making rela-
tionship which binds the successive stages of the career of a single
persisting object. When one first reflects upon this question, an
idea which might readily come to mind is that contiguous stages
of a single career must be qualitatively very similar and spatially
very close. Over an extended period an object may of course sig-
nificantly alter its qualitative makeup and its spatial location,
but it seems that such alterations occur continuously, i.e., by
small degrees. This may suggest the possibility of formulating
a very simple kind of analysis of the unity-making relationship
in terms essentially of two considerations: (1) continuity of quali-
tative change (which I will call, for short, "qualitative continu-
ity"), and (2) continuity of locational change (which I will call,
for short, "spatiotemporal continuity").
This very simple kind of analysis is much too simple, as I will
explain shortly. But many philosophers of the empiricist tradi-
tion have presented accounts of the identity of objects which
suggest just this simple analysis.
Certainly such an analysis seems to be suggested by the follow-
ing remarks of Russell:
Given any event A it happens very frequently that, at any neighboring
time, there is at some neighboring place an event very similar to A.
A "thing" is a series of such events. ... It is to be observed that in a
series of events which common sense would regard as belonging to one

7
8 THE CONCEPT OF IDENTITY

"thing," the similarity need only be between events not widely sep-
arated in space-time. There is not very much similarity between a three
months' embryo and an adult human being, but they are connected by
gradual transitions from next to next, and are therefore accepted as
stages in the development of one "thing."1
I take it that by an "event" Russell means here pretty much
the same as what I have been calling an "object-stage" (and what
philosophers sometimes call a "temporal slice" of an object).
Russell's remarks suggest that the unity-making relationship
which binds a succession of object-stages ("events") into a single
persisting object is essentially nothing more than spatiotemporal
and qualitative continuity. 2
A succinct formulation of what appears to be the same position
is expressed by C. D. Broad. In the course of discussing "the
durations of physical objects" Broad states: "A thing . . . is simply
a long event, throughout the course of which there is either
qualitative similarity or continuous qualitative change, together
with a characteristic spatio-temporal unity." 3 Apparently Broad
is saying that our criteria of identity for objects consist simply of
the two considerations of qualitative and spatiotemporal
continuity.
We have, then, as our first and simplest possibility, an analysis
of persistence which might be formulated as follows:
The Simple Continuity Analysis. A succession S of object-stages
corresponds to stages in the career of a single persisting object
if and only if:
(1) S is spatiotemporally continuous; and
(2) S is qualitatively continuous.
I call this "the simple continuity analysis" because it relies
exclusively on continuity considerations, whereas more compli-
i. Berlrand Russell, Human Knowledge: Its Scope and Limits (Simon &
Schuster, New York, 1948), p. 488.
a. Russell was actually aware that this account of persistence is incomplete,
as I will bring out in Chapter 4, Section I. Moreover the notion of spatio-
temporal and qualitative continuity as this figures in Russell's account may
imply a condition of causal continuity. I ignore this latter condition, but only
provisionally, until Chapter 4, Section IV (see also ftn. 14 there).
3. C. D. Broad, Scientific Thought (Routledge & Kegan Paul, London, 1949),
p. 393. See also p. 346ff.
 CONTINUITY §

cated analyses which will be discussed later rely on continuity

considerations in conjunction with considerations of other sorts.
We might consider how the simple continuity analysis applies
to a statement like "Something was a red table at time t and a
green table at time t'." According to the analysis this statement
is true if and only if you could trace a succession S of object-
stages such that S contains a red table-stage at t and a green table-
stage at t', and S is spatiotemporally and qualitatively continuous.
In the recent literature on the topic of identity the concept
of a "space-time path" is frequently employed in such a way as
to take over the analytic role played in the preceding discussion
by the concept of a "succession of object-stages." It is useful to
be able to shift freely from one idiom to the other. We may de-
fine a space-time path as a series of place-times, i.e., a series of
ordered pairs (p, t), where p is a region of space and t is a moment
of time. To say that the space-time path P is spatiotemporally
continuous means that where (p, t) and (p', t') are place-times in
P then if t is very close to t', p is very close to p'. And to say that
P is qualitatively continuous means that where (p, t) and (p', t')
are place-times in P then if t is very close to t', the object which
occupies p at t exemplifies qualities at t which are very similar
to the qualities exemplified at t' by the object which occupies
p' at t'.
In the idiom of space-time paths the simple continuity analysis
tells us that we can correctly trace an object's career by follow-
ing a spatiotemporally and qualitatively continuous path
through space-time. Hence the statement "Something was a red
table at t and a green table at t'" will be true if and only if you
could trace a path P through space-time such that P contains a
red table at t and a green table at t', and P is spatiotemporally
and qualitatively continuous.
The notion of a "succession of object-stages" could also be
defined along the same general lines as a "path," though I would
expect, and want, the former notion to bear some independent
intuitive force. We could define an object-stage as an ordered
pair (x, t), where x is an object and t a moment of time. To say
that the succession S contains, say, a red table-stage at t means
that some object x is a red table at t and S contains (x, t). It
will be noted that a "path" and a "succession" are distinguish-
able in some strict abstract sense, but I shall not hesitate in what
IO THE CONCEPT OF IDENTITY

follows to use these notions interchangeably where it may seem

suggestive to do so.

II. Qualitative Continuity

Before turning to a criticism of the simple continuity analysis I

want to examine the two notions of continuity which occur in
it. These notions of continuity will also figure as parts of other
analyses which will be considered later, so it will be worth get-
ting clearer about them.
Let me first raise a question about the notion of qualitative
continuity. An object's career exhibits qualitative continuity in-
sofar as the object either does not change qualitatively at all or
undergoes qualitative changes which are continuous. An object's
qualitative changes are continuous if at any given time the object
is very similar to the way that it is at neighboring times. If we
define a "small change" as a change which takes an object from
one qualitative state to a different but very similar state then
we can say that a continuous qualitative change is a change
that can be thought of as divided up into a series of small
changes.
But now what exactly do we mean by this? We may mean
either (a) that a continuous qualitative change is a change that
can be thought of as divided up into a series of changes as small
as you like, or (b) that a continuous qualitative change is a
change that can be thought of as divided up into a series of small
changes, but not necessarily into a series of changes as small as
you like. Let me call sense (a) the strong sense of "continuous
qualitative change" and sense (b) the weak sense. My question
is whether to interpret the condition of qualitative continuity
in our analysis as requiring continuity of change in the strong
sense or in the weak sense.
The difference between the strong and the weak sense of con-
tinuity of change can be brought out with the following example.
Let us assume, as seems plausible, that if two cats are alike in
every respect except that one has brown eyes and the other has
green eyes then these two cats can be said to be "very similar."
This implies that a change in which a cat passes from having
brown eyes to having green eyes, everything else remaining ex-
 CONTINUITY 1I

actly the same, would count as a "small change." Now suppose

that we have a cat which has suffered such a small change dur-
ing the time interval from t to t'; that is, the cat at t' is green-
eyed whereas at t it was brown-eyed, but everything else has
remained the same. If we require continuity of qualitative change
in the strong sense then we would require that the small change
suffered by the cat from t to t' should be further resolvable into
still smaller changes; i.e., that there should be a time between
£ and t' such that the color of the cat's eyes at that time is inter-
mediate between brown and green. But if we merely require con-
tinuity of change in the weak sense then it is not necessary that
the small change suffered by the cat from t to t' should be further
resolvable. It would be permissible for the cat's appearance to
"jump" noticeably, so long as the jump is a small one.
The question that I am here raising should not be confused
with a different and more famous one. A change is continuous
in the strong sense if it does not involve any noticeable, albeit
small, jumps like that from being a brown-eyed cat to being a
green-eyed cat. By a noticeable jump I mean, intuitively, a direct
transition from one qualitative state to another, where we have
an idea of what it would be like for there to be a state inter-
mediate between the two. Now the famous question is whether
a change which is continuous in the strong sense, one which does
not involve a noticeable jump, necessitates there being an infinite
number of qualitative states intermediate between any two dif-
ferent qualitative states. Kant apparently thought that it does:
"The question . . . arises how a thing passes from one state . . .
to another. . . . Between two instants there is always a time, and
between any two states in the two instants there is always a dif-
ference which has magnitude. . . . The new state of reality ac-
cordingly proceeds from the first wherein this reality was not,
through all the infinite degrees."4
As I understand Kant, the question to which he is here address-
ing himself is this: Given that a particular qualitative change
from the state Sx to the state S2 was continuous in the strong
sense, i.e., that it did not involve any noticeable jumps, how
are we to understand the nature of the change? Kant's answer

4. Immanuel Kant, Critique of Pure Reason, translated by Norman Kemp

Smith (Macmillan, London, 1963), p. 231 (8253-254).
12 THE CONCEPT OF IDENTITY

consists in positing an infinite number of states intermediate

between S1 and S2. To Kant's question there may be other plaus-
ible answers.5 But I am not here concerned with this question. I
am not concerned with the question "What is the ultimate nature
of a change which is continuous in the strong sense?" but rather
with the question "Should we require of an object that its
changes should be continuous in the strong sense?"
I think that it is a fairly common assumption among philos-
ophers that all of the qualitative changes that are typically suf-
fered by ordinary physical objects are continuous in the strong
sense.6 It may therefore be of some theoretical importance to
show that there is a certain range of perfectly ordinary changes,
generally ignored in philosophical discussions of continuity,
which are not continuous in the strong sense. I have in mind
cases in which an object either persists while having a part added
to it, or persists while having a part subtracted from it. Fre-
quently when an object changes in either of these two ways it
suffers a qualitative change which is not continuous in the strong
sense.
Consider, for example, what happens when a branch is
chopped off a tree some time between t and t'. Suppose that the
tree took up 30 cubic feet at t and 28 cubic feet at t', the fallen
branch having taken up two cubic feet. Now we are imagining a
case in which the branch was chopped off as a whole, and not
demolished piece by piece. It follows that the tree suffered a
noticeable jump with respect to volume. For the tree passed from
taking up 30 cubic feet at t to taking up 28 cubic feet at t'. When
did it take up 29 cubic feet? Obviously never.
It would be a mistake to think that a plea of vagueness could
somehow be invoked to disarm this argument. Admittedly the
question "When exactly did the tree lose the branch?" cannot be
answered with any definiteness. Our ordinary thinking about

5. For a discussion of Kant's question see Jonathan Bennett, Kant's Analytic

(Cambridge University Press, London, 1966), pp. 176-80.
6. Such an assumption seems implicit in Kant's discussion. Russell's attitude
about this is unclear from the previously quoted passage but the strong sense
seems implied in his discussion of continuity in "The Relation of Sense-Data
to Physics," in Mysticism and Logic (George Allen & Unwin. Ltd., London,
1917). P. 170.
 CONTINUITY *3

the case of a tree losing a branch is too vague to permit our pin-
pointing the exact moment at which the branch was lost. But
this is beside the point, the point being that our ordinary think-
ing about the case does dictate a perfectly definite answer to the
question "When did the tree take up 29 cubic feet?" namely,
the answer "never." And from this it follows that the change of
volume was not continuous in the strong sense.
Much the same considerations serve to show that frequently in
cases of part-addition and part-subtraction an object's altera-
tions are not strongly continuous with respect to other qualities,
such as color and shape. The present example is not especially
well-suited to illustrate the point that a change of parts will fre-
quently require a noticeable jump with respect to color. But to
make the example work let us imagine that the bark had been
stripped from that one branch before it was chopped off, so that
at t every part of the tree's surface was brown except for that
branch. Then at t the tree was, let us say, go percent brown, and
at t' it was 100 percent brown. When was it 95 percent brown?
Again, never.
An equally good case can be made for saying that the tree
suffered a noticeable jump with respect to shape. Assuming that
the tree started out at t with two branches and ended up at t'
with one, the successive shapes of the tree at t and t' can be
represented by Figures i and 2.

Fig. i Fig. 2
Tree at t Tree at t'
14 THE CONCEPT OF IDENTITY

Clearly there was no time between t and V at which the tree's

shape could be pictured in some way intermediate between
Figure i and Figure 2. For example there was no time at which
the tree's shape looked like Figure 3.

Fig-3
Does not represent tree

It follows that the change from the shape pictured in Figure i

to the shape pictured in Figure 2 was not continuous in the
strong sense.
This was an example of part-subtraction. But evidently the
same considerations will apply with equal force to many cases
of part-addition (and also to cases of part-replacement, in which
parts are both subtracted and added). If, for example, you add a
wheel to a car, the car has to suffer a noticeable jump with re-
spect to volume, weight, shape, and (generally) color distribution.
I should make it clear that I am not maintaining that all cases
of part-addition and part-subtraction involve changes which are
not continuous in the strong sense. On the contrary, it seems
evident that many such cases involve no element of discontinuity.
If the branch of a tree is gradually consumed by fire, without the
branch splitting off from the tree, then this would typically be a
case of part-subtraction in which there is no evidence of a notice-
able jump; here the various qualities of the tree, its volume,
shape, and so on, change in a way that can easily be regarded as
continuous in the strong sense. A case of part-addition which
 CONTINUITY 15

does not apparently involve any discontinuity is that of a tooth

gradually appearing and growing in a creature's mouth; here
the creature suffers no noticeable jump with respect to any quali-
ties. In these latter two cases a part is subtracted by gradually
diminishing, or a part is added by gradually growing. But where
a part is subtracted as a whole, as in the case of a branch splitting
from a tree, or where a part is added as a whole, as in the case of
adding a wheel to a car, then there is no recourse but to acknowl-
edge a noticeable jump in an object's qualities. For an object's
qualities are a function of the object's total content, of the total-
ity of parts that make it up, and where an object gains or loses
a part as a whole, then its content, and hence its qualities, must
undergo some degree of discontinuous change.
The upshot of this discussion is that we must interpret the
condition of qualitative continuity in the simple continuity
analysis as requiring continuity in only the weak sense. Accord-
ing to the analysis, then, a necessary condition for a succession
S of object-stages to correspond to the career of an object is
that any object-stage in S should be very similar to any tem-
porally neighboring object-stage in S. This weak requirement
of continuity is exceedingly vague, as vague as the idea of two
qualitative states being "very similar" to each other. But it does
seem plausible to assert, vaguely, that in at least the most typical
and standard cases of persistence some weak degree of qualitative
continuity is to be expected. Whether even this weak kind of
continuity is properly to be regarded as strictly necessary to an
object's persistence is a question to which I will return.

III. Spatiotemporal Continuity

Much the same points as were just made with regard to quali-
tative continuity apply as well to the notion of spatiotemporal
continuity. It may not be immediately evident that cases of part-
addition and part-subtraction have any bearing on the continuity
of an object's movements in space. But when one considers that
an object's overall location in space is determined by the loca-
tions of its parts it becomes clear that where parts are added as
a whole or subtracted as a whole the object's overall location
must suffer some degree of discontinuous change. In what follows
when I speak about the place which an object occupies at a given
l6 THE CONCEPT OF IDENTITY

time I will always mean that overall region of space which coin-
cides with the object at that time, i.e., the region which contains
at that time all and only the object's parts.
An object's career exhibits spatiotemporal continuity (and
corresponds, therefore, to a spatiotemporally continuous succes-
sion of object-stages, or, in the alternative idiom, to a spatio-
temporally continuous space-time path) if it either does not move
at all or moves continuously. If we define a "small movement"
as one in which an object passes from occupying one place to
occupying a different but very close place, then we can define a
continuous movement as one which can be thought of as divided
up into a series of small movements. And here, as before, we can
distinguish between different degrees of continuity. A strong
kind of spatiotemporal continuity is exhibited by an object only
if its movements are divisible into movements as small as one
likes. However, a weaker kind of spatiotemporal continuity would
be exhibited so long as the object's movements are divisible into
small movements, even if some of these movements involve a
small "jump" from one place to a very close place.
One way that we might try to explicate the notion of two
places being "very close" is in terms of the notion of two places
overlapping each other, where two places are said to overlap
each other if they have some part in common. In these terms we
would say that x's career exhibits the weak sense of spatio-
temporal continuity only if the place p which coincides with x
at a given time t overlaps places which coincide with x at times
slightly before and after t. This does indeed seem to be the abso-
lutely minimal requirement for any kind of continuity of motion.
Accordingly, the following definition seems reasonable.
Definition A. "x's career exhibits spatiotemporal continuity in
the weak sense" means: For any time I in x's career there is a
time interval around t such that for any tr in that interval the
place which x occupies at t overlaps the place which x occupies
at t'.i
To explicate the strong kind of spatiotemporal continuity we
have to be able to make sense out of the idea that places which
7. This definition, and the ones which follow, would have to be slightly modi-
fied in an obvious way to take account of the first and last moments of x's
career.
 CONTINUITY 17

are very close to each other (i.e., which overlap) may be more or
less close. How shall we assign degrees of closeness to places
which overlap each other? We may note that where p and p' are
identical with each other (and are thus, in a sense, maximally close)
they will overlap completely. But if p and p' are overlapping places
that are not identical with each other then there must be, besides
an area of overlap, an area in which they do not overlap. That
is, if p is not identical with p' then there must be a part of p that
is outside p', or a part of p' that is outside p, or both. Perhaps
we can say, for our present purposes, that the degree of closeness
between p and p' varies inversely with the extent to which they
do not overlap. We might straightforwardly measure the extent
of nonoverlap between p and p' by first measuring, say in cubic
feet, the amount of p which is outside p', then measuring the
amount of p' which is outside p, and then adding these two
measures together. The strong kind of spatiotemporal continuity
would then be one in which the extent of nonoverlap between
x's place at t and x's place at V can be made as small as you like
by choosing t' close enough to t. This idea, or a mathematically
idealized version of it, might be expressed as follows.
Definition B. "x's career exhibits spatiotemporal continuity in
the strong sense" means: For any time t in x's career, and for
any positive number n, there is a time interval around t such
that for any t' in that interval the extent of nonoverlap between
the place which x occupies at t and the place which x occupies
at t' is less than n.
So we have these two notions of spatiotemporal continuity,
the weak notion of definition A and the strong notion of defini-
tion B. A mistake that I want to warn against, corresponding to
the mistake already discussed in connection with qualitative
continuity, is that of assuming that the strong notion expressed
in definition B correctly characterizes, or is at least an apt idealiza-
tion of, the paths typically traced by ordinary objects. The truth
seems rather to be that whenever an object has a part added to
it as a whole, or a part subtracted from it as a whole, then its
career cannot be coherently thought of as spatiotemporally con-
tinuous in the strong sense B.
If, for example, a branch falls from a tree it seems that, at
least for purposes of mathematical idealization, we must be able
l8 THE CONCEPT OF IDENTITY

coherently to think of there being a final moment at which the

tree included the branch. (An alternative would be to think of
there being no final moment at which the tree included the
branch, but there being a first moment at which it did not in-
clude the branch. My argument would work essentially the
same way on that assumption.) Suppose, then, that the tree in-
cluded the branch at t but not after t. Then for any time t' after
t, no matter how close t' is to t, the extent of nonoverlap between
the place occupied by the tree at t and the place occupied by it
at t' would have to equal the extent of the fallen branch. In other
words, we cannot make the extent of nonoverlap between these
places as small as we like by taking t' sufficiently close to t. This
means that the tree's career was not spatiotemporally continuous
in sense B. On the other hand the tree's career was certainly
spatiotemporally continuous in the weak sense A, because the
places successively occupied by the tree did overlap to some
extent.
Leaving aside for the moment the question whether even the
weak degree of spatiotemporal continuity is strictly necessary
for an object's persistence, it seems safe to say that this weak
kind of spatiotemporal continuity is characteristic of at least
the most typical and obvious cases of persistence. Now, actually,
we can make a considerably stronger claim than this. Typically
the successive places occupied by an object do not merely over-
lap to some extent or other (which is all that the weak notion
of spatiotemporal continuity requires), but overlap to a rela-
tively large extent. More specifically we may note that, at least
as a general rule, an object does not persist through the addi-
tion or subtraction of a part, where the part added, or sub-
tracted, is so large as to have the effect of precipitously doubling,
or halving, the object's size. This means that, as a general rule,
the places occupied by an object at neighboring times will be
such that their extent of overlap is greater than their extent of
nonoverlap. This suggests the possibility of formulating a defini-
tion of spatiotemporal continuity which is intermediate in
strength as between the weakest sense A and the strongest sense B.

Definition C. "x's career exhibits spatiotemporal continuity in

the moderate sense" means: For any time t in x's career there
 CONTINUITY ig

is a time interval around t such that for any t' in that interval
the extent of overlap between the place which x occupies at t
and the place which x occupies at t' is greater than their extent
of nonoverlap.

Definition C is by no means the only intermediate notion of

spatiotemporal continuity which could in principle be formu-
lated. We might formulate a notion which requires that an ob-
ject's successive places overlap, say, by more than two-thirds.
But we can perhaps focus on definition C as presenting at least
one rather reasonable-looking moderating possibility.
Now we know that the strongest sense B cannot accommodate
typical cases of part-addition and part-subtraction, whereas such
cases are accommodated by both the weakest sense A and the
moderate sense C. Either of these latter senses may therefore be
regarded prima facie as figuring in our identity criteria. Are
there any considerations which might indicate that one of these
senses of spatiotemporal continuity has more relevance to our
identity concept than the other? I can think of one kind of case
which might, with some plausibility, be interpreted as indicating
a special relevance for the moderate sense C. Sometimes when
an object divides into fragments the relative sizes of the frag-
ments constitute our only apparent basis for deciding which, if
any, fragment to identify with the original object. A very simple
explanation of these cases can be provided if we assume that the
moderate notion of spatiotemporal continuity is the operative
one.
As an example of the sort of case that I have in mind imagine
that at time t we have a large puddle of water, which we will
call puddle i. At time t' the puddle is split in two, so that after
t' we are confronted with two puddles, puddle aa and puddle
2b. Let us imagine that puddle aa is considerably larger than
puddle ab. In such circumstances it would probably seem natural
to identify puddle i with the larger fragment aa rather than
with the smaller 2b. We could, no doubt, try to account for this
in a number of ways, but a very simple explanation, in terms
of spatiotemporal continuity, presents itself if we adopt the
moderate sense C.
This case may also help us to get clearer about the application
2O THE CONCEPT OF IDENTITY

of the three defined notions of spatiotemporal continuity. The

weakest kind of spatiotemporal continuity would be satisfied
whether we identified puddle i with the larger fragment aa or
the smaller fragment ab. Even the latter identification would
secure weak spatiotemporal continuity because any place occu-
pied by puddle 2b immediately after the moment of split t'
overlaps the place occupied by puddle i at t'. At the other ex-
treme, the strongest kind of spatiotemporal continuity would
not be satisfied whether we identified puddle i with the larger
fragment 2a or the smaller fragment ab. Even the former identi-
fication would fail to secure strong spatiotemporal continuity
because there is an irreducible extent of nonoverlap between
the places occupied by puddle aa after t' and the place occupied
by puddle i at t'. Thus we see that neither the weak notion nor
the strong notion would prompt the seemingly natural identifica-
tion of puddle i with aa rather than with ab.
But the moderate notion has precisely the effect of prompting
the natural-seeming identification. The moderate kind of spatio-
temporal continuity would be satisfied only if we identified
puddle i with aa and not with ab. Either identification would
allow us to think of the successive places occupied by the puddle
as overlapping to some extent or other, but it is only the identi-
fication of the original puddle with the larger fragment which
would satisfy the moderate requirement that the places suc-
cessively occupied by an object should be such that their extent
of overlap exceeds their extent of nonoverlap. (The identification
of the original puddle with the smaller fragment fails to satisfy
this requirement because we would then be thinking of the
puddle as persisting through the loss of a part, where this part
is so large as to have the effect of precipitously decreasing the
object's size by half or more.) This sort of case might then en-
courage us, at least tentatively, to interpret the requirement of
spatiotemporal continuity, as it figures in our analysis, in terms
of some such moderate notion as that formulated in definition C.
There are two more questions about spatiotemporal continuity
that I want briefly to consider. First, should we say that spatio-
temporal continuity entails temporal continuity? This would
mean that in order for an object's career to exhibit spatio-
temporal continuity it would have to persist over a continuous
 CONTINUITY 21

stretch of time and not go out of existence and then come back
into existence. Second, should we say that spatiotemporal con-
tinuity entails spatial continuity? This would mean that an object
whose career is spatiotemporally continuous must, at any mo-
ment of its existence, occupy a continuous region of space and
not exist in a (macroscopically) fragmented form. (We might
define a continuous region of space as one in which any pair of
points can be connected by a continuous curve lying wholly
within the region. Thus an object with holes in it may occupy
a continuous region in the sense here intended.) I do not think
that the previous definitions of spatiotemporal continuity neces-
sarily settle whether spatiotemporal continuity entails temporal
or spatial continuity. Those definitions (or a slight modification
of them) might be construed as satisfiable by an object irrespec-
tive of its temporal or spatial continuity, so long as the (perhaps
discontinuous) regions of space which the object occupies at
successive moments of its existence (moments which are perhaps
separated by intervals during which the object does not exist)
overlap appropriately. I think, however, that we can simplify our
overall account if we make the terminological decision that both
temporal and spatial continuity are necessary factors in spatio-
temporal continuity. Of course we can always separate these
factors out if it proves necessary.
So, summarily, the condition of spatiotemporal continuity in
the simple continuity analysis is satisfied by a succession S of
object-stages if and only if: first, each object-stage in S coincides
with a continuous region of space; second, S spans a continuous
stretch of time; and, third, the places which coincide with tem-
porally neighboring stages in S overlap sufficiently (perhaps by
more than half).8
8. Saul Kripke has pointed out that this account of spatiotemporal continuity
has the counterintuitive consequence that an object can be said to move
continuously if it jumps instantaneously from one place to a neighboring
place in such a manner that the place it first occupies extensively overlaps
the place it later occupies. This objection suggests that the account I offer
may be only a first approximation, and that the notion of "sufficient overlap"
may ultimately need to be refined in terms that are not purely quantitative.
The notion of spatiotemporal continuity will be taken up again in Chapter
5, Section II, and Chapter 6, Sections V and VII, where the puzzling inter-
dependence between object-identity and place-identity will be considered.
22 THE CONCEPT OF IDENTITY

IV. Is Continuity Necessary?

Turning now to an assessment of the simple continuity analysis
I must stress again that I am going to treat this analysis, as well
as the ones that will subsequently replace it, strictly from the
standpoint of our most ordinary conception of physical per-
sistence. These analyses, as I want to understand them, are at-
tempts to describe that ordinary conception. What we are seeking
at present is nothing more than an accurate description of our
ordinary identity scheme. The possibility, or desirability, of
revising that scheme is not now at issue, though this is a topic
that I will eventually address.
According to the simple continuity analysis the two conditions
of qualitative and spatiotemporal continuity are jointly neces-
sary and sufficient for a succession of object-stages to correspond
to stages in the career of a persisting object. Let us first consider
the necessity part of this claim. Is it true that qualitative and
spatiotemporal continuity are necessary for an object's
persistence?
We should note that there is one trivial and quite irrelevant
sense in which the continuity of a succession of object-stages
cannot possibly be necessary for it to correspond to stages in an
object's career. Suppose that S is a continuous succession which
corresponds to stages in the career of some object x. Then we
can certainly form a discontinuous succession S' by picking out
disjointed portions of S, e.g., all of x's Monday-stages. S' would
then be a discontinuous succession which could be said in a
sense to correspond to (some) stages in the career of the single
persisting object x. But this is obviouly not the sort of case we
are thinking about. When we say that a succession "corresponds
to stages in the career of an object" (or, for short, "corresponds
to a career," or "corresponds to an object"), we are always to
understand this to mean that the succession corresponds to, what
might more properly be called, the successive stages of an ob-
ject's career. This is to be understood as implying that, though
the succession may not correspond to any object's whole career
(from beginning to end), it must at least comprise all of the
stages of an object's career from some moment to some other
moment. Understood in this sense our question is not trivial: Is
it true, as the analysis claims, that a succession must be con-
 CONTINUITY 23

tinuous for it to correspond to the successive stages of an object's

career?
There is at least one kind of case which seems to show rather
decisively that this is not true. Many objects are thought of as
retaining their identities after being taken apart and put back
together again. But the careers of such objects must, it seems,
be radically discontinuous.
Let us imagine, for example, that a watch w is sent to the fac-
tory for repairs. The watch is completely disassembled on the
first floor of the factory and its various parts are sent to different
rooms on different floors of the factory. Eventually, we will
imagine, all of these same parts (perhaps oiled) are collected
on the second floor where they are put together to form the
watch w'. In such a case there seems to be no doubt that w' is
the same watch as w.
Though there is quite definitely an outright lapse of continuity
in this case it may not be entirely clear just what sort of con-
tinuity is lost. When the watch is initially disassembled and its
parts are still relatively proximate (e.g., they are all lying on
the same work table) there may be an inclination to say that
the watch still exists but in a fragmented form. If we say this
then we would so far forgo only spatial continuity but not yet
temporal continuity. Eventually, though, when the watch's
parts are already dispersed throughout the factory there is, I
think, no serious inclination to say that the watch still exists
(and that a single watch is simultaneously touching the third
floor and the tenth floor of a factory). It seems that we must say
that at some point the watch went out of existence, and then
later it came back into existence. So temporal continuity seems
certainly lost.
It might perhaps be argued that qualitative continuity is not
lost in this case, on the grounds that at the moment that the
watch comes back into existence it will be qualitatively very
similar to the way that it was at the moment that it went out
of existence. However we might settle this question about quali-
tative continuity (and this might depend on the details of the
case), it is clear that we cannot say, in the case as I described it,
that at the moment when the watch first comes back into existence
it occupies a place very close to the place it occupied at the
moment when it went out of existence. For the watch went out
24 THE CONCEPT OF IDENTITY

of existence on the first floor and came back into existence on

the second floor. Thus spatiotemporal continuity is very defi-
nitely lost (even apart from questions about spatial and temporal
continuity), because it is not true that the places successively
occupied by the watch overlap.
Details aside, it seems clear enough that in a case like this,
where an object is taken apart and put back together again,
continuity is lost. It follows that the simple continuity analysis
must be complicated at least to the extent of accommodating
such cases. Actually these cases do not seem very central (stand-
ard, paradigmatic), and the complication which they introduce
seems fairly superficial. Apparently the two continuity criteria
mentioned in the analysis need to be supplemented by, what I
will call (following Quinton),9 a compositional criterion. The
compositional criterion would allow us to say that x is identical
with y, even if the continuity criteria do not apply, so long as
all (or perhaps most) of PC'S parts (or perhaps x's major parts)
are identical with y's.
I will discuss this compositional criterion more fully in the
next chapter. For the moment I want to stress the point that
the compositional criterion, as it is here being treated, is by its
very nature merely a supplement to other criteria upon which
it must be dependent. The compositional criterion would allow
us to say that x is identical with y when we can say that the
parts of x are identical with the parts of y. But then we need
criteria in terms of which to analyze the latter identity judg-
ments about the parts of x and y. Compositional considerations
are thus dependent upon more primary considerations on the
basis of which compositional identity can be understood. As
far as our discussion now stands it may still be that these primary
considerations are exhausted by qualitative and spatiotemporal
continuity.
There emerges the general format of a somewhat more com-
plicated kind of analysis than the simple one so far considered.
Any analysis of persistence must start out by mentioning certain
primary noncompositional criteria as being (in some specified
combination) sufficient for an object's identity. Presumably

9. Anthony Quinton, The Nature of Things (Routlcdge & Kegan Paul, London
and Boston, 1973), p. 69. See also p. 6gff.
 CONTINUITY 25

these criteria will suffice for the most standard cases. The analy-
sis must then also provide for the application of the supple-
mentary compositional criterion in exceptional cases (such as
where an object is taken apart and put back together again).
The general condition which the analysis would imply to be
both necessary and sufficient for an object's persistence is that
either the (proper combination of) primary criteria apply or the
compositional criterion applies. We have seen that the simple
continuity analysis errs insofar as it ignores the compositional
criterion. The more fundamental question that remains is
whether that analysis at least gives a correct account of our
primary criteria. To say that it does would be to imply, at the
least, that qualitative and spatiotemporal continuity are jointly
sufficient for an object's persistence. Whether this is so is the
next question that I want to consider.

V. Is Continuity Sufficient?
According to the simple continuity analysis a sufficient condi-
tion for a succession S of object-stages to correspond to stages
in the career of a single persisting object is that S be both
qualitatively and spatiotemporally continuous. Is the analysis
correct in this respect? Let me first briefly consider one rather
obvious but, I think, relatively superficial demonstration that,
contrary to the analysis, continuity is not sufficient for persist-
ence. To say that continuity is sufficient for persistence suggests,
at least on the most obvious interpretation, that the career of
any given object is to be prolonged so long as tracing a con-
tinuous path allows. Now the point is often made that we some-
times judge one object to go out of existence and to turn into
(to be replaced by) a second object which comes into existence.
We make this sort of judgment in cases where a perfectly con-
tinuous path connects the terminal stages of the first object to
the initial stages of the second object. If continuity were in fact
a sufficient condition for persistence there would apparently be
no basis for our making this kind of judgment.
Here is an example. There is a machine which functions to
crush old cars until they turn into blocks of scrap metal. If a
car undergoes this process it is presumably correct to say that,
at some point in the process, the car went out of existence and
26 THE CONCEPT OF IDENTITY

was replaced by a block of scrap metal. But the transition from

the car-stages to the scrap-stages was completely continuous, both
qualitatively and spatiotemporally. Hence if we relied only on
continuity considerations we would have to identify the car
with the block of scrap metal and judge, not that the car went
out of existence, but rather that it persisted in a thoroughly
crushed form. This shows that continuity by itself is not sufficient
for persistence.
There are many other cases like this. A log which burns in
the fire may go out of existence and turn into a smoldering ash,
though there was no discontinuity in the process. A gold coin
may be destroyed by continuously melting it down. A table may
be continuously filed down until it no longer exists and is re-
placed by a mere lump of wood. None of these cases can be ex-
plained by reference to continuity considerations alone. In
general terms the objection is that the simple continuity analysis
cannot properly account for our judgments about one object
going out of existence and being replaced by another.
Now I think that this objection to the simple continuity
analysis is correct so far as it goes. The trouble is that the ob-
jection does not go nearly far enough, and consequently the
tendency among some philosophers to belabor it has the effect
of obscuring the really fundamental difficulty with the analysis.
To raise this objection gives the impression that continuity does
generally suffice for persistence, except in those rather tricky
and special cases in which we judge one object to go out of
existence and turn into another. The mistaken view that is
readily fostered (and which is, I believe, rather widely enter-
tained) is that simple continuity criteria do in fact give us rnore-
or-less all that we need for our ordinary identity judgments, but
these criteria need to be supplemented in some way to deal with
the tricky notion of one object turning into another. But I intend
to show that simple continuity criteria never suffice, that these
criteria do not give us even remotely what we need for our
ordinary identity judgments.
Let me try to clarify this point by drawing a distinction be-
tween two kinds of cases in which a judgment deviates from our
ordinary identity concept. (In talking of a "deviation" from the
ordinary concept I leave open, for the time being, whether
 CONTINUITY 27

such deviations may be in some sense legitimate or even desir-

able.) I will distinguish between drastic and nondrastic devi-
ations. An example of a nondrastic deviation would be one in
which someone judged the car which entered the crushing
machine to be identical with the block of scrap metal that
emerges. This judgment, I assume, deviates from our ordinary
conception since the car has, properly speaking, gone out of
existence. Nevertheless there is a quite obvious sense in which
this judgment is at least on the right track. The car is not strictly
identical with the block of scrap metal, but the car did at least
turn into the block of scrap metal. This deviation is somewhat
subtle (not to say controversial); certainly it is by no stretch of
imagination bizarre. A nondrastic identity-deviation is then one
in which someone judges x to be identical with y where the
strict truth (in ordinary terms) is that x went out of existence
and turned into y (or perhaps x went out of existence and turned
into something which went out of existence and . . . turned into
y), or vice versa.
By a drastic identity-deviation I mean one in which x is judged
to be identical with y where the truth (in ordinary terms) is
that x did not even turn into y (or turn into something which
turned into something which . . . turned into y), or vice versa.
A drastic identity-deviation (when it is made in optimal condi-
tions of observation) is likely to strike us as completely bizarre
and off the track. An example would be if someone observes an
ordinary moving car and judges that the car which exists at one
moment is identical with the back fender that existed at a
previous moment.
The earlier objection, which focused on the rather special
(and perhaps controversial) cases in which one object turns into
another, left the impression that, these cases aside, simple con-
tinuity criteria operate effectively. This implies that someone
who relied on simple continuity considerations would thereby
avoid any drastic deviations from our ordinary identity judg-
ments, though he might still be led occasionally to a nondrastic
one. But I intend to show now that someone who relied on
simple continuity considerations would frequently (indeed as
often as not) be led to drastic identity-deviations. This is a much
deeper indictment of the simple continuity analysis, since it is
28 THE CONCEPT OF IDENTITY

tantamount to showing that continuity considerations by them-

selves are totally ineffective as criteria for our ordinary identity
judgments.
The reason why simple continuity considerations are totally
ineffective is that there are literally innumerable space-time paths
(successions of object-stages), constantly encountered by us in our
experience, which, though they are spatiotemporally and quali-
tatively continuous, do not correspond even remotely to the
careers of any objects, as ordinarily conceived. This point seems
completely obvious the moment it is grasped. It requires, how-
ever, thinking in unaccustomed ways about space-time paths
which, from the point of view of our ordinary identity concept,
seem wholly strange and unreal. But unless we can force our-
selves to focus on such strange-seeming space-time paths our
analysis of identity will merely presuppose what it pretends to
explain.
Let me first describe schematically the sort of paths that I have
in mind; then I will give examples. If you consider any object
it will always be possible to trace an indefinite number of
qualitatively and spatiotemporally continuous space-time paths
which connect the whole object at one time to one of its parts
at a later time. Furthermore if you consider any part of any
object it will always be possible to trace an indefinite number
of qualitatively and spatiotemporally continuous space-time
paths which connect that part of the object at one time to some
other part of the object at another time. Yet there will generally
be no persisting objects, as ordinarily conceived, which corre-
pond to these paths; there is, in terms of our ordinary concep-
tion, generally no persisting object which combines stages of a
whole object with stages of its parts, or which combines stages of
one part of an object with stages of other parts. Hence, side
by side with the career of a whole and the careers of its parts
the simple continuity analysis would generate a menagerie of
pseudo-careers made up of scattered stages of the career of the
whole and the careers of its parts. Reliance on simple continuity
considerations would consequently yield the most drastic and
bizarre identity-deviations.
Let me try to broach this idea by way of the following example.
Imagine that a tree persists intact during the two day period
from Monday to Tuesday, and that nothing out of the ordinary
 CONTINUITY 29

happens to the tree during that period. Let S^ be the succession

of object-stages corresponding to the tree's career during those
two days. The tree of course will have a trunk. To be vivid
about this let us in fact imagine that this is a one-branched tree
consisting of nothing but a trunk and one branch. And let S2 be
the succession which corresponds to the trunk during that two
day period.
Now comes the somewhat weird part. I want to consider the
succession S3 which consists of the Monday-portion of St followed
by the Tuesday-portion of S2. S3, in other words, is a succession
which consists of the tree-stages of Monday followed by the
trunk-stages of Tuesday.
Let us consider whether S3 corresponds to the career of any
object, as ordinarily conceived. Very evidently it does not; S3
is in fact a mind-boggling path which we can barely get ourselves
to think about. In tracing S3 we have to follow the tree's career
on Monday and then suddenly jump on Tuesday to the trunk.
Remember that we are imagining a case in which nothing special
happened to the tree (and, specifically, the tree did not lose its
branch). In such a case there is, in terms of our ordinary con-
ception, no persisting object remotely corresponding to S3. If we
did try to think of an object corresponding to S3 then we would
have to say that this object coincided with the whole tree on
Monday and then shrunk in size so that it coincided with only
the trunk on Tuesday. In the circumstances that we are imagin-
ing there is (within the limits of our ordinary identity concept)
obviously no object which remotely fits this description.
But is S3 continuous? Well, S3 is certainly not strongly contin-
uous since it involves an element of discontinuity, both qualita-
tive and locational, in the jump from the tree on Monday to the
trunk on Tuesday. But we have already seen that strong con-
tinuity is not what the simple continuity analysis requires.
Moreover we saw earlier that the small element of discontinuity
suffered by a tree when it loses a branch does not disqualify its
career from exhibiting the degree of continuity required by the
analysis. In the case we are now imagining the jump in S3 from
the one-branched tree of Monday to the trunk of Tuesday seems,
on the face of it, to involve just that small and nondisqualifying
element of discontinuity. So it seems that S3 is sufficiently con-
tinuous. Hence S3> on the simple continuity analysis, ought to
gO THE CONCEPT OF IDENTITY

correspond to the ordinary career of an object. But S3 obviously

does not correspond to the ordinary career of an object.
It is, by the way, an unimportant feature of our example that
53 is a path which involves change, whereas Sj and S2 do not
involve change. For we might in fact imagine that the trunk's
(and tree's) color or size altered during that two day period, so
that Sj. and S2 also involve change. Indeed we might imagine
that some piece of bark fell off the trunk (and tree) on Monday,
so that S± and S2 even involve an element of discontinuity. All
that matters in our example is that the tree did not lose its
branch during those two days. And the difficulty is that whereas,
in terms of our ordinary identity concept, St and S2 correspond
to the careers of objects, S3 does not, though in point of con-
tinuity there seems to be no decisive difference between these
paths.
Imagine someone who stands in front of a one-branched tree,
to which nothing special is happening, in optimal conditions
of observation, and who makes the following statement: "There
is an object here that is now a trunk, and that same object was
larger a moment ago when it was a whole tree." That would
be what I call a drastic deviation from our ordinary identity
concept. But the simple continuity analysis does not even explain
why (how) this is a deviation.
We can now easily imagine even more complicated and bizarre
possibilities. We can consider the succession S4 which consists
of the tree-stages of Monday, followed by the trunk-stages of
Tuesday, followed by the tree-stages of Wednesday. Apparently
54 is again sufficiently continuous, and it therefore ought, on the
analysis, to correspond to an object which first got smaller and
then got larger. Evidently there is no limit to the kinds of
pseudo-careers that the analysis would generate in this fashion.
Someone might try to resist this point by arguing that S3 and
S4 are somehow not sufficiently continuous. There would be no
purpose in thrashing this out since we can simply change the
example slightly to suit this critic. Imagine that our tree con-
tains a very tiny twig at the end of its branch. Now consider
the portion of wood W which constitutes the whole tree except
for that tiny twig. Instead of S2 let us now refer to the succession
S2' which consists of the stages of W on Monday and Tuesday.
And instead of S3 let us construct S/ which consists of the
 CONTINUITY 31

Monday-portion of Sj^ (i.e., the tree-stages of Monday) followed

by the Tuesday-portion of S2'. In tracing S3' we follow the tree's
career on Monday and then jump on Tuesday to the tree minus
that little twig. Certainly this minute jump is not great enough
to disqualify S3' from being sufficiently continuous. But S/ does
not correspond to a persisting object, in terms of our ordinary
identity concept, any more than S3 did. It would still be a drastic
and wholly bizarre deviation from the ordinary conception if
someone judged, where nothing has happened to the twig, that
a tree has shrunk a little bit and no longer contains some twig
which it previously contained. But it seems quite definite that
the simple continuity analysis implies that S/ does correspond to
the career of a persisting object.
As before there are no limits to the complicating possibilities.
If we can trace a continuous path from the whole tree to the
tree minus a twig then we can trace a path back to the whole
tree again. And if we can trace a continuous path from the whole
tree to the tree minus one twig then we can surely go on from
that point to trace a continuous path to the tree minus two
twigs; and then eventually to the tree minus the branch; and
then back again; and so on. The general point is that we can
always move by continuous gradations from any object to any
of its parts, and from any of its parts to any other part. Con-
sequently if we relied on simple continuity considerations we
would have no tracing rule at all, no basis at all for judging, in
anything like ordinary terms, what is identical with what. This is
the fundamental and drastic inadequacy of the simple con-
tinuity analysis, by comparison with which any other inadequacy
seems scarcely worth mentioning.
I want to underscore this very central point about the insuf-
ficiency of continuity by considering another example, and in
a somewhat different light. We can sometimes clarify our under-
standing of how our language works by finding or constructing
a radically disparate language and then reflecting on what the
difference is between that language and our own. In that spirit
I want now to construct a language (or, really, some small seg-
ment of a language) which contains identity criteria radically
different from our own. When we reflect on what the difference
is between those criteria and ours we will be helped to appreci-
ate the total ineffectiveness of simple continuity considerations.
g2 THE CONCEPT OF IDENTITY

Consider, then, a language in which the word "car" is replaced

by the two words "incar" and "outcar." The criteria of identity
for incars and outcars can be indicated as follows. The term
"incar" will apply to any car that is entirely inside a garage,
and where a car is partly inside and partly outside a garage,
"incar" will apply to the segment of the car that is inside; cor-
relatively, "outcar" will apply to any car that is entirely outside
a garage, and where a car is partly inside and partly outside,
"outcar" will apply to the segment outside. When (as we would
say) a car moves from inside a garage to outside, the description
in that other language would be: "An incar moved towards the
exit whereupon it commenced to shrink in size until it even-
tually vanished; simultaneously with the shrinking of the incar
an outcar appeared at the outside of the exit, and gradually
grew until it attained the size and form of the original incar."
In this description the original object inside the garage (the ob-
ject which coincides with what we would call a car) is traced
in such a way as to render it identical with what is later a smaller
object inside the garage, and distinct from any object that is ever
outside the garage.
This language strikes us as very strange. I am not at the mo-
ment concerned with the question whether this language is
strange (or bad or wrong) in some absolute sense, or merely
relative to our conventional way of talking. I am not, that is,
concerned now with the question whether there could actually
be people who spoke like that, or whether, if there were such
people, their description of moving cars would be, in some
sense, less correct than ours. The crucial point for my present
purpose is that this language is, at the very least, strange relative
to our conventional way of talking. Clearly the incar-outcar
identity criteria deviate from our ordinary identity criteria. It
would certainly be a mistake for someone speaking our language
to identify what was first a whole car in a garage with what was
later a small portion of that car inside the garage. It is certainly
incorrect, at least in our language, to assert, in the ordinary cir-
cumstance of a car leaving a garage, that an object shrank in
size and vanished.
But why would this be a mistake in our language? What or-
dinary criteria of identity would be violated by tracing an ob-
ject's career along the path of the shrinking incar? Very evidently
 CONTINUITY 33

not continuity criteria. For it is strikingly obvious that if we

were to trace the path of the shrinking incar we would be trac-
ing a path that is perfectly continuous, both qualitatively and
spatiotemporally. Considerations of mere continuity do not give
us the slightest clue as to why it is that the path of the shrinking
incar does not, in our language, correspond to the career of an
object. The wrongness, indeed the strangeness, of the incar-outcar
criteria, from the point of view of our ordinary identity concept,
is thus a vivid indication of the ineffectiveness of simple con-
tinuity considerations to explain the nature of our identity
concept.
An absolutely minimal condition of adequacy for any analysis
of our identity concept is that the analysis imply that it would
be wrong to trace an object in accordance with identity criteria
of the incar-outcar variety. It is evident that these criteria are
merely one instance of a wholly general kind of aberrant criteria
(aberrant, that is, relative to our ordinary identity concept).
These criteria are aberrant insofar as they would permit us to
trace an object in such a way as to combine what (in our lan-
guage) ought properly to be regarded as stages of the object and
stages of some of its parts. A minimal condition of adequacy for
any analysis of our identity concept is that it at least imply the
wrongness of such drastic whole-part tracing confusions. The
fundamental error of the simple continuity analysis is that it
does not even satisfy this minimal condition.
 2
Sortals

I. The Sortal Rule

A COMPLETELY adequate analysis of persistence would have
to take account of complications involving the fact that objects
can be taken apart and put back together again, and also com-
plications involving the fact that one object sometimes turns
into another. But these are relatively peripheral problems which
pertain to cases that are not entirely typical. The more funda-
mental problem is to be able at least to characterize properly
those successions which correspond to persisting objects in the
most unexceptional cases. I want first to concentrate on this
problem and then return to the residual complications after-
wards.
We know that continuity considerations by themselves do not
suffice to rule out such aberrant paths as the one which com-
bined stages of a tree with stages of a trunk, or the one which
combined stages of a car with stages of car-parts. We want to
elicit some additional constraint which can be seen as operating
in our ordinary identity concept to rule out such paths. Now
one intuition that we may have about such paths, and why they
seem aberrant from the ordinary point of view, is that tracing
these paths involves some kind of illicit shift, some kind of rule-
violating loss of constancy. How can we characterize the nature
of this illicit shift?
Of course we cannot simply say that tracing an aberrant path
involves shifting from one object to another, since precisely

34
 SORTALS 35

what we are looking for is an analysis of what it is that consti-

tutes staying with the same object. Nor will it do, at the present
stage of analysis, to say merely that the aberrant paths involve
shifting from a whole object(-stage) to a part of an object(-stage).
For "whole" and "part," in at least one obvious sense, are rela-
tional terms that do not necessarily exclude each other. A whole
twig may be a part of a whole branch, which may in turn be a
part of a whole tree (which may, perhaps, in turn be a part of a
whole landscape). Evidently to try to refer simply to an illicit
shift from "wholes" to "parts" would not get us very far.
When one reflects upon the problem in these terms an idea
which is likely to suggest itself, and which will in fact turn out
to be very much worth developing, is that an illicit shift occurs
in the aberrant paths insofar as these paths combine an ob-
ject(-stage) of one sort with an object(-stage) of another sort. An
object may of course change in the course of its career, both
qualitatively and locationally, and these changes, so long as
they are continuous, may even be quite drastic. But the present
suggestion is that it is part of our concept of object-identity that
throughout all of its changes an object must at least remain
an object of the same sort. The constraint, therefore, which
a succession must satisfy, in addition to the simple continuity
conditions, in order for it to correspond to the career of an
object is that it consist of object-stages of the same sort. The
idea would then be that this constraint is not satisfied by such
aberrant paths as the one which combines tree-stages with
trunk-stages, or the one which combines car-stages with stages
of car-parts.
One difficulty which this suggestion immediately faces is to
explain what is meant by saying that two objects (or two object-
stages) are of the same "sort." We want to wind up saying that
the succession which combined car-stages with stages of car-parts
does not correspond to a persisting object in our language because
it involves combining object-stages of different sorts. But consider
that from the point of view of the incar-outcar language all of
these object-stages are mcar-stages, which would seem to imply,
perhaps, that from that point of view these stages are of the
same sort. Should we then say that whether or not one ob-
ject(-stage) is the same sort as another depends upon which
language we speak? This may be a helpful first move, and is
g6 THE CONCEPT OF IDENTITY

anyway quite harmless so long as we keep open, as before, the

possibility that some languages may be, in some sense, "better"
than others.
Recasting our earlier idea in more explicitly linguistic terms
let us say that such words in our language as "car," "tree," and
"trunk" are sortal terms. The constraint which a succession
must satisfy in order for it to correspond to a persisting object
(in our language) is that there be (in our language) a sortal term
F such that every object-stage in the succession comes under
(is an instance of) F. Hence a continuous succession of tree-stages
corresponds to the career of a persisting tree, because all of the
object-stages in this succession come under the sortal "tree";
and a continuous succession of trunk-stages corresponds to the
career of a persisting trunk, because all of the object-stages in
this succession come under the sortal "trunk"; and a continuous
succession of car-stages corresponds to the career of a persisting
car because all of the object-stages in this succession come under
the sortal "car." But the succession which combined Monday's
tree-stages with Tuesday's trunk-stages does not correspond to
any persisting object because there is no single sortal term F
which covers all of the object-stages in this succession. (Notice
that though a trunk is part of a tree a trunk does not, in the
relevant sense, come under the sortal "tree"). Again, the path of
the shrinking incar does not correspond to a persisting object
(in our language) because there is no single sortal (in our lan-
guage) which covers all of the object-stages in this path.
The analysis which we are trying to develop might now be par-
tially formulated as follows:
The Sortal Rule. A sufficient condition for the succession S of
object-stages to correspond to stages in the career of a single
persisting object is that:

(1) S is spatiotemporally continuous; and

(2) S is qualitatively continuous; and
(3) there is a sortal term F such that S is a succession of F-stages.
The sortal rule states only a sufficient, not a necessary, condi-
tion of persistence, because we want to leave room for the com-
positional criterion (and perhaps other elaborations of the rule).
We may summarize the rule by saying that it permits us to trace
 SORTALS 37

an object's career by following a continuous space-time path

under a sortal.1
We have still to clarify what it means to say that a term is a
"sortal." We certainly cannot say that every general term in our
language is a sortal since this would immediately defeat the
entire rationale of the sortal rule. If we said that "brown (thing)",
for example, is a sortal (so that any two brown things are, in the
relevant sense, things of the same sort), then the sortal rule
could no longer be relied on to disqualify the succession which
combined the Monday tree-stages with the Tuesday trunk-stages.
Such a succession might very well consist of only brown object-
stages, and would hence qualify under the sortal rule if "brown"
were a sortal. Or suppose that the term "in a garage," which
applies to any object in a garage, were counted as a sortal. Then
the sortal rule could no longer disqualify as aberrant the suc-
cession which combines stages of a car in a garage with stages
of car-parts in a garage.
We want to be able to say that such terms as "tree," "trunk,"
and "car" are sortals, but that terms like "brown" and "in a
garage" are not. Of course we immediately notice that the
former terms are nouns and the latter are not. But this purely
grammatical difference, though it may approximately coincide
with the logical (conceptual) distinction which we are seeking
to clarify, does not explain what that distinction is.
It is a mistake, however, to suppose that we need to be able to
explain what it means for a term to be a sortal before we can
understand the sortal rule. The proper way to look at this, rather,
is that the sortal rule itself defines (constitutes) what it means
for a term to be a sortal. A term is a sortal just in case the sortal
rule would allow us to trace a career under the term. A sortal
is thus a term which plays a distinctive role in our identity
conception, and to learn to speak our language involves finding
out just which terms play this role. A definition of "sortal"
might then be:
"The general term F is a sortal" means: It is a conceptual truth
(a rule of language) that any spatiotemporally and qualitatively

i. Cf. Wiggins, Identity and Spatio-Temporal Continuity, p. 35ff. The notion

of a sortal that I intend to develop in this chapter is, I think, essentially the
one employed by Wiggins.
38 THE CONCEPT OF IDENTITY

continuous succession of F-stages corresponds to (what counts

as) stages in the career of a single persisting F-thing.

Many typical nouns seem to qualify as sortals on this defini-

tion; for example, "tree," "trunk," "branch," "car," "fender,"
"dog," "eye," "mountain," "pebble." Many adjectives and verbs
do not qualify; for example, "red," "hot," "hard," "wet," "moves,"
"burns," "grows." But this grammatical test is by no means de-
cisive, even for the case of syntactically simple terms (which is
the only case to which the test might reasonably apply). A rather
obvious example of a noun which is not a sortal is "object," at
least when we take this word in the broad sense that I have
been using it. If "object" were counted as a sortal then it would
follow from the sortal rule that any continuous succession of
object-stages corresponds to a career, which is precisely the mis-
take which the rule is designed to avoid. I will consider other
examples of nouns which are not sortals presently. An adjective
like "canine," in the sense of "dog" (if there is such a sense),
would evidently be as much a sortal as the noun form. There
may even be some sortal verbs, but I do not know of any very
convincing examples.
When we consider syntactically complex terms we find that
any (conjunctive) combination of a sortal and a nonsortal yields
a sortal. Take "brown car," which combines the sortal "car"
with the nonsortal "brown." This complex term qualifies as a
sortal on the definition because, if it is a conceptual truth that
any continuous succession of car-stages corresponds to stages in
the career of a single persisting car, then it must also be a con-
ceptual truth that any continuous succession of brown car-stages
corresponds to stages in the career of a single persisting brown
car. It should be noted, however, that the sortalhood of "brown
car" does not imply that the terminus of a continuous succession
of brown car-stages corresponds to a brown car going out of
existence, for the career of a brown car whose color changes can
be prolonged under the more general sortal "car." I will return
to this complication about going out of (and coming into) ex-
istence in a later section.
Many (though not all) nonsortals have an important property
which may be seen as explaining why they cannot properly
function as sortals. The property is that these terms apply to
 SORTALS 39

objects that extensively overlap each other. Consider, for ex-

ample, what would happen if we tried to treat "brown (thing)"
as a sortal. We would then trace the career of a brown thing by
following a continuous succession of brown-stages. But this would
lead to precisely the tracing chaos which showed up in connec-
tion with the simple continuity analysis.
Remember the tree whose career was traced from Monday
through Tuesday. Let us imagine now that the tree is uniformly
brown. Sj was the succession of tree-stages during those two
days; S2 the succession of trunk-stages; S/ the succession of stages
of the tree minus a little twig; and S3 was the weird succession
which combined the Monday-portion of Sj with the Tuesday-
portion of S2, while S/ was the weird succession which com-
bined the Monday-portion of Sj with the Tuesday-portion of S/.
All of these are apparently continuous successions of brown-
stages. If we treated "brown" as a sortal we would have to count
each of these successions as corresponding to the career of a
single persisting brown thing. Moreover there is no limit to the
number of different successions of this sort that could be fabri-
cated by combining different stages of the mentioned successions.
Well, someone might ask, so what? If we did treat "brown"
as a sortal then we simply would count all of those overlapping
and crisscrossing successions as different careers of persisting
brown objects. What would be wrong with that?
I do not want to say that there necessarily would be anything
"wrong" with it. My present point is only that treating "brown"
as a sortal would lead to a certain consequence which seems
deeply inconsonant with the general tendency of our thought
and speech about persistence, and this explains, in a sense, why
"brown" is not treated as a sortal. It seems central to the way
that we think and speak about persistence that we should typi-
cally be able to pick out an object and go on to trace its career
unambiguously along some definite space-time path. But if we
picked out a brown object, say a brown tree, and tried to trace
its career under the covering concept "brown" we would not
be led unambiguously along any particular path, since we could
go on to trace any number of different continuous successions of
brown-stages.
The property of "brown" which renders it unsuitable to play
the role of a sortal is that brown objects (e.g., the brown tree
4<J THE CONCEPT OF IDENTITY

and the brown trunk) may overlap extensively (where one way
that objects can overlap extensively is for one of them to be a
large part of the other). I will say that any such term is dispersive,
where this is defined as follows:
"F is dispersive" means: It happens typically that different F-
things overlap extensively.
I want to leave this definition quite vague. The general idea is
that if F is dispersive it will frequently be possible to trace con-
tinuous paths which combine stages of one F-thing with stages
of another extensively overlapping F-thing. It will frequently
happen, therefore, that two continuous F-successions will partly
coincide and partly diverge, i.e., that they will contain the same
object-stages at one moment but different object-stages at another
moment. In our previous example Sj (which contains the tree-
stages of Monday and Tuesday) and S3 (which contains the tree-
stages of Monday and the trunk-stages of Tuesday) coincide on
Monday and diverge on Tuesday, while S2 (which contains the
trunk-stages of Monday and Tuesday) and S3 diverge on Mon-
day and coincide on Tuesday. It is the partial coincidence and
partial divergence of continuous F-successions which makes it
impossible to trace careers unambiguously under a dispersive
term F.

II. The Making of a Sortal

I am suggesting, then, that nondispersiveness is a necessary
condition for a term to be a sortal, but this is certainly not a
sufficient condition. I am not sure whether there are any syn-
tactically simple nondispersive terms which are not sortals, but
there is no trouble concocting complex terms which are both
nondispersive and nonsortals. Consider, for example, the term
"largest portion of a car in a garage," where this is understood
to apply to any whole car in a garage or, if the car is partly
inside and partly outside, to the inside portion. This term is, in
other words, the English counterpart of "incar." Evidently this
term is nondispersive, and we could in principle trace perfectly
unambiguous careers under it. But we do not trace such careers,
which shows that the term is not a sortal.
Or consider the disjunctive term "tree that is being rained
 SORTALS 41

upon or trunk that is not being rained upon," where this is to

be understood as applying to an object at a given time if and
only if either the object is a tree which is being rained upon at
that time, or the object is a trunk which is not being rained
upon at that time. This is a nondispersive term under which we
could in principle trace unambiguously such weird-seeming
tree-to-trunk successions as the one discussed earlier. (If it rains
on Monday but not on Tuesday then we get just the path dis-
cussed.) Of course this is not a sortal and we do not trace under it.
Returning to more ordinary examples, it may not be clear how
we should treat a shape word like "round" in the present con-
nection. But I would want to count "round" as dispersive, and
hence as definitely not a sortal. My reason for this judgment is
that a typical round thing, say a tomato, will extensively overlap
any number of round portions of matter which make it up. This
judgment might be resisted on the grounds that it seems odd to
apply the expression "round object" to some arbitrary portion
of a tomato. Now I might agree, perhaps, that at some level, and
in some sense, arbitrary round portions of tomatoes are not
properly called "round objects," or even "objects." It seems
reasonable, however, that at the present stage of analysis we
should rely on nothing but the widest sense of the word "object,"
as meaning, roughly, any continuous tract of matter. To rely
on any narrower sense of "object" at the present stage would
be simply to assume features of our concept of an object which
have yet to be explained. It seems sufficiently clear that at least
in this wide sense of "object," round objects frequently overlap
extensively, so that "round" is dispersive. The important point
here is that our concept of "round" could not by itself provide a
basis for unambiguously tracing careers.
It becomes clear why typical nouns like "tree," and "trunk,"
and "car" are peculiarly apt to function as sortals. These terms
are nondispersive in the extreme, for there are no remotely
typical cases in which two trees, or two trunks, or two cars ex-
tensively overlap. A sufficiently large portion of a tree (e.g., the
tree minus a twig) is to be sure something that, as we might
roughly put it, could have been a tree if it were separated from
the rest of the tree. (That is, if you took the twig away you
would be left with a tree.) None of these "potential trees," how-
ever, are trees. This is decisively evidenced by the fact that a
42 THE CONCEPT OF IDENTITY

term like "tree" is a paradigmatic count noun, which means

that there is typically a completely clearcut answer to the ques-
tion "How many trees are there in such and such a region?"
For example, it may be a clearcut truth that there is exactly one
tree in my backyard, which shows conclusively that no portion
of that tree is counted as a tree. The nondispersiveness of our
standard count nouns is what guarantees that we can employ
them as sortals and trace unambiguous careers under them.
By contrast, the dispersiveness of such so-called mass nouns
as "water," "wood," and "dirt" disqualifies these terms from
functioning as sortals. A term like "wood" is dispersive because
any stretch (quantity, bit) of wood will extensively overlap nu-
merous other stretches of wood that make it up. The ineffective-
ness of "wood" as a sortal can be brought out by considering,
with respect once again to our old example, how trying to trace
under "wood" would allow us to combine tree-stages and trunk-
stages (or tree-stages and stages of the tree minus a twig) in an
unlimited variety of ways. That mass nouns are not sortals will
figure in a later chapter as posing a problem about our concept
of the persistence of matter.
It may be noted, however, that mass nouns enter into various
nondispersive constructions that do apparently function as
sortals. Consider such expressions as "pool (or puddle) of water,"
"lump (or fragment) of wood," "pile (or heap) of dirt." These
are, or at least can legitimately be understood as, nondispersive.
If I say, "There are three pools of water on the floor," I am
evidently using "pool of water" in such a way that the parts of
a pool of water are not themselves pools of water. It seems
rather straightforwardly correct to trace careers under these
terms. If you trace a continuous succession of stages of pools of
water it seems to follow that you have kept your eye on the same
pool of water. Perhaps "pool of water," "lump of wood," and
"pile of dirt" do not really differ much from such standard
sortals as "river," "stick," and "mountain."
The sortal "pool of water" stands to the nonsortal "water"
in essentially the following way: "pool of water" is equivalent,
roughly, to "continuous stretch of water that is not part of any
larger continuous stretch of water." Now a more difficult ex-
ample to assess is a term like "continuous stretch of brown
that is not part of any larger continuous stretch of brown." Is
 SORTALS 43

there a sortal which corresponds to this term, a sortal which

stands to "brown" in the way that "pool of water" stands to
"water"? Where we have a continuous stretch of brown that is
not part of any larger continuous stretch of brown we have,
perhaps, a patch of brown. Is "patch of brown" (construable as)
a sortal?
Here is the sort of example to test our intuitions about this.
First, as a preliminary, consider that if you add a missing
drawer to your desk then something, viz. your desk, gets heavier.
Anyone who knows how to speak our language knows that this
is so. For anyone who knows how to speak our language im-
plicitly knows how to operate "desk" as a sortal, and knows,
therefore, that a continuous succession of desk-stages corre-
sponds to the career of a single persisting desk. If some later
member of such a succession is heavier than an earlier one it
follows that a desk got heavier. Now compare this case to the
following one. You have a brown desk and you place a brown
ashtray on top of it. (Perhaps the ashtray even sticks a little bit
to your desk.) Does anything get heavier in this process? Of
course the desk does not get heavier, and the ashtray does not
get heavier. But does anything get heavier? If "patch of brown"
is a sortal then it would follow that something does get heavier,
viz. a particular patch of brown. For in this example there is a
continuous succession of stages of brown-stretches-not-contained-
in-larger-brown-stretches, where later members of this succession
are heavier than earlier ones.
It may seem outrageous in this situation to assert, without
further ado, that something (let alone some object) got heavier.
On the other hand, if it is made clear that the thing being
referred to is a patch of brown, it does not seem clearly false to
assert this. There may even be exceptional circumstances in
which there would be a point in describing the situation in this
way. Perhaps we should say that "patch of brown" is a border-
line, or marginal, case of a sortal. Other examples of the same
sort might be "lump of hardness" and "patch of wetness."
I bring up these borderline sortals primarily to guard against
the error of construing the sortal rule as being perfectly exact.
The truth, on the contrary, is that the rule, and the correlative
notion of a sortal, can be no more exact than our ordinary con-
cept of persistence. The rule is merely a framework in terms
44 THE CONCEPT OF IDENTITY

of which we can, hopefully, describe and understand the nature

of our identity scheme. There is no denying, however, the marked
vagueness and amorphousness of that scheme. New sortals may
be brought into language, and old ones dropped out; it may
remain undecided whether a given term can function as a sortal;
terms may be ambiguous, in that sometimes they function as
sortals and sometimes not; and makeshift sortals may be adopted
to serve the needs of the moment. Indeed one positive sign of
the adequacy of the sortal rule is precisely that it helps us to
describe and explain the vagueness of our identity scheme.
A question which may naturally arise now is why it is that
some nondispersive concepts figure as standard sortals in our
language while others figure only marginally so, or not at all.
This question is actually part of a much larger one, as to why
our concept of persistence is what it is and not something else.
I shall have considerably more to say about this question later,
and in more than one context. A preliminary answer to the
specific question about sortals, which may seem at least super-
ficially satisfying, is that a nondispersive concept tends to figure
as a sortal in our language insofar as this concept is important
to us, from some practical or theoretical standpoint. To trace the
career of a tree in the ordinary way, or a car, or a desk, seems
more useful and relevant to our normal concerns than does
tracing the career of a patch of brown as such, or a lump of
hardness (not to mention tracing the career of an incar, or a
tree-to-trunk concoction). This answer obviously needs to be
elaborated, and it may actually turn out to have less explanatory
value than might initially seem.
There is one complication about dispersiveness which ought
to be aired before going further. I have suggested that a neces-
sary (but not sufficient) condition for F to be a sortal is that F
be nondispersive. Now it may be argued with some plausibility
that a considerably stronger condition than that is in order.
Consider the following definition:
"F is antidispersive" means: It cannot conceivably happen that
different F-things overlap extensively.
In order for a term F to be nondispersive it need only be the case
that F-things do not as a matter of fact typically overlap exten-
sively. But for F to be antidispersive it must be that F-things never
do, nor ever conceivably could, overlap extensively.
 SORTALS 45

It may be argued that for F to be a sortal it is not merely

necessary that F should be nondispersive, but also that F should
be antidispersive. For suppose that some sortal F were merely
nondispersive but not antidispersive. Then there would be a
conceivable (or perhaps even an actual but atypical) situation
in which two F-things extensively overlapped. It may seem to
follow that in that conceivable situation there would be con-
tinuous successions which combine stages of one F-thing with
stages of the other. There would then be continuous F-successions
which partly coincide and partly diverge. But, if F is a sortal,
we would have to judge these F-successions to correspond to
F-things whose paths partly coincide and partly diverge (and
hence, in a sense, to different F-things that temporarily occupy
the same place). Such a judgment, it may be maintained, ought
not to be admitted even as a possibility.2 And, the argument
would conclude, to rule this judgment out as a possibility we
must require that sortals be antidispersive.
This argument can, I think, be questioned on several counts.
For one thing it is not at all clear that wherever two F-things
"extensively overlap," even if this be in some rare and idio-
syncratic way, it necessarily follows that there is a "sufficiently
continuous" path which combines stages of one with stages of
the other. Both the notion of extensive overlap and that of
continuity (especially qualitative continuity) are much too vague
to allow for any such airtight connection. I did indeed assume
earlier that where F is dispersive, so that the typical case is for
F-things to overlap extensively, then stages of different F-things
could frequently be combined continuously. And this assump-
tion seemed completely plausible with respect to the dispersive
terms which I cited (e.g., "brown," "round," "wood"). (For
these terms, in fact, perfectly continuous paths can be traced
combining stages of different objects that come under them.)
But I would not necessarily assume that where F is nondispersive,
and two F-things extensively overlap in some rare case, a problem
would have to arise about continuously combining stages of
these objects. On the other hand if it could be shown (which
seems rather doubtful) that, for some seemingly standard sortal
F, it might happen that two incontrovertibly continuous succes-
2. Wiggins accepts the principle that, where F is a sortal, it cannot conceivably
happen that there are F-things whose paths partly coincide and partly diverge.
See Wiggins, ibid., p. 72, ftn. 44.
46 THE CONCEPT OF IDENTITY

sions of F-stages partly coincide and partly diverge, then I am

not convinced that we ought simply to rule out the possible judg-
ment that these successions correspond to F-things whose paths
partly coincide and partly diverge (and hence, in a sense, to
two F-things that temporarily occupy the same place).3
In any case I intend to leave open this rather vexing question
as to whether sortals must be antidispersive. Perhaps the follow-
ing consideration will suffice to show that this question is not
very urgent. Suppose that we have some seemingly standard
sortal F under which we apparently trace careers. But suppose
that it turns out upon reflection that F is not antidispersive. A
possible example of such a term might be "table," for it can
be argued (but not, I think, with overwhelming convincingness)
that one table might be made up of other tables in such a man-
ner as to warrant the judgment that two different tables ex-
tensively overlap. Actually it seems rather questionable that if
two tables are put together to make a third table we can simul-
taneously treat as tables the composite and its components. (Do
we then have three tables at a given moment?) Furthermore
even if it is proper to count all three as tables, so that we may
in fact be said to have a composite table extensively overlapping
a large component table, it still remains questionable, as I sug-
gested a moment ago, that we would be able to trace a (clearly)
continuous path which combined stages of the composite with
stages of its large component. But suppose even the worst pos-
sibility, that sortals must be antidispersive and that, as a con-
sequence, "table" cannot qualify as a sortal. Would we then be
left with the problem of explaining how we ordinarily trace the
careers of tables?
Not really. For even if "table" is disqualified as a sortal,
"standard (normal) table" is not. The latter term, at any rate,
is antidispersive, since it is inconceivable that two standard
tables should extensively overlap. (If even this seems unaccept-
able then, for the purposes of this argument, replace "standard
table" by the evidently antidispersive term "table that does not

3. Note that this judgment is not ruled out by the sortal rule. The latter says
that a continuous F-successioii corresponds to stages in the career of a "single
persisting object." I take this to mean that the succession corresponds to at
least one object, admitting the possibility that there may be more than one.
Even if this possibility cannot be realized in the kind of case under considera-
tion it is realized in other cases, as I will show in Section IV, below.
 SORTALS 47

extensively overlap any other table.") So we can say that we

ordinarily trace the careers of tables under the sortal "standard
table." And from this it seems a small step to saying that we or-
dinarily trace the careers of tables under the sortal "table" in
the sense of "standard table." This suggests, I think—at least
with respect to our most commonplace tracing procedures, which
is primarily what I want to continue to focus on—that the ques-
tion whether sortals need to be antidispersive, or merely non-
dispersive, makes no great difference. As to what tracing procedure
we might follow in an atypical case like that of the allegedly
overlapping tables, where arguably the sortal rule would not
suffice, this difficulty will be incidentally neutralized by the dis-
cussion in the next chapter, which will suggest that our de-
pendence upon the sortal rule is anyway less than absolute.

III. Coming into Existence and Going out of Existence

If we can assume that the sortal rule explains our primary non-
compositional basis for judging of an object's identity then it
should prove possible to give an account on this basis of our
judgments about when objects come into existence and go out
of existence (where the possibility is left open that an object
which goes out of existence might, via the supplementary com-
positional criterion, later come back into existence). A very
obvious kind of example in which an object is said to come into
existence is where various bits of matter come together, either
naturally or as a result of human design, to form the object.
And an obvious example of an object going out of existence is
where it is broken up or otherwise decomposed into fragments.
Thus a table comes into existence when a carpenter puts various
pieces of wood together in the appropriate form, and the table
might go out of existence when it is smashed to bits. How shall
we understand our thought about such cases in terms of the
sortal rule?
There are several wrong answers which we might initially be
tempted to give to this question. A trivial mistake would be
to suggest that wherever we have a continuous succession of
table-stages we can associate the beginning and end of this
succession with the coming into existence and going out of
existence of a table. Or, to put this in more general terms, the
idea would be
48 THE CONCEPT OF IDENTITY

(1) Where F is a sortal and S is a continuous succession of F-

stages, the beginning and end of S correspond respectively to
the coming into existence and going out of existence of an
F-thing.

(1) is just a trivial slip because it overlooks the possibility that

a continuous succession of F-stages might be merely some seg-
ment of a longer continuous succession of F-stages. Suppose that
some table was created in 1910 and persisted continuously until
it was destroyed in 1960. S might merely be the continuous suc-
cession of table-stages that corresponds to the segment of the
table's career from 1920 to 1930. Obviously the terminal points
of S do not then correspond to any table coming into existence
or going out of existence. The terminal points of a continuous
succession of table-stages do correspond to the coming into ex-
istence and going out of existence of a table only if the succession
is a longest continuous succession of table-stages, i.e., it is not
merely a segment of a longer continuous succession of table-stages.
A general principle which shows itself in this example is what
I will call the principle of prolongation. This principle says that
if F is a sortal and S is a continuous succession of F-stages then
no (proper) segment of S is such that its terminal points cor-
respond to the coming into existence and going out of existence
of an F-thing. (For simplicity we can confine ourselves to seg-
ments of S which both begin after S begins and also end before
S ends.) The principle says, in other words, that when we trace
an object's career under the sortal F we must prolong the career,
backwards and forwards in time, so long as tracing under F
allows. So we can never say, where F is a sortal, that one F-thing
went out of existence and was, without any loss of continuity,
immediately replaced by another F-thing that came into existence.
The principle of prolongation does not, I think, follow strictly
from the sortal rule, since the latter leaves open as at least a
formal possibility that one F-thing persists, corresponding to a
continuous F-succession, while other F-things, corresponding to
segments of that succession, come into existence and go out of
existence. Be this as it may, the principle is obviously called for
by the sortal rule, and seems quite plausible in its own right.
We saw that the answer to our question in the case of the
table is that the coming into existence and going out of existence
 SORTALS 49

of a table correspond to the terminal points of a longest con-

tinuous succession of table-stages. It may seem that we can easily
generalize this answer to cover all cases by emending (1) to read

(2) Where F is a sortal and S is a continuous succession of F-

stages, the beginning and end of S correspond respectively
to the coming into existence and going out of existence of
an F-thing if and only if S is not the segment of a longer
continuous succession of F-stages.

But (2) is still badly off the mark. As I pointed out earlier a
term like "brown table" is a sortal, since any continuous suc-
cession of brown table-stages must correspond to stages in the
career of a single persisting brown table. (2) would then have
us say that the terminal points of a longest continuous succession
of brown table-stages correspond to the coming into existence
and going out of existence of a brown table. This implies that
if you have a brown table and you paint it green the brown table
goes out of existence. This reductio ad absurdum is not, I have
found, always immediately appreciated. But surely it is absurd to
say that the brown table went out of existence. Suppose that
when the table is brown you make the prediction (vow) "That
brown table will never be touched by Miriam." Could you make
this prediction come true merely by painting the table green
before Miriam can touch it? Evidently not; evidently if Miriam
touches the table after it is painted green your prediction turns
out false, because the brown table that you referred to still per-
sists, though it is now the green table. An even more obvious
example, if one is needed, is the term "table in the living room,"
which is as much a sortal as "brown table." (2) would imply the
patent absurdity that if the table in the living room is about to
be moved into the dining room then the table in the living
room is about to go out of existence.
Moreover, in implying these absurdities (2) turns out to be
internally incoherent since (2) would lead us to violate the very
principle of prolongation which it expresses. If we had to say,
as (2) implies, that a brown table goes out of existence when it
is painted we would also have to say then that a table goes out
of existence (since it seems undeniable that "A brown table went
out of existence" entails "A table went out of existence"). But
50 THE CONCEPT OF IDENTITY

to say that a table goes out of existence when it is painted vio-

lates the principle of prolongation which (2) expresses, insofar as
the table's career, which is traced under the sortal "table," can
be further prolonged under that sortal.
It might now be suggested that the difference between the case
of painting a table, where nothing goes out of existence, and
the case of smashing it to pieces, where the table does go out
of existence, is that in the former case we can continuously pro-
long the career of the brown table by shifting from the sortal
"brown table" to the sortal "green table," whereas in the latter
case we simply lose the ability to prolong continuously the path
we are tracing, no matter how we might try to shift sortals.
That we cannot in any manner continuously prolong the path
of the table after it is smashed to pieces does seem a rather
plausible assessment, especially if we adopt my suggestion in the
last chapter that spatiotemporal continuity be interpreted in
the moderate sense. The principle of prolongation, as defined
earlier, required that a career which is traced under a sortal F
should not be terminated when it can be continuously prolonged
under F. The present suggestion is that the principle be strength-
ened to require that a career which is traced under a sortal F
should not be terminated when it can be continuously pro-
longed under any other sortal. The rule for terminating a career
would then be
(3) Where F is a sortal and S is a continuous succession of F-
stages, the beginning and end of S correspond respectively
to the coming into existence and going out of existence of
an F-thing if and only if S is not the segment of a longer
continuous succession of object-stages (where these object-
stages may come under various sortals).
(3) gets considerably closer to the truth than either (1) or (2),
and does perhaps accommodate the majority of typical cases
in which we distinguish between an object persisting through
change and an object going out of (or coming into) existence.
But (3) is still not correct. Whereas (1) and (2) were wrong in
ignoring, or not giving sufficient scope to, the principle of pro-
longation, (3) goes wrong in exaggerating this principle. It is
not true, as (3) implies, that we are always permitted to prolong
an object's career by shifting to a sortal other than the one
 SORTALS 51

under which we were tracing. Sometimes this is so but sometimes

it is not. The sorts of cases that need to be taken into account
are those that were mentioned in the last chapter to show that
an object can sometimes go out of existence only to be con-
tinuously replaced by another object. When a car is subjected
to the crushing machine it goes out of existence and is replaced
by a block of scrap metal. (3) would imply, however, that the
car's career ought to be prolonged by shifting from the sortal
"car" to the sortal "block of scrap metal."
There may in fact be some considerable resistance to admitting
that the car has to go out of existence just because it turns into
a block of scrap metal, and one important possible source of
this resistance will emerge in the next chapter. When we soberly
reflect upon this case, however, and keep in mind that we are
talking about the car and not about the material components
that make it up, it becomes sufficiently clear, I think, that the
car does go out of existence (though the material components
that make it up may of course persist). If you predict "That car
will never be touched by Miriam" then you can presumably
make this prediction come true by immediately putting the car
into the crushing machine. It will not matter if Miriam later
touches the block of scrap metal that comes out of the machine
because she would not be touching the car that you referred to.
What is evidently required, and what (3) does not accomplish,
is to explicate the rule on the basis of which it is legitimate to
prolong the career of a brown table by shifting to the sortal
"green table," where it is not legitimate to prolong the career
of a car by shifting to the sortal "block of scrap metal." We
want to understand the conceptual difference between the
legitimate shift from "brown table" to "green table" and the
illegitimate shift from "car" to "block of scrap metal." When
our question is put in these terms the correct answer fairly leaps
to the eye. There is after all a very obvious and special relation-
ship between "brown table" and "green table" that does not
obtain between "car" and "block of scrap metal." The former
pair of sortals, but not the latter, are qualifications or restric-
tions of a common sortal. "Brown table" and "green table" are,
as I will say, subordinate to "table." To say that the term F is
subordinate to the term G means that F's being truly predicable
of an object analytically entails G's being truly predicable of it.
52 THE CONCEPT OF IDENTITY

(This implies, as a degenerate case, that any term is subordinate

to itself.) The rule which has now emerged is that when an
object's career is traced under the sortal F its career is to be
prolonged by shifting to the sortal F' just in case F and F' are
both subordinate to some common sortal G. Thus:
The Sortal Rule Addendum. Where F is a sortal and S is a
continuous succession of F-stages, the beginning and end of S
correspond respectively to the coming into existence and going
out of existence of an F-thing if and only if S is not the segment
of a longer continuous succession of G-stages, for any sortal G
to which F is subordinate.
The sortal rule addendum properly explicates the principle of
prolongation which was already implicit in the original sortal
rule. The basic idea here is really quite simple. Any object's
career, from the moment of its coming into existence to the
moment of its going out of existence, must correspond to a
continuous succession of F-stages, where F is some highly general
sortal like "table," "car," or "tree." In the course of its career,
however, the object will pass through various transitory phases
that are marked off by such less general sortals as "brown table,"
"car in a garage," or "tree with snow on it." The gaining and
losing of these less general sortals do not affect the continuance
of the object's career so long as its career remains traceable under
the more general sortal. The object goes out of existence, how-
ever, when its career can no longer be prolonged under the
general sortal.
It will be noted that insofar as spatial continuity is regarded
as an ingredient of spatiotemporal continuity the sortal rule
addendum implies that an object can never persist in a spatially
discontinuous form (though, again, the rule allows for the possi-
bility that an object which goes out of existence because of
fragmentation may later come back into existence when its parts
are reassembled). As such the rule seems to provide one reason-
able, and especially simple, way of describing our ordinary
thought about what happens when objects are fragmentized. A
possible alternative, however, would be to suppose that there are
circumstances in which an ordinary object like a table might be
said to persist in a spatially discontinuous form (perhaps, for
example, where the table is only momentarily dismantled and
 SORTALS 53

then immediately reassembled). Presumably this supposition

would imply that the table might persist as a table in a spatially
discontinuous form. The idea, generally, would be that, for some
sortals F, there are special circumstances in which it is proper to
speak of there existing a spatially discontinuous F-thing. If this
alternative is favored all that is required is that for the purposes
of the sortal rule and addendum spatial continuity not be re-
garded as an ingredient of spatiotemporal continuity. The rule
would then imply that the career of a table be prolonged so long
as we can trace a temporally and qualitatively continuous succes-
sion S of (perhaps spatially discontinuous) table-stages such that
the (perhaps spatially discontinuous) places which coincide with
temporally neighboring stages in S overlap appropriately.
Wiggins draws a distinction, which is relevant to the present
discussion, between what he calls "substance sortals" and "phase
(or restricted) sortals."4 This distinction, along the general lines
that he explains it, might be defined as follows:
"F is a substance sortal" means: F is a sortal, and it is a con-
ceptual truth that if S is a continuous succession of F-stages,
and S is not a segment of a longer continuous succession of
F-stages, then the beginning and end of S correspond respec-
tively to the coming into existence and going out of existence
of an F-thing.
"F is a phase sortal" means: F is a sortal and F is not a sub-
stance sortal.
A highly general sortal like "table" seems to qualify as a sub-
stance sortal on this definition because, as we said before, it seems
to be a conceptual truth that the terminal points of a longest
continuous succession of table-stages correspond to the coming
into existence and going out of existence of a table. A less general
sortal like "brown table" would be a phase sortal, since it is not
a conceptual truth that the terminal points of a longest continu-
ous succession of brown table-stages correspond to the coming
into existence and going out of existence of a brown table. Phase
sortals will typically be complex expressions constructed from a
substance sortal and a qualifying adjectival expression. Hence
"brown table," "car in the garage," and "tree with snow on it."
4. Wiggins, Identity and Spatio-T'emporal Continuity, pp. 7, 89-30.
54 THE CONCEPT OF IDENTITY

There are, however, some clearcut examples of syntactically

simple phase sortals. One example is "kitten." It is clearly not a
conceptual truth that a kitten goes out of existence whenever a
continuous succession of kitten-stages terminates. This is because
"kitten" (roughly, "young cat") is subordinate to "cat," so the
principle of prolongation expressed by the addendum requires us
to prolong the kitten's career by shifting from "kitten" to "(older)
cat."
It is easy to see that if F is a sortal and F is not subordinate to
any other sortal then F must be a substance sortal. The sortal
rule addendum says in effect that the terminal points of a longest
continuous F-succession (i.e., a continuous succession of F-stages
that is not a segment of a longer continuous succession of F-stages)
correspond to the coming into existence and going out of exist-
ence of an F-thing unless the F-succession is a segment of a longer
continuous G-succession, for some sortal G to which F is subordi-
nate. But if F is not subordinate to any other sortal then there
can be no such sortal G. Hence the terminal points of a longest
continuous F-successiori must correspond to the coming into
existence and going out of existence of an F-thing, which qualifies
F as a substance sortal.
Someone might want to argue for the converse principle as
well, viz. if F is a substance sortal then F cannot be subordinate
to any other sortal. Instead of displaying this argument in general
terms I will illustrate it for the case of "dog" and "animal."
Assume that "dog" is a substance sortal and that "dog" is subordi-
nate to "animal." It might then be argued as follows that "ani-
mal" cannot be a sortal. We can conceive of a continuous succes-
sion S of animal-stages such that an initial segment of S contains
dog-stages and a later segment of S contains nondog-stages. S
might correspond, for example, to the imaginable situation of a
dog gradually changing into a cat. Since "dog" is assumed to be
a substance sortal we would have to say that the end point of the
segment of dog-stages in S corresponds to a dog going out of
existence (i.e., that the dog goes out of existence when it turns
into the cat). Since necessarily any dog is an animal ("dog" being
assumed subordinate to "animal") whenever a dog goes out of
existence an animal must go out of existence. We would then be
saying that it is possible for there to be a continuous succession
S of animal-stages which contains a segment the end point of
 SORTALS 55

which corresponds to the going out of existence of an animal.

(We would be saying, in other words, that it is possible for there
to be a situation in which one animal went out of existence and
was continuously replaced by another animal.) But if "animal"
were a sortal the principle of prolongation expressed by the sortal
rule addendum would prevent us from admitting this as a pos-
sibility. Hence "animal" cannot be a sortal.5
One rather serious defect of this argument, I think, is the
controversial status of the initial assumption that we can conceive
of a continuous succession of animal-stages which contains dog-
stages followed by nondog-stages. If a dog changed gradually
into a cat is it clear that we would have an animal during the
intermediary stage of this process (an animal that is, perhaps, no
particular sort of animal)? On the other hand if the dog changed
instantaneously into a cat it may certainly be doubted that we
have the requisite degree of qualitative continuity.
Even if this objection strikes someone as not very pressing with
respect to the case of "dog" and "animal," it would certainly
have to be taken seriously if the argument is to be generalized to
cover all cases. Another, somewhat more pedestrian, case that we
might consider is that of "shirt" and "article of clothing." Cer-
tainly a shirt might be cut and resewn to form an article of
clothing of a different sort, e.g., a scarf. Here we would presum-
ably want to treat "shirt" as a substance sortal and say that one
article of clothing (i.e., the shirt) went out of existence and an-
other article of clothing (i.e., the scarf) came into existence. But,
it is doubtful that this sort of possibility proves "article of
clothing" to be a nonsortal, since we might certainly question
whether we have in this case a continuous succession of stages of
articles of clothing, whether, that is, the tattered bit of cloth
found in the transition stage from the shirt to the scarf counts
as an article of clothing.
If we countenanced the principle that a substance sortal can-
not be subordinate to any other sortal then we might be left
floundering over which terms to count as sortals, and which of
these to count as substance sortals. If counting "dog" as a sub-
stance sortal actually forced us not to count "animal" as a sortal

5. Wiggins holds that "animal" is not a proper sortal, but his reasons may not
depend on the sort of argument just given. Cf. Wiggins, ibid., pp. 61-63.
56 THE CONCEPT OF IDENTITY

it is not at all clear what we want to say. For that matter we

might begin to wonder whether even "dog" is a sortal, since it
seems not implausible to treat terms like "terrier" and "spaniel,"
which are subordinate to "dog," as substance sortals. The doubts
over which terms to reckon as substance sortals do not trouble
me too much, since this notion does not figure in the sortal rule
or its addendum, as I formulated these. But it would certainly be
discomfiting to the present analysis if we had constantly to hesi-
tate over reckoning any term a sortal on the grounds that some
other term subordinate to it might plausibly be reckoned a
substance sortal.
My suggestion would be to avoid this situation by rejecting the
problematical principle that substance sortals cannot be subordi-
nate to other sortals. We can, I think, reasonably adopt a more
permissive policy towards the application of "sortal" and "sub-
stance sortal." A term can be reckoned a sortal, or substance
sortal, so long as there is no conceivable situation which would
constitute a relatively clearcut (i.e., nonborderline) case in which
the term fails to function as a sortal, or substance sortal. Thus
I would want to count all of the terms "terrier," "dog," and
"animal" as sortals on the grounds that there is, so far as I can
tell, no conceivable situation which would constitute a relatively
clearcut case in which a continuous succession of terrier-stages
(dog-stages, animal-stages) failed to correspond to the career of
a single persisting terrier (dog, animal). And I would also count
all of these terms as substance sortals on the grounds that there
is no conceivable situation which would constitute a relatively
clearcut case in which the terminal points of a longest continuous
succession of terrier-stages (dog-stages, animal-stages) failed to
correspond to the coming into existence and going out of exist-
ence of a terrier (dog, animal). We can conceive of borderline
cases (e.g., a dog changing into a cat) which, if they were to occur,
might possibly force us (perhaps on theoretical grounds) to make
new decisions as to which terms are sortals and which terms are
substance sortals (or even as to which terms are subordinate to
which). But until such decisions need to be made we can reason-
ably adhere to the permissive policy of counting all of these
terms as substance sortals. Other sortals such as "fat dog," "white
dog," "dog in the yard," and "puppy" remain as clearcut phase
sortals.
 SORTALS 57

IV. Identity, Predication, and Constitution

Earlier in the course of discussing the question whether sortals
need to be antidispersive I mentioned, and left open, the question
whether there could conceivably be, for sortal F, F-things whose
space-time paths partly coincide and partly diverge. Though this
question will remain open the following related proposition is
quite definitely true: For different sortals F and G it may happen
that the path of an F-thing and the path of a G-thing partly
coincide and partly diverge.
Consider again our one-branched tree on Monday. But this
time imagine that at the beginning of Tuesday the tree's branch
is actually chopped off, so that all that remains of the tree on
Tuesday is its trunk. We might certainly want to say in such a
case, especially if the branch was fairly negligible, that the tree
still persists on Tuesday, though it has been reduced to a trunk.
If S is the succession which corresponds to the tree's career from
Monday through Tuesday and S' corresponds to the trunk's career
during those two days, then S and S' diverge on Monday and
coincide on Tuesday. S, it may be noted, is similar in a way to the
weird concoction of tree-stages and trunk-stages discussed in the
earlier example, except that in the present case, where the branch
was actually chopped off, there is nothing weird about S, since S
is just the path of the tree. So here we seem to have a rather clear
case in which the path of an F-thing and the path of a G-thing
partly coincide and partly diverge (where F is the sortal "tree"
and G is the sortal "trunk").
This kind of case, however, gives rise to a certain difficulty. It
may seem perfectly legitimate in the case just imagined to make
each of the following three assertions on Tuesday:
(a) This tree is identical with (is one and the same object as) this
trunk.
(b) This tree was bigger yesterday.
(c) This trunk was not bigger yesterday.
The difficulty is that these three apparently true propositions
seem to be logically incompatible with each other. If it is true,
as (a) asserts, that the tree and trunk are one and the same object,
then it seems to follow that there is that one object which we can
refer to both as "this tree" and "this trunk." Well, was that object
58 THE CONCEPT OF IDENTITY

bigger yesterday or not? If it was bigger then (c) must be false

in asserting that it was not, and if it was not bigger then (b) must
be false in asserting that it was.
If the identity claim made by (a) does in fact commit us to
holding that there is some object on Tuesday which we can refer
to both as "this tree" and "this trunk" then there seems to be
no satisfactory way around this difficulty. It would not help us to
"relativize" our remarks about the object in question (i.e., the
object that can be referred to as both "this tree" and "this trunk")
by saying, for example, that the object was a bigger tree yesterday
but was not a bigger trunk yesterday, or that the object was the
same tree yesterday as a bigger object but was not the same trunk
yesterday as a bigger object.6 These maneuvers, whatever their
precise import might be, do not apparently help us at all to
answer the question "What was the object's size yesterday?" It
seems certain that this question must have an answer. If there is
that particular object which we can refer to today as both "this
tree" and "this trunk" then that object must have had some
definite size yesterday, and that size was either bigger or not
bigger than the size of the tree (the trunk) that is now presented
to us. We cannot possibly have it both ways. But if there is some
single object to which we intend to refer when we say both "This
tree was bigger" and "This trunk was not bigger" then we would
be trying to have it both ways.
Our difficulty is obviously not just with respect to the size of
the tree and the trunk. The general problem is that there may
be various properties which the tree had prior to Tuesday but
which the trunk did not have. It may be, for example, that the
trunk was never touched by human hands whereas somebody did
touch the tree by touching its branch. Or it may be that yesterday,
on Monday, the tree was partly white (the branch having been
white) but the trunk was completely brown. But if there is some
single object which we can refer to on Tuesday as both "this tree"
and "this trunk" then clearly that object was either once touched
6. Such relativizing maneuvers are suggested in P. T. Gcach, Reference and
Generality (Cornell University Press, Ithaca, N.Y., 1962), p. 39ff.; and are criti-
cized in Wiggins, Identity and Spatio-Temporal Continuity, part one, and in
Sydney Shoemaker, "Wiggins on Identity," Philosophical Review, 79 (1970),
530-35.
 SORTALS 59

by human hands or not, and it was either once partly white or

not.
The only sensible move to make here, I think, is to deny that
there is some single object which we can refer to on Tuesday as
both "this tree" and "this trunk." And to deny this is quite cer-
tainly to assert, on the contrary, that there is one object which we
can refer to as "this tree" and another object which we can refer
to as "this trunk." We must assert, in other words, that the tree
and the trunk are, at least in some important sense, not identical
with each other, that the tree is one thing and the trunk is an-
other thing. We can also acknowledge, however, that there is a
sense in which, as (a) correctly asserts, the tree and the trunk are
identical with each other. In a sense they are not identical, but in
a sense they are.
In line with a good deal of recent literature I will distinguish
these senses as "strict identity" and "constitutive identity."7 We
might define "x is constitutively identical with y" (or "x and y
constitute each other") as meaning "x and y occupy the same
place." It is in this sense that we can say that the tree and the
trunk are identical on Tuesday, for the tree and the trunk occupy
the same place on Tuesday. There is, at least with respect to the
most obvious cases, a rather straightforward connection between
"constitutive identity," in the defined sense of spatial coincidence,
and the intuitive notion of constitution (composition). Roughly
put, if two things are constitutively identical, in the sense of
occupying the same place, then they must be composed of the
same matter. But I do not want to enter into a discussion of
material composition until a later stage. For my immediate pur-
poses the definition of "constitutive identity," in terms of spatial
coincidence, suffices to draw the required contrast with "strict
identity."
I will not attempt to define "x is strictly identical with y," as
it is highly doubtful that this notion can be defined in any useful
way. But the meaning of strict identity, and its contrast with
constitutive identity, can be indicated in two related ways.
First, where a and b are singular terms which refer to individ-
7. See Wiggins, Identity and Spatio-Temporal Continuity, p. 1off.; R. M.
Chisholm, "Parts as Essential to Their Wholes," Review of Metaphysics, 26
(1973), 587ff.; and Shoemaker, "Wiggins on Identity," 531ff-
6o THE CONCEPT OF IDENTITY

ual objects the statement "a is strictly identical with b" implies
that there is some single object to which both a and b refer. On
the other hand, "a is constitutively identical with b" does not
imply this. We cannot say that the tree is identical with the trunk
in the strict sense because, as we saw, we cannot coherently sup-
pose that there is some object to which both "the tree" and "the
trunk" refer.
Second, it is strict identity, but not constitutive identity, which
satisfies the logical principle of substitutivity ("Leibniz's Law"),
according to which if x is identical with y all of x's properties,
past, present, and future, must be the same as y's. It seems self-
evident that we have a concept of identity which satisfies this
principle. It was this principle which I tacitly employed when
I argued in effect that the tree and the trunk cannot be strictly
identical since they have different past histories. We can say,
however, that the tree and the trunk are constitutively identical,
that they occupy the same place, without implying that their past
(or future) properties are the same.
If we conceive of the relationship between the tree and the
trunk in terms of the successions or space-time paths associated
with them it becomes very easy (in a sense, perhaps, too easy) to
understand how the difference between strict and constitutive
identity arises. There is certainly nothing difficult in the abstract
about the "Y"-shaped configuration of two successions of items
which differ up to a point but then share some segment in com-
mon. Looked at in this way the tree and the trunk are two succes-
sions which share their post-Tuesday segment. This way of think-
ing about the matter helps to relieve the sense of paradox in
saying that objects that are not strictly identical may occupy the
same place at once. The commonsense dictum that two things
cannot occupy the same place at the same time remains correct,
however, on the interpretation that if x and y occupy the same
place then x and y must be identical in at least the constitutive
sense.
If x and y are constitutively identical at a given time t then
their properties may differ radically both before and after t. This
is the essential contrast with strict identity. On the other hand,
if we consider only time t, since at that time x and y occupy the
same place (and are composed of the same matter), they will
evidently have to share their properties at that time (or, at least,
 SORTALS 6l

they will have to share at that time such straightforward proper-

ties as size, shape, color, location). Hence constitutive identity is,
so to speak, a weakened version of strict identity, which may ex-
plain why the ordinary locution "x is (one and the same as, iden-
tical with) y" can be used in both senses.
It will be seen that constitutive identity requires a temporal
qualification. The tree and the trunk are constitutively identical
on Tuesday but not on Monday. On the other hand it seems
evident (though I will not attempt to prove this) that an ascrip-
tion of strict identity cannot be temporally qualified. It would
make no sense to say that x and y existed on Monday and Tues-
day, that x and y were strictly one and the same on Monday, but
that x and y were not strictly one and the same on Tuesday.
This would make as little sense as saying that there was an object
that existed on Monday and Tuesday, and it was identical with
itself on Monday but not on Tuesday.
When the space-time paths associated with objects partly coin-
cide and partly diverge we have a case in which two (strictly)
different objects occupy the same place at one time, and do not
occupy the same place at another time when both of them exist.
Thus the tree and the trunk occupy the same place on Tuesday
and do not occupy the same place on Monday, though both ob-
jects exist on Monday. There is, however, another kind of case
in which different objects occupy the same place. It may happen
that the path associated with the object x is a segment of the
longer path associated with the object y. In this case x and y will
occupy the same place at every moment that x exists, but x and
y are not strictly identical since y exists at times when x does not
exist.
A possible example of such a case is the following. Suppose
that a lump of gold is made into a coin in 1940, and that the
coin persists until 1960 at which time it is melted down into a
lump of gold. In such a case we might certainly want to say that
the coin came into existence in 1940 and went out of existence
in 1960. We may also want to say that a single lump of gold
persisted throughout this entire period. The space-time path
associated with the coin is then a segment of the longer path
associated with the lump of gold.
If I held the coin in my hand in 1950 I might want to assert
the following three propositions:
62 THE CONCEPT OF IDENTITY

(d) The coin in my hand is identical with the lump of gold in

my hand.
(e) The coin in my hand did not exist before 1940.
(f) The lump of gold in my hand existed before 1940.

If we did not already have available the distinction between

strict and constitutive identity we might perhaps have been
tempted to maneuver around this case by denying either (e) or
(f). (To deny (e) would be to deny that "coin" is a substance
sortal, while to deny (f) would be to deny that "lump of gold"
is a sortal.) Given the distinction, however, it seems that the most
plausible expedient is to maintain both (e) and (f), but to inter-
pret (d) as asserting, not strict, but mere constitutive identity.
The coin and the lump of gold are strictly two objects, but they
occupy the same place at every moment that the coin exists.
I have presented two examples, of slightly different sorts, in
which we need to say that two strictly distinct objects occupy
the same place at once. Other examples of both sorts could evi-
dently be found. It is not entirely clear, however, just how com-
mon we ought to consider such cases to be. This will depend, in
part, on how broadly we employ the notion of an "object." A
point which must, I think, remain valid is that such cases occur
in a severely circumscribed and tightly controlled fashion, and
are in an important sense exceptions to the rule. It still remains
correct to say, as I did earlier, that it would seem deeply incon-
sonant with our intuitive notion of persistence to conceive of the
careers of objects as crisscrossing and overlapping in an endless
and unmanageably complicated variety of ways.
The distinction between strict and constitutive identity implies
a correlative distinction between two senses in which a sortal may
be said to apply to an object. When I hold the coin in my hand
in the previous example I might certainly want to say "This
lump of gold is a coin" (and, also, "This coin is a lump of gold").
But there is obviously a difficulty about saying this. For if the
lump of gold is a coin, then it seems that it should be correct to
call it "a coin." It should therefore be correct to say, in virtue of
the fact that the lump of gold existed before 1940, "A coin which
is now in my hand existed before 1940," or "There is now a coin
in my hand which existed before 1940," or "This coin existed
 SORTALS 63

before 1940." But this is precisely what we cannot say, for the
coin did not exist until 1940.
What is evidently required here is a distinction between two
senses of the sentence-form "x is a coin," corresponding to the
two senses of identity. In order for it to be correct to say "x is a
coin" in the strict predicative sense it must also be correct to say
"x is strictly identical with a coin." Since the lump of gold is not
strictly identical with any coin we cannot say that the lump of
gold is a coin, in the predicative sense. There is however another
sense of "x is a coin" which implies merely "x is constitutively
identical with a coin." It is only in this weaker constitutive sense
that we can correctly say that the lump of gold is a coin. The
constitutive, but nonpredicative, application of the term "coin"
to the lump of gold does not permit us to refer to the lump of
gold as "the (some, a) coin."
This distinction between the predicative and constitutive appli-
cation of a sortal to an object affects the earlier characterization
of what it means for one sortal to be subordinate to another. The
sortal F is subordinate to the sortal G if the predicative applica-
tion of F to an object entails the predicative application of G to
the object. Hence "dog" is subordinate to "animal" because if
something is, in the predicative sense, a dog (and can be referred
to as "the dog") it must be, in the predicative sense, an animal
(and can be referred to as "the animal"). But it is not sufficient
for F to be subordinate to G that the constitutive application of
F should entail the constitutive application of G. "Gold coin" is
not subordinate to "lump of gold" even though "x is (constitu-
tively) a gold coin" entails "x is (constitutively) a lump of gold."
This point is essential to our account, for if "gold coin" were
reckoned as subordinate to "lump of gold" the principle of pro-
longation would prevent us from saying that a gold coin goes out
of existence when it is melted down into a lump of gold that is
not a coin. (So I would say that part of knowing our language
consists in having in effect learned which sortals are subordinate
to which, and resultantly which sortals can be predicatively tied
to each other, and which sorts of objects can be strictly identical
with each other.)
It is now possible to characterize substance sortals in a manner
which might earlier have been open to misunderstanding. F is
64 THE CONCEPT OF IDENTITY

a substance sortal if the terminal points of a longest continuous

F-succession correspond to the coming into existence and going
out of existence of a single persisting F-thing. This implies the
important principle that, where F is a substance sortal, if F is
predicatively true of an object at any moment then F must remain
predicatively true of the object until it goes out of existence.
Hence a coin (i.e., something which is, in the predicative sense, a
coin) must continue to be a coin until it ceases to exist. This valid
principle about substance sortals must, however, be properly
distinguished from the false statement that, where F is a substance
sortal, if F is constitutively true of an object at any moment then
F must remain constitutively true of the object until it goes out
of existence. A lump of gold might be (constitutively) a coin at
one moment but not be a coin at a later moment that it exists.

V. The Compositional Criterion

The sortal rule tells us that spatiotemporal and qualitative con-
tinuity under a sortal is a sufficient condition for an object's per-
sistence. We want now to formulate a supplementary condition
which would allow us to judge, for example, that a watch which
is taken apart and goes out of existence may retain its identity
when it is later put back together. One important feature of this
sought after condition is that if x and y are judged to be identical
on the basis of compositional considerations then there must
presumably be some sortal F which is predicatively true of both
x and y. We want to allow for the possibility of a watch being
taken apart and later coming back into existence as a watch, or
a car being completely dismantled and later coming back into
existence as a car, but not that the watch should come back into
existence as a car, or vice versa. A moment ago I mentioned the
principle that if F is a substance sortal (like "watch" or "car")
then if F is ever predicatively true of the object x, F must remain
true of x until x goes out of existence. But it seems that we should
now strengthen this principle to read: If F is a substance sortal
then if F is ever predicatively true of x, F must remain true of
x at any moment when x exists (even if x should perhaps come
back into existence after going out of existence).
A first approximation to the compositional criterion might be:
 SORTALS 65

(1) Where x is an object that exists at time t1 and y is an object

that exists at a later time t2, a sufficient condition for x to be
identical with y is that some sortal is predicatively true of
both x at tl and y at t2, and some set of objects exhaustively
comprises both x at t1 and y at t2.
To apply the compositional criterion we pick out the F-thing x
at some early time t1 and the F-thing y at some later time t2
(where F is a sortal). We then consider whether x's parts at t1
are identical with y's parts at t2. This judgment about the identity
through time of the parts is made on the basis of the sortal rule,
upon which the compositional criterion is dependent.
The criterion as thus formulated, however, is too stringent in
one respect and too lax in another. We certainly want to be able
to say that if a watch is taken apart it can retain its identity
even if some small number of parts are replaced when it is put
back together. In this case there would be no single set of objects
which exhaustively comprises the watch before and after it is
repaired. We need to relax the condition to require of x at t1 and
y at t2, not compositional identity, but merely compositional simi-
larity. This latter requirement is roughly that some single set of
objects should comprise a major portion of both x at t1 and y at
t2.
On the other hand even if some single set of objects does
exhaustively comprise x at t1 and y at t2 it is not clear that this
would be sufficient to induce us to say that x is identical with y.
We may require that the set of objects that comprises x at t1 and
y at t2 should be similarly arranged in x at t1 and y at t2. If a
sweater is completely unravelled and the wool used to make a
sweater again then, unless we somehow had reason to think that
the wool was arranged the second time like the first, I think we
should not be much inclined to say that we had the same sweater.
A better formulation might then be:
(2) Where x is an object that exists at time t1 and y is an object
that exists at a later time t2, a sufficient condition for x to be
identical with y is that some sortal is predicatively true of
both x at t1 and y at t2, and some set of objects comprises a
major portion of both x at t1 and y at t2, and this set of ob-
jects is similarly arranged in both x at t1 and y at t2.
66 THE CONCEPT OF IDENTITY

What counts as a "major portion" of an object is evidently quite

vague, and many borderline cases can, and actually do, arise in
this connection. Brute considerations of bulk or volume will cer-
tainly matter considerably in determining what is a major por-
tion, but it may also matter whether the portion contains parts
that seem especially relevant to the object's being (called) the
sort that it is. If the car x is dismantled at t1 and some of its parts
are used to construct the car y at t2, then x's having the same engine
at t1 as y has at t2 would seem to count more in favor of saying
that x is identical with y than would the fact that x has the same
airconditioner at t1, as y has at t2.
If compositional similarity under a sortal, as expressed in (2), is
sufficient for identity then it follows immediately that so is com-
positional continuity under a sortal. Suppose that x is a car which
is dismantled on Monday and a major portion of its parts, in
similar arrangement, are used on Tuesday to construct the car y.
Suppose, further, that y is dismantled Tuesday night and a major
portion of its parts, in similar arrangement, are used on Wednes-
day to construct the car z. It may of course happen that no set of
parts comprises a major portion of both x on Monday and z on
Wednesday. (This possibility might be schematically represented
as follows. On Monday x is comprised of the parts A, B, and C;
on Tuesday y is comprised of the parts A, B, and C'; and on
Wednesday z is comprised of the parts A, B', and C'. If two parts
in common counts as a major portion we get the mentioned
possibility.) But it would still follow from (2) that x and z are
identical. For it follows from (2) that x and y are identical and
that y and z are identical, and (by the "transitivity of identity")
if x and y are identical and y and z are identical it must follow
that x and z are identical. This case could obviously be elaborated
in such a way that (2) forces us to judge that an object x, which
is picked out at t1, is identical with an object y, which is picked
out at t2, even where no parts are common to x at t1 and y at t2.
It is a quite unavoidable and generally satisfactory corollary of
(2) that an object may retain its identity through a drastic or
even total alteration of its parts, so long as this alteration takes
place by a continuous sequence of small changes, each small
change leaving the object with a major portion of the similarly
arranged parts that it had prior to the change. This condition
of "compositional continuity," it should be noted, may be satis-
 SORTALS 67

fied by an object even when its career is not temporally con-

tinuous, as in the case just discussed, in which an object goes out
of existence and comes back into existence.
It seems perfectly evident that the condition of compositional
similarity defined in (2), and the correlative condition of com-
positional continuity, depend outright on some prior notion of
persistence in terms of which we can understand what it means
to talk about the persistence of the parts which make up an
object. This is why I urged earlier that compositional considera-
tions must be construed as merely supplemental to more primary
criteria. There seems however to be a rather deeply rooted ten-
dency among philosophers to treat compositional considerations
as primary. Often, as in the case of Locke's treatment of the
identity of organic bodies, this tendency takes the form of the
suggestion that our concept of the persistence of familiar observ-
able objects is to be understood in terms of the idea that the
atomic particles which compose these objects are replaced only
gradually.8 Apart from leaving unanswered the question "And
what does the persistence of an atom consist in?" the glaring
difficulty with any such account is that our everyday concept of
familiar and observable cases of persistence cannot plausibly be
regarded as analyzable in terms of the highly theoretical concept
of the persistence of atoms. It cannot be that what I judge (and
observe) to be the case when I assert, for example, "This car has
persisted for the past few moments" is some theoretical fact
about the comings and goings of invisible atoms.
Once it is conceded that our concept of the car's persistence is
not primarily to be understood, via the compositional criterion,
in terms of the persistence of exotic atomic particles, there seems
little temptation to suggest instead that this concept is primarily
to be understood, via the compositional criterion, in terms of the
persistence of such smaller familiar objects as engines, fenders,
wheels, etc., where the persistence of these smaller objects is then
explained in the noncompositional terms of the sortal rule. Ad-
mittedly in order for something to be (called) a car it must pre-
sumably stand in various typical compositional relations to some
of these smaller objects. But if our concept of the persistence of

8. John Locke, An Essay Concerning Human Understanding, book 2, chapter

27, sections 3-6.
68 THE CONCEPT OF IDENTITY

an engine (fender, wheel, etc.) is primarily to be understood, via

the sortal rule, in terms of the idea of a spatiotemporally and
qualitatively continuous succession of engine-stages (fender-stages,
wheel-stages, etc.), then it seems equally correct to say that our
concept of the persistence of a car is also primarily to be under-
stood, via the sortal rule, in terms of the idea of a spatiotempo-
rally and qualitatively continuous succession of car-stages. There
would seem to be no point in treating our concept of the persist-
ence of a relatively larger familiar object like a car in some funda-
mentally different way from our concept of the persistence of a
relatively smaller familiar object like an engine or a fender.
Whether an object is large or small the sortal rule is to be re-
garded as defining its primary identity condition, and the com-
positional criterion is merely supplementary.
Nor is there any mystery as to why, given the primacy of the
sortal rule, we should allow compositional considerations to
function as a supplement. Looked at in a certain way an object
at any particular moment is nothing over and above the parts
that make it up, arranged in a distinctive way. It seems therefore
entirely natural, if not inevitable, that when we come across
those same parts, or most of them, similarly arranged we should
be inclined to say that we have the same object again. This
inclination, however, is coherent only against the background of
a more primary notion of persistence.
I want to consider now a certain rather intriguing difficulty
which besets the compositional criterion when it is formulated
in some such manner as (2) above. The criterion, thus formu-
lated, implies that compositional similarity, and hence compo-
sitional continuity, is sufficient for an object's identity in all cases.
The difficulty is that in some cases considerations of composi-
tional similarity or continuity yield incompatible identity judg-
ments. A famous case of this sort is that of "the ship of Theseus."9
Suppose that we start out in January with the ship x, which, let
us imagine, is made up entirely of wooden planks. We proceed
gradually to replace x's planks one by one until by December,
perhaps, we wind up with a ship y, such that none of the planks
which composed x in January compose y in December. Our cri-
terion dictates that x and y are identical on the grounds of com-

9. See Thomas Hobbes, Concerning Body, chapter 11, section 7.

 SORTALS 69

positional continuity, on the grounds, that is, that the original

ship retains its identity through each small compositional change.
But suppose now that someone collected together all of the
planks that were removed from the ship during that year, and
he used those planks in December to construct the ship z, arrang-
ing the planks in just the way they were originally arranged in
x in January. Our criterion obviously dictates that the original
ship x is identical with 2. In this imaginable case, then, our cri-
terion yields the incoherent judgment that the original ship x
is identical with the two different ships y and z.
My own somewhat ambivalent inclination when reflecting
upon this case is to judge that x is identical with y and not with
z.10 If this intuition is generally shared there would be two re-
lated ways to explain it. It will be noted that the judgment that
x is identical with y follows from the sortal rule as well as from
the compositional criterion, whereas the judgment that x is iden-
tical with z follows only from the compositional criterion. This
is because if we start out with the original ship x in January and
trace a spatiotemporally and qualitatively continuous succession
of ship-stages we wind up with y in December. Hence we might
say that x is identical with y and not with z because where the
sortal rule conflicts with the compositional criterion the former
rule, which we know to be primary, takes precedence. Or, for-
getting about primacy, we might simply say that the sortal rule in
conjunction with the compositional criterion outweighs the latter
standing alone, and this is what favors x's being identical with y
rather than with z.
The potential conflict between the sortal rule and the compo-
sitional criterion might induce someone to suggest that the sortal
rule ought to be weakened. Instead of expressing a logically suffi-
cient condition of identity, as it does in its present formulation,
the suggestion would be that it ought to express only a condition
which counts logically in favor of identity, but which might in
principle be defeated by competing compositional considerations,
or perhaps by other kinds of overriding considerations as well.
This suggestion for weakening the sortal rule certainly deserves
to be taken seriously. So long as the considerations which might

10. Wiggins, Identity and Spatio-Temporal Continuity, p. 37, shares this

intuition.
70 THE CONCEPT OF IDENTITY

override the sortal rule could be loosely specified in advance, and

seen as built a priori into our identity concept, we would still
have an acceptable kind of analysis of what our concept of iden-
tity consists in. Nevertheless I am inclined to resist any such
weakening of the sortal rule. I think it is reasonable, barring
some relatively clearcut counterexample, more simply to regard
the condition expressed by the sortal rule, that of spatiotemporal
and qualitative continuity under a sortal, as no less than logically
sufficient for identity. Certainly we seem to have no convincing
counterexample in cases like that of the ship of Theseus, at least
if we agree that in such cases it is plausible to make the judgment
which accords with the sortal rule.11
We might indeed wish to reformulate the compositional cri-
terion so as to accommodate cases like that of the ship of The-
seus, and to provide that in such cases the favored identity judg-
ment is the one that, accords with the sortal rule. One rather
straightforward way of accomplishing this, I suggest, is explicitly
to limit the application of the compositional criterion to just
those situations for which we had originally invoked the cri-
terion, i.e., to situations in which an object goes out of existence
and comes back into existence. Looked at in this way, the com-
positional criterion has no bearing on the case of the ship of
Theseus, since in that case when we trace the original ship accord-
ing to the sortal rule we have no occasion to judge that the ship
went out of existence.
Perhaps, then, the compositional criterion (clearly) applies only
in a case in which x goes out of existence at t1 and y comes
into existence at t2, and x's composition at t1 is appropriately
similar to y's composition at t2. This still does not get it quite
right, though. We need to rule out a more complicated variant
of the ship of Theseus case. Suppose that the ship x is completely
dismantled early in January, and that the ship y is constructed
later in January out of all of x's parts (say, wooden planks again),
arranged in the same order. Then x is identical with y. But now
suppose that y undergoes the process described earlier, in which
all of its planks are gradually replaced, and in December a ship
z is constructed out of those planks, arranged in the same old
order. Then we do not want to say that x is identical with z, even

11. Several other possible counterexamples will be discussed in Chapter 7.

 SORTALS 71

though x's composition when it went out of existence was ap-

propriately similar to z's composition when it came into exist-
ence. Here y's prior coming into existence with x's old parts
already preempted x's identity and left no room for a compo-
sitional claim in behalf of z for x's identity.
Taking these points into account, and making explicit now
the dependence of the compositional criterion upon the sortal
rule (including the addendum), a formulation of the criterion
might be:
The Compositional Criterion. Where (the sortal rule would
have us judge that) an object x goes out of existence at t1, and
an object y comes into existence at a later time t2, a sufficient
condition for x to be identical with y is that (the sortal rule
would also have us judge that):
(1) Some sortal F is such that F is predicatively true of x at t1
and F is predicatively true of y at t2; and
(2) Some objects are such that they comprise a major portion
of x at t1, and they comprise a major portion of y at t2, and
they are similarly arranged in x at t1 and y at t2; and
(3) There is no object z such that z comes into existence at a
time t' between t1 and t2, and conditions (1) and (2)
(substituting "z" for "y" and "t'" for "t2" in these condi-
tions) would have us judge that x is identical with z.
This completes my formulation of the sortal analysis of physi-
cal persistence. The overall analysis is to be understood as im-
plying that a necessary and sufficient condition for an object's
persistence is that the object's career exemplify either the primary
criterion specified by the sortal rule or the supplementary com-
positional criterion. In the next two chapters I will raise two
rather different kinds of questions about the adequacy of this
analysis. My general assessment of it, however, is highly favor-
able. It is not easy to conceive of any situation, actual or imag-
inary, involving standard physical objects, which would consti-
tute a relatively clearcut counterexample to this analysis.
 3
The Basic Idea
of Persistence

I. A Question about Sortal-Relativity

ONE IMPORTANT difference between the simple continuity
analysis (discussed in Chapter 1) and the sortal analysis (dis-
cussed in Chapter a) comes out when we consider the following
question: Can we analyze our concept of a physical object's
identity through time without taking cognizance of what sort
of object we are dealing with? Obviously the simple continuity
analysis, which makes no reference to sortal differentiations,
implies an affirmative answer to this question. That analysis
attempts a wholly general and sortal-neutral account of physical
persistence, and, as we saw, fails drastically in so doing. A nega-
tive answer to the question seems, on the other hand, indicated
by the sortal analysis. The latter analysis is, to be sure, quite
general in a sense, since the sortal rule is a single comprehensive
formula which is intended to apply to physical objects of all
sorts. Evidently, however, we can apply the rule to an object only
insofar as we do take cognizance of what sort of object it is, for
it is only then that we can properly trace the object under a
sortal. Indeed the essential idea behind the sortal rule was that
we need to divide objects into different sorts before we can
adequately analyze our identity concept.
In the present chapter I intend to argue that there is an im-
portant, albeit limited, extent to which we can analyze our
concept of an object's identity without taking cognizance of
what sort of object it is. Though the simple continuity analy-

72
 THE BASIC IDEA OF PERSISTENCE 73

sis was shown to be totally ineffective I will propose a modi-

fication of that analysis which preserves its sortal-neutral char-
acter. This proposal will not, however, imply a repudiation of
the sortal analysis. For it is only the latter analysis which, I will
maintain, can provide a relatively complete and accurate account
of our fully elaborated concept of persistence. The proposed
sortal-neutral analysis, on the other hand, will be able to provide
no more than a partial account of our concept of persistence,
but one which can plausibly be regarded as capturing the basic
kernel of that concept. Thus, on the view which I am now going
to develop, it will be equally important to appreciate both the
scope and the limits of a sortal-neutral account of persistence.
One extreme position which I want to contest is that our con-
cept of persistence is at its very roots dependent upon sortal
differentiations. It is this extreme idea which Wiggins seems to
be expressing when he says that there could not be "any usable
account of what it is, in general, to make a mistake or avoid a
mistake in tracing [an object] a. ... To trace a I must know
what a is."1 He then explains that to know "what an object is"
in the relevant sense is to be able to apply a special kind of term
to the object, viz. a sortal. It is only by reference to an applicable
sortal that we can understand what it means to trace the object.
Wiggins is apparently claiming that our identity criteria are
dependent upon sortals in a very radical way. When he says
that we can give no usable account of our identity criteria in
sortal-neutral terms, this seems to imply something much stronger
than simply that we could give no completely accurate account
in such terms. He seems to be implying that we could not even
formulate a usefully close approximation to our identity criteria
without appealing to sortals, that we could not even formulate
an account which works for the most part.2
This position strikes me as intuitively quite implausible. I am
prepared to believe that our sortal classifications affect our iden-
tity criteria in various significant respects. But should it not also
be possible to formulate some underlying general rule of identity
which cuts through those sortal classifications, a rule which it
1. Wiggins, Identity and Spatio-Temporal Continuity, p. 35.
2. It may possibly be that Wiggins does not mean to imply this extreme
position, but the fact that his account of persistence is from start to finish
inextricably tied to sortals leaves the strong impression that he does.
74 THE CONCEPT OF IDENTITY

would be correct to follow, if not always, at least almost always?

When we consider what it means to trace the careers of such
serially diverse objects as, for example, trees, apples, cars, and
the wheels of cars, certainly our intuitive impression is that,
though there might be certain nuances of difference between
the tracing rules which we follow in these cases, there must also
be a significant common denominator running through all these
cases, one which could supply a usable, even if not absolutely
accurate, handle on all of them.
The implausibility of the extreme contention that our identity
criteria are totally dependent upon sortal differentiations can
be brought out by reflecting upon the following obvious fact.
A person will frequently be able correctly to trace the career of
a new sort of object without requiring any information as to
what the identity criteria are for that new sort. Certainly the
simplest (though, perhaps, not the only possible) explanation of
this fact is that the person is applying the same criteria to the
new case that he has already learned to use in the old cases. But
this implies that, contrary to the extreme position, there are
usable general criteria which cut through sortal divisions.
We can, for example, easily imagine a child raised on a farm
who knows a substantial amount of English but who has never
seen or heard of a car, or, to make the case even purer, has never
seen or heard of any transporting vehicle. He is now shown a
car for the first time in his life, say, a blue and white car moving
across an open field. There is no doubt that he is immediately
in the position to say such things as "That big blue and white
thing (with the four round black things on the bottom) is moving
across the field." In these circumstances it would seem natural
to assume that the child is using the expression "that big blue
and white thing (with the four round black things on the bot-
tom)" to refer to the same object that we might refer to as "that
car." And, what is critical in the present connection, having re-
ferred (in his way) to the car he seems perfectly competent to
trace the car as it moves across the field. Moreover there seem
to be no very obvious limitations on his ability to reidentify the
car over longer periods or in more complicated circumstances. It
seems quite certain that his ability to reidentify the car would
not be in the least stymied by such typical alterations of the car
 THE BASIC IDEA OF PERSISTENCE 75

as its doors opening and shutting, its wheels turning, or its wind-
shield wipers starting to move. He seems, in short, capable of
getting on quite well without the allegedly special identity cri-
teria relativized to the sortal "car." This is evidence that in fact
no such radical sortal-relativization is necessary, that in fact the
general concept of the identity of an object provides, if not all,
at least a significant part of the identity criteria that we ordi-
narily need.
Or consider another example. Suppose that you know of some-
one, say an Eskimo, who has never before seen a tree. Imagine
that you bring this person to a tree and say to him (in his
language), "Keep your eye on that object." If when you say
this you move your hands and eyes in the appropriately sug-
gestive way, if you make the appropriate "sweeping gesture of
ostension," as Quine calls it,3 there is every likelihood that you
can get your Eskimo friend to focus on the tree, rather than on
one of its parts or on some larger portion of the landscape. On
the other hand there is no reason to expect that by focusing on
one isolated tree he would thereby immediately come to under-
stand the general sortal concept of a tree, a general concept, that
is, which he would apply to all and only trees. Presumably he
may remain quite ignorant of that sortal concept until he is
shown some fair selection of trees and, perhaps, hears some single
word applied to them. But would his ignorance of the sortal
prevent him from tracing the tree, from "keeping his eye on it"
in an essentially normal way? This seems wholly implausible.
Even if you proceeded to break a leaf off the tree, or rub some
dirt on it, or bend one of its branches, there seems little doubt
that your Eskimo friend, despite his sortal ignorance, would cor-
rectly reidentify the tree through these changes. (This could be
shown by the answers he gives, or the pictures he draws, in re-
sponse to the question "What happened to the object I asked
you to keep your eye on?") Again, the simplest explanation of
what is going on here is that he is employing a concept of iden-
tity which does not, at least in any radical way, depend upon
sortals.

3. W. V. Quine, The Roots of Reference (Open Court Publishing Co., Illi-

nois, 1974), p. 53.
76 THE CONCEPT OF IDENTITY

I am not suggesting that someone's sortal ignorance is likely

to have no effect at all on his identity judgments. It may seem
immediately plausible to expect that sortal ignorance might some-
times generate what I earlier called nondrastic identity-devia-
tions, i.e., conflations of cases in which x and y are identical
with cases in which x goes out of existence and turns into y.
The child who does not know the sortal "car" might very well
judge that the blue and white object which he picked out per-
sists in a flattened form after it is subjected to the crushing ma-
chine. And there may even be some rather special cases, as I will
show shortly, in which sortal ignorance can lead to drastic iden-
tity-deviations. What seems quite incontrovertible, however, is
that someone who is presented with a new sort of object will by
and large be able to trace the object in an essentially correct
manner, even though he knows of no sortal under which to trace
the object.
Certainly there is no serious possibility that someone's sortal
ignorance could lead him to trace any of the completely aber-
rant-seeming paths which showed up in earlier discussions. We
are entirely confident, for example, that the child who has never
before heard of a car would undoubtedly describe the scene of
a car moving out of a garage in terms of the essentially correct
idea of an object maintaining its size while moving from inside
to outside, and would not, in those circumstances, trace a
shrinking object along the path of an "incar." Nor could we
seriously wonder whether the Eskimo's ignorance of the sortal
"tree" might perhaps lead him to trace an object along a path
which combines tree-stages and trunk-stages (where we are
imagining, of course, that the tree was not actually reduced to its
trunk). The path of the shrinking incar, and the path which
combines tree-stages and trunk-stages, evidently clash with our
ordinary identity concept in some general way which does not
depend upon the role of sortals. This is presumably why we are
confident that no one (or, perhaps more cautiously, no one who
operates with our ordinary concept of identity), regardless of his
sortal ignorance, would trace an object along those aberrant
paths. It must then be possible to formulate an analysis of our
identity concept which is independent of sortal differentiations
at least to the extent of enabling us to explain, in sortal-neutral
terms, why such paths are aberrant.
 THE BASIC IDEA OF PERSISTENCE 77

II. The Basic Rule

We are essentially back at the problem that we faced at the
beginning of Chapter 2, but with a new twist. At that point we
realized that simple continuity considerations could not even
exclude the obvious kinds of aberrant paths just mentioned, and
we sought some other considerations which would accomplish
this. The solution which then suggested itself was the sortal rule,
i.e., that an object's career must be continuously traced under a
sortal. At present, however, we have found reason to impose an
additional constraint on our problem, a constraint that was not
satisfied by the previous solution. We now need to elicit some
rule of identity which is independent of sortal differentiations
but which is nevertheless adequate at least to exclude the most
obvious kinds of aberrant paths, so that someone could rely on
this rule to judge correctly of an object's identity in most ordinary
circumstances without having to apply any sortal to the object.
It is not to be anticipated, however, that the sortal-neutral
rule now being sought will supersede the sortal rule, for it
seems rather clear already, and will become more clear as we
go along, that only the latter rule can adequately explain various
nuances of our identity judgments. The relationship to antici-
pate between the sortal-neutral rule and the sortal rule is rather
that of the basic core of a concept to its more fully sophisticated
elaboration. The former rule must capture, in a way that the
latter does not, that most basic and elementary idea of per-
sistence which we confidently expect anyone to employ regard-
less of his sortal ignorance.
Let us consider again the Eskimo looking at the tree, which we
will now imagine undergoes no qualitative change at all during
this stretch of observation. Suddenly he turns to us and says,
"The object I was looking at was first rather wide and very
oddly shaped, but then it changed and now it is much thinner
and cylindrically shaped." As he says this perhaps he also ges-
tures with his hands in a way which suggests first the outline
of a whole tree and then the outline of a tree trunk. In other
words, he has in effect judged that the original tree is now iden-
tical with the trunk. This, we are quite certain, could never
happen. But our question now is why not? What is the rule of
identity which we expect him to be following, and which would
78 THE CONCEPT OF IDENTITY

preclude that incredible judgment? It cannot be a rule so simple

as to be definable purely in terms of continuity considerations,
since we know that any number of perfectly continuous paths
could in fact be traced from the tree at one moment to the trunk
at a later moment. Nor, on the other hand, can it be anything
as sophisticated as the sortal rule which we expect him to follow,
since by hypothesis he lacks the conceptual resources for the
application of that rule. That is our puzzle.
When we reflect on this puzzle we might initially be inclined
to respond like this: "Why would he judge that the object
changed in size and shape when nothing happened?' This re-
sponse is, I think, essentially on the right track, but it is not
helpful as it stands. For if what is meant is that "nothing hap-
pened to the tree," well, then that just begs the question out-
right, since precisely what we want to understand is how the
Eskimo is in the position to make any correct judgments about
what happened or did not happen to that persisting object.
And if what is meant is that "nothing happened period," this is
just wrong, since doubtless many things happened of which the
Eskimo was aware when he watched the tree, e.g., a cloud may
have moved behind the tree.
What we intuitively want to say, I think, is: "But nothing hap-
pened to make him judge that the object changed in those ways."
The key words here are "to make him judge." Our intuition
seems to be that when someone traces the career of an object he
will not countenance a change in the object unless he has to. We
imagine that the Eskimo fixes his attention on that tall oddly
shaped object and then traces its career by following a continu-
ous space-time path that is as stable, as unchanging, as he can get
it to be. Perhaps he did see a cloud move behind the tree. But
there was nothing in that, or in anything else, which would force
him to give up the stabilizing judgment "The object is still the
same size and shape" in favor of the change-countenancing judg-
ment "The object changed its size and shape." This is why we
are certain that he would make the former judgment and not
the latter.
The basic sortal-neutral identity rule which we confidently
expect to govern the Eskimo's thought might then be put
roughly: Trace an object's career by following a spatioternporally
 THE BASIC IDEA OF PERSISTENCE 79

and qualitatively continuous path which minimizes changes as

far as possible.
We just saw how this rule explains the Eskimo's judgment in
the case where the tree actually suffers no qualitative change
while it is being observed. But the same explanation is also
straightforwardly applicable to many cases in which the tree
does change. Suppose that while the Eskimo watches, a leaf is
broken off the tree. We are certain that this change in the tree
could not induce the Eskimo to judge that the tree has turned
into the trunk. The change-minimizing condition immediately
explains this. For the condition requires that the Eskimo's iden-
tity judgments should minimize change as far as possible. What
must be assessed, therefore, are degrees of change, and this assess-
ment is at least quite often perfectly clearcut. When the leaf
is broken off the tree the Eskimo has no choice but to counte-
nance some degree of change in the tree, a change which we can
perhaps loosely characterize as involving an element of size and
shape. But he is in no way forced to countenance the relatively
greater change in size and shape that would be entailed by judg-
ing that the tree has turned into the trunk. This is why he would
never make that judgment.
When we try to apply the change-minimizing condition to the
other aberrant path we needed to exclude, the path of the
shrinking incar, we come up against a fairly serious complica-
tion. We would like to be able to say that anyone, no matter
what his sortal ignorance, would trace a car leaving a garage
in the ordinary way, and not in the incar-outcar way, because
the ordinary way involves countenancing less change. This may
seem obviously correct on the grounds that tracing the car in
the ordinary way does not involve countenancing any such al-
terations in size and shape as is suffered by the shrinking incar.
But what should we say about the fact that the incar is rela-
tively more stable than the car in the following respect: the
incar remains wholly inside a garage whereas the car does not?
Since the path of the incar minimizes change in at least this
respect how exactly does the change-minimizing condition dic-
tate that the car be traced in the ordinary way?
A somewhat ad hoc answer to this question, which seems
nevertheless to ring true, is that stability with respect to a merely
80 THE CONCEPT OF IDENTITY

locational property like being wholly in a garage simply does

not count in assessing which of two paths minimizes change. This
answer would seem considerably less ad hoc if it could be gen-
eralized to read: None of an object's relational properties (i.e.,
no facts about how an object is related to other objects) count
in applying the change-minimizing condition. We might then
even try to explain why an object's relational properties do not
count by connecting this proviso to the intuitive idea that an
object must be in some sense self-complete, which implies, per-
haps, that an object's (rule of) identity must not depend upon
how it is related to other objects. The generalized proviso, how-
ever, despite its element of attraction, is too problematical. For
one thing it is not sufficiently clear in practice where to draw
the line between relational and nonrelational properties, since
the ascription of almost any property might plausibly be regarded
as entailing a comparison between objects. Furthermore, we will
see in the next section that some rather special properties which
are pretty clearly relational do apparently count significantly in
making a change-minimizing judgment.
I think, therefore, that our most promising approach is simply
to lay down the more specific proviso that locational properties
do not count in applying the change-minimizing condition. We
might define a locational property as one which can be signified
by an expression of the form "being in such and such a spatial
relation to such and such an object." Thus: "being inside (on
top of, to the left of, in contact with) a (the) garage (roof, red
thing)." This notion might be further clarified by reference to
specific cases, as the need arises. We can say, perhaps, that an
object's locational properties constitute at any moment the ob-
ject's most obvious and direct relations to other objects. Hence
a loose connection might still be upheld between the intuitive
idea that an object must be self-complete, that its identity cannot
depend upon any other object, and the proviso that when we
trace an object's career along a change-minimizing path we dis-
count mere locational changes.
My proposal, then, is that our most basic idea of the persist-
ence of an object (which coincides approximately, but not ex-
actly, with the more fully developed idea) can be analyzed in
terms of the following sortal-neutral rule of identity.
 THE BASIC IDEA OF PERSISTENCE 8l

The Basic (Change-Minimizing) Rule. A sufficient condition

for a succession S of object-stages to correspond to stages in the
career of a single persisting object is that:
(1) S is spatiotemporally continuous; and
(2) S is qualitatively continuous; and
(3) S minimizes change (discounting mere change of location).
The basic rule might also be supplemented by a suitable ver-
sion of the compositional criterion. I will not work out this
detail.
Let us try to get clearer as to what the change-minimizing
condition (3) amounts to when we consider this in terms of
successions of object-stages. A succession S minimizes change if,
roughly, any divergence from S would involve more change
than S does. That is, S minimizes change if, for any succession
S' which is such that S and S' partly coincide and partly diverge,
S' contains more change at the time of divergence than S does.
Suppose S is the succession corresponding to the tree from 3:00
to 3:10, and S' is the aberrant succession which combines the
tree-stages from 3:00 to 3:05 with the trunk-stages from 3:05 to
3:10. Then S and S' coincide from 3:00 to 3:05 and diverge from
3:05 to 3:10. The "time of divergence" would then be 3:05. We
want to say that S' is aberrant because it contains more change
at the time of divergence 3:05 than S does. What this means is
that if we compare S's object-stages at times very close to 3:05
we find that they are more similar to each other than are the
object-stages of S' around 3:05.
It should be noted that it is only comparisons around the time
of divergence that matter. S' would obviously still be aberrant
even if it should seem reasonable to judge that over the entire
interval from 3:00 to 3:10 S' contains on the whole less change
than S does. This might happen if the tree undergoes various
changes from 3:05 to 3:10 whereas the trunk is relatively un-
changing during that five minutes. Then S' would contain a rela-
tively larger change than S at 3:05 but relatively less change after
3:05. Still S' is aberrant and S is change-minimizing, because at
3:05, the time of their divergence, S contains less change. Intui-
tively put, the idea is that when we trace an object's career we
evidently do not countenance a needless change at a given mo-
82 THE CONCEPT OF IDENTITY

ment just to assure greater stability at later moments. (The im-

plications of such a tracing strategy are completely alien to us;
e.g., if there were a very stable object in the neighborhood we
would try to trace all paths to that object.)
The change-minimizing condition (3) of the basic rule might
then be put somewhat more strictly as:
(3') For any succession S', if S and S' partly coincide and partly
diverge and t is their time of divergence, then object-stages
in S at times very close to t are more similar to each other
than are object-stages in S' at times very close to t (discount-
ing mere locational similarity).

III. Limitations of the Basic Rule

The basic rule does, I think, adequately accomplish the task
originally set for it, viz. to express in sortal-neutral terms that
most elementary conception of persistence which anyone can
rely on regardless of his sortal ignorance. We have seen how the
basic rule yields (what we ordinarily regard as) correct identity
judgments in several specific cases, and these cases could obvi-
ously be multiplied without difficulty. It seems fairly clear, in
fact, that these cases are representative of the vast majority. The
application of the basic rule yields identity judgments which
coincide for the most part with the ones which we actually make,
and this is why we confidently expect anyone to trace an object
in an essentially ordinary way even if he is unable to apply any
sortals to the object. It is now necessary to appreciate the im-
portant correlative point, that the basic rule provides only a
good approximation to our operative identity scheme, but not a
wholly accurate account of it. There is in fact a certain general
disparity between the judgments of identity which we would
make if we relied entirely on the basic rule and the judgments
we actually make relying on sortal-relativized criteria. This is
why the basic rule represents only a partial analysis of our iden-
tity concept, and needs to be completed by reference to the
sortal rule.
The most obvious disparity between the basic rule and our
actual identity judgments shows up in connection with some of
the cases in which objects are said to come into existence or go
out of existence. The rule that is evidently suggested (if not
 THE BASIC IDEA OF PERSISTENCE 83

entailed outright) by the basic rule is that we should prolong

an object's career, backwards and forwards in time, so long as
tracing a continuous and change-minimizing path allows. Or,
to put this in a form corresponding to the sortal rule adden-
dum,4 the basic rule suggests that, where S is a continuous and
change-minimizing succession, the beginning and end of S corre-
spond respectively to the coming into existence and going out
of existence of an object if and only if S is not the segment of a
longer continuous and change-minimizing succession.
Evidently this rule will not explain the kinds of judgments
that we make about objects going out of existence and con-
tinuously turning into other objects. The basic rule would yield
the (conceptually) incorrect judgment that a car persists in flat-
tened form when it passes through the crushing machine, be-
cause we can trace a continuous and change-minimizing path
from the original car to the block of scrap that emerges. Or, to
take another case, if a table is gradually whittled away so that
it eventually goes out of existence and turns into a small lump
of wood, the basic rule would have us judge instead that the
table persisted through a decrease of size.
Though there seems quite definitely to be a disparity between
the basic rule and the judgments that we actually make in these
kinds of cases (i.e., cases in which an object goes out of existence
by continuously turning into another object), it should be noted
that in many (perhaps in most) cases the basic rule yields the
correct judgments about objects coming into existence and going
out of existence. Suppose, for example, that a table is precipi-
tously smashed to pieces. In such a case it is plausible to judge, as
I earlier maintained, that there is no way at all to prolong the
table's career along a continuous path. Here the basic rule would
correctly enjoin us to judge that the object has gone out of
existence.
Another kind of case in which the basic rule would lead to
our judging correctly about an object going out of existence is
if the object vanishes by merging indiscriminately into its en-
vironment. In such a case we cannot continuously trace the object
along a path which has any claim to being change-minimizing.
If, for example, a number of cars are melted down into a single

4. See Chapter 2 above, p. 52.

84 THE CONCEPT OF IDENTITY

indiscriminate mass then, though we could preserve continuity

by arbitrarily identifying each car with some specified portion of
the mass, there would be no plausible way to regard these iden-
tifications as change-minimizing since any number of other ar-
bitrary identifications would be on an equal footing. Thus, in
general, if there is no path which can be plausibly regarded as
the distinctive change-minimizing one we can only say that the
object has vanished and no longer exists. We might regard as a
special case of this sort one in which an object continuously
diminishes until there is nothing left of it, e.g., where an ice
cube melts away.
At the level of the basic rule, then, objects go out of existence
(by breakage, burning, etc.) either when they are precipitously
rendered into fragments or when they vanish by merging into
their environments. These ways of ceasing to exist do represent,
I think, the most fundamental sense in which an object might
be said to go out of existence. But then there are also the other
kinds of cases mentioned, where an object is said to go out of
existence and continuously turn into another object, which the
basic rule cannot apparently explain. Now this limitation of the
basic rule seems fairly marginal, and would in any case merely
lead to nondrastic identity-deviations (i.e., judging that x is
identical with y when the strict truth, in ordinary terms, is that
x went out of existence and turned into y). There is, however,
a far more fundamental limitation of the basic rule which I now
want to explain.
Though the rule would enable us successfully to trace objects,
parts as well as wholes, in a wide variety of circumstances, it
suffers from a general, and ultimately destructive, kind of vague-
ness. For there will be too many circumstances in which the rule
will not clearly guide us in choosing which of a number of paths
is to be treated as minimizing change. The kind of case which
gives rise to this problem can be characterized generally as one
in which, starting from a given object, we can trace continuous
paths in different directions each of which minimizes change in
a different respect. The few rather simple problem cases which I
will now examine in order to illustrate this point are not in-
tended merely to defeat the letter of the basic rule as formulated.
For there would be nothing to prevent us from trying to retain
the essential sortal-neutral character of the rule while emend-
 THE BASIC IDEA OF PERSISTENCE 85

ing it with various additional provisos (indeed the exclusion of

locational properties is already such a proviso). But I think these
examples will show that any attempt to force the basic rule to
work leads to indefinitely mounting complications, and there
is finally no indication that we can formulate for even the most
common cases a sortal-neutral procedure, accurately matching
our actual identifying practices, for unambiguously choosing be-
tween conflicting paths which minimize change in different re-
spects. This is the fundamental limitation of the basic rule and
the essential reason why we eventually need to invoke the sortal
rule.
To begin with a very simple example, imagine that you have a
red table and you decide to paint half of it black (perhaps in-
tending to finish the job later). As you apply the black paint the
red expanse which initially coincides with the full extent of the
table gradually diminishes in size as it is encroached upon by a
widening black expanse. Now if you tried to trace the table
according to the basic rule you would be faced with the follow-
ing choice. You might decide that the way to minimize change
is to preserve as far as possible stability of color. You would then
judge that the original wholly red table remained wholly red
while gradually diminishing in size until it is now only half of
a table. Or you might decide that the correct way to minimize
change is to preserve as far as possible stability of size. You would
then make the ordinary judgment that the table's color has
altered.
Perhaps it will be suggested that this is not really a hard con-
flict since in tracing the table in the ordinary way we preserve
not only stability of size but also stability of shape, so that we
thereby minimize change in the greatest number of respects. We
might then consider adding to the basic rule the proviso that
where two paths minimize change in different respects we must
choose the path which minimizes change in more respects. But
it seems doubtful that this proviso could really even cope with
the present simple case, let alone more complicated ones. For
suppose that the black paint significantly alters the texture and
temperature of the surface to which it is applied. How should
we then weigh up all the stabilities preserved by one possible
path against all those preserved by its competitor?
This might still be dismissed as too easy a case since I have
86 THE CONCEPT OF IDENTITY

omitted a possibly crucial stability which we preserve in the or-

dinary way of tracing the table, viz. its stability as a separately
movable thing (that is, roughly, a detached thing that tends to
move together with its parts). The property of being separately
movable does, I think, strike us as possibly deserving to bear
some special weight in tracing an object's career. We might then
consider adding to the basic rule the proviso that stability of
separate movability outweighs any other stability in making a
change-minimizing judgment. Note, incidentally, that separate
movability seems a pretty clearcut relational property. This is
one of the properties that I had in mind earlier when I said
that we do not want simply to exclude all relational properties
from being taken into account in change-minimizing judgments.
But, aside from the fact that we will soon come across cases
in which stability of separate movability is not favored in tracing
an object, it is immediately obvious that any special appeal to
this specific stability could not possibly provide a general solu-
tion to our difficulty. For we would still have no way of deal-
ing with a large variety of objects that are not, at least in any
perfectly straightforward way, separately movable: for example,
a fence, or a radiator, or a tree, or the wheel of a car, or, for that
matter, a table that, is securely fastened to something else (e.g., to
the floor). We certainly cannot rely on any general presumption
that objects either are, or must remain, detached from other
objects.
It might still be suggested that the kind of change-minimi/ing
conflict which could arise in the case of painting an object can
easily be resolved by simply adding a proviso to the basic rule
that minimizing change in size and shape weighs more than mini-
mizing change in other respects. However, this suggestion will
again fall far short of providing a generally applicable inter-
pretation of the basic rule, since the mentioned proviso could
certainly not qualify as a general principle, even if it works well
enough for the specific kind of example just considered. There
are many other examples in which our ordinary tracing proce-
dure shows no bias toward preserving size or shape. Consider,
for example, what happens when you add bumpers to a car
which previously had none. To make this vivid suppose that the
bumpers are imposingly large, prominently curvacious, and
colored conspicuously different from the rest of the car. Here our
 THE BASIC IDEA OF PERSISTENCE 87

ordinary identity criteria determine the judgment that the car

which was orginally bumperless is now bebumpered, with the
necessary implication that the car has altered somewhat in size,
shape, and color distribution. But at the level of the basic rule
we could just as well have traced the original car along a path
which perfectly preserved size and shape, as well as color distribu-
tion. We could have done this simply by judging that the original
car is now sandwiched between the bumpers but does not contain
them as parts. Our operative identity criteria, which are rela-
tivized to the sortal "car," determine us to trace a path which
might be said, in crudely unanalyzed terms, to preserve the
object's stability as a whole car. However, from the sortal-
neutral point of view of the basic rule there could be no decisive
reason to favor this stability over all the others that could be
preserved in the alternative way of tracing.
It is this last sort of case, in which objects are attached to each
other, that presents what is perhaps the most serious difficulty
for the basic rule.5 If the small object y is attached to the larger
object x to yield the composite object x-with-y the basic rule
will not clearly instruct us whether or not we are entitled to trace
a continuous path which would identify x-with-y with the orig-
inal x. (Presumably the identification of x-with-y with the
smaller y can be ruled out on grounds of insufficient continuity.)
Whether this identification of the composite object with the
larger original component is permissible can only be determined
by reference to the sortal under which we are tracing the ob-
ject: it will depend upon whether we can treat both x and
x-with-y as coming under the same sortal. Though there is un-
questionably a great degree of potential latitude in this decision,
and we can think up any number of borderline cases, at least with
respect to many typical and obvious cases a sortal concept will
provide a basis for deciding about the composition of an object
that is brought under the sortal. Thus we identify the car-with-
bumpers with the original car-without-bumpers, and accordingly
judge the car to have altered in size and shape, because we
treat both the object-with-bumpers and the object-without-
bumpers as coming under the sortal "car"; but we would almost.

5. Actually an exactly parallel difficulty arises when objects are detached

from each other. Cf. the discussion later about the trunk-tree.
88 THE CONCEPT OF IDENTITY

certainly not say that the car altered in size and shape if, say, a
small trailer or sled was attached to it, since we would not
normally think of a trailer or sled as entering into the composi-
tion of a car. Nor, pretty obviously, would we say that a tree has
altered in size or shape if some car bumpers are nailed onto it,
since the car bumpers would not naturally be treated as part
of the tree's composition. Here, at least, are judgments of iden-
tity that are in a fairly compelling sense sortal-relative.
It might be urged that these latter cases actually suggest a
more severe and perhaps fatally ubiquitous limitation of the
basic rule. For should we not say that even where the small ob-
ject y is merely brought into contact with the larger object x
but not attached to it (e.g., an ashtray is placed on a table) the
basic rule will not clearly instruct us whether or not we should
identify the composite x-with-y with the original x (e.g., whether
or not we should judge that a table gets bigger when an ashtray
is placed on it)? But this difficulty is, I think, not really serious.
To deal with cases in which y is merely brought into contact with
x but not attached to it we can safely lay down the general
proviso (which may no doubt admit of a few exceptions) that
tracing a change-minimizing path should not involve identify-
ing the cohesive x with the fragmented x-with-y. The property
of cohesiveness, by which I mean roughly being able to with-
stand various typical strains without coming apart, may seem
intuitively quite important in tracing an object's career (though
this too is apparently a relational property). But in cases where
y is attached to x there is apparently no property which we
could properly single out to resolve the change-minimizing con-
flicts which would typically arise.
I suggested earlier that someone, perhaps a child, who did not
know the sortal "car" could still pick out a car as, e.g., "the big
blue and white thing," and successfully trace its career in many
ordinary circumstances. We are now in the position to under-
stand why this is so. In many ordinary circumstances all that
the child will need to trace the car is the basic rule. But we can
now also take note of the kind of limitation which his ignorance
of the sortal will impose upon the child's tracing ability.
Imagine indeed that for the winter a small yellow sled is at-
tached to the back of the car. (It will not, I think, necessarily
matter whether or not the child knows the sortal "sled," but to
 THE BASIC IDEA OF PERSISTENCE 89

simplify let us assume that he does not.) How will the child
see this? Will he see it, in the way we do, as the original blue
and white thing maintaining its size and color but being at-
tached to the yellow thing? Or will he see it as the original
blue and white thing becoming bigger and partly yellow, i.e.,
as now containing the yellow thing as a part? The basic rule
provides no definite guidance. Tracing the first way, as we do,
preserves stability of size and color, but tracing the second way
preserves (what may seem important) stability of separate mov-
ability. (Note that this is a kind of case in which the ordinary
way of tracing does not favor separate movability.) Which of
these two ways of reidentifying the object might strike the child
as most natural is unclear. There is here a genuine illusion of
clarity because the grip of the sortal on our thought prevents
us from experiencing the conflict which could be generated in
such a case at the level of the basic rule. But that a real potential
for conflict does exist can scarcely be questioned once one stops
to compare this case, of attaching a sled to the car, with the case
of attaching bumpers. Our sortal-relative criteria determine a
relatively clear (though not, certainly, an absolutely exact) basis
for making discrepant identity judgments in these cases, but it
seems evident that from the child's sortal-neutral vantage point
there can be no relatively clear difference between the cases
(though there may be any number of obscure differences which
might point him in one direction or the other). Both cases
essentially leave the child with an option that is only properly
resolved by appealing to sortal-relative criteria.
It is worth noting that some cases of change-minimizing con-
flict correspond to, and in a way explain, a feature of our ordi-
nary identity criteria which was discussed in the last chapter,
viz. that sometimes in tracing an object we will branch off in two
different directions under two different sortals. These kinds of
cases were described as involving (strictly) two objects of differ-
ent sorts occupying the same place at once. If you chop off all
of a tree's branches so that all that is left of it is its trunk, then,
as we saw, in a sense the tree and the trunk are now one, but in a
strict sense they are not since they have different histories, the
tree having once been larger and the trunk never having been
larger than it is. If we trace the trunk-tree backwards in time in
order to determine its past we reach a conflict point where we
go THE CONCEPT OF IDENTITY

can trace along different paths in order to preserve different

stabilities. These different paths in fact correspond to tracing
under the sortal "trunk" and tracing under the sortal "tree."
Tracing under the first sortal (i.e., determining the history of
the trunk) allows us to preserve stability of size, whereas tracing
under the second sortal (i.e., determining the history of the tree)
allows us to preserve, roughly, boundary sharpness (i.e., the tree,
as compared with the trunk, always has a boundary which is
relatively more sharply discriminated from its environment).
In the last case our sortal-relative criteria resolve the conflict
in a compromise fashion, by allowing us to trace both paths
under different sortals. In other (perhaps in most) cases (e.g., the
case of painting the table and the cases of attaching the bumpers
or the sled to the car) the sortal criteria resolve the conflict by
choosing one path and discarding the other. But the basic rule
would leave us essentially stranded in all of these cases.
And that, I suggest, is the primary reason why we ultimately
need the sortal criteria properly to fill out our identity concep-
tion. What we have seen in the few cases examined, and could
see in any number of other cases, is that the sortal rule enables
us to trace objects through those junctures at which the basic
rule is helplessly vague. My view is that the sortal rule is in es-
sence nothing more than a clarification, a refinement, of the basic
rule. The excessively vague idea of tracing a continuous path
that minimizes change now gives way to the relatively clearer
idea of tracing a continuous path under a sortal concept. But
these two ideas are not logically independent. When we trace
under a sortal we are ordinarily tracing a path which incon-
trovertibly minimizes change; and even when we allow the sortal
to guide us through points of change-minimizing conflict we
nevertheless continue to trace a path which minimizes change
in a certain respect (i.e., in respect of satisfying the sortal con-
cept). The sortal, we might say, orients us toward an object in
terms of a specific viewpoint that clarifies for that object which
stabilities count, and which do not, in minimizing change.

IV. Refining the Basic Rule

I remarked a moment ago that in many cases the sortal rule re-
solves change-minimizing conflict by choosing one path and dis-
 THE BASIC IDEA OF PERSISTENCE 91

carding the other. Now this assessment seems evidently correct

at least if we limit ourselves to tracing under such standard
and relatively unquestionable sortals as "table" and "car." But
it might be suggested that the situation would be considerably
altered, in the direction of possibly expanding the legitimate
scope of the basic rule, if we allowed into play such marginal or
borderline sortals as "patch of red." For we might then construe
the basic rule as implying that in cases of change-minimizing
conflict we are allowed to trace paths in both directions, and
associate with these paths (strictly) different objects that tem-
porarily occupy the same place. And if we are sufficiently per-
missive in counting terms as sortals might it not turn out that
the basic rule, thus construed, yields judgments that conform
even in many cases of change-minimizing conflict with our
sortal-relative judgments? It will be instructive to examine the
implications of this suggestion.
If we interpret the basic rule in the manner just suggested then
in the case of painting the black table red we would be allowed
to trace both a path which preserves size and shape and a path
which preserves color. The first of these corresponds to tracing
an object under the standard sortal "table" and the second, it
is now being suggested, might be said to correspond to tracing
an object under the borderline sortal "patch of red." So it looks
as if the basic rule is working well.
In the case of adding bumpers to the car the basic rule, as now
construed, would allow us to trace both a path which preserves
size and shape and a path which preserves separate movability.
The second of these corresponds, we know, to tracing under the
standard sortal "car." Can we find some sortal, even a marginal
one, which would allow us, in terms of the sortal rule, to trace
the first path? This is not easy to answer. Possibly the term
"portion of a car other than the bumpers" might qualify. If we
did treat this term as a sortal then we could say that when we
add bumpers to a car some persisting object, viz. the portion of
the car other than the bumpers, maintains its size and shape
while being sandwiched in between the bumpers.
It must be noted (and this is a point to which I will return in
the next chapter) that if we do treat "portion of a car other
than the bumpers" as a sortal, then we have to be prepared to
allow that the "object" which we trace under this term can
92 THE CONCEPT OF IDENTITY

retain its identity while altering its material composition. (We

would have to distinguish, then, between that portion of the car
and the matter which, at a given moment, makes it up.) Sup-
pose, for example, that after you add the bumpers to the car
you also add a door that was missing. If you trace an object un-
der "portion of a car other than the bumpers" you would have
to say that this object (in contrast with the car) did not gain
any parts when the bumpers were added to the car, but that
(like the car) it did gain a part when the door was added. This
sounds very dubious, but is perhaps not entirely out of the
question.
In general it seems difficult to evaluate the possible sortalhood
of various constructions involving words like "portion" and
"part." Consider, e.g., "crumpled part of a shirt," "corner (por-
tion) of a table," "top (part) of a tree," "thin part of a piece of
clay." If we do trace persisting items under such terms we are
certainly not intuitively inclined to think of these items as per-
sisting objects, perhaps because their identities depend too bla-
tantly on their relations to the objects of which they are parts.
Some of these items we are inclined to treat as places on objects,
especially those whose identities depend on where they are lo-
cated in objects (e.g., the top of a tree is a place on the tree).
But the logic and status of these constructions out of "portion"
and "part" would have to be examined more closely than I can
now undertake.
If "portion of a car other than the bumpers" is allowed as a
marginal sortal we might go on to consider whether we can find
marginal sortals to cover the various other paths which showed
up in change-minimizing conflict (e.g., the path of the car-cum-
sled). Instead of pursuing the point in that direction, however,
I want to explain the decisive and important reason why this at-
tempt to enhance the scope of the basic rule is unsuccessful.
Imagine now the following variation of the case of painting the
table. On Monday you have a wholly red table. On Tuesday
you paint half of it black. Then on Wednesday you change your
mind and repaint it all red. And then, finally, on Thursday you
change your mind again and paint it half black. Consider the
following statements:
 THE BASIC IDEA OF PERSISTENCE 93

(a) Throughout these operations, some object maintained a

constant size from Monday through Thursday.
(b) Throughout these operations some object got smaller on
Tuesday, and then it got larger on Wednesday, and then it
got smaller on Thursday.
(c) Throughout these operations some object got smaller on
Tuesday, and then it got larger on Wednesday, and then it
maintained a constant size on Thursday.

(a) is the standardly correct description of what happened to

the table, (b), it is being suggested, is a marginally correct de-
scription of what happened to the patch of red. But (c) is very
definitely not a correct description of any object of any sort
whatever.
If, however, we construed the basic rule as allowing us to trace
both paths in any change-minimizing conflict (c) would be a
correct description. For there would then be nothing to prevent
us from resolving our conflict on Tuesday by choosing the path
which stabilizes color, and then resolving our conflict on Thurs-
day by choosing the path which stabilizes size. The reason why
we cannot do this is decisively beyond the scope of the basic
rule. The only way that we might try to emend the rule to deal
with this point is to assume that we have initially specified some
relatively clear list of distinguishable properties, and the rule tells
us that when we trace a given object's career we must consistently
resolve change-minimizing conflict for that object by stabilizing
the object with respect to some particular one of those proper-
ties (and that we are not allowed to shift from one property to
another in the course of tracing that object's career). But that
is in all essentials the sortal rule. For the essence of the sortal
rule, and its definitive refinement of the basic rule, is that we
associate with any object some specifiable sortal-property under
which its career is consistently traced. Hence this effort to force
a more accurate fit between the basic rule and our ordinary
identity judgments in effect transforms the basic rule into the
sortal rule. Insofar as the basic rule is allowed to retain its ele-
mentary sortal-neutral character we must conceive of it as not
containing anything so elaborate as a specifiable list of properties
94 THE CONCEPT OF IDENTITY

under some one of which any object must be consistently traced.

In whatever way we might then try to conceive of the rule's ap-
plication to change-minimizing conflict (either as requiring us
to choose one path or as allowing us to trace both paths) the
rule surely cannot provide a reasonably accurate account of our
ordinary judgments in such cases.
The primary disparity between the sortal rule and the basic
rule, then, is that these rules yield different results in cases of
change-minimizing conflict. This disparity, however, carries with
it the derivative and somewhat more obvious one discussed
earlier. The sortal rule implies that an object goes out of exist-
ence at the moment when we can no longer apply to it the sortal
under which its whole career is traced. We can, admittedly, shift
from one phase (restricted) sortal to another (e.g., from "red car"
to "green car") as an object passes through different phases, but
the object's whole career, from beginning to end, must be traced
under one substance (unrestricted) sortal (e.g., "car"), and the
moment this sortal can no longer be applied the object's career
must be extended no further. But this, as we noted earlier, can
easily happen in such a way that, from the point of view of the
basic rule, there is no reason at all to judge that the object has
gone out of existence. If a car, for example, is crushed into a
block of scrap metal, but suffers no discontinuous breakage in
the process, then the basic rule would have us say that it still
exists in a different form, whereas the sortal rule forces us to say
that, if it is no longer a car, it no longer exists.
The sortal rule is the operative one, but we can still quite defi-
nitely sense in such a case the latent pull of the basic concep-
tion. For one thing, people (philosophers not excluded) are sim-
ply not that quick to admit (if they ever do) that the car has to
go out of existence just because it ceases to be a car. Furthermore
we feel distinctly inclined in cases like that of the crushed car to
stretch the sortal as far as possible to keep our identity judg-
ment in line with the basic conception. ("Well, it's still a car
in some sense.") Eventually though, despite the understandably
opposite inclination, I think we must yield to the pressure of the
operative sortal criteria and say that we are no longer presented
with a car and therefore the car no longer exists.
A tendency to avoid in discussing these matters is that of im-
 THE BASIC IDEA OF PERSISTENCE 95

plying that the central role of our sortal criteria is precisely

this, to have us judge that an object ceases to exist when it ceases
to be the same sort. This makes it sound as if the sortal rule
strikes like an arbitrary bolt from above to drive objects to an
early doom. That an object ceases to exist at all is only a de-
pressing corollary of the logical conditions of its persistence
through change. And the central role of our sortal criteria is to
clarify and complete those conditions, which are only vaguely
and partially given by the basic rule. A somewhat earlier demise
is only the necessary price which an object sometimes pays for
having enjoyed a logically more refined mode of persistence.
The upshot of this discussion is that while the sortal rule pro-
vides a relatively more accurate account of our fully developed
identity concept, it is the basic rule which ought to be seen as
providing our most fundamental standard of what the persistence
of an object consists in. This interplay between the two rules
does not, perhaps, imply any rigid consequences, but does sug-
gest certain general tendencies, which are readily apparent in
our language. It suggests, for one thing, a general restriction on
the range of concepts which are apt to figure as sortals. In order
for a concept to be apt for sortalhood it must be, as we already
know, nondispersive.6 But it must also be such that when we
typically trace a continuous path under it we are tracing a path
which, at least from some intelligible viewpoint, minimizes non-
locational changes. This is why the introduction into our lan-
guage of a sortal like "incar" (or, even worse, a sortal like "tree
that is being rained upon or trunk that is not being rained
upon") would be, if not outright incoherent, at least conceptu-
ally jarring in the extreme. The reason is that a shrinking path
traced under "incar" fails to minimize nonlocational change in
any respect which seems remotely conceivable. Terms like "top
of a tree" or "middle of a table," which include a locational ele-
ment, may perhaps qualify as exceptions to this general tendency,
though, as noted earlier, such terms are at best marginal sortals,
and the items traced under them are not naturally regarded as
proper objects.
Another general tendency implied by the latent presence of the

6. Cf. above, p. 4off.

96 THE CONCEPT OF IDENTITY

basic rule is the logical pressure, mentioned earlier, for us to keep

our sortal-level judgments about "going out of existence" more
or less in line with the basic-level ones. It certainly strikes us as
being a rather fundamental feature of our conceptual scheme to
think of objects as persisting through far and away the large
majority of changes that they can suffer. Our language does not
proliferate highly specific substance sortals in a manner which
would force us to judge objects to go out of existence whenever
they change slightly. This is why even if someone has no idea
what sortal applies to an object he can, for the most part, con-
fidently trace it in accordance with the basic rule. So long as he
traces a path which unambiguously minimizes change he need
not be overly concerned with the merely remote possibility that
what he identifies as a persisting thing has, according to the
operative sortal criteria, gone out of existence and been replaced
by something else.
In explaining that the basic rule is fundamentally limited
with respect to cases of change-minimizing conflict, I have in
effect been suggesting two rather different kinds of points. The
clearest sense in which the rule is limited is that it cannot be
made to coincide exactly with our actual identity judgments.
But the rule seems also limited in the sense that if we did ac-
tually base our identity judgments on it, these judgments would
be considerably more vague and unwieldy than the judgments
that we actually base on the sortal rule. This is because the basic
rule, even when buttressed with additional provisos, must re-
main relatively problematical in its application to cases of
change-minimizing conflict. Now such cases may be in one sense
rare, since they perhaps occur in only a small proportion of the
times that make up an object's duration. (Consider the small
proportion of times in a car's total duration that the car is
painted, or a sled is attached to it, or a bumper is changed, or
anything else happens which might plausibly be regarded as
giving rise to change-minimizing conflict.) In another sense, how-
ever, such cases are common, since they occur repeatedly and
regularly. If we tried to rely entirely on the basic rule these cases
would regularly infect our thought and communication about
the identity of objects with a far greater degree of obscurity and
indefiniteness than our ordinary identity criteria tolerate.
 THE BASIC IDEA OF PERSISTENCE 97

V. Unity through Time and Space

In developing the several foregoing analyses of persistence I was
always imagining a situation in which an object has already been
picked out at a given moment, and what we wanted was an
explanation of the identity rule in terms of which the object's
career is traced through time. But another question that we might
ask is what the rule is for picking out objects at a given moment.
When does an aggregate (or collection) of matter presented at a
given moment count as a unitary object? This is a question about
an object's unity through space which might be seen as paral-
leling, at least to an extent, the question about unity through
time which has so far concerned us.
The question about spatial unity naturally leads to distin-
guishing several senses of "(physical) object" ("body," "thing,"
"entity"). There may possibly be a completely permissive sense
of the word "object" which applies in fact to any aggregate of
matter, however spatially discontinuous, and our question about
the nature of an object's spatial unity would evidently not arise
with respect to this sense (if there is such a sense). A criterion of
spatial unity first comes into play with respect to the less per-
missive, but still very broad, sense of "object" that applies to any
continuous portion (mass, bit) of matter. Hence:
(1) An aggregate of matter constitutes a single object, in the
sense of a single portion (mass, bit) of matter, if and only if
the aggregate is spatially continuous.
(1) explains how the parts of an aggregate or collection of mat-
ter must hang together in order to comprise a single object, in
one sense of that word. If we assume the geometrical notion of a
continuous curve we can define spatial continuity as follows: x
is spatially continuous if and only if any two parts of x can be
connected by a continuous curve every point of which touches x.
A somewhat simpler definition to the same effect is: x is spatially
continuous if and only if x is not exhaustively comprised of
two parts that neither touch nor overlap each other.
I am inclined to think that the broad sense of "object" de-
fined by (1) deserves to be called an ordinary sense of that word.
On the other hand there is no doubt that the word is standardly
98 THE CONCEPT OF IDENTITY

employed in a considerably narrower sense. Imagine that a child

sees two socks tied together and asks, "What is that thing?" We
would no doubt want to correct him by showing him that "it is
really two things, and not one thing." The way to understand
this, perhaps, is that though the two connected socks do consti-
tute one thing in a sense (i.e., in the sense of (1)), they do not con-
stitute one thing in the sense which allows for an answer to the
question "What is that thing?" since to answer this question
ordinarily involves telling what sortal applies to the thing. We
might say, therefore, that the socks do not constitute a standard
object, where this notion is explained as follows:
(2) An aggregate of matter constitutes a single object, in the
standard sense of the word, if and only if the aggregate is
spatially continuous, and some sortal applies to it.
I want to understand (2) in such a way that something is a
standard object only if some relatively clearcut sortal applies to
it. Hence cars, tables, and trees are standard objects, as are the
trunks and branches of trees. Perhaps we can also allow such
things as pools of water and splinters of wood. But I would not
count something as a standard object just because of the applica-
tion of some such highly questionable sortal as "patch of brown"
or "portion of car between the bumpers."
That some reidentifiable aggregates of matter may not be
standard objects (may not have any sortals apply to them) fore-
bodes a difficulty, viz. if no sortals apply to them how are their
careers traced under the sortal rule? I shall return to this ques-
tion in the next chapter.
It might now occur to someone to suggest that given the
concept of a standard object the simple continuity analysis can
be resurrected as at least an approximately correct analysis. That
analysis stated that any continuous succession of object-stages
corresponds to the career of a single object. We know that this is
completely wrong if we take "object" in the broad sense which
applies to any spatially continuous portion of matter. On the
other hand it may seem approximately correct to say that any
continuous succession of standard object-stages (i.e., stages of
standard objects) corresponds to the career of a single standard
object. Apart from the special case of one object turning into
another, exceptions to this principle may be fairly rare, e.g., in
 THE BASIC IDEA OF PERSISTENCE 99

the case of a tree with one small branch where an aberrantly

continuous succession can be traced combining stages of the tree,
which is one standard object, with stages of the trunk, which is
another standard object.
But this suggestion does not really show that the simple con-
tinuity analysis can give us any independent insight into the con-
cept of object-identity. For the idea of a standard object must
be explained (as in (2)) in terms of the idea of a sortal. Conse-
quently the simple continuity analysis, insofar as it aspires to
operate independently of the idea of a sortal, cannot avail itself
of the idea of a standard object. It can only be understood as
making the drastically mistaken claim that, in the broad sense
of "object," any continuous succession of object-stages corre-
sponds to a single object.
It is perhaps worth noting that there seems to be a sense of the
word "object" even narrower than "standard object." We may
be inclined to say that the trunk of a tree, and even perhaps the
attached bumper of a car, are in some sense not (whole) ob-
jects but "merely parts of objects." Here we seem to be using
"(whole) object" to mean something like "standard object that
is not part of any standard object."
I want to turn now to a consideration of various similarities
and differences between the unity through time of a standard
object and its unity through space. Both modes of unity seem
to involve (a) an element of continuity, and (b) an element of
sortal-coverage. The temporal unity of a standard object, as ex-
plained by the sortal rule, is constituted by spatiotemporal and
qualitative continuity under a single sortal. And the spatial
unity of a standard object, as explained by (2), might be said
to consist in spatial continuity under a sortal.
One difference that immediately meets the eye is that, given
what I just said, there seems to be nothing corresponding to
qualitative continuity in the case of an object's unity through
space. Now this point is actually quite tricky since it is by no
means clear what could be the spatial analogue of qualitative
continuity (i.e., continuity of qualitative change). Instead of
trying to work this analogy out I will simply make a few obser-
vations about an object's qualitative makeup.
Certainly an object need not be homogeneous with respect to
such qualities as color or texture. The cushion of a chair may
1OO THE CONCEPT OF IDENTITY

be soft and white while the legs of the chair are hard and black.
Nor does such qualitative heterogeneity require that some por-
tion of the chair exemplify qualitative states intermediate be-
tween the contrasting ones, in the sense in which grey is inter-
mediate between white and black. It may, of course, be that no
portion of the chair is semi-hard or grey. But notice, and here
is one tricky point, that obviously many portions of the chair,
like the chair itself, would have to be partly hard and partly
not hard, and partly black and partly not black. (So perhaps
we would have, as a necessary corollary of spatial continuity,
something analogous to qualitative continuity.)
The possible disanalogies between the continuity elements of
the definitions of temporal and spatial unity concern me less
than the fundamental difference in the sense in which these two
modes of unity involve sortal-coverage. The role played by sortals
in a standard object's unity through time turns out, when one
considers it carefully, to be quite different from the role played
by sortals in a standard object's unity through space. (2) says
that a continuous aggregate x constitutes a single standard ob-
ject if some sortal applies to x. But the sortal rule does not say
that a continuous succession S corresponds to a career if some
sortal applies to S as a whole. Rather the rule imposes a condi-
tion on all of the object-stages in S, viz. that they must all be
F-stages, for some sortal F. Now an object-stage is a (temporal)
part of an object's career, and the spatial analogue of an object-
stage would evidently be a (spatial) part of an object. The sortal
rule says that, where F is a sortal, any continuous succession of
F-stages (i.e., any continuous succession each of whose members
is a stage of some F-thing or other) constitutes stages of a single
persisting F-thing. The proper spatial analogue of this claim
would have to be something like:
(3) Where F is a sortal, any spatially continuous set of F-parts
(i.e., any spatially continuous set each of whose members is
a part of some F-thing or other) constitutes parts of a single
F-thing.
(3), however, is clearly false. Consider, for example, what (3)
implies if we take "car" as the sortal F: "Any spatially con-
tinuous set of car-parts constitutes parts of a single car." This
is false since the set in question might contain the parts of two
 THE BASIC IDEA OF PERSISTENCE 1O1

cars that touch each other (or that are even attached to each
other). Then the set would be spatially continuous, and each of
its members would be part of a car, but the set would certainly
not constitute parts of any single car.
The reason why (3) is false is that the spatial analogue of the
principle of prolongation does not generally hold. The principle
of prolongation said that, where F is a sortal, one F-thing can-
not go out of existence just to be replaced, without any loss of
spatiotemporal or qualitative continuity, by another F-thing.
The spatial analogue to this (if we may ignore qualitative fac-
tors) is the false principle that one F-thing cannot spatially be-
gin where another F-thing ends off. But certainly one car can
begin where another car ends off. In general where F is a sortal,
two F-things might certainly touch each other. (3) implies that
this can never happen (or, more strictly, that if it does happen
then the two F-things would have to be parts of some single
F-thing). There may perhaps be certain special sortals with re-
spect to which the principle expressed by (3) does hold. It may
be, for example, that if two pools of water come into contact
they necessarily form a single pool of water. But the principle
certainly does not hold with respect to most sortals.
The reason why the principle fails for most sortals is fairly
obvious. A typical sortal F is associated with the idea of a certain
kind of size, or a certain kind of shape, or a certain kind of
internal makeup which an object must exemplify in order to
qualify as an F-thing. There is therefore no reason to suppose
that when two F-things are brought into contact they will add up
to something which has that kind of size or shape or makeup.
On the other hand there seems typically to be no temporal
analogue of size or shape or makeup which is conceptually re-
quired of the career of an F-thing, at least none such as to allow
us to think of one F-thing being continuously replaced by another
F-thing.
This difference between a standard object's temporal unity and
its spatial unity is closely related to another one, which Quinton
expresses as follows: "The temporal parts of an enduring thing
would have been a perfectly good thing of that kind if they had
existed on their own, without the other phases which in fact
preceded and followed them, while this is very seldom true in
the analogous spatial case: the spatial parts of a thing, conceived
1O2 THE CONCEPT OF IDENTITY

as existing in spatial disconnection from each other, are not

things of the same kind."7
Let me try to clarify Quinton's point by reformulating it in
slightly different terms. To begin with, I will use the expres-
sion "part of an F-thing" (e.g., "part of a car") in such a way
as to apply also to a whole F-thing (so that, e.g., a whole car is
part of a car). And, similarly, I will use the expression "stage in
the career of an F-thing" (e.g., "stage in the career of a car") to
apply also to the whole career of an F-thing (so that, e.g., the
whole career of a car is a stage in the career of a car). Now where
p is part of an F-thing (at a given moment) I will say that p is an
intrinsic F-part (at that moment) if what qualifies p as part of
an F-thing does not depend upon p's relationship to other F-parts
outside it. A whole car is an intrinsic car-part because what
qualifies it as a car (and hence as a car-part) does not depend
upon its relationships to car-parts outside of it (though it would
depend upon its part-whole relationships to car-parts inside it).
Similarly, where s is a stage in the career of an F-thing s is an
intrinsic F-stage if what qualifies s as a stage in the career of an
F-thing does not depend upon s's relationships to F-stages at
earlier and later times. The contrast between an object's spatial
unity and temporal unity might now be expressed by saying that
whereas typically a part of an F-thing will not be an intrinsic
F-part, typically a stage in the career of an F-thing will be an
intrinsic F-stage.
This contrast is perhaps best regarded as one of degree. Cer-
tainly many large parts of an F-thing will qualify as intrinsic
F-parts. The portion of a car between the bumpers, for example,
is certainly an intrinsic car-part, since this portion would con-
stitute a car-part no matter how it was related to other car-parts
outside it. (If it were spatially disconnected from any car-parts
outside it, it would qualify as a whole car, and hence as a car-
part.) On the other side it seems not entirely clear that every
stage in the career of an F-thing will be an intrinsic F-stage. It
would seem at least arguable that in some cases an F-stage quali-
fies as an F-stage only because it is suitably connected to F-stages
at other times. Indeed this point seems trivially correct if we
7. Quinton, The Nature of Things, p. 77. Compare with Whitchead's notion
of a "uniform object" in The Concept of Nature (Cambridge University
Press, London, 1920), p. 162.
 THE BASIC IDEA OF PERSISTENCE 103

take into account such phase sortals as "car which is in the proc-
ess of moving from New York to California." Taking "H" as
the abbreviation for this sortal, it seems clear that some hour
long H-stage qualifies as an H-stage only insofar as it is suitably
related to earlier and later H-stages.
Even if we focus entirely on high-level substance sortals like
"car" and "tree," which is the more important case, it is perhaps
still possible to maintain that the application of such a sortal to
an object may depend in part on the object's behaving in certain
ways distinctive of that sort of thing (or, perhaps more plausibly,
on the object's not behaving in ways incompatible with that sort
of thing), so that to qualify under the sortal an object-stage
might have to be suitably related to other object-stages in such
a manner as to add up to the required form of (non-)behavior.
On the other hand it seems quite unclear to what extent such
criteria of behavior are absolutely essential to the application of
a typical substance sortal. Perhaps we can say, at any rate, that
the application of a typical substance sortal depends for the most
part on features of an object (indeed on "features" in something
like Strawson's sense)8 which could in principle be exemplified
by an object in any duration no matter how short. It would seem
therefore no great distortion to say that, for a typical substance
sortal F, an F-stage of even minute duration can more or less
qualify as an F-stage (can qualify at least as a prima facie F-stage)
quite independently of its relationships to earlier and later F-
stages.
The contrast that we then have is this. Where F is a typical
substance sortal, the only parts of an F-thing which can qualify
as intrinsic F-parts are those which are relatively large as com-
pared to the (whole) F-things of which they are parts. In contrast,
many intrinsic F-stages will be of relatively short durations as
compared to the (whole) careers of which they are stages; and
there seems indeed to be no definite lower limit on how brief an
intrinsic F-stage might be.
That an object-stage of relatively minute duration might plau-
sibly be regarded as (more or less) intrinsically qualifying as an
F-stage, quite independently of its relationships to earlier and
later F-stages, seems rather essential to the whole enterprise of

8. P. F. Strawson, Individuals (Methuen & Co. Ltd., London, 1959), p. 202ff.

104 THE CONCEPT OF IDENTITY

the sortal analysis. Surely the underlying point of that analysis

is that we can regard the concept of the unity through time of an
F-thing as in principle constructible out of the concept of the
interconnections between momentary F-stages, i.e., F-stages which
do not themselves depend upon any principle of unity through
time. But the analysis would seem rather blatantly circular if
the momentary F-stages in terms of which it is couched could
not be regarded as essentially intrinsic F-stages, for then these
F-stages would qualify as such only in virtue of their standing
to each other in just the unity-making relationship which the
analysis is supposed to explain. The analysis does in fact seem
illuminating because we apparently can regard the unity through
time of an F-thing in terms of the interconnections between
momentary intrinsic F-stages, in terms, that is, of the inter-
connections between intrinsic F-stages that are, if not literally
instantaneous, at least so brief that their durations can plausibly
be disregarded within the context of the analysis.
We saw before (vis-a-vis the falsity of (3)) that there is no gen-
eral formula, analogous to the sortal rule, which would express
the unity through space of an F-thing (for some typical sortal F)
in terms of the continuity connections between F-parts. An addi-
tional point which has now emerged is this. Even if we could
somehow express the unity through space of an F-thing in terms
of the idea of some more complicated connections between F-
parts (e.g., in terms of the overlap relations between F-parts)
such an exercise would be quite unilluminating as an explanation
of an F-thing's unity through space. Since relatively minute F-
parts will generally not qualify as intrinsic F-parts any reference
to minute F-parts as such would already presuppose that those
parts stand to each other in just the spatial unity-making rela-
tionship which needs to be explained. If, on the other hand,
the explanation took the form of merely expressing the unity
through space of an F-thing in terms of the interconnections
between relatively large intrinsic F-parts, the question would
remain glaringly left over as to what the unity through space of
a large intrinsic F-part consists in. It therefore seems quite hope-
less to seek a general sortal-relative analysis, modelled on the
sortal rule, of a standard object's unity through space in terms
of how its parts are interconnected. Here we seem to be left with
saying (as in (2)) that we have a unified standard object wherever
 THE BASIC IDEA OF PERSISTENCE 105

we have, for some sortal F, a unified F-thing, i.e., wherever we

have a unified table, or a unified tree, or a unified car, etc., etc.
Of course we might then go on and try to give an analysis, on a
case-by-case basis, of what the spatial unity of a table consists in,
of a tree, of a car, etc., etc.

VI. Articulation
This leads to one final question that I want to raise about the
similarities and differences between the spatial and temporal
unity of a standard object. In the case of temporal unity we have
in the form of the basic rule an elementary sortal-neutral concep-
tion which approximates at least fairly well to our full-blown
conception. Is there something comparable for the case of spatial
unity? Can we also provide an approximately correct general
account of when an aggregate of matter constitutes a standard
object, without presupposing the notion of a sortal?
That any such account is likely to be not a very close approx-
imation may seem pretty clear at the start. We imagined before
that a child might want to treat a pair of socks that are knotted
together as a unitary object. It seems unlikely that, without rely-
ing on the notion of sortal coverage, we could formulate any
general rule which would exclude such a case, and many others
like it.
On the other hand the general consideration which initially
led us to search for the basic rule, as a sortal-neutral account of
unity through time, seems to bear at least to some extent on the
case of spatial unity as well. Perhaps any continuous portion of
matter is a unitary object in some very broad sense, and could
reasonably be treated as such in some suitable circumstance (i.e.,
could be picked out and described as having a shape, size, loca-
tion, duration, etc.). But it seems certain that anyone, regardless
of his sortal ignorance, would be far more likely to treat some
portions of matter as units than others. Moreover we would
expect that the portions of matter which are especially apt to be
treated as units in sortal ignorance would coincide at least roughly
to those which we treat as standard (sortal-covered) objects. So it
appears that there is some sortal-neutral principle of unit-selection
which approximates in some rough manner to our fully developed
concept of the spatial unity of a standard object.
1O6 THE CONCEPT OF IDENTITY

This point can be illustrated by reference to the example we

once considered in which a child who does not have the sortal
concept "car" observes a car moving across a field. It seemed
perfectly natural to imagine that the child would pick the car
out as a unitary object. We would certainly not, however, imag-
ine that he would pick out some arbitrary portion of the car as
an object. If the car is anything like ordinary looking, he could
not be imagined to pick out the front and back halves of the car
as units (and to judge, "Two objects of roughly equal size and
shape and color are attached to each other and moving across
the field").
It seems, intuitively, that the car stands out from its surround-
ings in a way that its front and back halves do not, and this is
why the car, rather than either of its halves, is likely to be picked
out as a unitary object even by someone who cannot apply any
relevant sortals. The car, I will say, is articulated in a way (or to
a degree) that its front and back halves are not. Articulation, as
I want to try to understand this, is an elementary sortal-neutral
idea. If we can give an account of what makes for articulation
we may have at least some rough sortal-neutral approximation
to our concept of a standard object's unity through space.
The notion of articulation is closely related to a view that has
been developed by a number of psychologists, for example,
Kohler and Koffka.9 According to these theorists our sensory
fields tend "naturally" and "spontaneously" to be broken down
(articulated) into unitary objects in accordance with certain
general principles of "sensory organization." (These principles
are also alleged to explain why we tend to see objects as forming
distinctive kinds of groups or clusters, but this part of the theory
is not directly relevant.) From my present point of view the only
sense in which these principles of articulation (as I would call
them) need to be regarded as "natural" or "spontaneous" is that
they can operate at a relatively elementary level of knowledge,
and particularly at a level of sortal ignorance. What seems com-
pletely plausible intuitively, and is perhaps also confirmed by
experimental evidence, is that anyone (or, more cautiously per-
haps, anyone who speaks our kind of language) will tend, even
9. Wolfgang Kohler, Gestalt Psychology (Liveright Publishing Corporation,
N.Y., 1947), chaps. 5-6; K. Koffka, Principles of Gestalt Psychology (Harcourt,
Brace & World, Inc., N.Y., 1935), chaps. 3-7.
 THE BASIC IDEA OF PERSISTENCE 107

in a state of sortal ignorance, to see as unitary objects only a

select few continuous portions of matter, of the unlimited num-
ber potentially present to him. Why this is so (whether it is
"learned" or "innate") may at present be left open.
I should emphasize that what I call a "state of sortal ignorance"
in no way implies a lack of general concepts (e.g., of color, shape,
size). I am certainly not referring to a "preconceptual level," if
such an idea makes sense. Someone is in a state of sortal ignorance
with respect to a given object so long as none of the (perhaps
diverse and even sophisticated) general concepts that he can
apply to the object have the special status of sortalhood, as this
status was earlier characterized. (Roughly this would mean that
he can apply no ordinary nouns to the object, other than such
nonsortals as "object," "body," "thing.") What seems intuitively
plausible, I am suggesting, is that someone's sortal ignorance
with respect to a given object will typically not prevent the ob-
ject from presenting itself as an articulated unity, as something
that stands out from its surroundings.
Articulation is actually a large and intricate topic in its own
right, and I will confine myself here to the barest intuitive sketch,
drawing in part from Kohler and Koffka. Some of my remarks in
this section may be rather speculative, and of a psychological
nature. At the very end of the section, however, I will draw cer-
tain philosophical conclusions which seem fairly straightforward.
A leading articulation-making factor is boundary contrast. A
portion of matter seems to impress itself upon us as a unit insofar
as it is segregated, bound off, from its surrounding. This segrega-
tion is accomplished primarily by the fact that there is a qualita-
tive contrast between points on the object's surface and points
in the surrounding medium. We might consider as a maximally
articulated object a black billiard ball rolling across a green felt
table. Every point on the ball's surface contrasts dramatically
with either the air around it or the felt beneath it. Boundary con-
trast, it should be noted, is a matter of degree, as will be the case
with every other articulation-making factor. The ball contrasts
sharply with its surroundings, both visually and tactually, where-
as in other cases the contrast may be less sharp, or in only one
modality. The factor of boundary contrast is sufficient to explain
why the car stands out in a way that its halves do not, since the
halves have no contrast at the boundary where they meet.
108 THE CONCEPT OF IDENTITY

The qualitative homogeneity of the surface of the billiard ball

strikes us as enhancing its presence as a unit. Perhaps this is
because homogeneity of surface highlights boundary contrast. In
any case it is clear that homogeneity without contrast amounts
to nothing in the way of articulation, as is evidenced by arbitrary
portions of the billiard ball, which are as homogeneous as the
whole ball but not therefore articulated. Homogeneity, it will
easily be seen, is a matter of degree, and many objects may have
relatively little of it.
An object seems evidently to stand out more dramatically if it
is observed to move as a unit, i.e., to move together with its parts
but not together with anything else in its immediate vicinity.
Even if an object is stationary its articulation is seemingly en-
hanced insofar as we can think of it as tending readily (which
is a matter of degree) to move as a unit. Hence I will call this
articulation-making factor separate movability.
A closely related factor is dynamic cohesiveness, which is the
object's capacity (again a matter of degree) to hang together
when subjected to various strains. This factor seems intuitively
important. Someone in a state of relevant sortal ignorance might
be initially inclined to treat as a unit a table together with the
identically colored ashtray that rests on it, but this inclination
would probably be dispelled as soon as he discovered how easily
the ashtray separates from the table. Perhaps dynamic cohesive-
ness ought to be regarded as a necessary condition for separate
movability, but certainly not conversely. A tree has virtually no
movability but it is highly cohesive.
Kohler and Koffka mention as another articulation-making
factor regularity (or simplicity or symmetry) of shape. This con-
dition is not very easy to define or to assess. A possible example
of its application is the fact that some trunks of trees (e.g., trunks
of evergreens) seem to stand out rather prominently, perhaps
because their cylindrical shapes are in some sense simple.
There is one other rather obscure articulation-making factor
that I want to indicate, but not attempt to define. Certainly a
whole branch of a tree stands out as a unit much more than
some arbitrary portion of it, e.g., its outer half. But why is this?
Both the whole branch and its outer half enjoy boundary con-
trast except, respectively, where the branch meets the trunk and
where the outer half meets the inner half. And with respect to
 THE BASIC IDEA OF PERSISTENCE log

every other articulation-making factor they are evidently the

same. Why then the drastic difference in their articulation?
Intuitively the answer has something to do with the fact that
where the whole branch touches the trunk there is a joint, where-
as there is no joint, but rather a smooth continuation, where
the front half of the branch meets the back half. To explain
properly what this means would, if I am not mistaken, require
a quite elaborate account. I will limit myself here merely to
naming this articulation-making factor. I will say that the whole
branch forms a joint where it lacks boundary contrast, whereas
half the branch does not form a joint where it lacks boundary
contrast.
The articulation-making factors that I have mentioned are:
(a) boundary contrast, (b) qualitative homogeneity, (c) separate
movability, (d) dynamic cohesiveness, (e) regularity of shape, and
(f) joint-formation at boundaries lacking contrast.
Of these articulation-making factors the boundary require-
ments (a) and (f) seem most fundamental, at least in the sense
of defining the bare minimal conditions of articulation. It seems
reasonable to say that in general an object has virtually no articu-
lation at all unless most of its boundary exhibits some degree of
boundary contrast and it forms a joint wherever it completely
lacks boundary contrast. Certainly some arbitrary portion of a
branch, for example, does not stand out as a unit at all, despite
its perhaps being to a high degree homogeneous, cohesive, and
of regular shape. Perhaps (though this is almost certainly some-
thing of an oversimplification) we can regard the two boundary
requirements (i.e., boundary contrast and joint formation) as by
themselves determining the difference between an (at least some-
what) articulated object and a (wholly) non-articulated object,
and bring in the other articulation-making factors only for the
purpose of determining degrees of articulation. Looked at in this
way, a rough preliminary sketch of articulation might be
summed up in the following two propositions:

(1) A continuous portion of matter is articulated—i.e., stands out

as something apt to be treated as a unit-—insofar as most of
its boundary exhibits some degree of boundary contrast and
it forms a joint wherever it completely lacks boundary con-
trast.
11O THE CONCEPT OF IDENTITY

(2) An articulated object's degree of articulation is determined

by its degree of boundary contrast, qualitative homogeneity,
separate movability, dynamic cohesiveness, and regularity of
shape.
The sought after sortal-neutral approximation to our concept
of the unity through space of a standard object might now be
expressed in terms of the following principle.
The Principle of Articulation. In the vast majority of cases a
continuous portion of matter constitutes a unified standard ob-
ject if and only if it has a relatively high degree of articulation.
How much a "relatively high" degree of articulation is can
perhaps be roughly indicated by reference to some well-chosen
examples which exemplify different combinations of the articula-
tion-making factors. But this too might eventually be worked
out more carefully in the context of a fuller treatment of articula-
tion.
The principle of articulation can give us at best a rough ap-
proximation to our concept of the unity through space of a
standard object. There are any number of cases in which objects
are attached to each other to form a continuous portion of matter
which seems sufficiently articulated but is nevertheless not a
standard object. The two attached socks is one example. Another
possible example that came up earlier is a car-cum-sled, which is
not a standard object despite being perhaps as articulated as
many (ill-shaped) cars. Examples like these could be multiplied
indefinitely. Besides such examples of problematically over-
articulated nonstandard objects there will also be cases of prob-
lematically under-articulated standard objects, most obviously
where a standard object suffers a temporary lapse of articulation
because of being in some special setting (e.g., where a sock is tied
to another one just like it).
On the other hand, I think that the principle of articulation
can fairly be regarded as providing an importantly viable sortal-
iieutral perspective on the basic outlines of the standard objects
that figure in our identity scheme. The principle at least rules
out that vast ubiquitous background of continuous portions of
matter that seem straightforwardly lacking in sufficient articula-
tion. These would include, most obviously, arbitrary portions of
 THE BASIC IDEA OF PERSISTENCE 111

standard objects (e.g., half of a table), and continuous summa-

tions of such portions (e.g., half of a table-cum-half of an ashtray),
which lack the essential boundary requirements, and are there-
fore not articulated at all. The principle would also rather defi-
nitely rule out as insufficiently articulated typical summations of
whole standard objects (e.g., a table-cum-ashtray) which, though
they satisfy the minimal boundary requirements, are lacking in
any high degree of homogeneity, movability, cohesiveness, or
shape-regularity. In general the principal of articulation might
be said to draw the outlines of our standard objects in exceed-
ingly broad strokes, but these are still recognizably the essential
outlines of our standard objects.
The overall position which has now emerged is that our most
basic conception of the spatial and temporal unity of an object,
a conception which is refined and clarified by our sortals, is that
of a relatively articulated object whose career unfolds along a
continuous change-minimizing path. Now it will probably have
been noticed that several of the articulation-making factors had
already shown up earlier in the quite different connection of
change-minimi/ing conflict. There were a number of examples
in which it seemed intuitively important to stabilize these factors
when tracing a change-minimizing path. (In the case of painting
the table it seemed that it might be important to stabilize the
table's separate movability; in the case of placing the ashtray
on the table it seemed that it might be important to stabilize the
table's dynamic cohesiveness; in the trunk-tree case it seemed that
it might be important to stabilize the tree's sharpness of bound-
ary contrast.) This suggests the intriguing conjecture that there
may possibly be a deeper connection between articulation and
the change-minimizing condition, in that there may be some
general tendency to resolve change-minimizing conflict by stabil-
izing articulation-making factors. I shall not here attempt to
develop this suggestion, though it may be, I am led to believe,
consonant with a number of principles enunciated by psycholo-
gists.10
I stated earlier that for my immediate purposes the question
could be left open whether we are innately disposed to pick out
objects which satisfy the articulation-making factors or learn to
10. See Koffka's comments on Von Schiller's experiments and Metzger's experi-
ments, Principles of Gestalt Psychology, pp. 300-303.
112 THE CONCEPT OF IDENTITY

do so. The same question could be raised about our disposition

to trace objects along continuous change-minimizing paths. This
issue is notoriously difficult to clarify, and many questions con-
verge on it, particularly questions about the relationship be-
tween thinking and speaking. My tentative opinion (in general
agreement with that of Kohler and Koffka) is that there prob-
ably is some important sense in which human beings are in-
nately disposed both to pick out articulated objects within their
sensory fields and to trace those objects along continuous change-
minimizing paths.11 One relatively clearcut consequence of this
hypothesis is that there probably never has been, or will be, a
natural (first) language spoken by humans which contains an
identity scheme whose basic structure could not be captured in
terms of the principle of articulation for spatial unity, and the
rule of change-minimi/ing continuity for temporal unity. Cul-
tural divergence is to be looked for at the level of sortal refine-
ment but the basic structure of human thought about identity
is universal. That, at any rate, seems to me a highly plausible
hypothesis.
But it is, as far as I am concerned, essentially an empirical
hypothesis, to be dealt with in an empirically scientific manner.
The only a priori philosophical doctrines to which I want to be
committed, and for which I have tried to present arguments, are
these:
(i) We can provide a relatively accurate account of our (English
speaker's) concept of the unity through time and unity
through space of a standard object only by taking cognizance
of what sort of object we are dealing with, since sortal-
coverage (though in two rather different senses) figures in
both modes of unity;
(ii) However, a significant approximation to this account can be
presented without taking cognizance of what sort of object
we are dealing with, and this sortal-neutral approximation
might be regarded as capturing our most basic conception
of an object's unity through time and space.
11. Compare with Quine, who suggests that the unity of a body is something
that we are "innately predisposed to appreciate" (The Roots of Reference,
p. 54), and that "body-unifying considerations . . . are rooted in instinct"
(ibid., p. 55). The innateness question will be considered at length in Chapter
8.
 4
The Persistence of Matter

I. A Puzzle about Matter

EARLIER I expressed the opinion that any continuous portion
of matter, even one which does not come under any sortal (and
which does not, therefore, constitute a standard object), can
perhaps be called an "object" in one very broad, ordinary sense
of that word. This is certainly a debatable point, and I do not
intend to presuppose it in anything that follows. What seems
less debatable, however, and more important, is the following
proposition. Our ordinary concept of identity through time
applies, in at least many typical cases, to portions of matter
(whether or not these be called "objects") that do not come under
any sortals. And this immediately presents a difficulty, since
sortal-coverage is, according to the foregoing account, an opera-
tive condition for the application of our full-blown identity
concept.
Suppose, for example, that I specify some portion of wood
which partially makes up the uniform wooden table in front of
me. I might do this by momentarily laying my hand on the
surface of the table and thereafter referring to the hand-shaped
portion of wood which had momentarily extended directly down-
wards beneath my hand. Now that portion of wood is quite defi-
nitely not covered by any sortal, since we know that the disper-
sive term "wood" is not a sortal,1 and there is evidently no other
possible sortal which applies to the wood. Yet there seems no
1. Cf. above, p. 42.

113
114 THE CONCEPT OF IDENTITY

doubt that (whether or not we call it an "object") we do think

of that wood, of the particular portion of wood that was directly
under my hand, as having an identity of its own, as being spa-
tially related to other portions of wood that surround it, and as
in fact having almost certainly persisted for a longer time than
the table has. But what does this concept of persistence mean if
it cannot be understood in terms of the idea of a sortal-covered
path?
It will of course be tempting to suggest that the basic rule can
help us here. Can we not perhaps say that with respect to a
portion of matter which is not sortal-covered our only relevant
concept of persistence is the relatively primitive one provided
by change-minimizing considerations? However, this answer will
not work. We can indeed conceive of the basic rule as straight-
forwardly applicable to non-sortal-covered portions of matter
(to nonstandard "objects") wherever these portions of matter
are at least to some extent articulated. For example, the portion
of matter which constitutes two socks that are tied together
could typically be traced along a continuous change-minimizing
path. But there is something especially problematical about the
sort of case that I am considering in which a wholly non-articu-
lated portion of matter is referred to. It would seem that in
many (though perhaps not all) such cases the change-minimizing
condition is thoroughly inapplicable. Where a portion of matter
is wholly non-articulated it will often be the case that we could
trace continuous paths from the portion in any and all directions,
none of which could straightforwardly claim the unique status
of minimizing change in any relevant (i.e., nonlocational) re-
spect. This point seems rather clear in the particular case under
consideration. Since we are imagining that the hand-shaped
portion of wood being referred to is merely a wholly indistinct
portion of a larger uniform mass of wood, we cannot possibly
trace that portion along a change-minimizing path, for there will
be an indefinite number of paths extending away from that por-
tion, all of which contain qualitatively indistinguishable hand-
shaped portions of wood.
I do not wish to base my argument on the overly problematical
(though possibly defensible) general principle that the change-
minimizing condition can never straightforwardly apply to cases
of non-articulated portions of matter. All that I require for my
 THE PERSISTENCE OF MATTER 115

present purposes is the limited point that the change-minimizing

condition seems quite helpless at least with respect to some cases
of non-articulated portions, i.e., cases like the one under con-
sideration, involving some wholly undifferentiated portion of a
larger uniform mass.
It might still be urged, however, that even in the latter cases
change-minimizing considerations can work effectively so long as
we allow ourselves to take into account the location of a portion
of matter within the object that it partially constitutes. Perhaps
this should be treated as a special kind of locational considera-
tion, which does in fact count in tracing a career. The suggestion
would be, in effect, that we can state identity criteria for the
specified hand-shaped portion of wood in terms of its location
relative to the outlines of the table. Our concept of the identity
of that portion of wood would then be criterially determined
by the rule that you are referring to the same wood so long as
you are referring to the wood which occupies the same place on
the table (i.e., the same place relative to the outlines of the table).
But this is unquestionably wrong, because it is certainly not
the case that the same wood cannot alter its location on the table.
Suppose that yesterday I had filed the table down to make it
several inches shorter. Certainly this would warrant my now say-
ing, "The hand-shaped portion of wood which was directly
under my hand a moment ago was closer to the outlines of the
table yesterday than it was a day before yesterday." And quite
apart from anything akin to filing, it must surely be admitted
that it is at least conceptually coherent to suppose that my table
suffered from some chemical quirk which made some of its wood
regularly contract while the rest expanded, in which case pos-
sibly that portion of wood which was momentarily under my
hand regularly altered its size and its place on the table. No, the
location of the wood on the table quite definitely does not
criterially define its identity. It seems sufficiently clear, therefore,
that the change-minimizing condition will not explain the wood's
identity, even if we allow locational considerations into play.
And this leaves us with the puzzle as to what does explain the
wood's identity.
Perhaps this puzzle can be made a bit more stark by slightly
altering the image. Suppose that I break the table to pieces so that
I am left with a large number of wooden fragments. Holding
116 THE CONCEPT OF IDENTITY

one of these fragments in my hand I certainly want to say, "This

fragment of wood came out of the table." Now this is, so far, no
problem since we can in principle trace the fragment backwards
in time to some portion of the table, tracing, that is, along a
change-minimizing path or, if necessary, a path covered by the
nondispersive concept "fragment of wood."2 There is no prob-
lem here because, first, a fragment of wood is articulated, which
immediately invites a straightforward application of the change-
minimizing condition, and, second, "fragment of wood" can
function in a straightforward way as a sortal which covers a
tracing path. There is indeed the possible complication that,
depending on how the table broke, we might have to say that the
fragment is traceable back to some portion of a larger fragment,
which is in turn traceable back to some portion of a still larger
fragment, which is in turn . . . , which is in turn traceable back
to some portion of the table. Still, so long as "coming out of" is
transitive, there should be no major problem in explaining at
least what it means to say that the fragment came out of the
table.
The problem arises insofar as I also want to say something
else, viz. "This wood used to make up part of the table." (Not,
strictly, "This fragment of wood used to make up part of the
table," since my table was not made up of articulated fragments.)
But what can I possibly mean by this? What can I mean by
saying that this wood was in my table yesterday though it was
then merely a wholly indistinct portion of some larger uniform
mass? If I want to talk about that selfsame wood persisting from
yesterday to today I should, it seems, have criteria of identity
which explain what this means, which explain what the unity-
making relationship is that binds together the stages of that
persisting wood. But it seems that I have no such criteria of
identity, at least none provided by the idea of a change-minimiz-
ing or sortal-covered path.
Now the very natural first impulse upon hearing this puzzle
is to protest that the whole thing is really very obvious. (It is a
philosophically important feature of this puzzle that we find it
extremely difficult to take it seriously.) "Look, that wood that
you're holding in your hand came out of the table, right? If it

2. Cf. above, p. 42.

 THE PERSISTENCE OF MATTER 117

came out of the table it had to be in the table. Now it certainly

didn't jump into the table when we weren't looking. So it's been
in there all the time."
This seems like an outright evasion, for I was not asking for
a justification (for evidence) but for an analysis (for criteria). I
was not asking, "How do we know that the wood persisted before
it came out of the table?" which would imply that I already
understand what that state of affairs amounts to, but rather
"What can we mean by speaking of the persistence of the wood
before it came out of the table?" The challenge is to show me
the criteria of identity. (But we will see shortly that this natural
response, which apparently ignores the request for criteria, may
be in a certain sense quite apt.)
It is interesting to note that Russell was consistently sensitive
to the puzzle about the identity of matter which I am now dis-
cussing, even though he formulated it in the context of his pre-
viously mentioned inaccurate account of the identity of standard
objects. At one point Russell explains that ". . . continuity is not
a sufficient criterion of material identity. It is true that in many
cases, such as rocks, mountains, tables, chairs, etc., where the
appearances change slowly, continuity is sufficient, but in other
cases, such as the parts of an approximately homogeneous fluid,
it fails us utterly. We can travel by sensibly continuous grada-
tions from any one drop of the sea at any one time to any other
drop at any other time."3
Russell is wrong (or misleading) in stating that continuity is
a sufficient condition of identity for standard objects like tables
and chairs, since (as shown in Chapter 1) even in such cases
continuity considerations by themselves do not exclude aberrant
paths which combine stages of a whole with stages of its parts.
Yet Russell correctly perceives that there is an important differ-
ence between cases involving standard objects like chairs and
tables, and cases involving a merely undifferentiated portion of
some larger uniform mass, like a drop of water in the sea. The
difference, which Russell however never makes explicit, is that
considerations of sortal-covered (or change-minimizing) continu-
ity suffice for the former cases but not for the latter.
A problem about the identity of matter is most immediately

3. Russell, "The Relation of Sense-Data to Physics," p. 171.

118 THE CONCEPT OF IDENTITY

apparent in the sort of case that I considered, or the one that

Russell considers, involving a portion of matter which is neither
sortal-covered nor articulated, and which is evidently amenable
to neither the sortal rule nor the change-minimizing rule. But
the problem is actually a more general one, and pertains even to
portions of matter that are, at a given moment, both fully articu-
lated and sortal-covered. This is because it is always necessary
to distinguish between the identity through time of a portion of
matter and the identity through time of the articulated object
which, at a given moment, it makes up. Suppose that the portion
of wood x composes the whole table y today. Though the table
y's career can be traced along a path covered by the sortal "table"
(or along a change-minimizing path), x's career cannot be traced
in this manner, since x's career may not be the same as y's. If,
for example, a piece of wood chipped off the table yesterday
then, whereas the table y was larger yesterday, x was the same
size but composed only part of the table. (Hence the relation-
ship between x and y today is only one of constitution, but not
strict identity.) 4 This shows that even where a portion of wood
is covered by the sortal "table" (in the sense that "table" con-
stitutively applies to it), its career cannot properly be traced
along a sortal-covered (or change-minimizing) path.
A distinction between the identity of a portion of matter and
the identity of the articulated object which it makes up must be
acknowledged even in those cases where the only sortal under
which we can trace the articulated object that the matter makes
up is some term with roughly the force "articulated bit of such
and such matter" (e.g., "puddle of water," "fragment of wood").
Thus we have to distinguish between the identity of a puddle of
water (fragment of wood) and the identity of the water (wood)
which makes it up. This distinction is very obviously required
whenever a smaller articulated bit of matter is separated from,
or joined to, the larger articulated bit which we are tracing. For
example, we would not necessarily expect that in tracing a puddle
(or pool or expanse) of water we are thereby tracing the very
same water, since someone may have removed a glass of water
from the puddle, or added one. Or if a splinter of wood is sep-
arated from a larger fragment (or lump or chunk) of wood then

4. Cf. above, p. 59ff.

 THE PERSISTENCE OF MATTER 119

of course we no longer have the very same wood. But, to take

this point one step further, even where no smaller articulated
bit separates from, or joins, the larger articulated bit which we
are tracing this still does not give us a criterial guarantee that
we are tracing the very same matter. This is perfectly obvious in
the case of the puddle of water since we know that water may
evaporate out of, or condense into, the puddle. And a moment's
thought should convince us that even in tracing an apparently
intact fragment of wood it must always remain a conceptually
coherent possibility that some of the wood is vanishing (changing)
into thin air while some other wood is materializing into the
fragment. (Consider that a tree is in a way just a big chunk of
wood, whose matter is continuously changing.) So it seems that
the idea of a change-minimi/ing or sortal-covered path gives us
no criteria for the concept of the persistence of matter under
any circumstances. How then does this concept operate?

II. An "Ultimate" Kind of Persistence

It might be suggested that the appropriate move at this point
would be to search diligently for some perhaps complicated and
ingenious formulation of our identity criteria for matter (where
these criteria might vary significantly from one sort of stuff to
another). Before directly addressing this suggestion I want to
explain an alternative approach, which is the one that I hold to
be correct. My position, to put it somewhat incautiously at first,
is that we do not have any identity criteria for matter. But this
idea needs to be explained.
To begin with, it should be clear that the sort of identity
criteria that I have been discussing in this book are observa-
tional criteria. We have observational criteria for an identity
judgment if we are able to explain these criteria by reference
exclusively to conditions of a straightforwardly observable sort.
This seems tantamount to saying that observational criteria must
not go beyond the ordinary manifest properties of things (e.g.,
something's being a table or something's being red) and the
ordinary manifest relations between things (e.g., spatial and
temporal).
Now for the case of the persistence of an ordinary articulated
object we can draw a fairly clear distinction between criteria of
12O THE CONCEPT OF IDENTITY

identity and evidence of identity. The criteria are those observa-

tional considerations in terms of which we can, at least roughly,
analyze or define what our concept of the object's persistence
consists in. Identity criteria, on my understanding, may be ex-
ceedingly vague and allow for any number of borderline possi-
bilities. But we have criteria of identity for ordinary objects
insofar as in the most typical nonborderline cases our judgments
of identity about these objects are analytically entailed by
straightforwardly observable conditions. These conditions, I have
argued, basically amount to the requirement that an object's
career be traceable along a change-minimizing or sortal-covered
path. Evidence of identity, on the other hand, comprises facts
from which we can inductively conclude that the criteria are
satisfied. On the basis of directly tracing various objects along
change-minimizing or sortal-covered paths we arrive inductively
at generalizations (e.g., that a table tends, when left alone, to
remain qualitatively and locationally stable) which allow us to
judge that the identity criteria were satisfied in those cases where
we could not directly observe this. (So the fact that the table
which is present when we return to a room is qualitatively and
locationally similar to the one that we saw before leaving pro-
vides evidence that probably the criteria of sortal-coverage and
continuity were satisfied, and that we have the same table.)
But for the case of the persistence of matter I think we can
draw no such distinction between criteria and evidence of iden-
tity: here, in a sense, we have only evidence and no criteria. By
this I mean that the only way to characterize our general pro-
cedure for judging of the identity of matter is to say that we
reidentify matter in such a way as to arrive at the most coherent
and theoretically satisfying account of what we observe. In this
way we arrive at various principles which, both at the common-
sense level and at scientific levels, specify how bits of matter of
various sorts may be presumed to behave under different ob-
servable circumstances. (For example, one such principle might
be that, other things being equal, a non-articulated bit of wood
is presumed to maintain a constant location within a table.) But
these principles are both partial and provisional, and may, within
broad limits, be supplemented or revised in the light of scientific
progress. These principles cannot therefore provide an analysis
or definition of our concept of persisting matter.
 THE PERSISTENCE OF MATTER 121

This account implies that our concept of persisting matter,

even at the most commonsense level, incorporates something in
the way of a theoretical-explanatory posit of an underlying mode
of physical persistence which ultimately accounts for the observed
behavior of ordinary articulated objects. Even the most limited
level of common sense contains an abundance of relatively secure
and well-founded views about that underlying domain, though
it is ultimately for science to fill in the details. But no belief
about the observable manifestations of matter, however well
founded, is analytic of our concept of persisting matter. It must
always remain a conceptually coherent possibility that our ex-
planations have been faulty and that matter behaves differently
from the way that we think.
Though I am maintaining that we have no ordinary observa-
tional criteria of identity for matter it might still be possible to
provide some level of analysis or explanation of our concept of
persisting matter. We can, in fact, broadly distinguish between
two philosophical (and scientific) approaches to such an explana-
tion.
a. It might be held that the persistence of matter is ultimately
to be understood in terms of the persistence of particles such as
atoms, molecules, or electrons. These particles might be said, in
a somewhat extended sense, to be articulated by various unob-
servable properties in terms of which a sortal such as "atom"
can be defined. The persistence of an atom is then analyzed in
the standard way as depending upon the continuity of the atom's
path under the sortal "atom." (Alternatively we can say that the
atom's path minimizes change with respect to its unobservable
properties.) The persistence of a bit of matter is thus ultimately
analyzed in terms of the sortal-covered persistence of the particles
which make it up. (We can then say, if we like, that we have
"theoretical identity criteria" for matter since we can give an
analysis of the persistence of matter in theoretical terms.)
b. If we want to avoid a commitment to atomism we might
say simply that the persistence of matter depends upon some
unobservable relationship which binds the successive stages of
a single bit of matter. This relationship has sometimes been
called "genidentity."5 So we can, in a sense, explain the identity
5. See Rudolph Carnap, Introduction to Symbolic Logic (Dover Publications,
N.Y., 1958), p. 198. See also the remarks about genidentity in Hans Reichen-
122 THE CONCEPT OF IDENTITY

of matter by reference to that relationship. (Though this "expla-

nation" seems very thin indeed, there is still nothing to prevent
us from saying that in terms of the unobservable relationship of
genidentity we can state "theoretical identity criteria" for matter.)
The (b)-approach seems to me on the whole the more sensible
one. The (a)-approach does admittedly have the advantage of
allowing a closer analogy between the persistence of matter and
the ordinary sortal-covered persistence of familiar articulated ob-
jects. The (b)-approach, on the other hand, at least maintains a
formal analogy between these two modes of persistence, insofar
as in both cases a unity-making relationship (sortal-covered con-
tinuity for ordinary objects and genidentity for matter) can be
thought of as binding object-stages into successions which do not
generally crisscross or overlap. And the (b)-approach has the
decisive-seeming advantage of not forcing us to wed the concept
of persisting matter a priori to atomism. (The (b)-approach does
not exclude atomism since there is nothing to prevent the rela-
tionship of genidentity from turning out to depend upon the
sortal-covered persistence of atoms.) The a priori atomistic posi-
tion implied by the (a)-approach would have us reject on a priori
(conceptual) grounds various anti-atomic theories of matter that
have been influentially maintained in the history of science.6
This seems implausibly overbearing. It seems that our concept
of the persistence of matter ought to be seen as a priori accom-
modating the possibility that, even from the deepest theoretical
vantage point, portions of matter might persist as merely non-
differentiated parts of larger masses.
The contrast that I am trying to develop here between the
persistence of standard objects and the persistence of (portions
of) matter seems the more compelling, certainly, if one agrees
with me in rejecting an a priori atomistic analysis of our concept
of the persistence of matter. But I think that the point remains
essentially intact even if one opts for a priori atomism. There is
surely an important sense in which the persistence of such par-

bach, The Philosophy of Space and Time (Dover Publications, N.Y., 1958),
p. 270ff.
6. See the conflict between atomic theories and "continuum" theories as dis-
cussed in Stephen Toulmin and June Goodfield, The Architecture of Matter
(Harper & Row, N.Y., 1962), p. 64ff. and p. 158ff.
 THE PERSISTENCE OF MATTER 123

tides as atoms, molecules, and electrons cannot be understood in

straightforwardly observational terms. Obviously these particles
cannot be observed except perhaps by way of highly technical
and theory-imbedded apparatus. Moreover it would seem that we
cannot even soberly visualize such particles in terms of ordinary
qualities like color or texture. The conclusion seems therefore
warranted that if the only account that we can give of our con-
cept of the persistence of matter is in terms either of atomism
or genidentity, then our concept of this mode of persistence is
not properly regarded as analyzable in terms of ordinary obser-
vational identity criteria.
There is a philosophical tradition, loosely associated with the
word "substance," to the general effect that our concept of the
persistence of matter points to something beyond the reach of
what can be straightforwardly observed. (This is the gist of
Descartes's discussion of the wax in Meditation 2.) This seems
closely akin to the position which I am here advancing. My posi-
tion would also entitle us, I think, to say something else which
is close to the heart of the substance tradition, and this is that
the unity through time of matter, in contrast to that of familiar
articulated objects, is in a sense ultimate. This sense of ultimacy
derives from the two related points that the unity through time
of matter goes deeper than (because it is not analyzable in terms
of) the ordinary manifest properties and relations of things, and
this unity is posited as playing a central role in the ideally most
complete explanation of physical phenomena.7

III. Searching for Identity Criteria

This, then, is my view on the identity of matter. Let me return
now to the previously postponed suggestion that perhaps, con-
trary to my view, there really are observational identity criteria
for matter and we ought to look harder for them. I would not
expect this suggestion to induce a great deal of enthusiasm at the
present stage of the discussion. For we now have before us two
possible accounts, two possible models, for understanding the
nature of our judgments about the identity of matter. According
7. C£. Shoemaker's discussion of the connection between the substance tradi-
tion and the analyzability of identity judgments, in Self-Knowledge and Self-
Identity, pp. 57-63 and pp. 254-60.
124 THE CONCEPT OF IDENTITY

to the criterial model these judgments are to be seen as analyti-

cally entailed by some complex conjunction of straightforwardly
observable conditions. The model that I propose denies that
any such analytic entailments apply to these judgments. The
latter model must seem far more promising than the former in
light of the cases considered earlier (the hand-shaped bit of wood
in the table, the fragment of wood, the puddle, the tree). For
these cases certainly did not make it appear remotely hopeful
that we could specify even vaguely some observable conditions
which analytically entail the relevant judgments about the iden-
tity of matter.
To recapitulate briefly, we recall that the identity of the non-
(observably)articulated bit of wood in the table could evidently
not be analyzed in terms of any straightforward considerations
of sortal-covered or change-minimizing continuity. Nor, we saw,
did the wood's location in the table afford a criterial basis for
judging of its identity. And notice now the more general point
that our ability conceptually to divide an object like a table top
into such parts or portions as the part in the middle, the curved
part, that corner, that edge, and so on, will never give us the
sought-after criterial basis for the identity of the bits of matter
which compose the object, since it is by no means an analytic
truth that the same parts, in this sense, must be composed of the
same matter. Or, to consider a slightly different sort of case, when
a piece of clay is deformed we may be able straightforwardly to
observe the differing movements of such parts as, for example,
the top part, the middle part, the thin part, the bumpy part.
But none of this gives us a criterial basis for judging of the iden-
tity of the matter which makes up these parts.
This point reverts back to my observation in the preceding
chapter that if we want to trace a career under a term like
"portion of a car between the bumpers" then we must be careful
to distinguish between the item thus traced and the matter
which, at any given moment, makes it up. I noted then the
apparent difficulty in assessing the status as sortals of various
constructions out of "part" and "portion," and the correlative
difficulty in assessing the status as objects of the items traced
under such constructions. What seems sufficiently clear, however,
is that these items, whatever their precise status, cannot provide
 THE PERSISTENCE OF MATTER 125

us with identity criteria for the bits of matter which partially

compose an object. And once this is clear I think it is immediately
apparent that there are no remotely plausible candidates for such
a criterial basis.
The reasonable conclusion seems to be that we simply do not
rely on observational criteria when we judge of the identity of
some matter which partially composes an object. In a case like
that of the wood in the table what we do rely on (at a common-
sense level) is the simplifying assumption that the wood's loca-
tion in the table probably remains fairly constant. That is, we
rely on some such general simplifying principle as this: Other
things being equal the location of a non-articulated bit of matter
within an articulated object may be presumed to remain fairly
constant. But this principle provides nothing like a criterial guar-
antee, and there is no saying a priori in how many different ways
the principle might have to be augmented and reshaped both
by common sense and science. (There is no saying a priori in
how many different ways the "other things being equal" clause
would have to be filled in to yield the simplest and most coherent
account of the careers of different bits of matter.) In those various
cases where we judge that some non-articulated matter has al-
tered its location within an object (e.g., in the case of a piece of
clay that is being deformed, or in the case of a tree, or in the case
of a river) we rely on the most diverse evidential considerations
and ultimately on the best theory of matter we have available.
Even when we turn to the much simpler kind of situation, in
which the quantity of matter under consideration is fully articu-
lated, we find nothing like a criterial guarantee of our identity
judgments about the matter. Suppose, for example, that I am
holding a perfectly solid block of wood in my hand, stationary,
not squeezing it too hard, in ordinary atmospheric conditions,
with nothing observably weird going on (for example, there are
no observable changes in either the size or shape of the block).
In such circumstances I might venture the following identity
judgment: "The wood which now makes up this block is identical
with the wood which made up this block a moment ago." We are
searching for some observable conditions which might criterially
(analytically) entail the truth of this judgment. What needs to be
ruled out, among other things, is the possibility that some of the
126 THE CONCEPT OF IDENTITY

wood vanished (changed) into air (or into the flesh of my hand)
and/or that some additional wood materialized out of the air
(or out of my flesh) into the block.
What observable conditions could conceivably be such as to
entail analytically that this has not happened? Surely not the
conditions mentioned earlier (that the block was not squeezed,
that its size and shape remained constant, etc.). It might be
startling, but certainly not incoherent, for scientists to tell us
that under just those conditions some bit of wood turns into air
and vice versa. (The constancy of the block's size and shape
might be accounted for by positing suitable expansions and con-
tractions of the wood inside the block.) Is it not clearly hopeless
to seek some other observable conditions which somehow would
analytically entail the identity judgment?
If someone is not convinced that this is hopeless then let him
reflect upon the following point (which seems to me fairly deci-
sive). In order for it to be true to say "The wood which now
makes up this block is identical with the wood which made up
this block a moment ago" it must be the case that all the original
wood of the block is still in the block. This means that even if
some minute and possibly invisible speck of wood in the block
turned into air the identity judgment in question is, strictly
speaking, false. But it seems obvious beyond the need for further
argument that no straightforwardly observable conditions (no
facts about the ordinary properties and relations of things), how-
ever complex, could analytically guarantee that no such minute
speck turned into air. And since the identity judgment requires
just this guarantee it follows immediately that no observable
conditions can criterially guarantee the identity judgment.
Indeed the position that I am here defending seems evident
almost to the point of triviality the moment we remind ourselves
that the proposition "x (which exists now) is the same wood as y
(which existed before)" entails "Every (wooden) part of x, no
matter how minute, was a part of y." This entailment seems im-
mediately to render unobservable the identity through time of
matter. And, I might add, one can apparently say this without
being committed to any special views about the much debated
nature of the "observational"-"theoretical" distinction. Whatever
might be the ultimate epistemological status of the "straight-
forwardly observable" it seems sufficiently clear that in the sense
 THE PERSISTENCE OF MATTER 127

in which the persistence of a table, or a block of wood, is straight-

forwardly observable, the persistence of a portion of matter is
not, since the latter mode of persistence depends upon the clearly
unobservable condition that every minute part of the matter
remain the same.
We are perhaps inclined to ignore considerations about minute
and possibly invisible specks of matter in the context of a discus-
sion of our most commonsense concept of persisting matter. But
such considerations, once we are reminded of them, seem quite
definitely to have a legitimate bearing on the meaning of that
commonsense concept, and indeed to be in a way definitive of it.
If someone, say a child, could not appreciate the relevance, with
respect to his judgments about "same wood," of scientific theories
about minute and invisible specks of matter, this would seem to
be grounds for saying that he was not really employing the con-
cept of the identity of matter but was instead still at the more
elementary level of thinking only about, e.g., "same block,"
"same stick."
Admittedly a commonsense judgment about the identity of
some matter is not likely to stickle over details about minute
parts. From a commonsense point of view perhaps what I would
really want to say about the block of wood is that the wood which
makes it up now is more or less the same as before, but not neces-
sarily exactly the same. Be this as it may it would certainly seem
a mistake to try to suggest that though we have no observational
criteria for the judgment "x (which exists now) is the same matter
as y (which existed before)" we do nevertheless have such criteria
for the judgment "Some large portion of x is the same matter as
some large portion of y," or, colloquially, "x is more or less the
same matter as y." Surely our understanding of these latter judg-
ments presupposes our understanding of the former, presupposes,
that is, our understanding of what it means to say of a particular
bit of matter that it has persisted over some period of time. And
since, by hypothesis, we have no observational criteria in terms
of which to analyze or define what is meant by the former judg-
ment it follows that we cannot have observational criteria in terms
of which to analyze or define what is meant by the latter judg-
ments.
The inescapable conclusion seems to be that in judging of the
identity of matter, even in those cases where the matter happens
128 THE CONCEPT OF IDENTITY

to be articulated (e.g., in judging of the identity of the wood in

the block), we do not rely on identity criteria, but we rely instead
on some such simplifying principle as the following: Other
things being equal an intact articulated object (like an intact
block of wood) may be presumed to alter its material composi-
tion, if at all, only partially and very slowly. Common sense and
science augment this principle (fill in the "other things being
equal" clause) not on the basis of some a priori criterial strictures
but on the basis rather of the widest and in principle most un-
limited variety of facts about how things alter in size and weight,
decompose, mix together, and so on. The proper model here is
not that of criteria application but rather that of working to-
wards the most coherent theory of an underlying level of persist-
ing matter.

IV. Matter and Common Sense

That our commonsense concept of persisting matter involves the
positing of an underlying reality may seem unbelievable because
the concept strikes us as absolutely obvious and inevitable. In-
deed the premise that the concept is not criterially definable in
conjunction with the recognition of its utter obviousness could
naturally lead to the surmise that the concept must be in some
sense a priori. But I think that a more straightforward explana-
tion of why the concept is so obvious is that our experience with
ordinary articulated objects provides us incessantly with a liter-
ally overwhelming barrage of evidence that there exists underlying
matter. That is, our most immediate and surface-level observa-
tions of ordinary objects, whose persistence conditions are grasped
in terms of observational criteria, present us with a range of
facts which point unavoidably to the conclusion that these ordi-
nary objects are composed of, and are ultimately to be under-
stood in terms of, persisting items of a quite different sort.
The range of facts to focus on contains as an instance just the
sort of case discussed earlier, breaking a table to pieces. What
we cannot avoid noticing in such a case is that the fragments
which emerge from the table go together to add up to an object
of at least roughly the size, and even form, of the table. And this
is virtually a universal phenomenon: An object which contains
no (observably) articulated parts is split up into a number of
 THE PERSISTENCE OF MATTER 129

objects which add up to the original. How can one avoid trying
to explain this by invoking the idea that the smaller objects
which came out of the bigger one were in some sense already in
there prior to their articulation? This conclusion seems so inevi-
table that one could scarcely take seriously my question about
our basis for judging that (the wood in) the wooden fragment
had always been lurking non-articulatedly in the table. And as
far as the question where in the table it was, well, we adopt
initially the simplest and most obvious hypothesis that it was
always located right at the place in the table from whence it
came. Such assumptions as this constitute a rudimentary com-
monsense theory of underlying matter, and all that is now re-
quired is for someone like Thales to enter the scene and the
rest happens by itself.
But we can, I feel convinced, imagine what it would be like to
live in a world whose articulated objects did not display those
patterns of behavior which provide the primitive basis for our
concept of persisting matter. In such a world the concept would
have no use, at least not at a commonsense level. I will only
sketch this peremptorily. But imagine that whenever you break
a table to pieces the resulting fragments add up to five times the
size of the table. Imagine that if ever you start out with a basin
full of water and remove a glassful from it then when you pour
the water back the basin overflows enormously. Imagine that
whenever you dig a hole in the ground you wind up with a pile
of dirt which looks thirty times higher than the hole next to it.
These imaginings would have to be generalized indefinitely
before we reached the image of a world in which the concept of
persisting matter had no immediate application. But I can see no
reason to doubt that we can coherently broach this image. That
it is extremely difficult for us to do so shows how deep in our
experience the concept of matter penetrates; but that it is pos-
sible for us to do so shows that the concept is not, in the most ulti-
mate sense, unavoidable. If we lived in that imagined world we
would still have immediate use for the idea that "you can't get
something out of nothing," since to get, for example, a new
fragment of wood you would have to make one by separating it
out of something else. The idea for which we would have no
immediate use is that when you create a new articulated object
by separating it out of another object there was all along some-
130 THE CONCEPT OF IDENTITY

thing, the persisting matter, waiting there to be articulated. This

idea (which is, 1 think, in some sense really a strange idea)
would not impress itself upon us.
The essential simplification which our concept of matter con-
tributes at a commonsense level might be thought of as follows.
If we were to confine ourselves wholly to the observational level
of the basic (or sortal) conception we would be able to formulate
the following two kinds of laws. Type A laws would tell us what
tends to happen in the special case in which one articulated
object is created (or destroyed) by coming out of (or merging
into) another. For example: When an articulated wooden object
is broken to pieces the fragments which come out of it add up
to the original in size and weight (and perhaps shape). Type B
laws would tell us how articulated objects tend to behave in the
more general case in which no articulated object is created (or
destroyed) by coming out of (or merging into) another. An ex-
ample might be: Under ordinary conditions an articulated
wooden object tends to conserve its size, and weight, and shape.
Now it is a contingent fact, which we can imagine otherwise, that
these two types of observational laws are so related that they can
be subsumed under, and hence explained in terms of, relatively
simple type T laws which posit an underlying level of persisting
matter. An example of a type T law is: Under ordinary condi-
tions a portion of wood (whether articulated or not) tends to
conserve its size, and weight, and shape. This type T law allows
us to explain in a simple unified manner the two kinds of phe-
nomena which we would have had to treat separately if we lim-
ited ourselves to the observational level.
The explanatory application of a type T law, which describes
the behavior of underlying matter, presupposes the availability
of some principles ("correspondence rules," "bridge principles")
that connect observable facts about articulated objects with the
posited facts about matter. We might perhaps represent the most
rudimentary commonsense theory of matter as embodying such
principles in the form of the three presumptive principles which
have emerged in the course of this discussion, viz. (1) the pre-
sumption that the material composition of an articulated object
remains pretty much constant, (2) the presumption that the loca-
tion of a non-articulated bit of matter within an articulated
object remains pretty much constant, and (3) the presumption
 THE PERSISTENCE OF MATTER 131

that when a smaller articulated object comes out of (merges into)

a larger one then the matter which makes up the smaller one is
subtracted from (added to) the matter which makes up the larger
one. In terms of these three principles we are able to construct
a relatively coherent and satisfying picture of what happens to
the wood in a table both before and after the table is broken to
pieces. We can say that each fragment that comes out of the table
is composed of wood which was originally in the table (pre-
sumption (3)), that this wood probably partially composed the
table throughout its entire career (presumption (1)), and was
always located roughly at the place in the table from which it
eventually emerged (presumption (2)). And all of this happened
in accordance with the type T law mentioned before (i.e., the
law that a bit of wood tends to conserve its size, weight, and
shape).
I have been at pains to establish that our concept of the per-
sistence of matter is relatively theoretical as compared to our
concept of the observable persistence of an ordinary articulated
object, and that, as a consequence of this fact, we should be able
to imagine what it would be like to live in a world in which we
had a commonsense use for the latter concept but not the former.
A somewhat different and perhaps easier point, which reinforces
the previous one, is that we can imagine what it would be like
for our concept of the persistence of ordinary objects to function
normally while the concept of the persistence of matter is repu-
diated at the highest theoretical levels. (This is, I think, tanta-
mount to imagining a world in which ordinary objects in fact
persist without any matter in fact persisting.) It certainly seems
that scientists might tell us that there is no such thing as an
underlying level of persisting matter, but that some radically
different conception affords the ultimate explanation of physical
phenomena. (Indeed on some readings of contemporary physics
it may possibly be that scientists have told us this.) But this
would not (and should not) prevent us from continuing to judge
in the normal way about the identity of ordinary observable
objects.8
8. Compare with Sydney Shoemaker's criticism of Chisholm's position in "The
Loose and Popular and the Strict and Philosophical Senses of Identity," in
Norman S. Care and Robert H. Grimm, eds., Perception and Personal Identity
(The Press of Western Reserve University, Cleveland, 1969), pp. 108-9.
132 THE CONCEPT OF IDENTITY

It follows from the above account that I would regard our

concept of the persistence of ordinary articulated objects as de-
cisively more primary than our concept of persisting matter. This
implies a repudiation of the common procedure (e.g., in Locke's
discussions of identity) of stating identity criteria for ordinary
objects in terms of a prior notion of persisting matter.9 Often
this procedure will take the form of substituting for the require-
ment of spatiotemporal and qualitative continuity the require-
ment that an object's material composition can alter only grad-
ually. Now these requirements do in practice amount to virtually
the same thing, but it is nevertheless a major error of principle to
inject the concept of persisting matter into the very center of
our ordinary identity criteria. The clear and observational con-
cept of the persistence of an ordinary object deserves to be kept
relatively disentangled from the more difficult and theoretical
concept of persisting matter. We have, and have the right to
have, a concept of persistence which operates essentially at the
most straightforward observable level without much concern for
what may be happening in the theoretical depths.
This is not necessarily to rule out the possibility that our
relatively theoretical judgments about matter might marginally
influence our judgments about ordinary objects. Such influence
may be present in the following sort of case. Suppose that after
breaking a table to pieces I manage to glue the fragments back
together into a table. I should then probably want to say that it
was the same table again, basing myself presumably on the
judgment that "it's still the same matter." If this is correct then
we have here a counterexample to the overall sortal analysis,
insofar as we regard this analysis as implying that the identity of

9. Locke, An Essay Concerning Human Understanding. Chisholm's account

of persistence seems to contain a variant of the same (I think objectionable)
procedure. On his account our concept of the persistence of an ordinary object
apparently depends upon our prior concept of the persistence of what he
sometimes calls the "primary objects" which constitute ordinary objects, where
these "primary objects" seem to be pretty much the same as what I am calling
"portions of matter." Besides the previously cited works, see his Person and
Object (Open Court, LaSalle, I11., 1976), "Problems of Identity" in M. K.
Munitz, ed., Identity and Individuation (New York University Press, N. Y.,
1971), and "Identity Through Time" in H. E. Kiefer and M. K. Munitz, eds.,
Language, Belief, and Metaphysics (State University of New York, Albany,
N. Y., 1970).
 THE PERSISTENCE OF MATTER 133

a standard object must be determined either by the observable

sortal-covered continuity of the object, or by the observable
sortal-covered continuity of the (major) parts that compose the
object. In the present case the judgment about the identity of
the table would be based neither on the sortal-covered continu-
ity of the table, nor on compositional considerations about the
observable sortal-covered continuity of parts of the table. It seems
that we perhaps need to relax the compositional criterion by
allowing into play theoretical compositional considerations vis-
a-vis the identity of matter.
It should be remarked, however, that this possible counter-
example to the sortal analysis of the identity of standard objects
infects only the supplementary compositional criterion, but not
the primary criterion of sortal-covered continuity expressed by
the sortal rule. We have still found nothing to jeopardize the
condition of sortal-covered continuity as logically sufficient for
the identity of a standard object; and we can still uphold the
general idea that a logically necessary and sufficient condition
for the identity of a standard object is that it satisfy either the
primary sortal rule or the supplementary compositional criterion,
though the latter is perhaps seen now as infected by the identity
of matter.10 As regards the identity of matter, of course, my
whole argument has been to show that the sortal analysis (at
least at an observational level) is inadequate, and that theoretical
considerations are required to define the identity of matter.
A question which might be addressed to my account is whether
I would say that a continuous portion of matter can persist as
that identical matter when it is fragmented. To this I would
suggest that, as with all questions about the identity of matter,
it must be settled on theoretical grounds, which means ulti-
mately by the scientist. From the standpoint of elementary
physics and chemistry it seerns that various conservation laws
imply that a portion of matter does not go out of existence even
when it is fragmented. As to whether we should then say that
a fragmented portion of matter, which was once continuous, is
(still) in any sense an "object," this seems to be merely an in-
consequential point of terminology which we can settle as we

10. This formulation of the general identity conditions for standard objects
will be reconsidered in Chapter 7.
134 THE CONCEPT OF IDENTITY

like. And, of course, elementary physics and chemistry may not

give us the final answer.
Again, from my point of view it would be a question essen-
tially to be settled by scientific theory whether or not matter
can jump discontinuously through space (a possibility which
some contemporary scientists may take seriously). I can see no
decisive reason to rule this out a priori. This possibility, by the
way, carries with it a correlative possibility with respect to stand-
ard objects, assuming that the compositional criterion is now
properly regarded as embracing the identity of matter. Suppose
that a table vanishes into thin air, and that immediately after-
wards a table bearing all of the distinctive marks of the first
appears in a different place. Insofar as our theoretical-explanatory
needs might possibly induce us in such a case to hypothesize that
the matter which constituted the first table jumped discontin-
uously through space and now constitutes the second table, we
might also be entitled, via the compositional criterion, to judge
that the first table is the same as the second.
I want to note that Russell sometimes explains the identity
of matter in terms akin to the account that I have been pre-
senting. After posing a problem about the identity of a drop
of water in the sea in the passage last quoted he goes on to say:
"The characteristic required in addition to continuity is con-
formity with the laws of dynamics."11 If by "the laws of dynam-
ics" Russell means the body of scientific laws that make reference
to bits of matter then this seems fairly close to what I am saying.
Often, however, Russell seems also to be suggesting that the
condition of conformity to laws has a special bearing not only
on the identity of matter, but on the identity of standard ob-
jects as well.12 But it seems to me excessively unclear what this
could mean. Certainly even the most elementary laws of physics
do not make reference to any standard objects, as ordinarily
conceived. For example, the elementary law "For every bit of
matter x, x's mass never changes" is not satisfied by a car as or-
dinarily thought of, since a car's mass changes whenever a tire
is removed. All that this shows of course is that the physicist's
11. "The Relation of Sense-Data to Physics," p. 171.
12. Such a suggestion seems pretty clearly implicit in his discussion in Human
Knowledge, Its Scope and Limits (Simon & Schuster, New York, 1948), pp.
458-60.
 THE PERSISTENCE OF MATTER 135

laws do not refer to (do not quantify over) standard objects as

such, that a car is not (strictly identical with) a bit of matter.
The physicist explains what happens to the car on the basis of
what happens to the matter which makes it up. But then what
precise bearing do the physicist's laws have on our concept of
the identity of the car?
We can perhaps say (and this is what Russell seems sometimes
to have in mind) that the career of a car exemplifies, if not the
physicist's laws, at least some rough regularities in terms of which
the successive stages of the car can be thought of as to some
degree causally connected (in terms of which, as Russell some-
times puts it, the successive stages of the car can be thought of
as forming a "causal line").13 But whatever precisely this might
mean it seems that we could certainly say the same thing about
the career of a shrinking incar, viz. that it exemplifies some rough
regularities in terms of which its successive stages can be thought
of as to some degree causally connected. This very weak condition
of causal connectedness does not, therefore, even rule out drastic
part-whole tracing confusions. As such it seems to have no clear
role to play in an analysis of identity. 14
I am tentatively inclined to think that with respect to what
is most plausibly regarded as a commonsense level of thought
(as opposed to an expert or technical level) we should mention
the condition of conformity to law only in the context of a
discussion of the identity of matter. And here the idea is that
the identity of matter is determined by theoretical considera-
tions, and is ultimately clarified by the laws of science. As regards
standard objects the essential condition to mention is sortal-
covered, or change-minimizing, continuity.
Laws of physics (and chemistry), which describe the behavior
of matter, are of course not the only laws of science there are,
though it is often held that all laws are ultimately reducible

13. Ibid.
14. The relationship between causality and identity is examined further in
Chapter 7. But let me here stress one point: If it is held that causal connected-
ness is an essential aspect of the continuity of a body's history, then the
causal condition can be added to the other continuity conditions in the Simple
Continuity Analysis of Chapter 1, the Sortal Rule of Chapter 2, and the Basic
Rule of Chapter 3; the overall structure of my argument (including the con-
trast between matter and standard objects) remains essentially intact.
136 THE CONCEPT OF IDENTITY

to these. We can perhaps think of the identity conditions asso-

ciated with such essentially technical concepts as "bacterium"
and "cell" as determined by the theoretical needs of biology in
rather the way that the identity of matter is determined by the
theoretical needs of physics. But I think that we should resist
the possible suggestion that even such nontechnical notions
as "same tree" and "same cat" ought to be seen as ultimately
determined by how these notions can be made to figure in the
best laws of biology.15 This suggestion can plausibly be resisted
at least to the extent of maintaining that there is a nontechnical
sense (level, part) of the concept "same tree," or "same cat,"
which operates quite independently of biological theory, and
which is determined essentially by relatively straightforward
considerations of observable sortal-covered, or change-minimizing,
continuity. If it is agreed that our ordinary concept of the per-
sistence of a car is essentially unaffected by physical theory (e.g.,
by the principle of mass constancy), then what reason would
there be to suppose that our ordinary concept of the persistence
of a tree is somehow contingent upon biological theory? It seems
more plausible to characterize all of our ordinary concepts of
the persistence of standard objects, whether these objects be
man-made or natural, as operating in essential independence of
scientific-theoretic considerations, and as being essentially de-
finable in the relatively simple terms of sortal-covered, or change-
minimizing, continuity. For the case of the identity of matter,
on the other hand, we apparently cannot coherently distinguish
any sense of the concept which is not already implicated in
essentially theoretical considerations, considerations which are
then eventually elaborated by science.
In summary, I have discussed altogether three commonsense
ideas of physical persistence: the basic conception, the sortal
conception, and now the theoretical-explanatory conception of

15. Such a suggestion might be implicit in the approach to "natural kinds"

that one finds in Hilary Putnam, "Is Semantics Possible?" in H. E. Kiefer
and M. K. Munitz, eds., Language, Belief, and Metaphysics, reprinted in
Mind, Language and Reality (Cambridge University Press, Cambridge, Lon-
don, 1975); and in Saul Kripke, "Naming and Necessity," in D. Davidson and
G. Harman, eds., Semantics of Natural Language (D. Reidel, Dordrecht,
1972), republished as Naming and Necessity (Harvard University Press, Cam-
bridge, Mass., 1980).
 THE PERSISTENCE OF MATTER 137

persisting matter. The connection between the first two of these

was that the sortal conception serves primarily to clarify the
basic one. Can we see any other connections here? I think we
easily can. Perhaps any observational concept is amenable to
theoretical extension. But there seems to be a sense in which
our basic concept of persistence stands especially ready for this.
The core of the basic conception was the rule that an object's
career should be traced so as to minimize changes. But there
seems to be a very natural movement of thought from saying
"Minimize changes" to saying "Simplify changes." We can think
of this movement as passing through the intermediary step "Do
not countenance a change in an object unless it is necessary."
This can be interpreted to mean "unless it is necessary for trac-
ing the object in a continuous path," which gives us the basic
rule, and its eventual sortal clarification. Or it can mean "unless
it is necessary for providing the simplest explanation of the
phenomena," which gives us the general procedure for judging
of the persistence of matter.
As I see it, then, at the root of our concept of physical per-
sistence is the vague rule of minimizing changes. From this rule
there emanate two rather contrasting conceptions, on the one
hand the relatively concrete and definite conception of the
persistence of different sorts of articulated objects, and on the
other the relatively abstract and indefinite conception of persist-
ing matter.
 5
The Metaphysics of Persistence

I. Do We Need Persisting Objects?

ONE KIND of question that may motivate a philosophical exam-
ination of the concept of physical persistence is whether this
concept is indispensable to our thought about the world, or
whether we could, on the contrary, conceive of the world in some
radically different way. Let me first consider this question with
regard to the observable persistence of standard objects, like
cars, and tables, and trees; in later sections I will extend the
discussion to include the persistence of matter. (But it should be
understood that special problems revolving around the per-
sistence of persons are not to be treated until a later chapter,
except perhaps in an occasional parenthetical aside.)
It seems to follow immediately from the preceding account
that the persistence of a car, or a table, or a tree boils down
essentially to there being a continuously related succession of
car-stages, or table-stages, or tree-stages. This point would ad-
mittedly need to be complicated to accommodate the composi-
tional criterion. But there would seern to be nothing in this
complication to discourage the general conclusion that, for any
standard sortal F, our ordinary descriptions of the persistence
of an F-thing could be dispensed with in favor of descriptions
of F-stages and their interrelations.
There are in principle an indefinite number of possible con-
structions in terms of which ordinary talk about the per-
sistence of, say, a car could be replaced by talk about the

138
 THE METAPHYSICS OF PERSISTENCE 139

interrelations of car-stages. We could construct a language which

allowed us to refer only to car-stages that last for one year or
less, or to car-stages that last for as long as they maintain a con-
stant color, or to car-stages that satisfy any number of other
possible specifications. Each such construction would apparently
allow us to redescribe what we ordinarily express about per-
sisting cars in terms of descriptions of how car-stages (of this or
that specification) are related to each other.
The most obvious, and also most radical, construction along
these general lines is one in which we are allowed to refer only
to momentary car-stages. This construction strikes us as par-
ticularly challenging insofar as it seems to depart as far as pos-
sible from the ordinary idea of an enduring object. I do not,
however, want to get bogged down now over puzzles about
whether momentary car-stages are to be conceived of as literally
instantaneous or, if not, how long they last. Some of these puzzles
would be merely reformulations of perennial puzzles in the
philosophy of time, while others are perhaps specific to the
present case.1 I will assume that a "momentary" car-stage is of
relatively short duration as compared to the normal duration
of a car as ordinarily conceived of, but I will leave it open just
how short this is.
We can imagine, then, a language, let us call it the M-language,
within which only momentary object-stages (and their sets,
successions, etc.) are referred to and described. In this language
we could say "A car-stage is (spatially) in contact with a tree-
stage," or "There occurs a continuous succession of tree-stages,"
or "There occurs a discontinuous succession containing tree-
stages followed by car-stages." But we would not talk in ordinary
ways about the persistence of a car or a tree.
To suggest that we could in principle dispense with our
ordinary concept of the persistence of a standard object in favor
of the M-language means essentially this. For any ordinary
statement O which describes in ordinary terms the career of a
standard object, there is some statement M within the M-language
such that O and M are equivalent. But the kind of "equivalence"
that is to obtain between O and M must not be construed too

1. Cf. my earlier questions about the duration of an "intrinsic car-stage";

Chapter 3, pp. 102—4.
140 THE CONCEPT OF IDENTITY

stringently. Certainly we cannot expect that M should constitute

a straightforward translation of the ordinary statement O, in
the sense in which a French statement might constitute a trans-
lation of an English statement. Nor, even, do we want to require
that O and M should be analytically equivalent (i.e., that they
should analytically entail each other) in the strict sense in which,
for example, "x is a triangle" is analytically equivalent to "x is
a three-sided polygon." For one thing, an ordinary statement O
will suffer from some margin of vagueness, in virtue of which
certain states of affairs would render its truth-value indeter-
minate. And we have no reason to require that M should so
perfectly match up with O as to mirror exactly this margin of
vagueness. It would in fact be sufficient for our purposes if the
force of M approximates even roughly to that of O. The ques-
tion which concerns us is whether the M-language has as much
fact-stating power as our ordinary language, whether the M-
language would allow us to describe the world as we experience
it without, so to speak, missing out on anything. Certainly the
previous account of the nature of physical persistence would
suggest an affirmative answer to this question (at least insofar
as we continue to limit our attention to the observable persistence
of standard objects). That account would seem to imply that, in
terms of the idea of sortal-covered (or change-minimizing) con-
tinuity, we could in principle frame descriptions within the
M-language that are loosely, even if not strictly, equivalent to
our ordinary descriptions of persisting objects.
It might be suggested that the M-language does not really
dispense with the concept of a persisting object since even in
that language we can refer to successions of momentary stages,
and these successions, insofar as they contain elements that exist
at different times, are themselves in a sense persisting objects.
Now it is a relatively unimportant point of terminology
whether we apply the expression "persisting object," and kindred
expressions, to an M-level succession. The more important point
is that an M-level succession is not governed by any rules of
identity through time of the sort that govern our ordinary
thought. (I want to ignore any questions about identity through
space in the present discussion.) At the M-level any succession
of momentary stages, however the succession might be formed
(i.e., even if it combines what would ordinarily be thought of
 THE METAPHYSICS OF PERSISTENCE 141

as the stages of different objects), counts equally with any other.

This is the essential contrast with our ordinary conception, which
accords the special status of persistence, of unity through time,
to just those select successions of object-stages that are related
in some special way (e.g., by sortal-covered continuity).
The idea of the M-language is essentially the idea of a language
without rules of identity through time, a language in which
any succession of momentary object-stages is logically on a par
with any other. Such a language is, in one very important sense,
a language without the concept of persistence. And it appears
now that we could describe the world in terms of such a language.
A question might be raised about the intelligibility of the idea
of an "object-stage" within the M-language. Object-stages, it
might be objected, are abstract items (perhaps ordered pairs
of objects and times), which go together to make up another
kind of abstract item, the career or history of an object. But our
idea of these abstract items presupposes the idea of a concrete
persisting object. If we, therefore, try to do without our ordinary
concept of a persisting object we can no longer understand
what an object-stage is.
Now this objection is well taken insofar as it emphasizes the
kind of conceptual departure from ordinary thought which the
M-language would have to embody. From the point of view of
the M-language a "momentary object-stage" must be conceived
of as a concrete entity, which has the various properties (of
shape, color, etc.) which we ordinarily attribute to an object at
a given moment. We would therefore do better, perhaps, to
represent the M-language as containing such expressions as "suc-
cession of (momentary) trees," rather than "succession of (mo-
mentary) tree-stages." It may be admitted that ordinary language
does not permit us to think of a persisting object as temporally
divisible into concrete momentary parts. But there seems to be
nothing incoherent about a conceptual revision which would
permit this.
The legitimacy of thinking about an object as divisible into
temporal parts has been amply defended in recent literature.
Quine, for example, expresses this idea as follows: "A physical
thing—whether a river or a human body or a stone—is at any
one moment a sum of simultaneous momentary states of spatially
scattered atoms or other small physical constituents. Now just
142 THE CONCEPT OF IDENTITY

as the thing at a moment is a sum of these spatially small parts,

so we may think of the thing over a period as a sum of the
temporally small parts which are its successive momentary states.
Combining these conceptions, we see the thing as extended in
time and in space alike."2
For my present purposes there is no important distinction to
be drawn between the M-language and Quine's "space-time"
conception. The essential characterization of both conceptions
is that they contain no rules of identity through time, and are
therefore in an important sense free of the concept of persistence.
Quine's space-time idiom suggests the point that the momentary
items that figure in the persistence-free conception are to be
conceived of as concrete entities, on a parallel with the con-
creteness of the spatially small parts of objects. Moreover Quine's
space-time idiom contains the intrinsically interesting (though,
for my present purposes, not especially relevant) twist of regard-
ing the successions ("sums," as Quine called them) that are
formed from the momentary items as themselves concrete en-
tities that are extended in space and time. From Quine's space-
time vantage point reality is seen as comprised of concrete
space-time portions, whether continuous or discontinuous, that
can be characterized in many of the ways that we characterize
ordinary objects. The all-important point that I want to keep
in central focus is that the space-time portions of reality figuring
in Quine's scheme crisscross and overlap in every conceivable
way, and are not conceptually structured by the kinds of identity
rules that govern our ordinary concept of persistence.
There is evidently no prima facie connection between the idea
of a language that lacks rules of identity through time, a lan-
guage in which any succession of momentary items is logically
on a par with any other, and the idea of a language that lacks
the distinction between subjective experience and objective
physical reality. It is true that many philosophers of the past
who have tried to conceive of the physical world as ultimately
composed of momentary things were phenomenalists. Their
momentary things were sense data and the like, items that were
supposed to be mental (or private). But the kind of persistence-
2. W. V. Quine, Methods of Logic, grd ed. (Holt, Rinchart and Winston, Inc.,
N.Y., 1972), p. 222. For further discussion of "object-stages" and "temporal
parts," see Chapter 6, below.
 THE METAPHYSICS OF PERSISTENCE 143

free conception that I have been talking about is not intended

to be in the least bit tainted by phenomenalism. Certainly the
momentary things that figure in Quine's space-time conception
are supposed to be fully objective; and the conception as a
whole is to be an analytic restructuring of our ordinary thought
about the physical world, with no loss of any sense of objectivity.
Now there may possibly be some deep arguments to establish
that our distinction between subjective experience and objective
physical reality is inextricably linked to our conceptually struc-
turing physical reality in terms of rules of identity through
time. But such a linkage would certainly need to be established,
and not merely taken for granted. I do not in fact know of any
relatively clear argument which might establish this linkage.
And my impression is that many recent discussions of the
subjective-objective distinction simply take the linkage for
granted without seeing the need to establish it.
Such seems to be the case in Strawson's discussion of the con-
ditions of objectivity in his chapter "Sounds" in Individuals.
Strawson is there concerned with two questions. The first is
whether the conditions of a "non-solipsistic consciousness" can be
fulfilled within a purely auditory experience, that is, roughly,
whether a being whose experience was purely auditory could
make sense out of a distinction between subjective experience
and objective reality. The second question is whether there could
be enduring (persisting) and reidentifiable sound-particulars in
the purely auditory world. Strawson argues that an affirmative
answer to the first question would necessarily carry with it an
affirmative answer to the second:
For to have a conceptual scheme in which a distinction is made between
oneself and one's states and auditory items which are not states of one-
self, is to have a conceptual scheme in which the existence of auditory
items is logically independent of the existence of one's states or of one-
self. Thus it Is to have a conceptual scheme in which it is logically
possible that such items should exist whether or not they were being
observed, and hence should continue to exist through an interval during
which they were not being observed. So it seems that it must be the case
that there could be reidentifiable particulars in a purely auditory world
if the conditions of a non-solipsistic consciousness could be fulfilled for
such a world.3
3. Strawson, Individuals, p. 72.
144 THE CONCEPT OF IDENTITY

I want to focus on the second sentence in the above argument.

The sentence begins with the correct observation that a non-
solipsistic consciousness would necessarily require a conceptual
scheme in which certain items are thought of as existing while
unobserved. But then Strawson simply jumps at the end of the
sentence to the unargued conclusion that, since these items are
being thought of as existing while unobserved, they must be
thought of as continuing to exist while unobserved, i.e., they
must be thought of as enduring particulars. But why could not
the auditory scheme fulfill the conditions of a non-solipsistic
consciousness by including the idea of objective momentary
sounds, and successions of momentary sounds, which exist while
unobserved? How has Strawson established the necessity of en-
during sounds?
Of course, if by a "reidentifiable sound," a "sound that con-
tinues to exist," Strawson merely meant any succession of
sounds, then his conclusion would follow trivially. But it is
evident that this is not Strawson's intention. He is clearly trying
to establish the necessity of enduring sound-particulars in the
sense of particulars that are governed by rules of identity through
time, rules which are satisfied only by certain successions of
sounds, and which accord only to these successions the special
status of enduring particulars. But Strawson has failed to show
(at least in the quoted passage) that a scheme of objective par-
ticulars must necessarily include any such rules of identity
through time.

II. A Question about Spatiotemporal Continuity

Our ability to redescribe the world of standard objects in the
persistence-free terms of the M-language (the space-time lan-
guage) would depend upon our being able to express within
that language the idea that a succession of momentary things
is spatiotemporally and qualitatively continuous. The notion of
qualitative continuity seems to present no special difficulty since
the application of this notion requires nothing more than our
ability to make qualitative comparisons between the momentary
things which comprise a succession. But the notion of spatio-
temporal continuity does present a special difficulty.
In Chapter i, I defined several senses of spatiotemporal con-
 THE METAPHYSICS OF PERSISTENCE 145

tinuity, all of which revolved around the idea of two places over-
lapping each other. Roughly, a succession S was said to be
spatiotemporally continuous if the places which coincide with
temporally neighboring stages in S overlap, where the extent of
overlap that was required varied with each definition. Now to
say that two places overlap each other is obviously to say that
they overlap each other at a given time. That is, the places must
exist simultaneously. Of course when we say that two places
overlap each other we need not relativize this remark to some
specific time, since if two places ever overlap they always overlap.
The overlap relations, as well as the distance relations, between
places cannot possibly change. But the fact that we need not
specify a time must not obscure the point that for places to over-
lap they must exist together, i.e., at the same time.
For S to be spatiotemporally continuous the place p±, which is
occupied by the momentary thing in S at t±, must overlap the
place p2, which is occupied by the momentary thing in S at £ 2 ,
where ^ and t2 are neighboring times. This means that p^ and
p2 must exist together at some time, that we cannot think of pi
as existing only at i L and p2 existing only at t2. We cannot, that
is, think of p, and p2 as "momentary places," the counterparts
of momentary things. It is necessary that we should think of pt
and p.2 as ordinary persisting places, which exist not only at ij,
or at < 2 , but throughout an extended period, during which they
continually overlap. The upshot of this is that if the notion of
spatiotemporal continuity is to be explained along the general
lines discussed in Chapter i, then this notion involves essentially
the idea of a persisting place.
The difficulty now is that the idea of a persisting place seems
to presuppose the idea of a persisting object. Indeed the iden-
tity through time of a place must apparently be thought of as
relative to some specific object or system of objects. A succession
of lightning flashes that occur at the same place relative to the
earth occur at different places relative to the sun. This suggests
that our ordinary notion of "same place again" is in effect short
for "same place relative to such-and-such objects," where the
specified objects supply a coordinate system in terms of which
any place can be assigned a unique, and permanent, set of co-
ordinates. (Roughly, we are at the same place relative to such-
and-such objects if we are at the same distance and direction
146 THE CONCEPT OF IDENTITY

from the objects.) Consequently the M-language (the space-time

language), which excludes the ordinary idea of a persisting ob-
ject, cannot be coherently thought of as containing the idea of a
persisting place. How, then, can it contain the required notion
of spatiotemporal continuity?
I think that there are essentially two ways in which we might
respond to this question. The more moderate way would be
to concede the force of the question, and accordingly to limit
the scope of the persistence-free language; the more radical way
would be to concede nothing to the question. Let me first ex-
plain the moderate approach.
Suppose that we concede the point that the notion of spatio-
temporal continuity cannot be understood independently of
our concepts of persisting objects and persisting places. This
would imply that the world as we know it could not be ade-
quately described at the persistence-free level, that a thorough-
going reduction of our identity concepts in terms of the
persistence-free level is not possible. However, we could still
consider the following more limited possibility. We might desig-
nate some specific group of objects as providing our spatial
framework. The identity concepts associated with these objects
are not to be dispensed with in the persistence-free language
but are to be taken for granted. Then, presupposing the spatial
framework provided by these concepts, we could render all other
identity concepts in persistence-free terms.
We might, for example, designate buildings and their parts
as the objects that provide the framework. Presumably the class
of buildings and their parts will contain a sufficient number
and variety of rigid edges, movable rigid bodies, and whatever
else may be required to coordinate in the ordinary manner a
system of persisting places. Given a system of persisting places
we can define spatiotemporal continuity as in Chapter i. And
now the persistence of every standard object other than build-
ings and their parts can be rendered in persistence-free terms.
According to the moderate account it is incorrect to say, "We
could redescribe the world of standard objects in such a way
that, for every sortal F, our language would not employ the
concept of the persistence of an F-thing." But it remains correct
to say, "For every sortal F, we could redescribe the world of
standard objects in such a way that our language would not
 THE METAPHYSICS OF PERSISTENCE 147

employ the concept of the persistence of an F-thing." We are

not, that is, permitted to think of all our ordinary descriptions
of persisting objects as eliminable in toto by the persistence-free
language. We can, however, think of any specific range of such
descriptions as eliminable without loss. Given a background of
persisting buildings, and the spatial framework thereby provided,
we can dispense with persisting trees; and perhaps vice versa.
But some objects must remain to provide the framework. The
general category of a persisting physical object would thus be
seen as an irreducible and indispensable component of our
thought about the world, but the specific embodiments of this
category in our various concepts of specific kinds of objects
may be severely limited without detriment to the facts.
This moderate account strikes me as adopting an uncomfort-
able half-way position. If we can, so to speak, get rid of any kind
of persisting object that we choose then it seems unsatisfactory
to insist that we somehow cannot get rid of all of them together.
I think that there is in fact no compelling reason to limit the
scope of the persistence-free language in the way that the mod-
erate approach suggests. The kind of explanation of spatio-
temporal continuity that I gave earlier, in terms of a presupposed
framework of persisting places, is one natural-seeming explana-
tion of that notion within our ordinary conceptual scheme. This
need not prevent us, however, from regarding the notion in a
different light as it figures in the persistence-free scheme. In the
latter scheme spatiotemporal continuity might be treated as a
primitive idea which is explained ostensively, i.e., by exhibiting
examples of spatiotemporally continuous successions. There
seems in general to be no uniquely correct way to order our
ideas; and in the present case, at any rate, the alternative of
treating spatiotemporal continuity as primitive, rather than as
defined in terms of other concepts, seems quite plausible. Cer-
tainly we may be inclined to think of the spatial continuity of
an object at a given moment as directly observable and con-
ceptually simple. Why then should we not treat the spatiotem-
poral continuity of a temporally extended portion of reality in
the same light?4
4. Compare with Whitehcad's treatment of the continuity of an "event," in
The Concept of Nature, pp. 74-78. I defend the primitiveness of spatio-
temporal continuity further in Chapter 6, Sections V and VII.
148 THE CONCEPT OF IDENTITY

The analogy to spatial continuity may suggest a slightly dif-

ferent approach, much to the same effect. Spatial continuity
might be treated as a primitive concept, but it may also be de-
fined in several ways, one of which is this: "x is spatially con-
tinuous at time t" can be denned as meaning that, for any y and
z, if y and z exhaustively comprise x at t (i.e., y and z are parts of
x such that any part of x overlaps either y or z), and y and z do
not overlap at t, then y and z touch each other at t. (I am allow-
ing for the possibility that overlapping objects perhaps cannot be
said to touch each other.) Here we define spatial continuity in
terms of touching, though the reverse procedure seems equally
plausible.
The term "spatiotemporal contiguity" is sometimes used to
signify the spatiotemporal analogue of touching, to signify,
that is, the relationship which stands to spatiotemporal con-
tinuity in the way that touching stands to spatial continuity.
Hence "The succession S is spatiotemporally continuous" is
equivalent to "For any S, and S2, if S± and S2 are temporal seg-
ments of S which exhaustively comprise S, and S, and S2 do not
overlap, then S, and S2 are spatiotemporally contiguous to each
other." If we can assume the notion of spatiotemporal contiguity
in the persistence-free language then we can easily define spatio-
temporal continuity in terms of it.
Now the ordinary notion of one thing turning into (being
replaced by) another thing, in those cases in which the second
thing might be said to come into existence when the first thing
goes out of existence, is closely related to the idea of spatio-
temporal contiguity. The proposition "x went out of existence
and turned into y, which came into existence" (e.g., "The car
went out of existence and turned into the block of scrap metal,
which came into existence") can be roughly associated with the
proposition "A terminal segment of x's career is spatiotemporally
contiguous with an initial segment of y's career." To ascertain
just how close this association is would require a more detailed
examination of the ordinary notion of "turning into" than I
now want to undertake. But even a rather loose association en-
courages the judgment that, if the ordinary notion of "turning
into" can be treated as a primitive observational concept, and
this seems quite plausible, then it should also be possible to
treat spatiotemporal contiguity in the same spirit. And given
 THE METAPHYSICS OF PERSISTENCE 149

spatiotemporal contiguity we can, as indicated before, immedi-

ately define spatiotemporal continuity. (Hence the idea of a
spatiotemporally continuous succession can be loosely associated
with the idea of a succession which is such that any early segment
"turns into" a later segment.)
I conclude that the required notion of spatiotemporal con-
tinuity can be made available in the persistence-free language
without relying on a presupposed framework of persisting places.
Though it seems reasonable from the ordinary point of view to
connect the notion of spatiotemporal continuity to our ordinary
idea of a system of persisting places, severing this connection
appears to impose no overwhelming strain on our understand-
ing. It seems therefore that we can adequately redescribe the
world of standard objects in wholly persistence-free terms.

III. Identity Schemes

The persistence of any standard physical object can be regarded
as consisting simply in the occurrence of a succession of momen-
tary things, which is spatiotemporally and qualitatively con-
tinuous, and sortal-covered or change-minimizing. This is
obviously so if we accept the radical account, according to which
the notion of spatiotemporal continuity can be made intelligible
within a wholly persistence-free language. But even if we limit
ourselves to the more moderate account, and think of the
persistence-free language as presupposing a background of per-
sisting objects, in terms of which the required notion of spatio-
temporal continuity is explained, the conclusion still holds that,
given some relatively limited background of persisting objects,
the persistence of all other objects could be rendered in
persistence-free terms. At the very least, then, we can regard the
persistence of any specific object, or even any specific range of
objects, as consisting simply in the successive existence of suit-
ably related momentary things.
This conclusion may make us feel somewhat uneasy. We may
be inclined to say that if the persistence of a given object boils
down to nothing more than there being a certain kind of se-
quence of momentary things, then the object does not really
persist at all, and our ordinary way of thinking about the matter
is simply wrong. But the inclination to say this, however natural
150 THE CONCEPT OF IDENTITY

it may be, must, I think, be resisted. The inclination stems

basically from the difficulty that we have in acknowledging
that the world can be described, with equal completeness and
correctness, in more ways than one. "If you could correctly de-
scribe the world, or some part of it, without mentioning any
persisting objects," we are inclined to argue, "then there must
not really be any persisting objects there to be mentioned."
This argument, however, must be rebuffed by insisting on the
point that the correctness of one mode of description does not
necessarily preclude the correctness of some radically different
mode of description. If the preceding account is accepted it
would not follow that the world really consists of only criss-
crossing sums of momentary things, but no genuinely persisting
objects. The proper way to understand this, rather, is that what
we ordinarily describe correctly in terms of the idea of a persist-
ing object could also be described correctly in terms of the idea
of a sum of momentary things. The compatibility of these two
modes of description can be brought out by saying that what
would be thought of in the persistence-free language as a certain
kind of sum of momentary things is precisely what we call "a per-
sisting thing."
The adequacy of the persistence-free language (whether this is
understood in the radical sense or the moderate sense) is not,
then, to be construed as in any manner denigrating the correct-
ness of ordinary language. The persistence-free language is not
an "ideal language" that gives us a glimpse into the world-as-
it-really-is which is somehow hidden by ordinary forms of ex-
pression. The primary interest of the persistence-free language
is that it affords a vantage point which is outside our ordinary
identity scheme, and from which, therefore, we can gain a deeper
insight into the character of that scheme. Such a vantage point
may also have a certain intrinsic philosophical appeal, in that
it provides a view of reality which is less structured, and hence
in a sense less artificial, than our ordinary one. The relative non-
structuredness of the persistence-free language, as compared to
ordinary language, derives from the fact that in the former
language any succession of momentary things is logically on
a par with any other, whereas in ordinary language certain suc-
cessions, but not others, are accorded the special status of con-
stituting a unitary object. But there is no question of our having
 THE METAPHYSICS OF PERSISTENCE 151

to decide which language gives us the uniquely correct view of

reality.
If we define an "identity scheme" as a system of rules which
determines which space-time portions do, and which do not,
qualify as unitary objects, then we can perhaps say that the
persistence-free language (which we can henceforth imagine on
the model of Quine's space-time language) contains a maximally
permissive identity scheme. In a sense it contains no identity
scheme at all. From the space-time vantage point, "[a]ny arbitrary
congeries of particle-stages, however spatiotemporally gerry-
mandered or disperse, can count as a physical object."5 "Each
[object] comprises simply the content, however heterogeneous,
of some portion of space-time, however disconnected and gerry-
mandered."6 Because of its extreme permissiveness the space-time
language provides a kind of neutral ground to which a philoso-
pher may naturally retreat when reflecting about identity schemes.
Adopting an identity scheme may be compared with attaching
a particular sense to the expression "(same) object" and logically
equivalent expressions. In the maximally permissive persistence-
free scheme any sum of momentary stages can correctly be
brought under the heading "one and the same object." The
definitive feature of this scheme is that a statement of the form
"Some object was A at i t and some object was B at a later time
t2" (where A and B are terms that attribute qualities or spatial
relations) entails both "Some object (i.e., the sum of the A-stage
at t1 and the B-stage at t2) was A at t± and B at t2" and "Some
object (i.e., the A-stage at tj) was A at <t and non-existent at t2."
These entailments obviously do not hold given the ordinary
sense of "(same) object."
The fact that the space-time language contains no rules of
identity through time, so that in that language any space-time
portion counts equally with any other, is what leads me to de-
scribe that language as being free of the concept of persistence.
As intimated earlier, however, I could not strenuously object to
someone's favoring a different terminology, according to which
that language is said to contain, rather than no concept of per-
sistence, a concept of persistence different from ours. At the other
5. Quine, The Roots of Reference, p. 54.
6. W. V. Quine, Word and Object (The M.I.T. Press, Cambridge, Mass., 1960),
p. 171.
152 THE CONCEPT OF IDENTITY

extreme, I could not object to someone's describing that language

as being devoid not merely of the concept of persistence, but
also of the very concept of an object (or body or entity). Often
philosophers have characterized the space-time language (or
kindred languages) as breaking reality down, not into objects,
but into processes or events (or, in Strawson's suggestive expres-
sion, process-things).1 These terminological nuances are less im-
portant than grasping the various logico-grammatical analogies
and disanalogies between the space-time language (what I choose
to call the "persistence-free language") and ours. Though I have
not attempted to spell out these analogies and disanalogies in
any detail, evidently the essential analogy is that the referential
apparatus of the space-time language, as well as ours, centers
upon items that are thought of as standing in spatial and tem-
poral relation to each other, and as bearing ordinary sensible
qualities like size, shape, color, and texture. And the essential
disanalogy is that the space-time scheme contains no rules of
identity through time.
The space-time language provides a maximally permissive
identity scheme containing no rules of identity through time.
We can also conceive of identity schemes different from our
ordinary one which do contain rules of identity through time,
rules different from the ordinary ones. This point can be illus-
trated by considering a somewhat generalized version of the
previously discussed incar-outcar language. Let us now try to
imagine a generalized in-out language which contains a great
many strange-seeming descriptions on an analogy with the incar-
outcar one. We can build up to the idea of this language by
associating various kinds of objects with surroundings that are
especially significant with respect to these objects, in rather the
way that a garage is especially significant with respect to a car.
Thus the language might allow one to refer to (and reidentify
accordingly) inpigs and outpigs, depending on whether the ob-
ject is inside or outside a pen, onbooks and offbooks, depending
on whether the object is on or off a bookshelf, onapples and
offapples, depending on whether the object is on or off a tree,
and so on. The details of the language do not matter, so long
as we have the general idea of a language containing many terms

7. Strawson, Individuals, p. 56. Cf. Quinc, Word and Object, p. 171.

 THE METAPHYSICS OF PERSISTENCE 153

which stand to our ordinary terms in rather the way that "incar"
and "outcar" stand to "car." Such a language could be said to
contain a different identity scheme from ours.
Now it is surely an incontrovertible fact, which reflects a fea-
ture of our ordinary use of language, that when a car leaves a
garage no object gets smaller in the process. Yet in the in-out
language speakers describe the situation of a car leaving a
garage by saying "An object, namely the incar, got smaller."
This shows that they must be using the sentence "An object got
smaller" differently from the way we do. It would perhaps not
be incorrect to say that they use every word in that sentence
(and perhaps indeed every nonlogical word in their language)
differently from us. But it seems more natural to pin the differ-
ence on their use of "(same) object," and to say that they use this
expression differently from us, that they operate with a different
concept of (the identity of) an object.
Whereas I am inclined to characterize the space-time language
as containing no concept of persistence, it seems natural, surely,
to characterize the in-out language as containing a concept of
persistence, but a different one from ours. This is because that
language is analogous to ours in containing rules of identity
through time, though these rules are, at least to a significant
extent, different from ours. So the space-time language, as I want
to look at this, gives us a persistence-free scheme, while the in-out
language gives us an example of an identity scheme containing
a different concept of persistence from our ordinary one.
Could there possibly be people who speak either of these
strange languages, the persistence-free language or the in-out
language, as their natural language, who learn one of these as
their first language and who think in terms of it? My conjecture
would be that this is probably an empirical impossibility, since,
as the discussion of Chapter 3 suggested, these languages would
contravene our basic concept of persistence, which I am inclined
to regard as in some sense instinctive. But I hold that there is
no logical inconsistency or incoherence in the idea of people
speaking such languages, in people operating with those different
concepts of the identity of an object. Nor, I believe, is there any
sense in which these languages, if they were spoken, would give
a less correct description of the world than our ordinary
language.
154 THE CONCEPT OF IDENTITY

This point may be especially hard to accept with respect to the

in-out language. Insofar as this language does contain (what
would naturally be regarded as) judgments of persistence, these
judgments may appear to be simply logically incompatible
with our ordinary judgments. If, as we ordinarily judge, noth-
ing gets smaller when a car leaves a garage, how can they be
correct in judging "An incar got smaller"? The futility of this
sort of objection, however, can be brought out by turning it on
its head. Since the incar-outcar description is (was introduced
as being) merely an unfamiliar sentence used to describe the
familiar facts, given that those facts obtain how could the sen-
tence possibly fail to assert a truth? Again, what I think needs
to be overcome here is the idea that there must be some uniquely
correct way of describing reality.
There is at some level the inclination to say this. "Yes, there
could conceivably be people whose conventions of language en-
joined them to utter the words 'An incar got smaller' in a
situation in which a car leaves a garage. But then what they
would have to mean by uttering these words is simply that a
car left a garage."
This seems to suggest that they somehow could not mean what
they say but must instead mean what we say, as if they merely
uttered the words of their strange language while secretly think-
ing in English. But if it is conceivable that they should speak
that language then it surely must also be conceivable that they
should internalize it, i.e., think in terms of it. Since, by hy-
pothesis, our statement "A car left a garage" is roughly equiva-
lent to their statement "An incar moved toward the exit of a
garage and then diminished in size until it vanished, while
simultaneously an outcar appeared at the outside of the exit
and gradually grew to the size and form of the original incar,"
we can indeed say that their statement merely expresses the fact
that a car left a garage. But by the same token we can also say
(shifting now to their language) that the English statement
merely expresses the fact that an incar moved toward the exit
of a garage and then diminished in size, etc. There is no lack
of symmetry here between the descriptions in the two languages.
It is easy to make the mistake of supposing that the incar-
outcar description is incorrect because it would play havoc with
 THE METAPHYSICS OF PERSISTENCE 155

science. How, it might be asked, could a physicist explain the

strange loss of mass of a shrinking incar?
But the truth is that the physicist could easily explain this in
precisely the way he explains the loss of mass of an ordinary car
which has a part subtracted from it. The physicist could say that
the mass of an incar decreases because the matter which makes
it up at one moment is identical with only part of the matter
which makes it up at a later moment (the rest of the matter going
into the outcar). As noted once before even the most elementary
laws of physics (e.g., the principle of mass constancy) do not
refer (as such) to ordinary objects like cars, but refer instead to
the matter which makes up these objects. There is therefore no
reason to require that these laws refer (as such) to incars and
outcars, nor less that these laws be rendered in some such terms
as "inmatter" and "outmatter." It may indeed be conceivable
that the hypothetical physicists of that language could develop
some suitable in-out substitute for our ordinary concept of per-
sisting matter. This possibility, or a variant of it, will be discussed
in the next section. My present point is the much simpler one
that our ordinary physics, in terms of our ordinary concept of
persisting matter, can explain, without the slightest difficulty,
all of the strange-sounding expansions and contractions of incars
and outcars.
In short, if our ordinary statements about cars gaining and
losing parts can be accepted as correct and essentially theoret-
ically innocent descriptions of the observable phenomena, then
the incar-outcar statements can be accepted in precisely the same
spirit. Our ordinary identity scheme and the in-out scheme
would be merely two conceptual devices for framing different
but equivalently correct descriptions of what is in some sense
the same observable phenomenon.
The metaphysical attitude which I have been expressing is
familiar from much recent literature. It is an attitude that one
associates with the later Wittgenstein, and with one sense of
the word "relativism." A capsule summary of this attitude is
aptly expressed by Urmson toward the end of his book Philo-
sophical Analysis: "If two sentences are equivalent to each other,
then while the use of one rather than the other may be useful
for some philosophical purposes, it is not the case that one will
156 THE CONCEPT OF IDENTITY

be nearer to reality than the other. . . . We can say a thing this

way and we can say it that way, sometimes; if we can it may be
helpful to notice it. But it is no use asking which is the logically
or metaphysically right way of saying it."8
I judge Urmson's dictum to be difficult, perplexing, and
essentially correct. As it pertains to the present topic of persis-
tence the dictum implies that an object's persistence is no less
"real" or "genuine" just because statements which describe the
object as persisting could in principle be replaced by equivalent
statements in which the object's persistence does not figure.

IV. "Real" and "Fictitious" Persistence

I argued in the last chapter that the persistence of a portion of
matter, in contrast to the persistence of a standard object, is
in a sense ultimate and unanalyzable. Now it may seem to fol-
low from that account that what I have been saying in the pres-
ent chapter about the persistence of standard objects ought not
to be said about the persistence of matter, that our concept of
the persistence of matter, at least, could not be eliminated, or
altered, without our thereby losing the ability to describe the
world correctly. Perhaps this conclusion does follow; I confess
to feeling considerably less than confident about this point. But
I am inclined to think that it does not follow, and I now want
to state why.
It may be that there are strictly two somewhat different ques-
tions to be considered. One is whether our ordinary concept of
persisting matter could be eliminated in favor of a persistence-
free conception. The second question is whether that ordinary
concept could be eliminated in favor of some alternative con-
ception of persistence, something, say, on the order of an "inmat-
ter"-"outmatter" conception. I will focus on the first question,
which seems a bit more tractable, though what I say would, I
think, carry over rather directly to the second question as well.
The persistence of matter was argued to be "ultimate" in the
sense that we cannot state observational identity criteria for mat-
ter. But I now want to suggest that there is nothing in this to
8. J. O. Urmson, Philosophical Analysis (Oxford University Press, London,
1956), p. 186.
 THE METAPHYSICS OF PERSISTENCE 157

prevent us from coherently regarding a persisting portion of

matter as made up of temporal parts, and regarding its persist-
ence as consisting in nothing more than its momentary parts
standing in some unobservable relationship. We might give this
relationship the name "genidentity," along the lines of one of
the approaches that I sketched in the preceding chapter. (A com-
pletely parallel point could be made, perhaps with even greater
plausibility, if we adopt the alternative a priori atomistic ap-
proach to matter.) We can then redescribe the persistence of a
bit of matter in terms of the equivalent idea of the occurrence
of a succession of momentary things that are genidentical with
each other.
In general, I would suggest that we need to distinguish be-
tween these two questions: (a) "Can we coherently regard the
persistence of x as consisting in the occurrence of a succession
of momentary things that stand in some distinctive unity-making
relationship?" and (b) "Can we give an account in relatively ob-
servational terms of what that unity-making relationship is?"
My overall position commits me to holding that where x is a
standard object the answers to both (a) and (b) are affirmative,
whereas where x is a portion of matter the answer to (b) is nega-
tive. But I am now maintaining that a negative answer to (b)
does not imply a negative answer to (a).
I think, in fact, that it is a general a priori truth that the
answer to (a) must always be affirmative, no matter what kind of
entity x is. This seems to follow from the bare idea of persistence
through time. If x is an entity that persists through time then,
no matter what sort of entity x is, we can make intelligible to
ourselves a conceptual revision which allows us to redescribe x
as made up of temporal parts. That we can "make it intelligible"
is perhaps only another way of saying that we can draw many
clear and persuasive analogies to the revisionary idea of the
entity's having temporal parts from our ordinary idea of an
object's spatial parts, and also from our ordinary idea of the
temporal parts of a process, such as a game or a battle. In the
face of these analogies I can see no point in someone's insisting
that the contemplated redescription (and this is all that it would
be, a redescription of the accepted facts) is somehow illegitimate
or incomprehensible. But, now, if we have gotten ourselves to
158 THE CONCEPT OF IDENTITY

think about x as made up of temporal parts then we can, indeed

must, think of those parts as standing in some relationship which
constitutes x's unity through time. This conclusion remains valid
even in a case in which the unity-making relationship for x must
be regarded as nonobservational.
Our concept of the persistence of matter is, on my view, some-
thing on the order of a theoretical extension of our concept of
the observable persistence of standard objects. What we mean
even at a commonsense level by "the persistence of matter" is
something to the effect of "that underlying mode of physical
persistence which can ultimately provide the simplest and most
coherent explanation of the observable phenomena." It is there-
fore scientific theory which eventually fills in the detailed facts
about the behavior of matter. But once the scientist presents us
with those facts there cannot, I think, be any factual or meta-
physical error in redescribing them in persistence-free terms, in
terms of the idea of a genidentical succession of momentary
things. There is, at least on my intuition, no way to attach any
sense to the question "How is it really in the world? Does matter
persist or are there merely momentary things related to each
other by genidentity?" These are two ways of saying the same
thing. If we describe the world in terms of persistence we can say
that the underlying persistence of matter is what ultimately ex-
plains the observable persistence of standard objects. And if we
describe the world in persistence-free terms we can make the
correlative and equally correct remark that the underlying rela-
tionship of genidentity is what ultimately explains the observ-
able relationships of sortal-covered or change-minimizing con-
tinuity. There is no question of fact at stake here.
It will be objected perhaps that what is wrong with the per-
sistence-free rendition of the facts about matter is that this
rendition would be less simple and less coherent than the ordi-
nary formulation. This assessment may indeed seem to follow
directly from my view that our concept of persisting matter is
essentially the concept of a mode of persistence in terms of which
it would be possible to formulate theoretically best explanations.
But there are difficulties here. What needs to be borne in mind
is that our ordinary concept of persisting matter presupposes
our general commitment to the concept of persistence, our gen-
 THE METAPHYSICS OF PERSISTENCE 159

eral commitment, that is, to describing the world in terms of a

category of temporally extended units that do not merely corre-
spond to arbitrary successions. Given this general constraint our
concept of persisting matter implies that these units are to be
selected in such a way as to maximize simplicity and coherence.
What is at issue at present, however, is precisely the possibility
of removing this general constraint by repudiating the concept
of persistence altogether. It is not clear that an overall descrip-
tion of reality in persistence-free terms would have to be less
simple than an overall description in terms of persistence. (Con-
temporary science, on some readings, may in fact favor a kind
of persistence-free description.)
More specifically, it would seem that our ordinary concept of
persisting matter presupposes, and is indeed an outgrowth of,
a background of ordinary descriptions of standard objects. It is
against this background that an identity judgment about mat-
ter is assessed. Now let S and T be sets which contain all of the
true statements that could be formulated in ordinary terms about,
respectively, standard objects and persisting matter; and let S'
and T' be the persistence-free reformulations of these statements.
It may seem clear that the composite view S-plus-T is simpler arid
more coherent than S-plus-T'. It is less clear that S-plus-T is
simpler and more coherent than S'-plus-T'. It is not clear, in other
words, that a thoroughgoing and consistent repudiation of per-
sistence would constitute a theoretical loss.
Let us, however, suppose for the sake of argument that a per-
sistence-free description of reality would be less simple and less
coherent than a description in terms of persistence. It would
seem to follow that there is a theoretical gain in describing real-
ity in terms of persistence. But this would have to be regarded
as a mere gain of elegance, rather than a gain of truth. Of two
logically incompatible hypotheses it seems correct to say that,
other things being equal, the one that is simpler and more co-
herent is more likely to be true. But where two propositions are
logically equivalent, so that the truth of either entails the truth
of the other, it could make no sense to say that one is more likely
to be true than the other. Since, as I am assuming, statements
about successions of genidentically related momentary things are
logically equivalent to (are merely reformulations of) statements
l6o THE CONCEPT OF IDENTITY

about persisting matter, it could make no sense to say that the

latter are more likely because they are simpler.9
A lot more needs to be said about these points, much of which
baffles me. The conclusion which this discussion suggests, how-
ever, is that the persistence of matter, despite its "ultimacy," can
be legitimately redescribed in persistence-free terms.
(Let me state briefly, without argument, how I would want to
connect these ideas to the persistence of persons. Here too I would
urge that we distinguish between the question (a) "Can we
coherently regard the persistence of a person as consisting in the
occurrence of a succession of momentary person-stages that stand
in some distinctive unity-making relationship?" and the ques-
tion (b) "Can we give an account in relatively observational terms
of what that unity-making relationship is?" I am inclined to
think that we can give such an account, so that both (a) and (b)
should be answered affirmatively.10 But even if it is the case, as
many philosophers have held, that our concept of the persistence
of a person cannot be analyzed in terms of identity criteria, so
that (b) must be answered negatively, I would still say that (a)
should be answered affirmatively. For it would still be legitimate
for us to regard a person as made up of person-stages that stand
in some distinctive relationship. We can, if we want, give this
relationship a name, say "person kinship."11 This way of think-
ing about the identity of a person remains legitimate even if we
have to construe the unity-making relationship of person kin-
ship as in some sense simple and unobservable.)
The "relativistic" attitude toward persistence which I have
been expounding in these sections runs counter to an important

9. Compare with Reichcnbach's distinction between "simplicity as a criterion

of truth" and mere "descriptive simplicity," in The Philosophy of Space and
Time, pp. 34-35. A similar idea is expressed in Rudolf Carnap, Philosophical
Foundation of Physics (Basic Books, Inc., N.Y., 1966), pp. 83-85.
10. The kind of account that I have in mind is suggested in Derek Parfit,
"Personal Identity," Philosophical Review, 80 (1971), 3-27. For a discussion
of the issue of personal identity, see below, Chapter 10. (The argument in this
paragraph is elaborated in Chapter 10, Section II.)
11. Compare with Quine's use of "river kinship" and "water kinship" in
From a Logical Point of View (Harvard University Press, Cambridge, Mass.,
1961), p. 66. My view would be that river kinship is observable in a way that
water kinship is not.
 THE METAPHYSICS OF PERSISTENCE l6l

tradition. Many philosophers, notably Butler, Reid, and more

recently Chisholm, have held the view that the persistence of
ordinary objects like tables and trees is in some sense "fictitious"
(or "imperfect" or "loose"), as compared to the "real" (or "per-
fect" or "strict") persistence of other entities, such as, perhaps,
persons or bits of matter.12 Shoemaker has illuminated this view
in his book Self-Knowledge and Self-Identity. A passage in which
he summarizes his explanation of what underlies the view is
worth quoting at length:
There is a common inclination to say that if the persistence of a [par-
ticular kind of object] through time can be regarded as consisting simply
in the occurrence of a temporal succession of momentary states or events
that are empirically related to one another in certain ways (by resem-
blance, spatiotemporal contiguity, and so on), or in the successive exist-
ence of momentary things . . . , then what is called the persistence of
[that kind of object] is not really the persistence of anything. Where
persistence can be so regarded, one is inclined to say, the unity attrib-
uted to those sequences that are regarded as histories of persisting things
does not derive from anything intrinsic to the sequences themselves, but
is somehow imposed on them by conventions of language. In such cases
it seems to be only our need to have economical ways of talking about
the world that leads us to describe any sequences at all as histories of
persisting things, and only our practical or theoretical interest in certain
kinds of sequences that leads us to single out these, and not sequences
of other kinds, to be described in this way. Now if someone thinks that
the persistence of some things is of this sort, and that the persistence of
other things is not, then it will be natural for him to mark this distinc-
tion by saying that only the latter is real persistence involving real
identity.13
Up to a point I am quite sympathetic to the Butler-Reid-
Chisholm doctrine, as Shoemaker here explains it, since I would
also want to emphasize a distinction, of roughly the sort sug-
gested by that doctrine, between two kinds of persistence, be-
tween the criterially determined persistence of standard objects
and the "ultimate" persistence of matter. As I see this distinc-
ia. Joseph Butler, "Of Personal Identity," in The Whole Works of Joseph
Butler, LL.D. (Thomas Tegg, London, 1836), pp. 263-70; Thomas Reid, "Of
the Nature and Origin of Our Notion of Personal Identity," in Essays on the
Intellectual Powers of Man, essay III, ch. Ill, sec. II; and Chisholm in all the
previously cited works.
13. Shoemaker, Self-Knowledge and Self-Identity, pp. 37-38.
l62 THE CONCEPT OF IDENTITY

tion, however, it is a potentially serious distortion to express

it by saying that the persistence of a standard object, as com-
pared to that of a portion of matter, is somehow less than "real"
(or that a standard object persists in only a "loose" sense, whereas
a portion of matter persists in a "strict" sense).
This seems to me a distortion, summarily, for two reasons, the
second of which is somewhat clearer than the first. The first rea-
son is that even where the persistence of an entity, such as a bit
of matter (or, on some views, a person), is not governed by
observational identity criteria, we can still, or so I am inclined
to think, legitimately regard that persistence as consisting in
the occurrence of a succession of momentary things that are re-
lated in some unobservable way. As I see it, there is not, and
cannot conceivably be, any "intrinsic" unity through time, in
the sense of a unity that cannot be adequately redescribed in
different terms. Nor can there be any relationships that are
"intrinsically" unity-making, in the sense of a relationship that
cannot be coherently separated from its unity-making role in our
identity scheme. Hence the persistence of any conceivable entity
can be construed as a "mere succession," if we want to look at it
that way.
But, second, even if it were correct to say that the persistence
of a standard object can be redescribed in terms of the idea of
a succession, in a way that the persistence of a bit of matter (or
a person) cannot be so redescribed, it would still be unreason-
able to conclude that the former kind of persistence is "unreal"
(or "loose"). That the persistence of an object can be redescribed
in different terms does not discredit the descriptions that we
actually employ. There seems no question that our ordinary de-
scriptions of standard objects provide one essential paradigm of
our use of the ordinary concept of persistence. The persistence
of a standard object is therefore properly regarded as an espe-
cially clear and obvious kind of persistence; here we have as
"real" a kind of persistence as we know of.

V. Can We Justify Our Identity Scheme?

Though I have professed myself to be a "relativist" about per-
sistence I am not, in at least one important sense of the word, a
"conventionalist." As a relativist I hold that our identity scheme
 THE METAPHYSICS OF PERSISTENCE l6g

is not the only one that could in principle be employed in mak-

ing true statements about the world. But, as I stated earlier, I
am inclined toward the empirical speculation that our ordinary
identity scheme, or at least the basic core of that scheme, is in-
stinctive to human beings. My conjecture would be that, as a
matter of contingent fact, each of us enters the world innately
disposed in some manner to interpret experience in terms of our
basic idea of persistence, in terms, that is, of the idea of persisting
objects whose careers unfold along continuous change-minimiz-
ing paths. So I would not suppose that our identity scheme is a
"mere convention," in the sense of something that we could
easily decide to alter. In fact I doubt whether we could in any
way get ourselves, or our descendants, to perceive the world in
terms of an identity scheme radically different from the one pres-
ently employed by a speaker of English.
I shall not at present attempt to develop these rather vague
speculations about our "instincts" and "innate dispositions."14
Instead I want to raise certain questions about a common as-
sumption among philosophers which may seem to render such
speculations philosophically superfluous. I think that it is often
taken for granted that, at least from a philosopher's point of
view, the interesting explanation of why we operate with our
identity scheme, rather than with another one, is that it is reason-
able for us to operate with the scheme that we have. Sometimes
this position will take the bald form of the claim that no other
identity scheme could allow us to make true statements about the
world. Against this I have already argued. But I think that
even many philosophers who would allow that reality could in
principle be described in terms of some other identity scheme,
often take it for granted that the philosophically relevant ex-
planation of why we operate with this particular identity scheme
is that, given our human needs and purposes, this scheme is the
reasonable one for us to have.
I want to question whether anything of this sort is correct. I
will suggest that there is reason to doubt that there is any im-
portant sense in which our identity scheme can be said to be
especially right, or reasonable, or practical, or convenient, or
anything of that sort. This is not to suggest that our identity

14. This issue will be addressed in Chapter 8.

164 THE CONCEPT OF IDENTITY

scheme may be especially wrong, or unreasonable, or impractical,

or inconvenient. Rather I doubt that any such assessments really
make much sense, or that they have any fundamental bearing on
explaining why we think about persistence the way that we do.
The kind of assumption that I want to call into question
seemed implicit in one of the remarks quoted previously from
Shoemaker. Our reason for treating certain successions as per-
sisting objects, Shoemaker suggests, is "our need to have economi-
cal ways of talking about the world," and "our practical and
theoretical interest" in certain kinds of successions rather than
others.15
Let us consider the principle
(i) We tend to treat a kind of succession as corresponding to
the career of a persisting object when that kind of succes-
sion is especially important or interesting to us.
It is not immediately clear what it means to say that a kind of
succession is important to us. This may mean that it is often
important to us whether that kind of succession occurs. Or per-
haps it means that, given the occurrence of that kind of succes-
sion, it would often be important to us what the properties and
interrelations are of the items that comprise the succession. But
in whatever way we make this out I think that if we consider
the principle carefully we can see that it really has no plausi-
bility at all. More often than not we are especially interested in
facts about the successive stages of different objects that may
be related in various ways. More often than not, therefore, we
are especially interested in successions that do not correspond
to unitary careers.
Consider, for example, all of the following sorts of successions:
(a) a succession which consists in the stages of every car that I
have every owned during the periods when I owned them; (b)
a succession which consists in a stage of the car x when x leaves a
particular parking space, followed by a stage of the car y when y
enters that space immediately afterwards; (c) a succession which
consists in a stage of an object x immediately prior to x's impact
with the object y, followed by a stage of y immediately after the
impact; (d) a succession which consists in the stage of a tree while

15. Shoemaker, Self-Knowledge and Self-Identity, pp. 37^38.

 THE METAPHYSICS OF PERSISTENCE 165

it is blooming, followed by the stage o£ an apple that grows and

falls from the tree.
I think it is unnecessary to multiply these examples, or to make
them out in any detail, in order to see that there is really no
evident connection between the idea of a succession that is im-
portant to us and the idea of a succession that corresponds to a
persisting object. There is no apparent truth to the principle that
we generally treat important kinds of space-time portions of real-
ity as unitary objects.
I think this principle may have a specious plausibility because
we can easily confuse it with a different and much more limited
one. Suppose we take the basic rule of identity as given, and
suppose we take it for granted, furthermore, that this rule is to
be refined in terms of the introduction of a list of sortal con-
cepts under which the careers of objects are traced. The question
then arises why we introduce just the sortal concepts that we
have. Part of the answer to this question, I have already sug-
gested, lies in the idea that the sortal refinement must not issue
in an overly drastic departure from the basic conception. Hence
our sortals must be nondispersive, they must allow us to trace
paths that minimize nonlocational change, and they must not
force us to terminate careers in ways that conflict too often with
the basic rule. But even granted these constraints imposed upon
our sortal introductions by the basic rule there would still be in
principle a great deal of leeway, and the question persists why
we have just the sortals that we have. Now to this question it
may be quite plausible to suggest
(a) We tend to treat a property as a sortal when (besides satisfy-
ing the constraints imposed by the basic rule) that property
is especially interesting or important to us.
In other words we tend to resolve change-minimizing conflict
by tracing paths in such a way as to stabilize important prop-
erties. I am prepared to believe that some such principle as
this is acceptable, though I find myself unable to make even
this much out with any clarity. But this limited principle, which
already takes for granted the general structure of our identity
scheme, is not at issue. A philosopher who holds that our iden-
tity scheme is to be justified by reference to our human needs
and interests must be able to show that the general structure
l66 THE CONCEPT OF IDENTITY

of that scheme, the fact that it includes the basic rule and the
sortal rule, can be explained in these terms. This he cannot show
by appealing to any such limited principle as (2). Nor can he
show it by appealing to the false principle (i) that we generally
treat important successions as persisting objects.
Now actually it is not clear that this discredited principle (i)
is the one that is crucial to the case which this philosopher would
want to make. We need to distinguish between saying
(i) We tend to treat a kind of succession as corresponding to
the career of a persisting object when that kind of succession
is especially important or interesting to us
and saying
(3) We tend to treat a kind of succession as corresponding to
the career of a persisting object when it is especially im-
portant (or convenient or useful) for us to treat the success-
sion in that way.
Though it seems tempting to equate (i) with (3) there is really
no obvious connection between these principles. There is no
obvious reason to suppose that it would be especially important
for us to treat a succession as a persisting object when (or only
when) the succession is especially important to us. Though (i)
seems quite definitely unacceptable it might still be maintained
that (3) is correct. And it is surely (3) that is most directly rele-
vant to the claim that, given our needs and interests, our iden-
tity scheme is the reasonable one for us to have.
Is it, then, important (or convenient or useful) for us to think
about persistence in the way we do, to treat as persisting objects
just those successions that we do so treat? Here we should per-
haps distinguish between theoretical importance and practical
importance.
The discussion in the last section suggested that there are
difficulties in maintaining even that our concept of persisting
matter affords a gain to theory. What can be said with reasonable
confidence is that given our general commitment to describing
the world in terms of persistence, our concept of persisting mat-
ter attempts to realize that commitment in the theoretically
optimal fashion. It seemed unclear, however, whether we could
say unqualifiedly, and without already presupposing our com-
 THE METAPHYSICS OF PERSISTENCE 167

mitment to persistence (and, specifically, our commitment to the

ordinary persistence of standard objects), that our concept of
persisting matter is theoretically advantageous.
Be this as it may, it seems quite definitely impossible to main-
tain that our ordinary concept of the persistence of standard
objects is especially advantageous to theory. There seems to be
no special theoretical point in our reidentifying cars, and tables,
and trees in the way that we do rather than any number of other
possible ways. We say that the same car persists with different
parts when parts are added and subtracted, though the physicist's
description of the world may mention no entities that persist
with new parts. Here, with respect to the identity of a standard
object, scientific-theoretic considerations are apparently not rele-
vant as arbiters of the truth. Our ordinary description of the
persistence of a car is, on my view, a literally (and "strictly")
true description of the observable facts. But there appears to be
no special theoretical gain in our describing the facts the way
that we do rather than in some other way (e.g., by requiring of
an object's identity that no parts are added or subtracted). In-
deed from the standpoint of scientific theory the most that could
safely be said about our ordinary identity scheme is that it might
possibly have been worse.
This does not, as far as I am concerned, show that there is any-
thing especially bad about our identity scheme, but only that
there is nothing especially good about it from the point of view
of our scientific-theoretic needs. Those needs are apparently not
to be regarded as the primary determinants of our thought about
persistence. We might perhaps try to imagine a purely intellec-
tual creature, given our sensory intake, who confronts his ex-
perience in terms of the exclusive disposition to seek unity-mak-
ing principles which would yield the simplest laws of nature.
Such a propensity seems quite alien to our own overall perspec-
tive. It is unclear whether the exclusively theorizing creature
could have any basis for adopting even a concept of persisting
matter. It seems certain that he could have no basis for adopting
our scheme of standard objects.
So at least the major brunt of our ordinary identity scheme, the
part of it that deals with standard objects, cannot apparently be
justified in terms of theoretical purposes. Can we perhaps say,
however, that there is some decisive practical gain in the way
l68 THE CONCEPT OF IDENTITY

that we think about the persistence of a standard object? A

point which is often made, and which may be suggested by
Shoemaker's remarks about "our need to have economical ways
of talking about the world," is that if we tried to describe the
world in terms of a different identity scheme, we would have to
use many more words than we now do to convey the kinds of
information that we typically need to convey.
Now there really seems to be no very close connection between
the number of words that we need to describe a given situation
and the kind of identity scheme that is being employed in the
description. Suppose that the in-out language contained the rela-
tional term "ancestor" which functioned in such a way that the
statement "x is an ancestor of y" is true if and only if either x is
identical with y, or some finite sequence of in-out transforma-
tions leads from x to y. Hence, the statement "The incar x is an
ancestor of the outcar y," a case in which x and y are assumed to
be nonidentical, would in effect assert that either the outcar y
came out of the incar x, or y came out of an incar which came
out of an outcar which came out of x, or . . . , etc. Would this con-
struction not perhaps allow speakers of the in-out language to say
many of the things that we typically say about cars in a reason-
ably limited number of words? Where we say of a car, for
example, "It broke down five times last year," they could perhaps
say, "Its ancestors broke down five times last year" (where, as
follows from the previous definition, any incar or outcar is de-
generately an ancestor of itself). Even here the in-out sentence
is longer by one word than its English counterpart, and certainly
more difficult cases would have to be considered. But there is al-
ways the possibility of introducing additional abbreviations into
the in-out language which would not apparently affect the es-
sential character of their identity scheme. (Eor example, they
might have one word for "broke down.") Given the fairly un-
limited prospects for such abbreviations it seems entirely un-
clear that there could not be an in-out scheme which might allow
for roughly as "economical" a way of talking as our ordinary
way.
Again, suppose that Quine's space-time language contained
such expressions as "longest car-continuum" to denote a contin-
uous succession of momentary car-stages that is not a segment of
any longer continuous succession of momentary car-stages (to
 THE METAPHYSICS OF PERSISTENCE 169

denote, that is, roughly what in the ordinary scheme is the whole
career of a persisting car). Given such constructions, in con-
junction with other well-placed abbreviations, it seems that we
might be able to express with a comparable number of words
many of the things that we ordinarily express. It would in fact
be quite difficult to understand how the space-time language
could necessitate our using more words than we now use. That
language, with its wholly permissive identity scheme, would al-
low us to single out any space-time portion of reality as a unit,
including of course the select few space-time portions which our
ordinary identity scheme accords the special status of unity. The
space-time language might in principle contain any number of
abbreviated, or even syntactically simple, expressions which de-
note various space-time portions, including, among others, the
ones that we ordinarily talk about. The permissibility of singling
out units in addition to the ones that we ordinarily single out,
would seem to promise, if anything, the prospect for a possible
gain of brevity, not a loss.
In short, I am unable to see any relatively clear connection
between our identity scheme and our supposed need for brevity.
But actually there is, to begin with, something faintly embar-
rassing in philosophers' continually seeming to tell us that the
underlying rationale of our language is to enable us to talk less.
The fact is of course that people like to talk, apparently as much
as possible, and a breath saved here and there is simply not
credited by us as enhancing the felicity of our condition. There
may be, I suppose, an outer limit to human garrulity. If it re-
quired three hours of steadfast oration to convey that a car
broke down this would probably not be good. But there seems
no reason to suppose, or to hope, that our language is, or ever
will be, even close to the realization of some philosophical ideal
of maximally condensed talk. From the standpoint of taciturnity
the most that could safely be said about our ordinary identity
scheme is that it might possibly have been worse.
Word-count aside, can we conceive of any other practical gain
that we might derive from talking about persistence in the way
we do rather than some other way? Suppose that there were
people who talked about (and thought about and experienced)
the world in terms of some generalized in-out language, a lan-
guage in which many (though perhaps not all) objects stood
170 THE CONCEPT OF IDENTITY

to our ordinary objects in rather the way that incars and out-
cars stand to cars. Would it follow from the fact that they
talked this way that their lives would be essentially different
from ours? Would it follow that their actions and attitudes would
be different from ours?
In one sense this would have to be so. In a sense, someone who
spoke that language could not intend (or wish or fear) that a car
would leave a garage, and perhaps could not therefore be said
(intentionally) to bring it about that a car would leave a garage.
But of course he might have the equivalent intention (or wish
or fear) that an incar would go out of existence and be replaced
by an outcar. The relevant question is whether this difference
in the way that he thinks about things would have any practical
implications. Would his actions and attitudes be different from
ours in the sense that when they are, so to speak, translated into
English they fail to match our own actions and attitudes?
I suggest that there is no a priori answer to this question. It
is a priori possible that the lives of the in-out speakers would
be in the relevant sense exactly like ours. This would mean
roughly that if there is a situation S such that a typical speaker
of our language who found himself in S would entertain an in-
tention (wish, fear) which he could express, for example, in the
words "I intend (wish, fear) that the car will leave the garage,"
then a typical speaker of the in-out language who found himself
in S would entertain an intention (wish, fear) which he could
express in the words "I intend (wish, fear) that the incar will go
out of existence and be replaced by an outcar." It would mean
that when the rules which define their practices and institutions
are translated into English what comes out are the rules which
define our practices and institutions. (So it would perhaps be a
law of their land that if you own an incar or outcar then, unless
you sell it, you automatically own its immediate "descendant.")
It is a priori possible that their lives would be just like ours.
But it is also possible that their lives would be very different
from ours. Perhaps the rapid fluctuations of the identities of ob-
jects, as they experience this, would express itself in some distinc-
tive way in their attitudes and behavior. (Maybe they would be
saintly nomads who flit from place to place without possessions
or property rights, or maybe crazed hoarders compulsively staked
out against the ins and outs of their fate.) Perhaps all of their
 THE METAPHYSICS OF PERSISTENCE 171

practices and institutions would seem radically alien to us. There

is no (a priori) way of saying. Nor less is there any way of saying
that, if there were such alien practices and institutions, they
would be less practical, convenient, or useful than ours.
(It is possible, though problematical, that these remarks would
remain valid even if the discussion could be extended to include
personal identity. Perhaps we can imagine an in-out language
which affects the concept of a person and correlative personal
pronouns. Someone might be called an inperson or an outperson,
and be reidentified accordingly, depending upon whether he or
she is, let us say, inside or outside the village. It may still be
conceivable that their actions and attitudes would be in the
relevant sense just like ours. This would mean that in a situation
where one of us would announce with equanimity "I am about
to be taken out of the village," one of them would announce with
equanimity "I am about to cease to exist and be replaced by an
outperson."16
But it may also be conceivable that their actions and attitudes
would be in the relevant sense very different from ours, in that
they might be, in terms of their concept of personal identity,
as predominantly self-regarding as we are in terms of our con-
cept. This would mean that one of them who is about to be car-
ried out of the village would judge with horror "I am about to
cease to exist (and be replaced by an outperson)," and attach to
those words all of the dismay, and resistant behavior, that we
ordinarily attach to the prospect of personal extinction.17
It may well be that these fantasies about inpersons and out-
16. That is, " . . . and be replaced by an outperson who will remember every-
thing about me." For the kinds of revisions in the concepts of memory, inten-
tion, and related notions that would be required in the inperson-outperson
language, see Parfit, "Personal Identity," and Sydney Shoemaker, "Persons
and Their Pasts," American Philosophical Quarterly, 7 (1970), 269-85.
17. The connection between our concept of personal identity and our sense
of self-interest is explored in Chapter 10. This issue is discussed in Parfit,
"Personal Identity"; in Bernard Williams, "The Self and the Future," Philo-
sophical Review, 79 (1970), 161-80, reprinted in Bernard Williams, Problems
of the Self (Cambridge University Press, London, 1973), pp. 46-63; and in
Shoemaker's comments in "The Loose and Popular and the Strict and Philo-
sophical Senses of Identity." (See also Shoemaker's remarks there about the
connection between our concept of physical identity and our interests in ob-
jects.)
172 THE CONCEPT OF IDENTITY

persons are in some fundamental way incomprehensible. Insofar

as the fantasies are entertainable, however, and entertainable
in the two forms mentioned, they suggest that it may not even be
possible to draw any evident connections between our concept of
personal identity and our actions and attitudes. But this point
is much clearer with respect to standard objects, other than per-
sons, which are the only cases that I am seriously treating.)
My conclusion is that there is no evident sense in which it is
especially practical (or useful or convenient) for us to think about
the persistence of standard objects in the way we do rather than
some other way. There is no evident benefit or gain which our
identity scheme seems peculiarly adept at securing for us. Insofar
as our concepts may be said to enter into our intentions and
attitudes there is a trivial sense in which we could not do or feel
any of the things that we ordinarily do or feel except by having
our ordinary concepts. In a trivial sense, therefore, we need our
ordinary concepts to do and feel all of the ordinary things that
we like to do and feel. So we can say that thinking (and acting
and feeling) in the ordinary way serves our need to think (and
act and feel) in the ordinary way. This near-tautology obviously
does not explain, or justify, why we (need to) think in the
ordinary way.
Looked at in one way, then, the attempt to justify our iden-
tity scheme by reference to our practical needs is futile because
those needs, as conceptualized by us, already presuppose the
identity scheme. (For example, in order for someone to think "I
need to move the car out of the garage," he must be operating
with the ordinary concept of the persistence of a car.) On the
other hand, to the extent that it makes sense to distinguish be-
tween our needs as such and our ways of conceptualizing them,
it is no longer clear how our identity scheme is especially adept
at serving these needs.
The impulse to assume that ours is (for us) the best of all pos-
sible identity schemes sometimes takes an evolutionary turn.
Quine, who holds (as I do) that our identity scheme is "rooted
in instinct,"18 drops the casual remark that "man and other
animals are body-minded by natural selection; for body-minded-
18. Quine, The Roots of Reference, p. 55.
 THE METAPHYSICS OF PERSISTENCE 173
19
ness has evident survival value in town and jungle." This is
precisely what is not evident to me. To be "body-minded," in
Quine's sense, is to experience objects in terms of our ordinary
unity-making principles. A creature who experienced the world
in terms of the in-out language (or in terms of Quine's space-time
language) would not be body-minded. I would want to see how
Quine could make out that such a creature would be less fit to
survive than we. I doubt that this can be made out. We can say
that in the struggle to survive we discover ourselves to be two-
eyed body-minded survivors, and that being two-eyed and body-
minded is, other things remaining equal, evidently better for
survival than various other possibilities (for example, being to-
tally blind or totally unconscious). As to whether we (or crea-
tures otherwise like us) would have necessarily been in any sense
worse off with some different number of eyes, or some different
identity scheme, I think no one can say.
My aim in this section has been to stress, perhaps at some risk
of exaggeration, the seemingly implacable primitiveness (non-
derivativeness) of our ordinary commitment to our identity
scheme. At the level of common experience we think about per-
sistence in the only way that we know how, in the only way
that makes sense to us, perhaps in the only way that is psycho-
logically possible for us. This, I think, is essentially the only
"justification" that we can give for thinking the way that we do.

19. Ibid., p. 54.

This page intentionally left blank
 PART TWO

MINDS AND BODIES

This page intentionally left blank
 Introduction to Part Two

THIS PART attempts, first, to clarify and defend some of the

views already presented, and, second, to open up some addi-
tional questions about the nature of identity. Each chapter is
designed to be an essentially self-complete essay, though there
is inevitably a considerable amount of cross-reference.
The first two chapters deal directly with several objections to the
previous views. Chapter 6 addresses in detail the crucial objec-
tion that an analysis of our concept of bodily identity is neces-
sarily circular because the concept is more fundamental, both
metaphysically and epistemologically, than any concepts in terms
of which the analysis might be couched. In Chapter 7 I take
up Shoemaker's suggestion that causal connectedness is necessary
for identity, and the radical suggestion that there are no logically
sufficient criteria of identity. Here I also show how Putnam's no-
tion of a stereotype might be applied to an analysis of identity.
Several times in Part One, I expressed the conjecture that our
concept of bodily identity is innately determined. This is elabo-
rated and defended in Chapter 8. The question of innateness
leads to a consideration of the essential connection between the
issue of "unity" and the issue of "similarity," a connection which
is pursued in Chapter 9. The notion of a natural kind is promi-
nent in recent literature, and in the latter chapter I explore vari-
ous points of connection between that notion and what I call a
"natural unit."
The topic of personal identity, which I studiously avoided in
Part One, is now addressed in Chapter 10. This chapter extends
177
178 THE CONCEPT OF IDENTITY

various issues of relativism, conventionalism, and innatism into

the realm of personal identity. In this discussion I adopt the
rather extreme device of assessing at length an utterly alien con-
ception of personal identity; the device will not, I hope, overly
tax the reader's indulgence for the philosophy of the weird.
Running through a number of these discussions is my pre-
occupation with the issue already broached in Chapter 5 of what
the status is of our ordinary identity concept. There I considered
the claim that:
There are compelling reasons for us to describe the world in
terms of ordinary objects.
And now in Chapter 6 I go on to consider the claim that:
Ordinary objects are basic,
and in Chapter 9 the claim that:
Ordinary objects are natural units.
These are three ways of conferring upon ordinary objects a
special and exalted status.
The three status claims are evidently not unrelated, but they
do carry rather different philosophical associations. The first claim
suggests that we could cite some ordinary reasons in support of
our identity concept, theoretical or practical reasons akin to
those—e.g., of probability or efficiency—which we ordinarily
give in support of a belief or practice. I have already criticized
this position, and will argue against it again in somewhat dif-
ferent contexts in Chapters 8 and 10.
The claim that ordinary objects are "basic" can be taken in
two senses. From a metaphysical standpoint the claim suggests
that the ordinary concept of an object cannot be analyzed or de-
fined in terms which do not already presuppose the concept.
From an epistemological standpoint the claim suggests that our
knowledge of the world depends on our knowledge of ordinary
objects. The metaphysical claim does not seem to me convincing;
some of the central issues here have partially emerged in Chap-
ter 5, and will be clarified and developed in Chapter 6. As re-
gards the epistemological claim it is necessary to distinguish be-
tween two questions. We can compare the status of an ordinary
object to the status of the momentary stages of an object; or we
 INTRODUCTION TO PART TWO 179

can compare the status of an ordinary object to the status of other

successions of stages, successions, that is, which do not correspond
to what we ordinarily conceive of as persisting objects. Whereas
I will suggest in Chapter 6 that the stages may be epistemologi-
cally more basic than the objects, I would certainly hold that
the objects are epistemologically more basic than other succes-
sions; obviously this is so if, as I think, our minds are innately
determined to synthesize the stages into ordinary objects. Note
that from the metaphysical standpoint this distinction between
the two comparisons seems inconsequential: ordinary objects, it
seems, are more basic metaphysically than the other successions
if and only if they are more basic metaphysically than the stages.
What is suggested by the claim that ordinary objects are "nat-
ural units"? When philosophers talk about "natural kinds" they
seem to imply that there is an objective distinction, apart from
our human attitudes and practices, between kinds and artificial
constructions. The analogous claim with respect to objects is
that there is an objective distinction between ordinary per-
sisting objects and other successions of stages, a distinction that
can be drawn without reference to our attitudes or practices.
But that claim seems almost trivially correct; and certainly it
does not confer any special status on the objects. That ordinary
objects are objectively distinguishable from the other succes-
sions surely does not exalt the objects above the other succes-
sions. The objects are, I think, exalted and "natural" only in the
psychological sense that it is natural for us to conceive of the
world in terms of such objects. If it is held that there is an ob-
jective distinction between natural kinds and artificial classes—
a position which (as I explain in Chapter 9) is denied by certain
nominalists—then it may perhaps also seem plausible to regard
the natural kinds as metaphysically basic, as presupposed in any
adequate conception of the world. But this connection between
"naturalness" and "metaphysical basicness" is, I shall maintain,
not plausible for "natural units." Though there is an objective
enough distinction between the natural units—i.e., the ordinary
objects—and other successions, the special status of the ordinary
objects seems to be essentially subjective, essentially a function
of how we think.
And it is not just physical things that seem to lack an objec-
tively or metaphysically exalted status but persons too; or so I
l8o THE CONCEPT OF IDENTITY

argue in Chapter 10. But here especially the psychological cen-

trality of our concept of identity seems strikingly evident. One
aspect of my view, then, is that our concept of identity, in its
application both to bodies and to persons, suffers from a certain
kind of metaphysical arbitrariness. That theme has already been
sounded at the end of Part One, and will be amplified in this
part. But in the ensuing chapters I want also to lay stress on the
correlative point that our concept of identity is psychologically
not arbitrary at all; there are probably deep psychological con-
straints which determine that just this concept should structure
our understanding and knowledge. If our identity concept dis-
appoints us as metaphysicians, it may yet fulfill our expectations
as philosophical psychologists and epistemologists.
 6
Foundations of Identity

IN THE recent literature a number of philosophers, including

myself, have attempted to analyze our concept of bodily iden-
tity in terms of the interrelations between the successive momen-
tary stages of a body. This kind of analysis implies that there
is a conceptual connection between bodily identity and various
conditions which might be satisfied by a succession of body-
stages. These conditions have often been called our "criteria"
of bodily identity.
These criteria of identity have typically been regarded as
essentially comprising two kinds of elements: an element of
continuity, and an element of sortal-coverage. The idea is that
there is a special class of "sortal" terms, which can perhaps be
specified by a list, such that if F belongs to this class, then the
identity through time of an F-thing can be (more or less) equated
with the continuity of a succession of F-stages. The kinds of
continuities that have generally been most stressed are spatio-
temporal and qualitative, though other continuities have some-
times been mentioned, typically in a derivative role. As regards
the condition of sortal-coverage, in my own work this has been
seen as based upon a more rudimentary condition of "minimi/-
ing change."1 I shall ignore these various complications in the
present chapter; but everything I say here about "sortal-
covered continuity" could be said just as well about "change-
minimizing continuity," and might apply regardless of what pre-
cise form the continuities take.
i. See above, Chapter 3.

181
l82 THE CONCEPT OF IDENTITY

There are two general kinds of questions that might be raised

about this analysis of bodily identity. It might be questioned,
obviously, whether there is the required conceptual (or semantic
or a priori) connection between bodily identity and sortal-covered
continuity. But even if this connection is granted it might still
be questioned whether we have what can properly be called an
analysis or criteria of bodily identity. For it might be insisted
that the analysis or criteria of a concept must be more basic
than the concept itself. And what can be more basic than our
concept of bodily identity? Bodies, it might be said, are more
basic than body-stages, and bodily identity is more basic than
any relationship between body-stages. Hence the purported anal-
ysis moves in the wrong direction: It moves from the more basic
to the less basic.
This latter question about analytical priorities is the only issue
I want to discuss in the present chapter. For the purposes of this
discussion I will therefore simply take it for granted that there
is in fact the required conceptual connection between bodily
identity and sortal-covered continuity. Since I suspect that doubts
about this connection often conceal concerns about the priorities,
a discussion of the latter issue may interest even those who enter-
tain such doubts.

I. Metaphysical Priorities
and Epistemological Priorities
I want, first of all, to draw a tentative distinction between two
kinds of philosophical issues concerning "basicness": There are
issues of epistemological basicness and issues of metaphysical
basicness. A kind of thing is epistemologically basic if our
knowledge of it belongs to the "foundations of knowledge"; or,
more generally, one kind of thing is epistemologically more
basic than a second if our knowledge about the second kind of
thing derives from our knowledge about the first kind of thing.
What is epistemologically basic depends on our sense organs
and other aspects of our human situation. But the issue of meta-
physical basicness is supposed to depend not at all on our human
situation. Something is metaphysically basic if it "ultimately
exists," if it belongs to the "ultimate structure of the world," if
it figures in the "ideal description of reality." I think it is im-
 FOUNDATIONS OF IDENTITY l8g

mediately apparent that the metaphysical notion is more proble-

matical than the epistemological one; and it is equally clear that
the metaphysical notion looms very large in the history of
philosophy.
That there is at least a prima facie distinction between these
two notions can be brought out by considering our attitude
toward the particles of physics. Certainly these particles are
epistemologically nonbasic, for whatever we know about them
we derive from our knowledge of ordinary bodies. Yet we may
be strongly inclined to regard these particles as metaphysically
basic, as belonging to the most ultimate level of reality.
It may not be easy to give an equally convincing example of
the reverse situation, something which strikes us as epistemologi-
cally basic but metaphysically nonbasic. The sort of example to
look for is one in which a complex "gestalt" seems to present
itself directly to our experience, though we are still inclined to
regard it as metaphysically derivative of its constituents. Of
course ordinary bodies are arguably just such examples, but
this is precisely the controversial issue we are about to consider.
A possibly less controversial example would be a song, which
we may want to regard metaphysically as merely a construction
of certain kinds of notes in certain kinds of relations, though
our recognition of the song seems quite direct.
There are of course many examples which may strike us as
both metaphysically and epistemologically nonbasic. Many typi-
cal processes, such as an economic depression, are likely to strike
us in this way.
As regards our analysis of bodily identity I think it is im-
portant to consider the issues of both metaphysical and episte-
mological basicness. Perhaps the first issue relates more to the
word "analysis," and the second to the word "criteria." Certainly
the word "analysis," in the traditional use of a philosopher like
Russell, suggested a movement toward metaphysical ultimates.
The word "criteria" perhaps carries no such suggestion. In its or-
dinary nonphilosophical use there is not even the suggestion that
criteria must be conceptually connected to that of which they
are criteria. In current philosophical usage, however, "criteria"
(as opposed to "evidence" or "symptoms") conveys the idea of
conceptual connectedness, while retaining the ordinary associa-
tion with the epistemic basis for a judgment. Criteria of bodily
184 THE CONCEPT OF IDENTITY

identity would, then, have to be conditions which are con-

ceptually related to identity, and upon which we can base our
judgments of identity.
It might therefore be said, perhaps, that a question about the
metaphysical basicness of bodily identity challenges the possi-
bility of analyzing bodily identity, whereas a question about the
epistemological basicness of bodily identity challenges the possi-
bility of stating criteria for bodily identity. I will first discuss
the metaphysical question, and then later turn to the episte-
mological question.

II. Body-Stages
The proposed analysis of bodily identity in terms of the sortal-
covered continuity of body-stages may give rise to several issues
about metaphysical priorities. I shall address only two of these,
which are perhaps the most prominent: an issue about the
status of body-stages and an issue about the status of spatio-
temporal continuity.2 1 think it has seemed to some philosophers
that the analysis must be misguided in principle, since body-
stages are derivative of bodies and spatiotemporal continuity is
derivative of bodily identity. These philosophers might agree
that there are conceptual connections between bodily identity
and the spatiotemporal continuity of body-stages (that perhaps
bodily identity is essentially equivalent to the sortal-covered
spatiotemporal and qualitative continuity of body-stages), but
they would say that this is because both the notion of a body-
stage and the notion of spatiotemporal continuity are deriva-
tive of the notion of bodily identity, not the other way around.
Let us first consider the question about body-stages. There may
be a metaphysically innocuous construal of the notion of a body-
stage for which an issue of metaphysical priorities need not even
arise. By a body-stage we might simply mean how a body is at a
given lime. The analysis, on this construal, shows us how to
translate ordinary statements about bodily identity in terms of
the relations between the descriptions-at-a-moment of a body.
Where F is a sortal, and t and t' are neighboring times such that

2. Both o£ these issues were discussed briefly, and from a less general stand-
point, in Chapter 5.
 FOUNDATIONS OF IDENTITY 185

the body x exists at t and the body y exists at t', the analysis
tells us that x = y if x is F at t, y is F at t', and the continuity
conditions prevail. On this construal the analysis refers only to
ordinary persisting bodies; no metaphysically dubious "momen-
tary things" are brought into play. It is not that descriptions of
bodies are analyzed in terms of descriptions of certain other
items, but rather that one kind of description of bodies, that
pertaining to persistence through time, is analyzed in terms of
another kind of description of bodies, that pertaining to con-
tinuity and sortal-coverage.
It may seem unclear how the continuity conditions can be
understood in the context of this construal. These conditions
require that a succession of body-stages should be related in
certain ways. Does this not imply that body-stages are treated as
special kinds of things to be related in these ways? But no, it may
be the bodies themselves which are related in these ways. We
might take "Qxtyt'" as signifying a four-termed relationship,
where x and y are bodies which may or may not be identical,
and t and t' are neighboring times. We can read "Qxtyt'" as
"x at t is qualitatively continuous with y at ('." (Compare with
"x at t is bigger than y at t'," where it may be that x = y.)
Similarly we can take "Sxtyf" for "x at t is spatiotemporally
continuous with y at t'." Now the analysis tells us that, where F
is a sortal, x = y if ~Fxt, Fyt', Qxtyt', and Sxtyt'. There is still
no reference here to anything but ordinary bodies (and times).
Of course there may be a major problem in understanding what
these continuity relations amount to, especially the relation of
spatiotemporal continuity, but that is another problem, to which
I will return.
If body-stages in the analysis are supposed to be literally in-
stantaneous, then the innocuous construal might seem to require
the idealization of "neighboring instants." At least this is sug-
gested by my formulation of the analysis in the previous para-
graph. I doubt, however, that this is a decisive problem; the
idealization could probably be dispensed with at the cost of
some complications.
Be this as it may, the fact is that most proponents of the sort
of analysis being considered have not adopted the innocuous
construal. Body-stages are typically construed as momentary
things, distinct from persisting bodies. The analysis is interpreted
l86 THE CONCEPT OF IDENTITY

as showing how the persistence of a body boils down to the inter-

relations between these other kinds of items.
What are these other kinds of items? If we try to stay fairly
close to common sense we might answer that body-stages are
parts of a body's history. Each body has a history, which can be
thought of as comprised of momentary segments or parts; ac-
cording to the present suggestion these are what we are calling
"body-stages."
Given this interpretation of what a body-stage is, it is not
difficult to appreciate the critic's complaint that the analysis
moves in the wrong direction. For it may seem obvious that
bodies are more basic than their histories or the parts of these
histories.
Are there any arguments, any general considerations, which
we can adduce in favor of this judgment? If we ask why the
existence of an economic depression strikes us as less basic than
the existence of a person, the first answer that might come to
mind is that it is logically possible for there to exist persons
without there existing any depressions, but there could not
possibly exist depressions without there existing any persons.
This kind of answer can apparently not help us to establish
that bodies are more basic than their histories. For just as it is
logically impossible for there to be body-histories without
bodies, it is equally impossible for there to be bodies without
histories.
The answer may anyway be of dubious value. Our judgment
about the relative basicness of depressions and persons cannot
depend simply on the fact that "There are depressions" entails
"There are persons," but not vice versa. "There are things which
are red" entails "There are things which are red or yellow,"
but not vice versa. This does not lead us to say that being red
is less basic than being red or yellow.
Perhaps what we really wanted to say about the case of eco-
nomic depressions is this. It seems logically impossible that some-
one should have the concept of an economic depression without
having the concept of a person, whereas it is possible that some-
one should have the latter concept without having the former.
This is perhaps why the existence of an economic depression is
less basic than the existence of a person. This approach may
also give us the intuitively right line on the case of "red" and
 FOUNDATIONS OF IDENTITY 187

"red or yellow." It seems on the face of it impossible that some-

one should have the concept "red or yellow" without having the
concept "red," but not vice versa. And our intuitive judgment
indeed is that being red or yellow is less basic than being red.
How will this approach apply to bodies and body-histories?
It seems evidently impossible for someone to have the concept
of the history of a body without having the concept of a body.
A problem arises, however, in that it may seem equally im-
possible that someone should have the concept of a body without
having the concept of a body-history. Can one conceive of a body
without conceiving of something which has different properties
at different times, something, that is, which has a history? But
perhaps it can be answered that while it is necessary to conceive
of a body as changing in various ways, it is not necessary to
reify this; it is not necessary to conceive of a distinguishable
item called the body's history, which itself has various parts or
stages.3
I am not sure how convincing this answer is. Let us note, how-
ever, that for the purposes at hand it may be quite unnecessary
to press this point. The critic of our analysis need not show
that body-histories are less basic than bodies; it is enough for
him to show that body-histories are not more basic than bodies.
A legitimate piece of analysis, at least by the critic's standards,
must move from the less basic to the more basic. This the analysis
fails to do, if body-stages are not more basic than bodies. And
that they are not is perhaps shown by the fact that to have the
concept of a body-history, or the concept of a stage of such a
history, necessarily depends upon having the concept of a body.
The test for metaphysical priorities which I have been dis-
cussing might be called the concept-dependence test. It amounts
to this:
(i) If it is logically impossible for someone to have the con-
cept of F without having the concept of G, then the concept
of F is not more basic than the concept of G;
(a) If in addition to the condition stated in (i) it is logically
possible for someone to have the concept of G without hav-
3. Compare with P. F. Straw-son's suggestion that our concept of an animal
may depend on our concept of being born, but not on our concept of partic-
ular births, in Individuals, p. 42.
l88 THE CONCEPT OF IDENTITY

ing the concept of F, then the concept of F is less basic

than the concept of G.
I think it is clear that the concept-dependence test captures
the essential point that the critic is trying to make. He thinks
that there is a uniquely correct way to order our concepts, an
ordering that corresponds to the true logical or metaphysical
structure of the facts. The concept-dependence test merely ex-
presses the idea that it cannot be correct to order our concepts
with F preceding G if it is not even logically possible that
someone should have the concept of F without having the
concept of G. Hence it cannot be correct to treat the (concept
of the) histories of bodies, or the (concept of the) stages of these
histories, as more basic than the (concept of the) bodies them-
selves.
(Throughout this discussion I allow myself to be rather care-
less about a distinction that in other contexts might be crucial,
viz. the distinction between saying that (a) the concept of F
is more basic than the concept of G, and saying that (b) particu-
lar instances of F are more basic than particular instances of G.
It is the (a)-claim that figures most directly in the present
argument.)

III. Temporal Parts

If body-stages are construed innocuously then the analysis does
not even purport to explain the persistence of bodies in terms
of the interrelations between momentary things. And if body-
stages are merely segments of the histories of bodies then the
analysis does not appear to move in the right direction. This does
not, however, exhaust the main possibilities. In fact many phi-
losophers regard body-stages in neither of the two ways just
mentioned.
Many philosophers regard body-stages as the "temporal parts"
of bodies. We are invited to conceive of these temporal parts on
the analogy of a body's ordinary spatial parts. The temporal parts
are said to bear the qualities and relations that we would
ordinarily ascribe to a body at a given moment. On this account
the momentary things which figure in the analysis have colors,
shapes, textures, and stand to each other in various spatial
relations.
 FOUNDATIONS OF IDENTITY 189

In some literature we are presented with the issue whether

there are such things. The question is sometimes put: Do bodies
have temporal parts? But it seems to me that the question, so
put, is verbal: the philosopher who says that there are temporal
parts is using language differently from the philosopher who
denies that there are such things. I am not here endorsing any
kind of positivist view to the general effect that issues of ontology
are always verbal. Each issue has to be taken separately, on its
own merits. Indeed I will in the next section want to turn to an
issue about temporal parts which I am not inclined to regard as
merely verbal, viz. an issue about the metaphysical basicness of
temporal parts. All I am now suggesting is that the stark ques-
tion "Are there such things as temporal parts?" is verbal.
Let me underscore this point, so that my position is not mis-
understood. I have been assuming throughout this discussion that
we are often prepared to talk about the existence of certain kinds
of things (e.g., economic depressions) which we might then want
to regard as metaphysically nonbasic. Consequently when a phi-
losopher talks about the existence of temporal parts I assume
that he has not (as yet) committed himself to anything about meta-
physical priorities. (He may indeed not even acknowledge any
such notion as "metaphysical prorities.") And it is only this
"pure" existence question about temporal parts that I regard as
verbal.4
Certainly the question is a priori. We could not begin to under-
stand a philosopher who said the following: "It seems probable
that in the actual world bodies have temporal parts. For ex-
ample, when a tree grows the chances are that it has early parts
and later parts, and the early parts are smaller than the later
parts. But we can imagine a different situation. We can imagine
a situation in which a tree grows without its having any temporal

4. I do not of course regard as verbal all "pure" existence questions (i.e., ques-
tions about existence which do not broach on issues of basicness), but only
those which succumb to the kind of argument I am about to give. (Very
roughly these would be questions about the existence of F's where both parties
to the dispute agree on certain facts which only one party regards as logically
equivalent to the existence of F's.) For a similar approach see G. A. Paul's
classic paper "Is there a Problem about Sense-data?" Supplementary Pro-
ceedings of the Aristotelian Society (1936), reprinted in A. Flew, ed., Logic
and Language, ist ser. (Basil Blackwell, Oxford), 1951.
igO THE CONCEPT OF IDENTITY

parts." That remark would be utterly incomprehensible. It is

clear that the philosopher who says "The early parts of a tree
are smaller than its later parts" regards this statement as logi-
cally equivalent to "The tree was first small and then large." And
the philosopher who denies the statement about the tree's tem-
poral parts refuses to treat these statements as equivalent. My
suggestion is that the source of this disagreement, whether the
philosophers acknowledge it or not, is that they have adopted
different ways of talking. (As in all verbal disagreements the
disputants may not simply have adopted different uses of lan-
guage, but they may also be tacitly disagreeing about which is
the ordinary use of language; this is still not a "substantive
disagreement" in the relevant sense.)
Now I am not suggesting that any issue about logical equiva-
lence must be verbal. People often make substantive mistakes
about what is equivalent to what. If someone says that "There
are twenty-seven times eighteen objects" is equivalent to "There
are five hundred eighty-six objects" we would not assume that
his mistake is merely verbal; we would not assume, that is, that
he is merely making a strange use of language, and that in his
idiolect the statements really are equivalent. We would expect
rather that his general methods of arithmetical calculation would
reveal that these statements are not equivalent in his idiolect.
Again, we would suspect someone of a substantive mistake
if he says that "a knows that p" is equivalent to "a believes p
with good reason and p is true." Presumably this person's re-
sponse to Gettier examples would belie this equivalence, and
would show that, even within his own idiolect, his remark was
mistaken.
Let us consider, however, what we would say if this person
responded to Gettier examples by insisting that these are cases
of knowledge. Perhaps we would want to present him with
more examples, and develop these examples from various dif-
ferent angles; and perhaps also give him some time to think
about it. But suppose that after all of this he persists in his
evaluation of the examples. Eventually I think we would have
to judge that in his idiolect "knowledge" is equivalent to "true,
rational belief." In this case his mistake was verbal, merely re-
vealing that his use of language is strange.
 FOUNDATIONS OF IDENTITY igi

A reasonable rule of thumb seems to be this: If someone claims

that two statements are equivalent in his idiolect, and this is
borne out by his formal calculations and/or his responses to
particular examples, then there is a strong presumption that the
statements are indeed equivalent in his idiolect.
Let us now return to the philosopher who claims that "The
tree grew larger" is equivalent to "An earlier part of the tree
is smaller than a later part." All of this philosopher's responses
to particular examples, as well as his formal calculations, indi-
cate that in his (philosophical) idiolect the statements are equiv-
alent. It seems indeed clear that this philosopher has adopted
a way of talking in which any statement of the form "The body
x is A at t and B at t"' is equivalent to the statement "The
body x contains the temporal parts y and z such that y exists
only at I and z exists only at t', and y is A and z is B."
We can approach this point from a slightly different direc-
tion. Suppose that we explicitly introduce a new way of talk-
ing, which we might call "the language of temporal parts." We
can introduce this language informally by saying that we are
going to treat time on the analogy of space, and then giving a few
examples to show how this works. This might suffice to teach
the language to anyone who cares to learn it. In principle we
should be able to introduce the language more formally, by
stipulating a range of transformation rules which equate various
English statements with their counterparts in the language of
temporal parts. One such rule might indeed be that any English
statement of the form "x is A at t and B at t'" can be trans-
formed into "x contains the temporal parts y and z such that y
exists only at t and z exists only at t', and y is A and z is B." I
do not want to minimize the formal difficulties that might arise
in a rigorous presentation of the new language. But I doubt
that anyone could deny that this language is in principle in-
telligible and consistent (if English is).
Once this point is granted I think it is quite impossible to fail
to see that the philosophical exponent of temporal parts has
in effect adopted this new language of temporal parts; and that
the philosophical antagonist of temporal parts refuses to speak
the new language, but carries on in ordinary English. So this
is an exemplary case of a verbal dispute. (It follows from my
iga THE CONCEPT OF IDENTITY

account that the exponent of temporal parts may be trivially

mistaken, if, that is, he is tacitly claiming that his is the ordinary
use of language; but he may be entering no such claim.)
The dispute is verbal, but not therefore entirely trivial. Here
we should recall the lesson continually emphasized by John
Wisdom, that a metaphysician's use of language, and especially
his misuse of language, is often designed to reveal unnoticed
aspects of reality, unnoticed analogies and disanalogies.5 Cer-
tainly we can say that the language of temporal parts discloses a
startlingly new perspective on the world.
Philosophers sometimes express their antagonism toward tem-
poral parts by insisting that the expression "the temporal part
of x at t" merely refers to the ordered pair (x, t). But this seems
wrong, in just the way that it would be wrong to maintain that
"the depression of 1929" refers to a set of people (i.e., the people
involved in the depression). We say that the depression began in
1929, but we do not say "A set of people began in 1929." By the
same token, if we speak the language of temporal parts we will
say "The part of the tree at t was short, and straight, and
smooth," whereas we would not ascribe such properties to an
ordered pair. How precisely to characterize the relationship be-
tween an object and the various sets which might be said to
"correspond" to it, is one of the great mysteries of metaphysics;
and I am not suggesting anything to dispell this mystery. But
this problem is not peculiar to temporal parts; and it remains
sufficiently clear that a temporal part cannot be straightforwardly
identified with any set-theoretical item.

IV. A Question of Priorities

If there is a legitimate issue to raise here it must be this: Assum-
ing the notion of a temporal part, which is metaphysically more
basic, a body or its temporal parts? Is a body merely a "logical
construction" out of its temporal parts, or is it rather the tem-
poral parts which are merely "logical constructions"? We may
5. Sec the essays in John Wisdom, Philosophy and Psychoanalysis (University
of California Press, Berkeley and Los Angeles), 1969, especially "Metaphysics
and Verification."
 FOUNDATIONS OF IDENTITY 193

well begin to wonder whether our notion of metaphysical basic-

ness is sufficiently clear to make good sense out of these questions.
Whereas the stark question about the existence of temporal parts
is merely verbal, the deeper issue about basicness may be too
obscure.
(In the metaphysical tradition a number of closely related
questions about the basicness of temporal parts might be form-
ulated, for example: Do statements about temporal parts have
the logical form which their superficial grammatical form sug-
gests? Or is the language of temporal parts merely a "code"
which allows us to express in distorted form what is better ex-
pressed in ordinary terms? On the other hand is it perhaps our
ordinary statements about persisting things which distort the
true logical structure of the facts, and which cannot be taken at
face value? All of these questions are, I take it, essentially varia-
tions on the issue of metaphysical priorities, an issue which in
the present context is coming to look exceedingly problematic.)
The proposed analysis of bodily identity is, on the present
construal, couched in terms of the language of temporal parts.
The critic cannot simply insist that the analysis must not be
couched in such terms. His objection must rather be that, even
given the language of temporal parts, the momentary things
which the analysis talks about do not have the required meta-
physical priority over persisting bodies.
How can he show this? In terms of the concept-dependence
test considered earlier, the key question here would seem to be
whether it is logically possible that someone should have the
concept of a momentary thing without having the concept of a
persisting body. Why should this not be possible? Momentary
things, as now construed, have various sensible qualities and
stand to each other in various spatial relations; and persisting
bodies are constituted by distinctive kinds of successions of these
things. Why should it be logically impossible for someone to
have the concept of the momentary things without having the
concept of these distinctive kinds of successions?
Our question might be seen as arising from two steps. First
we introduce the language of temporal parts, a language in terms
of which we can render all of our ordinary English statements.
Then we ask whether we can imagine someone who speaks only
1Q4 THE CONCEPT OF IDENTITY

a segment of this language, i.e., the segment which refers only

to momentary things.
Here perhaps we have nothing to appeal to but our intuitions
about the possibilities of conceptualization. And these intuitions
seem to be especially flimsy in the present case. It may seem
immediately obvious than no one could possibly have the con-
cept of the history of a body without having the concept of a
body; or even that no one could possibly have the concept of an
economic depression without having the concept of a person. I
do not doubt that serious questions could be raised even about
these examples. But it does not seem even prima facie obvious
that no one could possibly have the concept of a momentary
thing without having the concept of a persisting body.
Suppose it is granted that the critic cannot show that bodies
are more basic than momentary things. The fact is, however,
that the proponent of the analysis, if he is to meet the critic's
standards of analysis, must maintain the opposite, that momen-
tary things are more basic than bodies. And how can he show
this? Perhaps we have reached a standoff.
It may seem, however, that we should be able to develop an
argument in behalf of the priorities set by the analysis. For there
is surely a pervasive tendency in metaphysics to regard a thing's
parts as more basic than the thing. A general argument in be-
half of the primacy of parts over wholes can perhaps be formu-
lated in terms of the concept-dependence test. Suppose that we
divide bodies up into two great classes, those which are of size
N or greater, and those which are less than N. Then it may
seem that our concept of the larger objects necessarily depends
upon our concept of the smaller objects, but not vice versa. If
this is so then it follows that the existence of the smaller ob-
jects is more basic than the existence of the larger ones. Assum-
ing that this holds for any size N, we seem to have the general
result that the smaller the object the more metaphysically basic
it is.
Whatever may be the merits of this argument with respect to
spatial parts, it runs into obvious problems when applied to
temporal parts. The argument requires two assumptions: first,
that the concept of a momentary thing does not necessarily de-
pend upon the concept of a persisting body; and, second, that
the concept of a persisting body does necessarily depend upon
 FOUNDATIONS OF IDENTITY 195

the concept of a momentary thing. Even if the first point is con-

ceded, the second looks hopelessly wrong. For unless we adopt
the language of temporal parts, which is a philosophical crea-
tion, we apparently do perfectly well without the concept of a
momentary thing.
Something of the argument may yet be salvageable. For we
need not limit ourselves to comparing the category of persisting
bodies to the category of momentary things. We may instead
compare the more general category of persisting items of all
sorts to the more general category of momentary items of all
sorts. Of course the analysis I have been discussing pertains to
ordinary bodies, and does not apply directly to various other
kinds of persisting items, such as persons, or places, or the theoret-
ical particles of physics. But it might be maintained with some
plausibility that our concept of a persisting body is the most
central application of our general category of persistence through
time. So if it has been conceded that the concept of a momen-
tary thing is in principle independent of the concept of a per-
sisting body, then it might not be difficult to maintain that the
general category of momentary items is in principle independent
of the general category of persistence. On the other hand it may
be argued that the general category of persistence is not in turn
independent of the general category of momentary items. For
though the thought of persistence through time obviously does
not require the concept of the concrete momentary things which
figure in the analysis it may seem to require at least the concept,
of a moment of time or a momentary event, or some other kind
of item which will represent the category of the momentary.
Along these lines it might perhaps be argued that the general
category of momentary items is more basic than the general
category of persisting items, so that at least in this sense the
analysis moves in the proper direction.
Though this argument does seem to me to have some weight,
it is obviously less than decisive. And I think it is rather doubt-
ful that some other argument will strike us as settling this is-
sue. (We might perhaps attempt an argument from Occam's
razor. But first of all it seems thoroughly unclear whether sim-
plicity considerations should favor persisting things or momen-
tary things; and moreover it is not even clear how such considera-
tions can provide us with the required insight into why one kind
ig6 THE CONCEPT OF IDENTITY

of thing—or one way of talking—is more basic than another.)6

Indeed the slippery twists and turns of this whole debate may
encourage the response that we should simply admit that our
intuitions about metaphysical basicness are quite tenuous and
inconclusive, and that perhaps the very notion of metaphysical
basicness is not to be taken very seriously. I will return to this
negative note after first addressing the question of spatiotemporal
continuity.
V. Spatiotemporal Continuity
The standard treatment of spatiotemporal continuity consists
in defining this notion on the basis of bodily identity. This
obviously presents a problem for the proposed analysis of bod-
ily identity on the basis of spatiotemporal continuity.
The standard treatment actually contains two separable parts.
In the first part, place-identity is defined in terms of bodily
identity. The definition, roughly put, states that a place at one
time is identical with a place at a second time if the first place
and the second place stand in the same relations of distance and
direction to a framework-defining system of persisting bodies.
Place-identity is thus relativized to the choice of a framework, and
presupposes the notion of a persisting body.
In the second part, spatiotemporal continuity is defined in
terms of place-identity, roughly as follows. A body is said to move
in a spatiotemporally continuous manner if it occupies closely
neighboring places at closely neighboring times. Given the pre-
vious definition of place-identity, what this amounts to is that
at closely neighboring times the body stands in very similar
spatial relations to the framework-defining system of persisting
bodies. Or to recast this idea in terms of momentary things, a
succession of momentary things is spatiotemporally continuous
if temporally neighboring elements of the succession stand in
very similar spatial relations to the framework-defining system
of persisting bodies. On this account the proper ordering of our

6. It is on the face of it quite absurd to be told that a mere change of lan-

guage, which allows us to make statements logically equivalent to those made
in the first language, could have the effect of generating an ontological prob-
lem of "overpopulation." Is this "problem" supposed to explain why the
second way of talking is necessarily parasitic on the first? Or why the second
way of talking misrepresents "true logical form"?
 FOUNDATIONS OF IDENTITY 197

concepts evidently must take bodily identity prior to spatio-

temporal continuity.
Both parts of the standard treatment can be challenged. Obvi-
ously the relativization of place-identity to a choice of frame-
work presupposes the rejection of the doctrine of absolute
motion. That doctrine would have to treat place-identity as
basic, for the whole point of the doctrine is that place-identity
does not depend upon the comings and goings of bodies. If place-
identity is treated as basic, then our analysis of bodily identity
may proceed smoothly. The ordering of concepts might be:
place-identity, then spatiotemporal continuity, then bodily
identity.
Though we seem to have some rather strong and, I think, not
easily expungible intuitions in favor of absolute space, science
has presumably taught us that there is no such thing. And this
scientific lesson has obviously not gotten us to repudiate our
concept of bodily identity; indeed the lesson does not even ap-
pear to impose any great strain on our concept of bodily identity.
So it could not be right for the analysis to base .our concept of
bodily identity on absolute place-identity.
Let us assume, therefore, that the standard treatment of rela-
tivized place-identity is accepted. We are still left with two alter-
natives: We can define spatiotemporal continuity on the basis of
bodily identity, as in the standard treatment; or we can define
bodily identity on the basis of spatiotemporal continuity, as
suggested in the analysis. On the first approach we arrive at the
ordering: bodily identity, then place-identity, then spatiotem-
poral continuity. On the second approach we have the ordering:
spatiotemporal continuity, then bodily identity, then place-
identity.
I think it is not easy to decide which is the correct ordering.
And we may indeed wonder, as we did with respect to the issue
of body-stages, whether it even makes good sense to suppose that
there is "the correct ordering." If there is a metaphysical issue
here, to be distinguished from various epistemological and psy-
chological issues, it cannot be settled on the basis of any con-
tingent facts about human concept formation. The crucial
question would seem to be whether it is at least logically possible
that someone should have the concept of spatiotemporal con-
tinuity without having the concept of bodily identity. My own
ig8 THE CONCEPT OF IDENTITY

intuitions certainly do not decisively indicate a negative answer

to this question.7 And if there is no compelling reason to regard
spatiotemporal continuity as less basic than bodily identity, the
justification for regarding it as more basic might simply be a
corollary of the earlier argument for regarding the momentary
things, and hence their interrelations, as basic.
It must be understood that the question here is not whether
every case of spatiotemporal continuity must be a case of bodily
identity. No one is claiming this. There are many kinds of cases
of spatiotemporal continuity which are certainly not cases of
bodily identity. The most obvious case is where a spatiotem-
porally continuous succession combines the early stages of an
object with the later stages of its parts. Another kind of case is
where we judge an object to go out of existence and be replaced
by another object, e.g., where a gold coin is melted down and
replaced by a lump of gold. But according to the standard defi-
nition all of these cases of spatiotemporal continuity can only
be understood against some background of persisting bodies
which define a spatial framework.
The case of something being replaced by something else does,
however, suggest a problem for the standard treatment. If we
can conceive of a body going out of existence and being replaced
by another body, can we conceive of this happening simul-
taneously to every body in the universe? This seems to be
logically possible, and may even be empirically possible. But the
standard treatment would rule this possibility out. To say that
one body replaces the other must surely imply a spatiotem-
porally continuous connection between the initial stage of one
body and the terminal stage of the other. But if this happens
at a certain time to every body in the universe then at that time
there would be no framework of persisting bodies, and hence,
according to the standard treatment, there could be no spatio-
temporal continuity, and consequently no replacement of bodies
by bodies, contrary to the hypothesis. This is, I think, a fairly
serious problem, which argues again for the basicness of spatio-
temporal continuity.
We should briefly consider another alternative to those so far
7. An affirmative answer might be suggested by the tempting comparison
between spatiotemporal continuity and spatial continuity; cf. above, pp. 147-
49-
 FOUNDATIONS OF IDENTITY 199

mentioned. It is tempting to suggest that both spatiotemporal

continuity and bodily identity derive from something more basic
than either, viz. the successive spatial relations between the
momentary things. Any momentary thing x has what we might
call a global context. The description of x's global context would
include such statements as this: "x is 30 thousand miles from a
green spherical thing," "x is touching a hard blue thing," etc.
The description of x's global context would be a description of
the whole momentary universe from, so to speak, x's standpoint.
If x is a momentary thing that exists at t, and y is a momentary
thing that exists at t', it is intuitively obvious what could be
meant by saying that x's global context is very similar to y's.
This would imply that if x is 30 thousand miles from a green
spherical thing, then y is approximately 30 thousand miles from
something which is approximately green and approximately
spherical.
It might now be suggested that we can define spatiotemporal
continuity in terms of the continuity of global contexts. A suc-
cession of momentary things is spatiotemporally continuous, ac-
cording to this definition, if temporally neighboring elements of
the succession have very similar global contexts.
This definition will immediately raise an epistemological prob-
lem, insofar as it apparently implies that spatiotemporal con-
tinuity is not observable (since one cannot presumably be said
to observe something's whole global context). Apart from this,
the definition has some rather counterintuitive metaphysical
implications. Perhaps the most glaring of these pertains to imag-
inary cases in which there is radical duplication of objects.
Max Black has imagined a universe which consists exclusively
of two spheres that are descriptively indiscernible, i.e., that are
alike with respect to all general qualitative and relational char-
acteristics.8 It seems that we can also imagine these spheres as
descriptively indiscernible relative to one time, but discernible
relative to earlier and later times. For example, they might first
differ in color, then become indiscernible, then differ in color
again. In order for us to make sense out of this possibility we
must be able to distinguish between the following two cases.

8. Max Black, "The Identity of Indiscernibles," in Problems of Analysis

(Cornell University Press, Ithaca, N.Y., 1954), pp. 80-92.
200 THE CONCEPT OF IDENTITY

D ± : The universe contains nothing but two spheres, and these

spheres never differ in any respect except color. One of the
spheres is initially red while the other is initially yellow. The
sphere which is initially red undergoes a continuous change of
color, in which it passes from red to orange to yellow. While
this is taking place the other sphere also undergoes a continuous
change of color, in which it passes from yellow to orange to red.
There is a time t during this period when both spheres share the
exact same shade of orange, so that at t the spheres are descrip-
tively indiscernible.
D 2 : The universe contains nothing but two spheres, and these
spheres never differ in any respect except color. One of the
spheres is initially red while the other is initially yellow. The
sphere which is initially red undergoes a continuous change of
color, in which it passes from red to orange and back to red
again. While this is taking place the other sphere also under-
goes a continuous change of color, in which it passes from yellow
to orange and back to yellow again. There is a time t during
this period when both spheres share the exact same shade of
orange, so that at t the spheres are descriptively indiscernible.
Dj says that the sphere which was initially red winds up yellow,
and the sphere which was initially yellow winds up red, whereas
D2 says that the sphere which was initially red winds up red, and
the sphere which was initially yellow winds up yellow. Intuitively
it seems clear that we can distinguish between these two cases.
But given the previous definition of spatiotemporal continuity
in terms of contextual continuity, it would seem to follow that
we cannot make the distinction, at least not on the basis of
spatiotemporal continuity. For suppose that Dt is true. Then we
can derive D2 by combining the pre-t stages of one sphere with
the post-i stages of the other. And there would be no considera-
tions of spatiotemporal continuity to block this, assuming the
previous definition. For the succession which combines the pre-t
stages of one sphere with the post-f stages of the other would be
spatiotemporally continuous, given that definition, since there
is evidently no lack of continuity of global context. Intuitively
we want to say that this succession will lack spatiotemporal con-
tinuity, which is why the truth of Dj precludes the truth of D2.
I am therefore disinclined to accept the definition of spatio-
 FOUNDATIONS OF IDENTITY 2O1

temporal continuity in terms of contextual continuity. Of course

this definition would have been congenial to our analysis of
bodily identity in terms of spatiotemporal continuity, since it
would ground spatiotemporal continuity in something other
than bodily identity. But, as I have already indicated, I can
see no decisive reason why spatiotemporal continuity needs to
be grounded in anything.

VI. Analyzing Bodily Identity

All in all, I would conclude that a modestly plausible case can
be made out for the metaphysical priorities set by the analysis.
At least these priorities seem no less plausible than those ad-
vanced by the critic.
We might be tempted to draw a more general and negative
conclusion from this whole discussion. Perhaps the discussion
suggests that the notion of "metaphysical basicness" makes no
real sense and has no real use. If this is so we must jettison the
notion of "analysis" in its traditional sense, and with it the tra-
ditional sense of such kindred notions as "logical construction,"
"logical simplicity," and "logical form." Perhaps none of these
notions can really be separated from various contingent facts
about human concept formation and human knowledge.
Though I have some considerable sympathy for this negative
attitude, I think we might reasonably adopt it with respect to
some issues without necessarily adopting it across the board. Our
intuitions about metaphysical basicness may be dim but they
are not necessarily benighted. And where we have a clear and
decisive intuition (as perhaps with the case of economic depres-
sions) then I can see no reason to reject it. But with respect to
the sorts of issues I have been raising about bodily identity it is
doubtful that we do have any such clear intuition.0
9. Here as elsewhere in this book I challenge, though only selectively and
partially, various traditional doctrines of "metaphysical basicness" and
"analysis"; but I draw short of waging a full-scale or head-on assault on these
doctrines. For a general critique of such doctrines, and a historical perspec-
tive, see Urmson, Philosophical Analysis.
Some of our strongest intuitions about metaphysical basicness, I think, relate
to the relative priority of "natural kinds" over "artificial classes." See Chapter
9, especially Section V, and ftn. 24. In the present discussion I am in effect
maintaining that we have no comparably strong intuitions about the meta-
202 THE CONCEPT OF IDENTITY

So I would say this: If to give an "analysis" of bodily identity

means to explain this notion in terms which seem to be de-
cisively more basic in some noncontingent metaphysical respect,
then I think it is doubtful that we can ever give an analysis of
bodily identity.
But we might of course give an "analysis" in various less
stringent senses. We can try to depict the logical interconnec-
tions between (the "logical geography" of) the concept of bodily
identity and various other concepts, such as sortal-coverage and
continuity. We can try to determine whether bodily identity is
equivalent to some combination of these other concepts. And
we can try, as I have been trying, to decide whether these other
concepts are, if not decisively more basic, at least not decisively
less basic than bodily identity. And of course we might also
examine the epistemological relations between these concepts,
which is indeed the topic I next want to consider.

VII. Epistemological Priorities

I continue to assume in this discussion that bodily identity is
essentially equivalent to the condition of sortal-coverage in con-
junction with the continuity conditions. It is important to be
quite clear that this assumption does not by itself commit us to
holding that a person's judgments about bodily identity must be
derived from some prior judgments about sortal-covered con-
tinuity. In general, if a proposition is logically equivalent to a
conjunction of certain other propositions, it does not follow
that a judgment about the first proposition must be derived from
judgments about the other propositions.
This point was already implicit in an illustration mentioned
earlier. It may seem plausible to say that the playing of a par-
ticular song is equivalent to the playing of certain kinds of notes
in a certain kind of relationship. The proposition "That was the
Star Spangled Banner," for example, might be regarded as
essentially equivalent to a complicated conjunction of proposi-

physical priority of "natural units" (i.e., bodies). I would maintain this nega-
tive attitude even with respect to the special case of persons; see Chapter 10,
especially Section II. (See also the last two paragraphs of the Introduction to
Part Two.)
 FOUNDATIONS OF IDENTITY 203

dons about individual notes (together perhaps with certain

facts about the social status of the song). But this does not imply
that the ordinary person's judgment "That was the Star Spangled
Banner" derives from judgments about the notes. Most people
cannot even recognize notes.
So even on the assumption that bodily identity is equivalent to
sortal-covered continuity, it must remain an open possibility that
judgments about bodily identity are not based on judgments
about continuity or sortal coverage. And from a phenomeno-
logical standpoint this possibility seems immediately appealing.
At least a superficial exercise of phenomenological inspection
does not seem to reveal that our judgments of bodily identity
derive from any other kinds of judgments. As I look around me
the perceptual judgments that seem to form directly in my mind
are such as "The pen is moving on the paper," "The cigarette
is burning down," "The cup remains stationary." These are
evidently judgments about how bodies persist and change. It is
not clear that I make any judgments at all about continuity or
sortal coverage, let alone that I derive from these my judgments
about the vicissitudes of bodies.
Yet it is a rather common philosophical assumption that our
judgments of bodily identity must derive from such "criteria"
as sortal-covered continuity. Indeed, as I noted earlier, a standard
use of the philosophical expression "criteria of bodily identity"
seems to imply not only that judgments about bodily identity
are logically equivalent to judgments about the presence of the
criteria, but also that the former judgments are inferentially
derived from the latter. The first point I take for granted here;
but I am questioning the second assumption.
I think that this second assumption frequently derives from
a certain fallacious argument, which I will call the argument
from immediacy. It might be put as follows: "Obviously we
cannot see into the future, or into the past. This means that a
perceptual judgment, properly speaking, can only describe what
is happening at a particular moment. So a perceptual judgment
cannot possibly be about identity through time. Rather our
judgments about identity through time must be inferred from
various observed facts about particular moments, which facts
constitute our criteria for these judgments."10
10. Cf. Quinton, The Nature of Things, pp. 58-59.
204 THE CONCEPT OF IDENTITY

Let me temporarily grant the argument's premise, viz. that a

perceptual judgment can only describe what is happening at a
particular moment. The main point I want to make is that this
premise cannot sustain the argument's conclusion. If a perceptual
judgment can only describe what is happening at a particular
moment then it would indeed follow that a perceptual judg-
ment cannot possibly be about identity through time. But it
would also follow that a perceptual judgment cannot be about
spatiotemporal continuity, which is a relationship between bodies
(or body-stages) at different times. So it would follow from the
argument's premise that spatiotemporal continuity, which is
supposed to be a criterion of identity, is itself unobservable. This
contradicts the argument's conclusion that we infer bodily iden-
tity from observable criteria.
The argument wants to have it both ways. It wants to say both
that (i) perceptual judgments cannot be about identity through
time, and (2) perceptual judgments can be about conditions
which logically entail identity through time. But the same con-
siderations which might induce us to accept (i) will induce us
to deny (2). I single out spatiotemporal continuity for special
attention because, first, this condition is most widely accepted
as a criterion of identity, and also because with respect to this
condition the conflict between (i) and (2) is most glaring.
If the premise of the argument were correct then there could
be no observable facts which logically entail bodily identity.
Even the mere temporal ordering of events would be prob-
lematical, though perhaps it could be said that we order event
A as temporally prior to event B if the observation of B is ac-
companied by the memory of A. But a mere sequence of tem-
porally ordered momentary observation-reports could not possibly
entail any facts about identity through time, for it could not
even entail that a succession of body-stages is spatiotemporally
continuous.
Given the premise of the argument the most basic level of
objective knowledge would consist of temporally ordered facts
about the qualities and spatial relations of momentary body-
stages. From this we could perhaps deduce certain facts about
qualitative continuity and continuity of local (as opposed to
global) context. At this point some kind of mysterious leap
would have to take place, perhaps inspired by instinct, to the
 FOUNDATIONS OF IDENTITY 205

level which includes both judgments about bodily identity and

judgments about spatiotemporal continuity. There is no argu-
ment here at all for regarding the former judgments as criterially
based upon the latter.
So even if we accept the premise of the argument from imme-
diacy, the conclusion does not follow. And I do not in fact think
that we should accept the premise. The truism that "we cannot
see into the future or past" does not compel us to deny that we
can, in the most literal and proper sense, observe facts which
relate different moments of time, facts which can perhaps be
characterized, in one traditional jargon, as belonging to a
single "specious present." So we need not rule out a priori the
possibility that we simply observe how bodies persist and change.
Let us consider an example in which it seems straightforwardly
correct to say that a judgment is based on criteria (in the philo-
sophical sense of "criteria" under discussion). I might judge that
a figure is an octagon on the basis of (a) observing that it is a
polygon, and (b) counting eight sides. Its being an octagon is
equivalent to its satisfying the two conditions. And my judgment
that it is an octagon consciously derives from my prior judg-
ment that each condition is satisfied.
I think it is clear that our perceptual judgments of bodily
identity are not typically based on criteria in this way. When
we observe a body, our judgments about its persistence and
change are not consciously derived from any prior judgments
about the satisfaction of some identity conditions. In this sense,
our judgments about bodily identity, at least in optimal condi-
tions of observation, are not based on any criteria whatever.
But we also need to consider a weaker and more nebulous
sense of "basing a judgment on criteria." Some of the murkiness
of this issue can be brought out immediately by comparing the
case of judging that something is an octagon with the case of
judging that something is a triangle. It seems correct to say that
I can simply see that something is a triangle, and that I need
not consciously derive this judgment from any prior judgments
that the figure is a polygon and that it is three-sided. On the
other hand it may also seem correct to say that I see that it is
a triangle insofar as I judge it to have three sides and to be a
polygon. Perhaps we want to say that the perceptual judgment
that it is a triangle is based on an unconscious, or implicit, in-
aof) THE CONCEPT OF IDENTITY

ference from the judgments that it is three-sided and that it is

a polygon.
Now an interesting possibility, I think, is that our perceptual
judgments of bodily identity also involve an unconscious or
implicit inference from criteria. Though I cannot venture any
general explanation of the nature of "unconscious" or "implicit"
inference from criteria, I suggest that the primary kind of
situation in which it may seem correct to apply this notion might
be roughly characterized as follows. A person S judges that a
proposition p is true, where there exists a certain set C of condi-
tions such that: (i) p is logically entailed by the proposition that
the conditions in C are satisfied; and (2) S does not consciously
or explicitly judge that he perceives the conditions in C to be
satisfied; but (3) if S's attention were redirected, perhaps by his
being asked certain questions, then S would explicitly judge
that he perceives the conditions in C to be satisfied.
To the extent that these three clauses apply it may seem correct
to say that in a sense S's judgment that p is true is based on the
criteria C; that S's judgment that p is true is based on the un-
conscious or implicit inference from the judgment that each
condition in C is satisfied.
Clause (3) requires that the criterial conditions should be
accessible to S, in the sense that S's perceptual and conceptual
abilities would enable him on that occasion to form the con-
scious perceptual judgment that the conditions obtain. Though
this requirement is rather vague, and may admit of degrees, it
seems obviously to be met in the case of recognizing a triangle.
If the case is typical we can easily induce the explicit judgments
that the thing is a polygon (or a closed figure) and that it has
three sides. This is why we may be inclined to say that recogniz-
ing a triangle is implicitly based on these judgments.
I think it can now be argued that judgments of bodily iden-
tity ought to be compared to the case of recognizing a triangle.
One source of resistance to this proposal may be that the condi-
tions of identity seem rather technical, and therefore inaccessible
to the typical observer. But this is, I think, a rather superficial
point. First of all, for our present epistemological purposes a
body-stage can be understood in terms of the essentially common-
sensical notion of the stage of a thing's history. Of course the
typical observer will not be readily induced to judge "There goes
 FOUNDATIONS OF IDENTITY 2O1/

another sortal-covered continuous succession of body-stages." But

he can be readily induced to make judgments which, in a
rudimentary fashion, amount to much the same thing. Suppose
he is observing a pen. Then we can easily get him to attend to
the fact that at any time during this stretch of observation the
body he observes is a pen; that the qualities of the body he
observes at any time are very similar to what the qualities were
of the body he observed at the just previous time; and that the
location of the body he observes at any time is very close to what
the location was of the body he observed at the just previous
time. These facts amount in our philosophical terminology to his
having observed a succession of body-stages which all come under
the sortal "pen," such that the succession is both qualitatively
and spatiotemporally continuous. So I think it is fair to say
that the conditions of bodily identity are in the relevant sense
accessible to the typical observer. And this argues for the con-
clusion that typical judgments of bodily identity are implicitly
or unconsciously based on the conditions as criteria.
If we accept this conclusion we need not deny that people can,
in the most strict and proper sense, perceive how objects persist
and change. Just as we might want to say that someone can
perceive that a figure is a triangle insofar as he implicitly infers
this from the conditions of triangularity, so we can say that
someone perceives the persistence and change of a body insofar
as he implicitly infers this from the criterial conditions of bodily
identity.
There is, to be sure, an important difference between perceiv-
ing a triangle and perceiving bodily identity. Almost anyone
can state what the conditions of triangularity are, whereas no one
except perhaps a few philosophers (who even disagree among
themselves) can state what the conditions of bodily identity are.
But it seems clear that there are many cases in which we recog-
nize a complex kind of object or phenomenon on the basis of
perceiving certain features, without our being able to state
authoritatively what the features are that define that kind of
object or phenomenon.
An interesting example of this sort, which may indeed be very
similar to the case of bodily identity, concerns our judgments
about how things form into groups. When we look at the following
figure we are likely to see pairs of dots and triplets of crosses,
2O8 THE CONCEPT OF IDENTITY

rather than various other possible combinations of these ele-

ments. Gestalt psychologists have attempted to formulate the
principles of grouping which in effect define what we mean in
this kind of context by a "group," or "cluster," or "arrange-
ment." These principles are complicated and controversial. It
may seem evident that in the exhibited case the conditions of
proximity and similarity play some role in determining our
perception of the group-units; but it is by no means obvious how
to combine or weight these conditions, nor what other condi-
tions might be relevant. But though we may be unable to state
precisely what the conditions of group-unity are, I think that
few would doubt that our perception of the groups is in some
manner derived from the implicit recognition of these conditions.
I am suggesting that the conditions of bodily identity may play
very much the same kind of role in determining our perception
of how bodies persist and change.
I believe that the only serious objection to this position stems
from the widely accepted view that judgments of spatiotemporal
continuity must derive from judgments of bodily identity. This
view seems to follow directly from the standard treatment of
spatiotemporal continuity, according to which spatiotemporal
continuity must be relativized to a framework-defining system
of persisting bodies.
Now there are some philosophers who appear to think that
our judgments of spatiotemporal continuity must always de-
pend on judgments of bodily identity, and also that our judg-
ments of bodily identity must always depend on judgments of
spatiotemporal continuity.11 Such a view seems quite baffling.
How could these two kinds of judgments feed oft each other in
that way? If the standard treatment of spatiotemporal continuity
is accepted then at least our judgments about the identities of
the framework-defining bodies cannot possibly derive from
judgments of spatiotemporal continuity. And if this is so the
obvious conclusion to draw is that our judgments of bodily
identity need never derive from judgments of spatiotemporal
continuity. As I have repeatedly emphasized such a view of the
epistemological priorities may be fully compatible with the
11. See especially Strawsori, Individuals, p. 26.
 FOUNDATIONS OF IDENTITY 309

conceptual point that bodily identity is essentially equivalent to

sortal-covered continuity.
But I am inclined to think that this is the wrong view of the
epistemological priorities. I have already suggested that there are
some metaphysical problems with the standard treatment of
spatiotemporal continuity. From an epistemological standpoint
this approach is even more obviously vulnerable. Certainly it
seems plausible to say that we can observe spatiotemporal con-
tinuity, that we can observe that a body moves continuously
through space. And surely our ability to observe a case of spatio-
temporal continuity does not depend on their being any
framework-defining bodies within the scope of our observation.
We can imagine someone who observes a single body in total
isolation from all other bodies. (We can even imagine that the
observer suffers from some abnormality which prevents him from
perceiving his own body.) Surely the isolated body could be
observed to move continuously, to move without any discon-
tinuous jumps. How can we make sense out of this if, as in the
standard treatment, spatiotemporal continuity must be relativized
to some framework-defining bodies?
Perhaps there is some way to get around this question. But I
think the most obvious and plausible response is to reject the
standard treatment. Spatiotemporal continuity can be regarded
as an observable phenomenon that is as primitive, as indefinable,
as color, or shape, or contact.
This does not necessarily imply that our perception of spatio-
temporal continuity is in no sense based on something. Perhaps
all of our perceptions are based on the having, or even on the
implicit recognition, of our subjective sense data, as so many
philosophers have thought. And perhaps, as noted earlier, there
is even an objective level of perception, more basic than either
spatiotemporal continuity or bodily identity, which embraces
only momentary facts. These issues need not be settled here. I
am only arguing for the position that our perception of spatio-
temporal continuity need not derive from any judgments about
bodily identity.
If this position is accepted then nothing stands in the way of
saying that our perception of bodily identity does derive from
judgments of spatiotemporal continuity, as one criterion of
bodily identity. And if nothing stands in the way of saying this
21O THE CONCEPT OF IDENTITY

then the analogy to other cases (such as that of group-unity)

seems to indicate that we should say this.
I conclude this discussion of the epistemological aspect of
bodily identity on much the same tentative note that concluded
my previous discussion of the metaphysical aspect. Certainly I
have presented no very conclusive argument for the epistemolog-
ical priorities I recommend. But I think a fairly good case has
been made for the position that our perception of bodily iden-
tity derives from the implicit recognition of the conditions of
bodily identity, so that these conditions can rightly be called
criteria of identity.
 7
Matter, Causality,
and Stereotypes of Identity

I. Optimal Cases
I AM inclined to maintain—with various reservations to be dis-
cussed—that the following is a conceptual truth: If a succession
of car-stages is spatiotemporally and qualitatively continuous,
then it constitutes some portion (perhaps the whole) of the his-
tory of a persisting car. This principle, to put it somewhat less
technically, states that if there exists a car at one moment, and
there exists a car at the next moment, and there exists a car at
the moment after this, and so on, and the car that exists at any
moment is located very close to where the car at the next moment
is located, and also the car that exists at any moment has qualities
that are very similar to the qualities of the car at the next mo-
ment, then this is one and the same car that exists at each of
these moments.1 I will call this principle SQ, after the conditions
of spatiotemporal and qualitative continuity. Of course I would
suggest the corresponding principle for various other sorts of
things, such as trees, rocks, and tables. The principle, relativized
to each sort of thing, would state that any spatiotemporally
and qualitatively continuous succession of stages of that sort
of thing constitutes some portion of the history of a thing of
that sort.
To say that SQ is a conceptual truth means, trivially, that its
truth follows from the nature of the concepts involved, or, if
i. Here as elsewhere I allow the loose but intuitive notion o£ a "next" or
"neighboring" moment.

211
212 THE CONCEPT OF IDENTITY

one prefers, from the meanings of the relevant words. Another

way to put this is that we cannot so much as conceive of what
it would be like for SQ to turn out to be false. I will have some-
thing more to say later about the notion of a conceptual truth,
and its connection to the notion of metaphysical necessity.
I place one general limitation on my commitment to SQ as a
conceptual truth. I do not necessarily deny that there could be
a case which seems to satisfy the SQ conditions but which is
nevertheless a borderline case of the identity of a car. In such a
case a decision might be made as to what to say, and some might
decide against the identity claim. But according to my suggestion
SQ is a conceptual truth up to borderline cases: There is no
case, actual or possible, which both clearly satisfies the SQ con-
ditions and also clearly is not a case of identity.
Before proceeding let me mention a weaker version of the SQ
principle, which may strike some readers as immediately more
attractive. The weaker version would say that it is a conceptual
truth that the presence of the SQ conditions constitutes prirna
facie evidence, though not conclusive evidence, for an identity
claim.2 Toward the end of this chapter I will broach an analysis
which would sustain this weaker version. For now, however, I
want to try to show that even the stronger version may be
viable.
Assuming the general notion of a conceptual truth, there are
two kinds of objections that might be raised against my claim
that SQ is a conceptual truth. First, it might be maintained
that, in addition to the two conditions mentioned in SQ, certain
other conditions are necessary for the identity of a car. Two
conditions which have often been mentioned in the literature
are compositional continuity and causal continuity. The first
condition would require that the matter which makes up the
car at any moment be almost the same as the matter which makes
it up at the next moment. And the second condition would
require that the car's qualitative state at each moment be caus-
ally related to its qualitative state at the next moment.
2. It may be noted that even the weaker version of SQ could sustain, with
suitable modifications, the general approach to identity presented in Part
One. (For example, we could define F as a sortal if it is a conceptual truth
that the presence of an F-succession satisfying the SQ conditions constitutes
prima facie evidence for the identity of an F-thing.)
 MATTER, CAUSALITY, AND STEREOTYPES OF IDENTITY 213

The first objection to SQ is an objection of detail; it does not

question the possibility of stating some nontrivial combination
of conditions which suffices to guarantee conceptually the iden-
tity of a car. A philosopher might, however, deny even this gen-
eral possibility. Such a philosopher might insist that the identity
of a car is "primitive" and "indefinable," and therefore cannot
be entailed by any combination of spatiotemporal, or qualitative,
or compositional, or causal conditions. (Of course even this phi-
losopher would admit that we could trivially define some new
expression, e.g., "car kinship," in such a manner as to make it
tautologous to say that a succession of kindred car-stages consti-
tutes a persisting car.)
Let me first address this extreme form of objection. I will
try to describe a case, which I will call the optimal case, in
which every possible source of doubt about a car's identity has
been removed. Doubts might arise about the degree of spatio-
temporal continuity that is typically required for a car's identity.
So let us assume that in the optimal case we have a very high
degree of spatiotemporal continuity. Similar doubts might arise
about the degree of qualitative continuity required- But also
doubts might conceivably arise about how much a car can
change, even continuously, and still be the same car. So let us
assume that in the optimal case we have not merely a high degree
of qualitative continuity, but a high degree of qualitative sta-
bility, i.e., any two car-stages in the succession are qualitatively
very similar to each other. For the same reason let us assume that,
in the optimal case we have a high degree of compositional
stability, i.e., any two car-stages in the succession are composed
of almost the same matter. And we will assume that in the
optimal case there is a high degree of causal connectedness be-
tween the car-stages in the succession.
In sum, in the optimal case we have a succession of car-stages
which is to a high degree spatiotemporally continuous, which
is highly stable both qualitatively and compositionally, and
which has a high degree of casual interconnectedness. Could
anyone seriously deny that it is a conceptual truth that the
optimal case is a case of identity? Such a denial would seem on
the face of it extremely paradoxical. If we have a succession of
car-stages which satisfies all of the conditions of the optimal
case, then it seems that we cannot conceive of anything more
214 THE CONCEPT OF IDENTITY

that might be required for this succession to qualify as a per-

sisting car. Of course a philosopher might respond that some-
thing more is required, viz. that it be the same car. But this
response, barring some special explanation, seems simply
unintelligible.
So I think that reflection on the optimal case will convince
most people that it is indeed possible to specify some (nontrivial)
conditions which are logically sufficient for the identity of a car.
The only question then is what these conditions are. I suggest
that one set of logically sufficient conditions is the SQ condi-
tions. Note that I do not suggest that these conditions are
necessary, but only that they are sufficient. The objection that
now concerns me implies that the SQ conditions are not sufficient
because certain other conditions are necessary, i.e., the conditions
of compositional and causal continuity. I suggest that these
latter conditions are not necessary. Note again that I do not
necessarily deny that these conditions may be sufficient in their
own right.
Before pursuing this issue let me introduce another kind of
case which I will call the nearly optimal case. In the optimal
case we had qualitative and compositional stability; in the nearly
optimal case we merely have qualitative and compositional
continuity. This means that in the latter case two car-stages in
the succession may be qualitatively very dissimilar or composed
of entirely different matter, so long as temporally neighboring
car-stages are qualitatively similar and composed of almost the
same matter.
There may be, at least prereflectively, the impulse to maintain
that continuity of qualitative or compositional change is not
enough, but that there is an upper limit on the degree of change
in these respects that a car can suffer while retaining its iden-
tity. (There is, for some reason, not even the slightest impulse
to impose such a limit on how much a car can move and still
be the same car.) Of course if the car changes so much that it
ceases to be a car then it will indeed not qualify under any of
the conditions I have so far been discussing. All of these condi-
tions are relativized to the sortal "car"; it is a succession of car-
stages that is always presupposed. But even given that we have
a succession of car-stages there may still be the impulse to hold
that, in order to have a case of identity, this succession must
 MATTER, CAUSALITY, AND STEREOTYPES OF IDENTITY 215

be to some extent stable with respect to its qualities or material

composition.
I think that this impulse is for most people rather decisively
quelled by simply considering a few examples in which we ob-
viously do not require stability. For example, the material com-
position of a living thing, such as a tree, may alter entirely over
a period of time without that thing's going out of existence. And
so long as a tree remains a tree, there is apparently no general
limit on how much it can change with respect to its qualities of
size, or shape, or color.
There is a more general consideration which seems to show
that if stability suffices for identity so must continuity suffice.
Let x be a car that is picked out at f t , y a car that is picked out
at t2, and z a car that is picked out at t3, where tlt t2, and t3 are
neighboring times. Suppose that x is the same car as y because
(in addition to the other conditions being satisfied) x at tt is
almost like y at £ 2 with respect to qualities and composition, and
that y is the same car as z because y at t2 is almost like z at ( n with
respect to qualities and composition. Then, by the transitivity of
identity, x must be the same car as z even if x at ^ is not almost
like z at ts with respect to qualities and composition. In other
words, if a car can retain its identity through a small qualitative
or compositional change, it seems that it must be able to retain
its identity through any sequence of such small changes, even
if they add up to a large change. But a sequence of small changes
is precisely what we mean by a continuous change.
It might be suggested that this is just another case of the
familiar sorites argument. It is like arguing that, since a poor
person who is given a penny remains poor, therefore no sequence
of such gifts could ever make the person rich. But there is an
important difference between these cases. In the typical sorites
argument we can at least in principle stipulate a cut-off point,
e.g., the point at which getting one more penny will change the
person to being not poor. The fact that there is in practice no
such definite cut-off point might then be regarded as merely re-
flecting the vagueness of our concept of poverty. But in the
identity argument it seems that there cannot be even in principle a
cut-off point, for the transitivity of identity precludes our saying
that x = y and y = z but x ^= z.
Perhaps there are some subtle maneuvers to escape this argu-
2l6 THE CONCEPT OF IDENTITY

3
ment. But I think it is more plausible to assume that if sta-
bility suffices then so does continuity. And from this it follows
that, since the optimal case is a case of identity, so is the nearly
optimal case. In the nearly optimal case we have the four kinds
of continuities, each to a sufficiently high degree. I will not
attempt to specify what constitutes a "sufficiently high degree";
perhaps it is not possible to specify this in any useful way.
Now my argument will proceed as follows. The conjunction of
the four continuity conditions suffices for identity; this is the
nearly optimal case. It seems plausible to assume that any con-
dition which is either necessary or sufficient for identity is some
combination of the four continuities. Suppose that it can now
be shown that neither compositional continuity nor causal con-
tinuity (nor their disjunction) is necessary for identity. It then
would follow that the other two conditions, i.e., the SQ condi-
tions, suffice.

II. Compositional and Causal Continuity

It is quite easy to think of examples which at least strongly sug-
gest that compositional continuity is not necessary for the iden-
tity of a car. Suppose that in the next issue of Scientific Amer-
ican we read that whenever a car backfires the majority of its
subatomic, particles simultaneously go out of existence and are
replaced by other particles. This would amount to a massive
discontinuity of material composition. Would this announce-
ment induce us to say that, as it has turned out, cars go out of
existence when they backfire? I think not. We would say rather
that it has turned out that cars wind up with new subatomic
particles when they backfire.

3. One interesting possibility, noted by Alan Brody, is that the identity of a

car depends on its qualitative and compositional similarity to how it was
when it was first created. This suggestion would allow us to stipulate a cut-
off point, and hence even cast a kind of doubt on the optimal case (for even
a minute change might be too much relative to the cut-off point). Certainly
this suggestion could not apply to many things, such as trees; and I doubt
that it really has much plausibility even for cars.
 MATTER, CAUSALITY, AND STEREOTYPES OF IDENTITY 217

I think we can imagine a more extreme possibility.

Could we not reject the concept of matter and build a pure field
physics? What impresses our senses as matter is really a great concentra-
tion of energy into a comparatively small space. . . . A thrown stone
is, from this point of view, a changing field. . . . There would be no
place, in our new physics, for both field and matter, field being the
only reality.*

Einstein and Infeld, from whom this passage is quoted, put

this forward only as a possibility. But suppose that such a pos-
sibility were generally accepted by scientists. If the concept of
matter is rejected then apparently the condition of continuity
of material composition can never be satisfied. But surely we
would still continue to assert that a stone can be thrown (cf. the
quoted passage), or that a car can move down First Avenue. Evi-
dently the identities of such bodies do not in general depend
upon any facts about the existence or persistence of underlying
matter.
It might be questioned whether Einstein and Infeld should
have expressed their speculation in the words "The concept of
matter is rejected (i.e., there is no such thing as matter)," rather
than in the words "Matter has turned out to be very different from
what people thought." But I think this is like asking whether
it is correct to say "There is no such thing as phlogiston," rather
than "Phlogiston has turned out to be very different from what
people thought." Our concept of matter is defined by its role
in our most general theory of the underlying structure of physical
reality. If this theory changes in ways which seem very drastic
or essential than it may be natural and correct to conclude
"There is no matter."5 But even if we draw this conclusion we
will surely continue to assert that cars move down First Avenue.

4. Albert Einstein and Leopold Infeld, The Evolution of Physics (Simon and
Schuster, New York, 1961), pp. 242-43.
5. If "There is matter" simply means "There are material bodies" then the
existence of stones and cars would trivially entail the existence of matter.
That, however, is not the sense of "matter" at issue. The question is whether
such observable bodies as cars and stones are made up of some underlying
material substance which can be said to persist and change independently of
these observable bodies.
2l8 THE CONCEPT OF IDENTITY

Our judgments about the identity of a car do not in general de-

pend on our theories about the underlying structure of reality.
There may still be a question to raise about compositional
continuity as a necessary condition of identity. Suppose it is
granted that a body's ultimate composition (e.g., at the atomic
or subatomic level) can alter discontinuously without the body's
going out of existence. It might still be suggested that a body's
observable composition cannot change discontinuously. This
would imply that most of a body's observable parts at any
moment must continue to be parts of the body at the next
moment.
This suggestion is difficult to assess, for it is not clear how a
body's observable parts could alter discontinuously without there
also being a lapse of spatiotemporal and qualitative continuity.
So it is difficult to see how this suggestion could threaten the
claim that the SQ conditions suffice. I am able to think of one
possible kind of example. Suppose that we have a wicker chair
composed of observable strands of wicker. Because of moisture
in the air the strands gradually meld together. Perhaps it would
eventually be correct to say that the wicker strands have been
destroyed. This might happen (i.e., the process of destruction
might be completed) at the same moment to all the strands. So
here we would have a massive discontinuity in the chair's ob-
servable composition without there being any lapse at all in
spatiotemporal or qualitative continuity. Presumably in this
case we have little inclination to deny that it is still the same
chair. So at least in this kind of example the proposed composi-
tional condition seems not to be necessary.
I turn now to the condition of causal continuity. There are
several considerations which seem to suggest that this is not a
necessary condition of identity. Perhaps most obviously, if this
were a necessary condition of identity then it would seem to be
logically impossible for there to be such a thing as miraculous
survival. Suppose that because of some internal force a car is
about to explode, dispersing its particles to the four corners of
the Earth, but that God intervenes with a miracle and keeps
the car intact. Surely this does not strike us as a logical contra-
diction. But a miracle is by definition a suspension of causal
laws, so that presumably the car's post-miracle stages are not
causally determined by its pre-miracle stages. The causal require-
 MATTER, CAUSALITY, AND STEREOTYPES OF IDENTITY 219

merit would seem to render God impotent in effecting miraculous

survivals, which is perhaps one of His major traditional functions.
Miracles aside, there actually exists in the world a kind of
physical thing (I do not say physical body) which moves and
changes without satisfying the causal requirement. I have in
mind shadows. It is certainly correct to say that we can observe
a shadow moving across the floor or changing its shape. But the
stages of a shadow are causally independent of each other, each
shadow-stage being the effect of the body which has the shadow.
It will be said perhaps that shadows are not to be compared
to bodies with respect to their identity conditions. (It might also
be questioned whether the SQ conditions suffice even for shad-
ows in all circumstances, for example, where shadows tem-
porarily merge.) I agree that this comparison should not be
pushed too far. Still I think that reflecting on the phenomena
of shadows can help to raise a question in our minds as to
whether our thought (and experience) of identity through time
depends essentially on a causal condition.
Sydney Shoemaker has presented an example which he thinks
demonstrates that causal continuity is a necessary condition of
bodily identity.
Suppose, contrary to fact, that the following remarkable machines are
possible. The first is a table canceller; if you have set its controls to
pick out a certain location, then pushing a button on the machine will
cause any table at that location to vanish into thin air. The second is
a table producer; if you have set its controls so as to pick out a certain
location, then pushing a button on the machine will cause a table to
materialize out of thin air at that location, and the properties of that
table will depend on the setting of the machine and on nothing else.
. . . [N]ow we set the controls of the machines so that the location
picked out on both is that of my dining room table, and we push both
buttons simultaneously. Assuming that the controls of the table producer
are set to produce tables of the shape, size, and color of my present
dining room table, it will look as if nothing has happened. There will be
a spatiotemporally continuous series of table-stages, and it will appear
to the casual observer as if the same table has persisted throughout. But
knowing the powers of the machines, we know that this is not so. If t
is the time at which the buttons were pushed, then the nature of the
table-stages that occurred after t is due to the pushing of the button
on the table producer at t, and not at all due to the properties of the
table that was there before t; given that the button was pushed, we
22O THE CONCEPT OF IDENTITY

would have had such a table there after t even if there had been no
table, or a very different table, there before. It seems plain that in this
case one table has been replaced by another. 8
I think Shoemaker has drawn a too hasty conclusion. Suppose
that the way the machines work is as follows. The table canceller
disperses the subatomic particles of the table at any target loca-
tion. On the other hand the table producer operates on the
particles at a target location (or if there are insufficient particles
in that location, draws from particles at the closest available
location) and rearranges them into the specified form of a table.
What happens now if both machines are simultaneously activated
upon a certain target location which contains a table of the speci-
fied form? Presumably the table producer in effect prevents the
table canceller from dispersing the particles. So we are left after
t with a table containing exactly the same particles arranged in
(more or less) the same way. When the story is filled out in this
way is it not plain that the table after t is the same table as the
one before t? I think that no one would doubt this, even though
the causal requirement seems not to be satisfied. (The causal re-
quirement seems not to be satisfied because "the nature of the
table-stages that occurred after t is due to the pushing of the
button on the table producer at t, and riot at all due to the
properties of the table that was there before t"—Shoemaker's
characterization of the absence of causal continuity. If we are
asked to explain why there is a table with certain properties at
the target location after t then it will apparently be no part of

6. Sydney Shoemaker, "Identity, Properties, and Causality," in P. A. French,

T. E. Uehling, Jr., H. K. Wettstein, eds., Midwest Studies in Philosophy, vol.
4 (University of Minnesota Press, Minneapolis, 1979), p. 326.
There are a couple of minor problems in Shoemaker's exposition. Why does
he require both machines to be activated? Even if just the table producer is
activated it would still follow that "given that the button was pushed, we
would have had such a table there after t even if there had been no table,
or a very different table, there before." He must hold that if just the table
producer is activated then the pre-i stages would be, though not a necessary
cause, at least part of a sufficient cause for the post-/ stages, and this would
suffice to satisfy the causal requirement.
Why does he suppose that if both buttons are pushed simultaneously then
the net result would be the existence of a table at the target location? He
must be assuming that the producer machine is in some sense stronger. That
should be made explicit.
 MATTER, CAUSALITY, AND STEREOTYPES OF IDENTITY 281

this explanation to say that there was a table with certain prop-
erties there before t.)7 If this is correct then Shoemaker's kind
of example may serve very nicely to demonstrate that causal
connectedness is not a necessary condition of identity.
Shoemaker is probably thinking of another version of the
example. In the second version the table canceller annihilates
not just tables but the matter which composes them, whereas the
table producer creates tables ex nihilo. Would we say in this
second version that the table after t is not identical with the
table before tf
In the second version we may perhaps grant that the table after
t is not composed of the same matter as the table before t. Even
this conclusion is by no means straightforward, since there is no
obvious reason to assume that the best scientific theory of matter
would necessarily imply that, say, the history of an electron must
be causally connected. But perhaps we can imagine that in the
second version of Shoemaker's example our simplest theoretical
explanation of what transpired at t would include the judgment
that the matter which made up the table was replaced by differ-
ent matter. (This judgment can be made to seem more plausible
if we imagine that the original particles are replaced by particles
with somewhat different properties at somewhat different loca-
tions.) But even if we do grant that the table after t is not com-
posed of the same matter as the table before t, that could not be
our reason for denying that this is one and the same table, for
we have seen reason to think that continuity of material com-
position is not a necessary condition of identity. 8 Nor could our
reason be the lack of causal continuity, if the first version of the

7. If it is held that the causal requirement is still satisfied because the com-
position of the table after t is caused by the composition of the table before
t (I think a dubious formulation), then change the example as follows: The
table producer draws particles randomly from neighboring areas (simultane-
ously dispersing the particles in the target location if they are riot drawn),
and it happened by chance that the particles were drawn from the original
table, so that we wind up with a table composed of the same particles (ar-
ranged by chance in roughly the same way). There still seems to be no doubt
that this is the same table, though the causal requirement seems quite defi-
nitely not to be satisfied.
8. As Shoemaker himself agrees: see his "The Loose and Popular and the
Strict and Philosophical Senses of Identity," in Care and Grimm, Perception
and Personal Identity, pp. 108-9.
222 THE CONCEPT OF IDENTITY

example shows that this is not a necessary condition of identity.

But perhaps our reason could be that the disjunction of these
conditions is necessary for identity; perhaps there must be either
compositional continuity or causal continuity.
Let me not exaggerate my opposition to this suggestion; it. may
be correct. But I am not convinced. Even in the second version
of Shoemaker's example my intuition is that we perhaps can
treat the tables as identical. The most that I could unreservedly
concede is that this is a borderline case. Perhaps it would be
appropriate in this case to say, "In a sense it's the same table and
in a sense it isn't." It does not seem to me clearly unacceptable
to say, "By an incredible fluke the table survived, because both
buttons were pushed at the same time." If someone claims that
his intuition says that this clearly is unacceptable I would have
to wonder whether he has properly distinguished between the
question "Is it the same table?" and the question "Is it the same
matter?"9
I am aware that my intuition about this example may not be
universally shared. And I will eventually offer a kind of com-
promise account which may accommodate differing intuitions
about such a case. For the moment, however, I will simply con-
clude that the SQ conditions may suffice at least up to borderline
cases.
Let us now inquire what other combinations of the continu-
ities might suffice for identity. First of all we may wonder whether
spatiotemporal continuity by itself (of course relativized to the
sortal) might suffice even without qualitative continuity. If we
have a spatiotemporally continuous succession of car-stages which
suffers some massive qualitative discontinuity would we perhaps
still judge it to be the same car? I will not attempt to settle this
question.10 The possibility remains open then that the SQ con-
ditions can be simplified to the single condition of spatiotemporal
continuity.
On the other hand it is clear that qualitative continuity by
9. I wonder if Shoemaker was really trying to show that, the SQ conditions
are not sufficient for the identity of matter. But. that point can be conclu-
sively established without appealing to any science fiction examples: see
above, pp. 113-19.
10. For a discussion of this question see Quinton, The Nature of Things,
pp. Gy-tw).
 MATTER, CAUSALITY, AND STEREOTYPES OF IDENTITY 22g

itself does not suffice for identity. We might have two cars Cr
and C2 which are qualitatively indistinguishable. A succession of
car-stages which combines stages of Ct with stages of C2 would
be qualitatively continuous but would obviously not correspond
to a single car.
It might be suggested that this is because the hybrid succession
must compete with both of the successions corresponding to C,
and C2, and these win out in virtue of having other continuities
in addition to the qualitative one. But suppose that C^ was
destroyed before C 2 was created (out of different material). Then
we could prolong the Cj-succession in a qualitatively continuous
manner by combining it with the C^-succession. Obviously we do
not do this; we do not identify the car that is destroyed with the
car that is later created, even though there is no other identifica-
tion competing with this one. So it is clear that qualitative con-
tinuity by itself amounts to nothing as a sufficient condition of
identity.
It seems equally clear that causal continuity does not in general
suffice for identity. We can imagine a machine which creates an
exact copy of a given car out of new material, destroying the
original car in the process. Here there may be the most intimate
causal connection between the final stages of the original and the
initial stages of the copy. Still there is no inclination to identify
the two cars. In this example there was both causal and qualita-
tive continuity; so we see that the conjunction of these two con-
ditions does not suffice for identity.
What about compositional continuity? I think a case can be
made for saying that this condition, or some close variant of it,
does suffice for identity. A car can be taken apart and put back
together again, retaining its identity through a lapse of spatio-
temporal continuity. In this case we can appeal to continuity of
observable composition. But it seems plausible that the more
general condition is continuity of material composition, whether
observable or not. In a famous science fiction example (from the
TV show "Star Trek") bodies are "beamed" aboard a spaceship.
The observable facts are that some apparatus on the spaceship
is in some manner set to the location of the body, whereupon
the body vanishes into thin air, and a short while later a similar
body materializes on the spaceship. I think that our understand-
ing of this example is probably premised on the assumption that
224 THE CONCEPT OF IDENTITY

compositional continuity is satisfied, though at a nonobservable

level. We can conceive of the body that vanishes as being iden-
tical with the body that later materializes, because we assume
that (most of) the particles of the first body somehow wind up
on the spaceship, where they are again arranged into the form of
that body.
In these examples it is perhaps also required that the matter
which made up the original body is later arranged in the recom-
posed body in pretty much the same way as in the original. This
might incidentally assure some fairly high degree of qualitative
continuity. Furthermore in these examples, and other typical
examples, compositional continuity is accompanied by at least
some degree of causal continuity: typically there will be some
causal connection between the properties of the body prior to
decomposition and its properties after recomposition. So per-
haps these additional factors should be understood as implicit in
the condition of compositional continuity.11
The compositional condition complicates our concept of bodily
identity in two very important ways. First of all, it relates the
identities of such observable bodies as cars, and stones, and trees
to theoretical speculations about the underlying reality. This is
why our concept of the identity of a car can in principle ac-
commodate such exotic possibilities as a car disappearing in one
place and reappearing in another place (cf. the "Star Trek" ex-
ample). There is an important sense in which our concept of the
identity of matter is indefinable; it is doubtful that there are any
definite a priori constraints on the identity of matter. 32 The
compositional condition thus invests even the identity of a car
with a derivative dimension of indefinability.
The second important complication is that the compositional
condition can generate conflict cases, i.e., cases in which the con-
dition directs us in contradictory ways. The most famous ex-
ample of this sort is the case of the ship of Theseus.13 In that
case the compositional condition would permit us to make the
judgment that the original ship suffered a continuous and total
11. The condition of compositional continuity might also have to be com-
plicated by reference to the importance of the parts that are preserved in a
compositional change. See above, p. 66.
12. Cf. above, Chapter 4.
13. See above, pp. 68-71.
 MATTER, CAUSALITY, AND STEREOTYPES OF IDENTITY 225

replacement of its material, and also to make the contradictory

judgment that the original ship was later reconstituted by its
original material. Perhaps we might favor the first judgment on
the grounds that this is bolstered by the SQ conditions as well.
(Note how the compositional condition can conflict both with
itself and with the SQ conditions.)14 But I am inclined to doubt
that there is a definitively "correct" answer to this question. We
can say that the original ship is in a sense identical with one
ship, and in a sense identical with the other ship. So in this
kind of case we seem to have at worst only a borderline counter-
example to the SQ principle.
But I cannot deny the possibility of constructing some exotic
conflict examples in which the pull of the compositional condi-
tion might strike many as decisively stronger than that of the
SQ conditions. One such example might be developed along the
following lines. Assuming that our theory of matter might co-
herently allow for the discontinuous motion of matter (which
I am inclined to think is possible), then we can imagine a situa-
tion in which we are able to predict how matter will jump dis-
continuously from one place to another. In this situation it might
be commonplace to rely on the compositional condition to pre-
dict how standard objects like cars jump discontinuously from
place to place. Suppose now that in this situation there are two
very similar cars, and we are able to predict that the matter
which makes up one car will jump discontinuously to the place
of the second, while at the same instant the matter which makes
up the second car jumps discontinuously to the place of the first
(both portions of matter maintaining their forms as cars during
this process of displacement). Here the compositional condition
dictates the judgment that these cars retain their material com-
position while instantaneously exchanging places, whereas the SQ
principle implies that the cars stay where they are while in-

14. It is sometimes supposed that even the SQ conditions can give rise to
conflict cases if, e.g., a car splits into two like an amoeba. But that is an
error. If a car splits into two there will be a moment during the process of
splitting when we have a monster object, followed at a later moment by the
two resultant cars. First of all it is not clear that the monster object qualifies
as a "car," so there may not even be sortal coverage. And certainly the jump
from the monster object to each resultant car will not be spatiotemporally
or qualitatively continuous.
226 THE CONCEPT OF IDENTITY

stantaneously exchanging their matter. In this kind of example

I think that many people (if indeed they can accept the premise
of the example) might feel that the compositional condition de-
cisively predominates. If such an example can qualify as clearcut
(i.e., nonborderline) then I shall have to relinquish my commit-
ment to the SQ principle to that extent.
And there are other important complications that I pass over,
such as the possibility of taking into account degrees of SQ or
compositional or causal continuity in attempting to resolve a con-
flict case. In general I have no very firm position about these
conflict cases, except to suggest that by and large they seem to me
to be left indeterminate by our ordinary identity concept.
The discussion up to this point suggests the following analysis
of the identity of a car:
Analysis A. A succession of car-stages constitutes a persisting
car if and only if either it satisfies the SQ conditions or it
satisfies the condition of compositional continuity.
This analysis, as well as those which follow, should be under-
stood as containing the implicit proviso that (with the reservations
noted) a case is indeterminate if the specified conditions yield
conflicting judgments.
(A possible emendation of Analysis A, which I have been only
mildly resisting, is to add the condition of causal continuity to
the first disjunct, so that the two sufficient conditions of identity
will be (i) SQ continuity together with causal continuity, and
(2) compositional continuity. I resist this to the extent of sug-
gesting that the emended and unemended versions may differ
only in borderline cases.)15
15. Analysis A corresponds substantially to the position adopted in Part One;
on the possible emendation see Chapter 4, fin. 14.
It should be noted that even given the emendation conflict cases of the sorts
I have mentioned still arise. The example of the ship of Theseus remains
unaffected. And in the more exotic example last considered we can easily
imagine that the cars interact in such a manner that some properties of both
of them after the "jump" causally depend on some previous properties of
both of them. (Think of how the temperatures of two objects in contact
causally depend on the prior temperatures of both of them.) Indeed we can
even imagine that some relationship between the cars causes the instantaneous
displacement of their matter. In this case the condition of causal continuity
 MATTER, CAUSALITY, AND STEREOTYPES OF IDENTITY 227

III. Stereotypes of Identity

I now want to consider a somewhat different approach to these
issues. In a number of publications Hilary Putnam has presented
a distinctive pattern of analysis which seems to apply to a wide
range of concepts.16 It is worth considering how Putnam's kind
of analysis might apply to the concept of identity through time.
Putnam's basic idea is that often the concept of F is associated
with a stereotype of F, which fixes the reference to a hidden
structure. The stereotype is defined by some superficially ob-
servable characteristics, and the concept refers to the hidden
structure via the stereotype.
The term "lemon," for example, is associated with a stereo-
type consisting of a certain characteristic shape, color, texture,
taste, and perhaps observable origin and growth. But the concept
of a lemon is not to be equated with the stereotype of a lemon.
The connection between the concept and the stereotype is rather
to be understood along the following lines:
x is a lemon if and only if x has that (rigidly designated)
underlying structure distinctive of most local cases satisfying
the lemon-stereotype.
Suppose that S represents the relevant underlying structure com-
mon to most of the stereotypical lemons that human beings might
have encountered. Then the above definition implies that some-
thing is a lemon if arid only if it has that structure S.
This account has two kinds of interesting consequences, one
about the actual world and another about possible worlds. As
regards the actual world, there may be objects that are lemons
even though they do not satisfy the stereotype, and objects which
are not lemons even though they do satisfy the stereotype. It all
depends on whether the object has the hidden structure S, i.e.,

would be satisfied by the judgment that accords with the SQ principle, viz.
the judgment that the cars stood still and exchatiged their matter. We would
then have both SQ and causal continuity pitted against compositional (and
causal) continuity; many will still feel that compositional continuity wins.
16. Hilary Putnam, "Is Semantics Possible?" "Explanation and Reference,"
and "The Meaning of 'Meaning,' " all reprinted in Mind, Language and
Reality.
228 THE CONCEPT OF IDENTITY

the structure which is common to most of the objects around here

which satisfy the stereotype.
Because "that hidden structure" in the definition is taken
rigidly, when we consider whether an object in some counter-
factual situation is a lemon we have to ask whether that object
has the structure S, i.e., the structure which in the actual world
corresponds to the stereotype. If there would have been a situa-
tion in which all of the objects which satisfy the lemon-stereotype
have the hidden structure S' rather than S, then none of those
objects would have been lemons (though they would have had
all of the superficial properties of lemons).
Thus according to this account the statement "Most of the
objects around here which satisfy the lemon-stereotype are lem-
ons" is a conceptual truth, i.e., it follows from the definition of
"lemon."17 But the statement is not metaphysically necessary
since, as we just saw, there is a possible situation in which it is
false. The statement is, in Kripke's terms, an a priori contingency.
On the other hand if S turns out to be the structure shared
by most of the local stereotypical lemons, then the statement
"Any lemon has S" is metaphysically necessary, for in any pos-
sible world whether something is a lemon depends on its having
S. But the statement is not a conceptual truth: it does not follow
from the concept of a lemon that S is the related structure. The
statement, if true, is a posteriori necessary.
Let us now see what happens if we try to apply this pattern
of analysis to our concept of the identity of a car. The first point
I would suggest is that we take the SQ conditions (relativized to
the sortal "car") as defining the stereotype of car-identity. The
SQ conditions seem to provide us with our basic picture of the
persistence of a car. The difficult question is how exactly to re-
late the stereotype, the picture, to the general concept. If we try
to conform as closely as possible to the preceding analysis of
"lemon" we wind up with this:
Analysis B: A succession of car-stages constitutes a persisting
car if and only if it has that (rigidly designated) underlying
17. There are several complications here which I ignore. One rather obvious
question is what we should say if it turns out that there is no underlying
structure (nor even some few underlying structures) corresponding to the
stereotype. Putnam suggests that in this case the concept is simply equivalent
to the stereotype; see "The Meaning of 'Meaning,'" pp. 240-41.
 MATTER, CAUSALITY, AND STEREOTYPES OF IDENTITY 22g

structure distinctive of most local cases satisfying the stereotype

of car-identity (where the stereotype is denned by the SQ
conditions).
What should we say this underlying structure is? Assuming
some standard notions of physics (and ignoring for the moment
the kind of comment made by Einstein and Infeld), it seems
plausible to identify this underlying structure with continuity of
material composition. That is, in the stereotypical case of car-
identity we have a succession of car-stages, each stage consisting
at the underlying level of a swarm of particles of matter, where
closely neighboring stages contain almost the same particles ar-
ranged in almost the same way. If we assume that compositional
continuity is the relevant underlying structure of the stereotypi-
cal cases, then Analysis B implies that a given succession of car-
stages constitutes a persisting car if and only if it satisfies the
compositional condition.
How does this compare to Analysis A? In that analysis there
were two independently sufficient conditions of identity, the SQ
conditions and the compositional condition. On Analysis B the
story is more complicated and in many ways more interesting
(though not perhaps more accurate).
According to Analysis A it is a conceptual truth that any suc-
cession of car-stages which satisfies the SQ conditions constitutes
a persisting car. This is not a conceptual truth according to
Analysis B. What is a conceptual truth according to the latter
analysis is that most local successions which satisfy the SQ con-
ditions constitute persisting cars.18 Analysis B allows for the
exceptional case in which, though the stereotypical SQ condi-
tions are satisfied, the relevant underlying structure is absent,
and hence the case cannot qualify as one of identity.
According to Analysis A the statement "Any succession of car-
stages which satisfies the SQ conditions constitutes a persisting
car" is not only a conceptual truth but also metaphysically neces-
sary: the statement holds in any counterfactual situation. This
unqualified statement is not, as we just saw, a conceptual truth on
Analysis B; nor obviously is it metaphysically necessary. But even
18. Here again we would require the stipulation mentioned in £tn. 17, viz.
that if it should turn out that there is no relevant hidden structure the
concept of persistence collapses into the stereotype.
230 THE CONCEPT OF IDENTITY

the qualified statement "Most local successions of car-stages which

satisfy the SQ conditions constitute persisting cars" is not neces-
sary on Analysis B, though it is a conceptual truth; the analysis
implies that this statement is a priori contingent. For imagine a
counterfactual situation in which the cases which satisfy the SQ
conditions do not satisfy the condition of compositional con-
tinuity. Assuming that compositional continuity is the under-
lying structure related to the stereotypical cases in the actual
world, none of those counterfactual cases satisfying the SQ con-
ditions would qualify as cases of identity.
So two ways in which Analysis B departs from Analysis A is
that according to the former analysis only the qualified statement
is a priori, and even this statement is not metaphysically neces-
sary. And there is a third difference. According to Analysis A it
is both a priori and necessary that a case of compositional con-
tinuity is a case of identity. According to Analysis B this is not
a priori; for it is not a conceptual truth that compositional
continuity is the hidden structure related to the stereotypical
cases. It may yet turn out that some other structure (e.g., of the
sort discussed by Einstein and Infeld) has that status. But if
compositional continuity is in fact the relevant structure then it is
metaphysically necessary that all (and only) cases of composi-
tional continuity are cases of identity. According to Analysis B
the connection between compositional continuity and identity is
a posteriori necessary.
It should be clear that Analysis B does not have the implau-
sible consequence that our ordinary judgments about bodily
identity must in general depend upon some specific theory about
the underlying reality, in general we can base ourselves on the
conceptual truth that at least most of the stereotypical cases we
encounter are cases of identity. The analysis only requires that
our concept of identity should incorporate the vague idea of
there being a distinctive kind of hidden structure related to the
stereotype. Detailed theories about the nature of that structure
(e.g., theories about the persistence of matter) only come later,
after we have built up sufficient knowledge of the stereotypical
cases.
It is evident that Analysis B cannot sustain all of the intuitions
that I tried to promote earlier. The analysis will depart from
 MATTER, CAUSALITY, AND STEREOTYPES OF IDENTITY 2gl

these intuitions with respect to cases in which the SQ conditions

are satisfied but the underlying structure is atypical. The second
version of Shoemaker's example was such a case. Here the SQ
conditions were satisfied but the underlying structure was as-
sumed to involve a wholly atypical lapse of compositional (and
causal) continuity. Analysis B would decisively disqualify a judg-
ment of identity in this case, whereas my intuition is that such
a judgment would at least be borderline correct.19
Let me therefore suggest an emendation of Analysis B, still
somewhat in the spirit of Putnam's approach, but more in line
with my intuitions:
Analysis C: A succession of car-stages constitutes a persisting
car if and only if either it satisfies the stereotype of car-identity
or it has that (rigidly designated) underlying structure distinc-
tive of most local cases satisfying the stereotype (where the
stereotype is defined in terms of the SQ conditions).
Analysis C implies that it is both a priori and necessary that
any case which satisfies the SQ conditions is a case of identity
(with the exception of conflict cases which remain indeterminate).
In this respect it is just like Analysis A. The difference between
the two analyses concerns the second condition. In Analysis A
this condition is specified as compositional continuity, whereas
in Analysis C the condition is specified to be whatever is the hid-
den structure generally characteristic of the stereotypical cases.
It cannot be determined a priori what the nature of that hidden
structure is. If this structure is compositional continuity then it
will follow from Analysis C that it is metaphysically necessary
that any case of compositional continuity is a case of identity.
But this is only an a posteriori truth.
The difference between Analysis A and Analysis C is perhaps
not very great. But Analysis C, in contrast to Analysis A, invests
our identity concept with a kind of open-endedness or indefinite-
ness which may seem intuitively appealing. Suppose it turns out
that the theoretically correct description of the underlying struc-
19. Note that Analysis B would not even allow it to be a conceptual truth
that the presence of the four continuities suffices for identity (since these
continuities may not typify the stereotypical cases), a consequence which I
think many (including Shoemaker) would find unacceptable.
232 THE CONCEPT OF IDENTITY

ture of cars and other observable bodies is not in terms of per-

sisting matter, but is rather in terms of some field structures (as
in the Einstein-Infeld remark), or perhaps in terms as yet not
conceived of by anyone. The second condition of Analysis C
would enable our identity concept to absorb these new theoreti-
cal structures, whatever they might turn out to be. Whether the
structure corresponding to the stereotype is continuity of ma-
terial composition, or some facts about fields, or some facts as
yet undreamed of, that structure is according to Analysis C a
sufficient condition of bodily identity.
The element of indefiniteness in Analysis C can be brought
out by considering again the science fiction fantasy in which
bodies are "beamed" aboard spaceships. I suggested earlier that
we understand this fantasy by conceiving that the particles which
make up the beamed body travel (perhaps discontinuously) to the
spaceship. And I think it is indeed plausible to suppose that most
people would conceive of the case in that way. But what about
a small child, or even an extremely uneducated adult, who may
have no clear idea, or perhaps no idea at all, of there being any
such things as invisible particles which make up a body? Surely
it would seem that such a person could still understand the
fantasy. Analysis C can account for this fact. All that is required
for understanding the fantasy is that one should vaguely conceive
of there being some kind of underlying reality characteristic of
the stereotypical cases, which is in some manner present in the
beaming case. It is unnecessary that one should have any definite
ideas about what that underlying reality is like.
I think our ordinary concept of the identity of a car might
plausibly be placed somewhere between Analysis B and Analysis
C. The latter analysis is more elementary, less theoretical, than
the former; for in the latter analysis the superficially observable
SQ conditions, which define the stereotype, suffice for identity,
whereas in the former analysis only the associated hidden struc-
ture suffices. I think our ordinary concept of the identity of a car
may be said to vacillate vaguely between the relatively ele-
mentary sense defined by Analysis C and the relatively theoretical
sense defined by Analysis B. And this kind of vagueness may
characterize many of our concepts. For example, our concept of
a lemon may vacillate between a more elementary sense in which
the presence of the stereotypical properties suffices and the more
 MATTER, CAUSALITY, AND STEREOTYPES OF IDENTITY 233

theoretical sense defined earlier, in which only the presence of

the associated hidden structure suffices,20
If this is correct then it follows that the SQ conditions suffice
at least for typical cases. Where the SQ conditions are satisfied
but the relevant hidden structure is absent, then we have a case
for which the two competing senses, the more elementary and the
more theoretical, yield different judgments, and perhaps either
judgment is acceptable.
We can say something, I think, about the kinds of conversa-
tional contexts which are likely to encourage someone to employ
the more elementary sense, or the more theoretical sense. If a
philosopher presents us with a complicated example, and asks us
to consider what identity judgment is appropriate in that ex-
ample, then since we know that philosophy is a serious business
that aims at deep truths, in considering the example we are
naturally going to be propelled in the direction of the more
theoretical sense approximating to Analysis B. Indeed the solem-
nity of the philosopher's question "Is x really identical with ji?"
has often propelled a response at a level of theoreticalness much
beyond that of analysis B. I have in mind those philosophers
(such as Reid, and Butler, and Chisholm)21 who have maintained
that "strictly speaking" bodily identity requires identity of mat-
ter, so that, for example, a car cannot survive the loss of a hub-
cap. This response, I would say, simply replaces our ordinary
concept of the identity of observable bodies, like cars and tables
and trees, with the quite different concept of the identity of the
matter which makes up these bodies. On the other hand if we
respond to a philosopher's identity-question by employing the
relatively theoretical sense of Analysis B, then I think we are
still within the vague bounds of the ordinary concept. We are
merely pushing the concept in a certain direction, in the theo-
20. It may be possible to combine Putnam's relatively theoretical definition
of "lemon" with the elementary definition of "the persistence of a lemon."
If S represents the hidden structure associated with the lemon-stereotype then
Putnam's definition implies that a succession of lemon-stages must be a
succession of stages of bodies with the structure S. We can still say, in accord-
ance with Analysis C, that a sufficient condition for a succession of lemon-
stages to constitute a persisting lemon is that it satisfy the SQ conditions.
There are to be sure complications here, which could be worked out in
different ways.
21. Cf. above, pp. 161-62.
 THE
234 CONCEPT OF IDENTITY

retical direction induced by the solemnity of the question. That

may be alright so long as we do not conclude that there is some-
thing wrong with employing our identity concept in the more
elementary sense of Analysis C. Indeed in the most typical con-
versational contexts, which tend to be casual and theoretically
carefree, the more elementary sense may well predominate.
If philosophizing has a tendency to propell us toward the more
theoretical sense of bodily identity then obviously scientific theo-
rizing will have the same effect. A biologist who theorizes about
the persistence and change of a tree will perhaps be likely to
employ a concept of the identity of a tree corresponding to Anal-
ysis B (with the sortal "tree" replacing the sortal "car"). Again I
would maintain that there are more casual levels of discourse
about trees for which the more elementary identity concept may
legitimately predominate.
It would of course be possible to define an identity concept
even more elementary than that corresponding to Analysis C.
The most obvious possibility is a concept which is simply equiva-
lent to the stereotype, a concept, that is, for which the SQ condi-
tions are not only sufficient but also necessary. A somewhat more
complicated possibility would add the sufficient condition of con-
tinuity of observable composition, but still with no regard at all
for the underlying reality even as a sufficient condition. It may be
that these concepts can be said to operate at some very primitive
levels of thought; but it seems fairly clear that they are not domi-
nant at any ordinary level (though for all practical purposes
there is a quite negligible difference between either of these con-
cepts, especially the latter, and the ordinary concept which ac-
knowledges the underlying reality). A more important point is
that the discussion of this whole chapter has assumed a concept
of identity relativized to such sortals as "car," "table," and "tree."
However, the basic core of our identity concept can be under-
stood without appeal to such sortals; at the most basic level the
condition of minimizing change takes the place of sortal cover-
age.22 In terms of the notion of a stereotype we might say that
our most basic stereotype of bodily identity is the picture of a
succession of body-stages which satisfies the SQ conditions while
minimizing change. At a level of knowledge which allows for the

22. See above, Chapter 3.

 MATTER, CAUSALITY, AND STEREOTYPES OF IDENTITY 235

application of sortals, however, this most basic stereotype gives

way to the more refined sortal-relativized stereotypes.
To summarize: I have considered in this chapter three closely
related analyses of our concept of bodily identity. One may
regard these as providing competing accounts, or, I think more
plausibly (at least for the last two analyses), as characterizing
different senses of identity, each with some legitimate claim to
ordinary usage. In any case the point which emerges clearly, and
which is common to all of these analyses, is that there is the most
intimate connection between our concept of the identity of a
specified sort of body and the idea of a spatiotemporally and quali-
tatively continuous succession of body-stages of that sort. It is
implicit in the identity concept that the presence of a succession
satisfying those conditions constitutes at least a prima facie basis,
and perhaps even a conclusive basis, for making a judgment of
identity.
 8
A Sense of Unity

PHILOSOPHERS HAVE often raised questions about our con-

cept of the unity of a thing. Most typically what is sought is
an analysis of what our concept of unity consists in. The answer
to this question commonly takes the form of citing various con-
ditions that seem to provide a definition of our judgments of
unity. These conditions may be said to constitute our criteria
of unity, our criteria of identity.
The question I want to raise in the present chapter is some-
what different from this typical one. Suppose that we have al-
ready ascertained what our criteria of unity are. Then I want
to ask why it is that we employ just those criteria rather than
others. What determines us to base our judgments of unity on
just those conditions?

I. Criteria of Unity
Let me present an example to illustrate the difference between
these two kinds of questions: the one I am asking and the more
typical one. Suppose that you have a tree in your backyard and
that next to the tree there stands a table. Common sense would
judge that the tree is a single object and the table is a single
object. Each of these objects is of course composite; the tree, for
example, is composed of a trunk, some branches, twigs, leaves,
and so on. Now something that common sense would definitely
not judge is that there is a single object that is composed of the
tree together with the table. If I am, say, touching the tree and

236
 A SENSE OF UNITY 237

you are touching the table, common sense would not say that
there is some single object that you and I are both touching.
But why not? Why should we not say that the tree and the table
add up to a unitary object?
At one level the answer to this question would consist in citing
relevant criteria of unity through space, i.e., criteria that deter-
mine whether or not an aggregate of matter can properly be said
to add up to a single thing. In the case under discussion two
relevant criteria would seem to be spatial connectedness and
dynamic cohesiveness. Generally an object must be spatially
connected, in the sense that any of its parts can be connected
by a continuous curve whose points all touch the object. And
generally an object must be cohesive, in the sense that all its
parts tend to remain together under various pressures. I do not
mean to suggest that these two conditions (connectedness and
cohesiveness) are strictly necessary for an object's unity in all
imaginable circumstances; nor am I suggesting that they are
sufficient for unity. But these conditions are pretty likely to fig-
ure in any general analysis of an object's unity through space,
and with respect to our simple example they seem enough to
rule out the tree-cwm-table as a unitary object.1
At one level, then, we can explain by appealing to criteria why
the tree and the table do not add up to one object. The question
I want to raise, however, is why these criteria function the way
they do. What is it that induces common sense to base a judg-
ment of unity on the particular conditions of connectedness and
cohesiveness? Why allow those conditions to dictate the matter?
Why does common sense not choose some other criteria of unity,
criteria that might allow for the judgment that the tree and the
table compose a single object (where this object happens to be
disconconnected and noncohesive)?
Let me extend this example a bit, so as to bring identity
through time into play. The tree is not just spatially composite;
it is also, in a sense, temporally composite. Insofar as the tree
persists through time, it (or its history) can be thought of as
comprising a succession of temporary stages, where these stages
can be delimited in any number of ways. (For example, we can

i. On our criteria o£ unity through space, see above, Chapter 3, Sections V

and VI.
238 THE CONCEPT OF IDENTITY

think of the tree as combining an early stage in which it is

short followed by a later stage in which it is tall, or as combining
stages in which it is in bloom with stages in which it is not in
bloom.) And the same can be said for the table; this too is tem-
porally composite. If, however, we were to combine in thought
some early stage of the table with some later stage of the tree
we would not, at least by the lights of common sense, arrive at a
unitary persisting thing. In the case I am imagining, where a
tree and a table are situated together in a normal way, com-
mon sense could not even take seriously the idea that some single
persisting thing is first a table and then a tree. But why not?
Why should we not judge in this case that there is a single per-
sisting object that combines a table-stage and a tree-stage?
Again, the answer at one level consists in citing criteria, in
this case criteria of unity through time. Two criteria that seem
to suffice for the case (though they do not suffice for all cases)
are qualitative continuity and spatioternporal continuity.2 If
we tried to think of there being a single object that is first a
table and then a tree we should have to say that this object
changed discontinuously, as regards both its qualities and its
location. Our criteria of unity through time do not (in general)
allow us to say this.
And again, my question is: What induces common sense to
credit those particular criteria of unity through time? Why not
choose other criteria which might accommodate the judgment
that a table changed discontinuously into a tree?
There are philosophers, notably W. V. Quine, who in fact rec-
ommend a revision in our commonsense notion of an object
which would have precisely the effect of accommodating the
judgments that I have just instanced as conflicting with our or-
dinary criteria of unity. In terms of Quine's revised concept of an
object we would indeed say, in the imagined example, that there
is at a given moment some object that is composed of the table
and the tree, and that there is over a period of time an object
that is first a table and then a tree. On Quine's proposal an
object "comprises simply the content, however heterogeneous, of
some portion of space-time, however disconnected and gerry-

2. Cf. above, Chapters 1-3, passim, and Chapter 7.

 A SENSE OF UNITY 2gQ
3
mandered." Any space-time portion of reality qualifies as an
object, in Quine's terms. But this technical notion of an object
is crucially different from the ordinary notion, as Quine himself
amply stresses. In terms of the ordinary notion only a select few
space-time portions qualify as objects, namely, those which sat-
isfy our criteria of unity. It is the ordinary notion that concerns
me. Quine often marks off the ordinary notion from his technical
one by using the word "body" for the ordinary notion. Hence
he says: "Man is a body-minded animal."4 In these terms what
I am asking for is an explanation of why common sense is body-
minded.
What is at stake in this question is not merely the use of two
or three words (such as "body," "object," "thing"), but a whole
way of thinking. Exactly how to characterize that way of think-
ing is itself an essential part of the philosophical difficulty. But
one can say, to begin with, that the category of a body seems to
constitute for common sense the primary way of breaking up the
world into units. And this category is denned in terms of various
specific and complicated criteria of unity, for example, spatial
connectedness, dynamic cohesiveness, spatiotemporal and quali-
tative continuity. The question, then, is why common sense
should divide reality up in just that particular way.

II. Unity and Similarity

A possible answer to this question, which I want to consider
and defend, is that we think of the world in terms of our criteria
of bodily unity because we are innately disposed to think in this
way. According to this hypothesis, a sense of bodily unity is part
of our inborn constitution, and this is what determines us to
interpret our experience in the way we do.
This hypothesis has something in common with Kant's view
about the a priori category of substance. I want to stress, how-
ever, two differences between the hypothesis under consideration
and Kant's view. First of all, Kant had little, if anything, to say
about specific criteria of bodily unity. In fact Kant's category

3. Quine, Word and Object, p. 171.

4. Quine, The Roots of Reference, p. 54.
240 THE CONCEPT OF IDENTITY

of substance is not equivalent to the commonsense idea of a

body. Ordinary bodies, like trees and tables, are created and de-
stroyed, but Kant's idea of substance referred to the underlying
matter that was supposed to persist forever. The second difference
is that Kant maintained that his category of substance is a neces-
sary ingredient of understanding. This necessity claim is no part
of my hypothesis. My hypothesis claims only that, as a matter
of contingent fact, human beings are innately disposed to inter-
pret their experience in a certain way.
This hypothesis is more closely related to some of Chomsky's
ideas about innate grammatical schemata. And it is even closer to
the views expressed by gestalt psychologists like Wolfgang Kohler,
who have maintained that, as a matter of empirical fact, our
sensory fields are "naturally" and "spontaneously" organized in
terms of distinctive kinds of units. 5
I want to broach this idea by way of an analogy. I want to
compare the idea of an innate sense of unity to the more familiar
philosophical idea of an innate sense of similarity. It has been
persuasively argued by Quine, and also by Anthony Quinton,
that our grasp of general concepts must ultimately be rooted in
an innate tendency to classify objects in certain definite ways.6
In order for a child to acquire the use of a general term, he
must be able to extrapolate from observed cases of the term's
application to new cases. This extrapolation evidently requires
that the child have some basis for deciding which new cases go
together with the observed cases. At least with respect to the
most elementary vocabulary, the basis for this decision would
apparently have to be innate. The idea here is not that our fully
developed scheme of classifications depends on nothing but our
primitive classificatory impulses. Perhaps the scheme is eventually
affected by various practical and theoretical needs. At bottom,
however, there must be the innate tendency to classify things
in certain ways rather than others.
Quine sometimes refers to this innate tendency as an innate
"sense of similarity"; sometimes he refers to it as an innate
5. Kohler, Gestalt Psychology, chaps. 5 and 6.
6. See W. V. Quine, "Natural Kinds," in Ontological Relativity and Other
Essays (Columbia University Press, N.Y., 1969), p. n6ff; Word and Object,
p. Sgff.; The Roots of Reference, p. 19; and Quinton, The Nature of Things,
pp. 261-65.
 A SENSE OF UNITY 241

"quality space." It should be borne in mind that the first expres-

sion ("sense of similarity") is not meant to imply any special
views about the possibility of reducing properties to similarity
relations; and the expression "quality space" applies not just to
qualities properly speaking, since all properties, including rela-
tional properties, would have to be treated in the same manner.
The general point is simply that we are innately disposed to
classify in certain ways rather than others.
Now for common sense the most basic thought about physical
reality is the thought that some specified body has some specified
property. The first ingredient of this thought (the specification
of a body) is linked to our criteria of bodily unity, and the sec-
ond ingredient (the specification of a property) is linked to our
principles for classifying bodies. We have just seen that, accord-
ing to Quine and others, the classificatory ingredient is rooted in
the innate disposition to classify in distinctive ways. One can
scarcely resist the speculation that perhaps the other ingredient,
that related to our criteria of unity, is likewise rooted in the
innate disposition to adopt certain criteria of unity rather than
others. The general scheme we then wind up with is this: As our
innate sense of similarity stands to our principles for classifying
bodies, so does our innate sense of unity stand to our criteria of
unity for bodies.
It will be instructive to try to make out what Quine's reaction
might be to this proposal. Some of his remarks may certainly
seem to suggest that he too believes in an innate sense of unity.
He says that "body-unifying considerations, though complex, are
rooted in instinct,"7 and he refers to our "instinctive body-
mindedness."8 But this is pu/zling, since the view standardly
attributed to Quine is that, besides such obvious general facul-
ties as perception, intelligence, and motor behavior, quality space
is essentially the only innate endowment that can confidently be
related to the process of learning language. Not that Quine is
at all adamant about this; he seems quite open to other possibil-
ities, even perhaps to some of Chomsky's suggestions.9 But it would
certainly be extremely odd to attribute to Quine the unheralded
7. Quine, The Roots of Reference, p. 55.
8. Ibid., p. 56.
9. See Quine's "Philosophical Progress in Language Theory," in H. E. Kiefer
and M. K. Munitz, ed., Language, Belief, and Metaphysics, p. 6.
242 THE CONCEPT OF IDENTITY

position that working side by side with the language-learner's

innate quality space is the quite distinct disposition to adopt
certain identity criteria. That is in fact the position that I want
to maintain, but I have to doubt that this is Quine's position.
Actually if we look more closely at that section in The Roots
of Reference from which I previously quoted, we find that when
Quine refers to our "instinctive body-mindedness" he probably
does not mean to introduce an innate disposition distinct from
our sense of similarity. Rather he seems to be suggesting that
our body-mindeclness is itself the result of our innate sense of
similarity.
Thanks to [the child's] instinctive body-mindedness, he is an apt pupil
when the general terms are terms for bodies. He is able to appreciate
not only that the second-order similarity of a dog to a dog exceeds that
of a dog to a rabbit, but also that the latter in turn exceeds that of a
dog to an apple or buckle. . . . And then there is the yet slighter degree
of second-order similarity, residing in just those very general body-
unifying considerations that preserve the identity of each dog, each
rabbit, each apple, each buckle, in short each body. This would be a
second-order similarity basis for the child's ostensivc learning of the
general term "body" itself, or "thing," to take the likelier word.10
What Quine seems to be saying here is that our "instinctive
body-mindedness" is actually nothing more than our disposition
to appreciate the complicated similarity relations that obtain be-
tween those space-time portions of reality which we count as
bodies.
But there is something wrong here. To operate with the ordi-
nary concept of bodily unity is not just a matter of appreciating
various similarities between those portions of reality which qual-
ify as bodies. Imagine someone who did not operate with the
ordinary concept, but who operated instead with that technical
notion of an object which, as I mentioned earlier, Quine ulti-
mately favors over the ordinary notion. Someone who operated
with this revised concept would be treating all portions of reality,
whether disconnected or whatever, as units on a logically equal
footing. But certainly he might very well appreciate the relevant
similarities between those select portions of reality which com-
mon sense dignifies as bodies. To be body-minded, in the way
10. Quine, The Roots of Reference, p. 56.
 A SENSE OF UNITY 343

that common sense is body-minded, is to adopt an ontology that

excludes all of those portions of reality which do not qualify as
bodies. Common sense simply does not credit such portions of
reality. Our commonsense adoption of this exclusionary ontology
cannot be regarded as merely a corollary of our disposition to
appreciate certain similarities.
Perhaps someone will be tempted to suggest that we exclude
portions of reality other than bodies because our sense of similar-
ity provides no basis for comparing or contrasting such portions
of reality, and hence we cannot classify them in any way. But
this is wrong. If we did credit such portions of reality as units we
certainly could classify them in various ways. If, say, there is a
brown table and a brown tree in my backyard and there is a
brown table and a brown tree in your backyard, then we could
say that my table-cwm-tree is similar to your table-cwm-tree at
least with respect to the property of being brown, or, even more
obviously, with respect to the property of being a table-cwm-tree
(i.e., the property of being exhaustively composed of a table and
a tree). In these respects both items could be said to contrast
with any table-CMm-tree that is not brown, or with any chair-
cwm-tree. Of course we do not ordinarily draw any such com-
parisons and contrasts. This is because we do not ordinarily
credit any such unit as a table-cwm-tree. But that fact is in no
way explained by our sense of similarity.
It is unclear, then, what connection Quine intends to educe be-
tween our body-mindedness arid our sense of similarity. I think
that part of the trouble here is that Quine does not distinguish
between two questions. One is a question about why our lan-
guage is the way it is; the other is a question about how our
language is learned. The first question is: "Why does ordinary
language contain just these particular criteria of unity?" The
second question is: "How do children learn these criteria of
unity?"
It is the first question that I raised at the outset of this chapter.
It was this question that I also expressed by asking why it is that
common sense is body-minded. Now what we have just seen is
that Quine certainly offers no answer to this question. There is
no way that a sense of similarity can be seen as delivering com-
mon sense into body-mindedness.
On the other hand Quine may have provided a viable answer
244 THE CONCEPT OF IDENTITY

to the second question, about how a child acquires the criteria

of unity implicit in our language. The child must learn to dis-
tinguish between those portions of reality which do, and those
which do not, satisfy the criteria. This he may be able to do,
so long as his classificatory impulses are attuned to the complex
considerations that enter into these criteria. That is, if the
child's sense of similarity reveals a general contrast between
what qualifies in our language as a unitary body and what does
not, he may be in the position to imbibe the rule that only the
bodies are referred to as units, the rest being excluded. Thus he
may be able to pick up the body-minded way of talking. There
are many complications here, of course, as Quirie readily admits.
But the general idea may seem workable. What cannot be ex-
plained along these lines, however, is why the language, which is
being passed on to the child, contains just those criteria of unity.

III. Conventionalism
The answer to this question that I am advocating is that our
language contains those criteria of unity because of our innate
disposition to see the world in a certain way, where this disposi-
tion must be distinguished from our sense of similarity. Now
one possible alternative to this answer would be to maintain that
there is in fact no reason why our language had to contain just
those criteria of unity, but that this is nothing more than an
arbitrary convention of language. Our ordinary body-minded-
ness, according to this "conventionalist" position, is merely one
scheme for conceptually dividing the world into units, and any
number of other schemes might have done just as well. The
scheme that we have gets passed on from generation to genera-
tion, in the manner suggested by Quine.
At the very outset of Kohler's discussion of the topic of unity
in his book Geslalt Psychology, he peremptorily dismisses this
conventionalist alternative in the following words:
On the desk before me I find quite a number of circumscribed units or
things: a piece of paper, a pencil, an eraser, a cigarette, and so forth.
The existence of these visual things involves two factors. What is in-
cluded in a thing becomes a unit, and this unit is segregated from its
surroundings. In order to satisfy myself that this is more than a verbal
affair, I may try to form other units in which parts of a visual thing
 A SENSE OF UNITY 245

and parts of its environment are put together. In some cases such an
attempt will end in complete failure.11
I think we may assume that a case of "complete failure" in
Kohler's terms would occur if we tried to see a tree, or some part
of it, as forming a unit together with a nearby table, or some part
of it. Kohler's line of reasoning seems to be as follows. I cannot
get myself to see the tree and the table as forming a unit, though
I can of course easily utter the words "The tree and the table
form a unit." This shows that a judgment of unity is "more than
a verbal affair," more than an arbitrary linguistic convention.
Unity is something that we experience; it is, as Kohler says a
few sentences later, a "visual fact."
The conventionalist is not likely to be convinced by this argu-
ment. The issue is not whether we experience unity; obviously
we do. As I look around me I can see that some portions of the
scene add up to a unitary object and some do not. This the
conventionalist would not deny. His suggestion, however, is that
the way that I experience unity is determined by the arbitrary
conventions of my primary language, i.e., the language I habitu-
ally speak and in terms of which I think. Of course I cannot al-
ter my experience merely by mouthing some strange sentence
(e.g., "The tree and the table make up one thing"), because it is
my primary language that matters. Hence Kohler's stark dichot-
omy between "visual facts" and "verbal affairs" does not speak
to the issue.
Kohler's failure to address the possible influence of language
on our experience is a flaw in his whole treatment of sensory
organi/atiori. Some of his most impressive observations pertain
to the way that we see things as forming groups or clusters, a
phenomenon which he sometimes refers to as the formation of
"group-units." Though this phenomenon of "group unity" is riot
directly relevant to the topic of bodily unity under discussion,
it may serve to highlight the problematical relationship between
our language and the structure of our experience. Kohler points
out that, when we look at Figure i, we see two groups of dots,
each group containing three dots. We do not see three groups
of two dots each, or two groups divided in some other way.
What we see very clearly are two groups of three dots each. Here
11. Kohler, Geslalt Psychology, pp. 137-38; my italics.
346 THE CONCEPT OF IDENTITY

Fig. i

the phenomenological datum is so striking that one is immedi-

ately inclined to share Kohler's assumption that our language
can have nothing to do with the matter.12
But is this really so clear? Consider that in order for us to de-
scribe our experience of Figure i we must make use of the Eng-
lish word "group" (or some equivalent word like "cluster,"
"collection," "arrangement"). Two groups are what (we say) we
experience. We must, then, have learned at some point what
the criteria are for applying the English word "group," where
these criteria presumably coincide with just those "principles
of grouping" which gestalt psychologists try to elicit. Is it not
possible that, in learning how to use the word "group," we
learned these principles of grouping, and that, had we spoken a
different kind of language, we might have experienced Figure i
quite differently? I am not saying that this is plausible; I doubt
that it is. But the question needs to be focused on properly.
I would like to apply some of these points to the topic of
unity through time, a topic which Kohler essentially ignores.
There is a comment that Sydney Shoemaker makes about unity
through time which seems to parallel Kohler's dismissal of con-
ventionalism:
It is a striking fact that motion, though it involves the persistence
through time of the moving object, is often directly observed rather
than inferred. . . . And I think it is partly because there is an experi-
ence of motion that spatiotemporal continuity occupies the central role
it does as a criterion of identity. . . . It does not seem to be just a
matter of convention that we use spatiotemporal continuity as a cri-
terion of identity. On the contrary, when I see motion (as opposed to

12. Sec Kohler, ibid., p. 142.

 A SENSE OF UNITY 247

inferring it) there seems to be no way in which I could describe what

I see except by saying "It (or: something) is moving," and in saying
this I imply the persistence of something through time.13
What does Shoemaker mean when he says that what I see could
only be described in terms of the ordinary proposition "It (or:
something) is moving?" Here the expression "what I see" has the
characteristic kind of ambiguity that Wittgenstein discussed un-
der the heading of "seeing as." 14 Of course if I am asked to de-
scribe the way 1 ordinarily experience a moving object I must
employ just those ordinary notions which are constitutive of my
ordinary experience. But the conventionalist would hold that,
had we learned a different kind of language, with different iden-
tity criteria, we might have experienced a moving object in a
radically different way (under a radically different "aspect").
Shoemaker's comments merely dismiss this possibility.
To flesh out the conventionalist's point we would have to
sketch an alternative language, one which did not make use of
our ordinary identity criteria, and explain how the world might
have been experienced in terms of that language. Strange lan-
guages, of the general sort required, are of course not un-
known to philosophical literature. An alternative language that
is especially relevant to the present purpose is one that employs
Quine's technical notion of an object, a notion I have mentioned
several times before. From the vantage point of that "space-time
language," any space-time portion of reality, however discon-
nected or noncohesive, qualifies as a unitary object. The space-
time language is, in an important sense, a language without
criteria of unity. In the sense that a club that allows anyone to
join has no criteria of membership, a conceptual scheme that
allows any space-time portion to qualify as a unit can be said
to have no criteria of unity. What would the world look like
from the space-time standpoint?
Well, in a sense everything remains the same. If there is a
mouse moving across the floor, then, from the space-time vantage
point, there is that particular spatially and temporally extended
chunk of reality which corresponds to common sense's moving

13. Shoemaker, Self-Knowledge and Self-Identity, pp. 203-4.

14. Ludwig Wittgenstein, Philosophical Investigations (The Macmillan Co.,
N. Y., 1953), pp. 193-214.
248 THE CONCEPT OF IDENTITY

mouse. But that chunk of reality would be seen as crisscrossing

and overlapping a myriad of other "objects," such as an object
that consists of an early stage of the mouse together with a later
stage of a nearby table, or an object that consists of an early
stage of the mouse together with a later stage of the mouse's
head. Hence that space-time chunk of reality which corresponds
to common sense's moving mouse would not stand out as it does
for common sense, because it would be seen as embedded in a
reality swarming and whirling with an endless number and vari-
ety of objects corresponding to every shift in one's attention.
There would scarcely be, from this standpoint, a "moving ob-
ject" in anything like the ordinary sense.
This description of the space-time standpoint pretty quickly
fades into impressionism. But that is to be expected on any ac-
count since we are trying to construct a language that goes
against the deepest habits of our thought. What the conven-
tionalist would insist is that, if we had learned that language
in childhood as our primary language, we could have seen the
world that way.
Now my aim is not to defend conventionalism, but only to
try to get somewhat clearer as to what the issues are. Actually,
as I stated earlier, I favor the position that our criteria of unity
are rooted in an innate disposition, what I called our "sense of
unity." This seems to be Kohler's position, and it is also, I be-
lieve, implicit in Shoemaker's remarks. I think that what is
really behind Kohler's and Shoemaker's abrupt dismissal of the
conventionalist idea is the intuition that this idea is too incredi-
ble to be taken seriously. And it does indeed seem to me intui-
tively incredible that, had I only been trained to speak differ-
ently, my experience of unity might have been completely
different. Still, it would be better not to have to rely entirely on
this intuition, but to have some argument to lean on.
One possible argument against conventionalism would be of
an anthropological sort. If it could be established that every lan-
guage known to us contains criteria of unity essentially like ours,
then this would evidently be a problem for the conventionalist.
Of course, if a philosopher holds, as Quine apparently does, that
how we translate a foreign language (how we choose a "transla-
tion manual") is itself (something like) a convention, then it
 A SENSE OF UNITY 249

seems that we could never hope to establish, as a factual mat-

ter, that our criteria of unity are universal. But I am assum-
ing, as I think most philosophers would, that it is a factual
question, and in principle an answerable one, whether other
people operate with a concept of bodily unity like ours. If, then,
an affirmative answer to this question were forthcoming, this
would seem to be a strong objection to the view that our criteria
of unity are merely arbitrary conventions.15
Another line of attack against the conventionalist, which I will
take up presently, might draw on speculations about how chil-
dren see the world before they learn a language. Before pursuing
that point, however, I want to consider another possibility that
is found in Kohler's discussion.

IV. An "Empiricist" Explanation

Interestingly enough, although Kohler immediately dismisses the
conventionalist approach, he addresses himself at length to the
"empiricist" (or, as Kohler calls it, the "empirist") explanation
that our judgments of unity are based on previous learning.16
What "previous learning" must mean here, if this position is to
be distinguished from conventionalism, is learning about the
world rather than learning about language. The empiricist posi-
tion, in the sense relevant to Kohler's discussion, is that the cri-
teria of unity in our language are neither innately determined
nor arbitrarily conventional, but are rather the result of our
having derived these criteria from something that we learned
about the world. As against this, Kohler wants to argue that
our criteria of unity are innate rather than learned and that,

15. But perhaps not a decisive objection: see Hilary Putnam's suggestion that
perhaps all languages have a common origin, in "The 'Innateness Hypothesis'
and Explanatory Models in Linguistics," Synthese, 7, i (March 1967), p. 18.
16. For his (somewhat unclear) distinction between the two terms "empiri-
cist" and "empirist," see Kohler, Gestalt Psychology, p. 113. I will use the
more familiar word "empiricist" to signify the third alternative to the con-
ventionalist and innateness positions that have been discussed, though obvi-
ously an "empiricist," in some more general sense, could easily hold either
of these latter two positions. (Cf. Quine's remarks about "empiricism" in the
passage cited in fin. 9, above.)
250 THE CONCEPT OF IDENTITY

therefore, our ordinary experience of unity is elementary rather

than derivative.
The reference to previous learning can still be interpreted in
two ways. It might mean that each human being must derive the
criteria of unity from something that he learns about the world;
or it may mean that prehistoric people over the millenia derived
the criteria from what they learned about the world, and then
passed these criteria down to us through our language. On either
interpretation the essential difficulty is to explain how our cri-
teria of unity could have been derived from anything learned
about the world.
A crude but not unfamiliar explanation is depicted by Kohler
in the following words: "Since early childhood we have often
observed that sets of sensations which have approximately the
same color, and differ in this respect from their environment,
tend to behave as units, i.e., to move and be moved, to appear
and disappear, at the same time. Such is the case with stones,
with papers, with plates, with shoes, with many animals, with
the leaves of plants. . . . It is only an example of the well known
generalizing power of memory if, as a result of such experiences,
we treat all homogeneously colored areas as units."17
Kohler's own critique of this position is quite complicated and
relies on a variety of somewhat specialized phenomenological
data. But it seems that his discussion overlooks the obvious and
decisive objection to the proffered explanation.
The explanation is premised on our having often observed that
certain "sets of sensations . . . tend to behave as units." Of course
one would immediately like to know how the word "sensation"
is being used here. But, even more to the point, we need to ask
how the idea of something "behaving as a unit" can possibly
operate within this explanation. Something behaves as a unit if
its parts move together, appear arid disappear together, and, in
general, are causally interdependent. This is essentially the con-
dition that I earlier called "dynamic cohesiveness." So the idea
is that we treat certain conditions (such as color homogeneity)
as criteria of unity because we have learned to associate those
conditions with dynamic cohesivenes. But why, then, do we treat
dynamic cohesiveness as a criterion of unity? How did we ever

17. Kohler, Gestalt Psychology, p. 141.

 A SENSE OF UNITY 251

learn that? Evidently not the slightest gesture is being made to

answer this question.
Furthermore, to make matters much worse, the criterion of
dynamic cohesiveness itself presupposes various other criteria of
unity. To judge that the parts of a thing move together we
must be able to trace those parts through time. For this we re-
quire various criteria of unity through time (e.g., qualitative
and spatiotemporal continuity). Hence the proffered explana-
tion of how we derive our criteria of unity really presupposes
many of these criteria from the start.
When one reflects on these difficulties it becomes clear that
there cannot be any remotely straightforward way of explaining
how our criteria of unity might have resulted from previous
learning. The sort of "empiricist" who believes in such an ex-
planation must describe an elementary level of experience which
does not already presuppose our ordinary criteria of unity. He
must then go on to explain how at that level we learn some-
thing about the world which somehow gets us to adopt these
ordinary criteria. It is far from clear how one could even begin
to formulate such an account.
Traditionally, sense-data languages and the like were often
taken to provide a level of experiential judgment more elemen-
tary (more "immediate") than that of common sense. It is not
clear, however, that sense-data descriptions can have any bear-
ing on the present topic. For one thing, these descriptions require
criteria of unity for sense data, and these criteria seem generally
to be simply borrowed haphazardly from our ordinary criteria
of bodily unity. Futhermore, the sense-data maneuver relates to
the "problem of the external world," which is not the problem
we are discussing. When I originally introduced the case of the
table and the tree, my question was not "How can I be sure
that this is not a hallucination?" or "How do I know that this
sort of scene ever exists unperceived?" but rather "How do I
know what criteria to employ in dividing the scene into units?"
This latter question pertains to the external world, and tradi-
tional sense-data dialectics about how we get to external reality
seem quite irrelevant to answering the question.
We might try to think of the space-time language as providing
a level that is elementary in a sense relevant to the present dis-
cussion. For, as I explained earlier, the space-time language
252 THE CONCEPT OF IDENTITY

might be said to contain no criteria of unity, since at that level

we would not exclude any portion of reality from qualifying as a
unit. Perhaps we might then focus on the following question.
Suppose that we (as infants or as prehistoric people) started out
by experiencing the world from the space-time vantage point.
Could we at that level have discovered something about the
world which would have taught us to adopt our ordinary criteria
of bodily unity? That is, could we have discovered something
that would have induced us to convert to commonsense body-
mindedness?
It seems fairly clear that purely theoretical motives could not
provide this inducement. Neither science nor metaphysics, the
two repositories of good theory, have ever been much enthralled
by our commonsense criteria of unity. This relates to the point
that I made earlier vis-a-vis Kant: that the commonsense concept
of a body cannot be equated with the scientist's or metaphysi-
cian's concept of underlying matter. A creature with purely theo-
retical needs, who started out without any criteria of unity,
might possibly develop some concept of matter conservation, or
he might plunge directly into the physics of fields and space-
time manifolds. At any rate, he could scarcely have any reason
to take a detour through the specific conceptual concoction that
constitutes commonsense body-mindedness.
A more likely suggestion would be that practical rather than
purely theoretical needs might motivate the adoption of our
ordinary criteria of bodily unity. Such a suggestion seems rather
common in pragmatist literature. For example, William James
says: "But what are things? Nothing. . . but special groups of
sensible qualities, which happen practically or esthetically to in-
terest us, to which we therefore give substantive names, and
which we exalt to this exclusive status of independence and
dignity." 18 The hint of phenomenalism in James's remark is not
relevant to the present issue. What is relevant is James's sugges-
tion that our practical (and aesthetic) interests are what induce
us to "dignify" certain portions of reality as unitary things.
Translated into the model I am considering, James's idea is that,

18. William James, The Principles of Psychology (Henry Holt and Co., N. Y.,
1890), I, p. 285.
 A SENSE OF UNITY 253

if we had originally started out with the space-time standpoint

(i.e., without any criteria of unity), our practical concerns would
eventually convert us to body-mindedness.
I do not think that James's suggestion can withstand careful
reflection. Our concerns and interests are almost never directed
toward objects taken one at a time, but are directed rather
toward a multiplicity of objects related in various complicated
ways. A typical concern, as ordinarily conceived, would be to
alter the relations between objects, say, to bring them closer to-
gether or farther apart, to attach or detach them, to replace one
with the other, and so on. But any such concern might be said to
embrace that whole portion of reality which includes those ob-
jects or their relevant stages. If, for example, I want to bring a
chair nearer to a table, the target of my concern might be de-
scribed as that space-time chunk of reality which includes the
table and the chair (or perhaps their stages during the period
when their increased proximity is of concern to me). One might
say roughly that my concern here is to replace a wider or less
compact table-cwm-chair with a narrower or more compact table-
cwrn-chair. Of course that is not how I would ordinarily describe
my concern. When I ordinarily describe my concern I do so by
conceptually dividing reality in the ordinary way; I talk about a
unitary chair and a unitary table, and not about any such thing
as a table-ciim-chair. This is because the ordinary concept of
unity is already given, and I describe the target of my concern
in terms of that concept. But there is apparently nothing in the
concern as such which could explain why I operate with that
concept of unity, why I divide reality up in just that way. And
this same point could be made for virtually any example one
cares to consider. Contrary to James's suggestion, there appears
to be no clearcut connection between our interests or purposes
and our concept of unity.
It is perhaps true that we tend to classify objects, especially
artifacts, in terms of their aptness to fulfill human purposes. But
this is a comment about our classifying tendencies, not about our
unifying tendencies. The units are already given, and we classify
them from a practical standpoint. There are all kinds of space-
time portions which, if we only treated them as units, we could
easily classify in practical terms as well. James's idea is that an
object is a portion of reality which is especially spotlighted by
254 THE CONCEPT OF IDENTITY

our needs and interests. But the fact is that our human concerns
striate space-time in a manner that crisscrosses and overlaps the
space-time paths of objects in every imaginable way. There seems
to be nothing in our practical standpoint which can account for
the specific concept of unity we have.
It seems to me that there is no way that the "empiricist" posi-
tion under discussion can overcome these difficulties. If we im-
agine ourselves as having started out without our ordinary cri-
teria of unity, there is nothing that we might have learned about
the world, or about our practical relationship to the world,
which could have yielded those particular criteria. In criticizing
the empiricist position, I have been using the word "learning"
in a rather loose and intuitive way. Certainly the prospects for
the position would be immediately diminished if one limited
learning to the sort of stimulus-response model that Quine some-
times seems to favor. But my criticism of the position does not
rely on any particular analysis of what learning consists in. Tak-
ing the concept of learning in what seems to be the broadest sense
relevant to this discussion, as roughly equivalent to "inference"
or "rational derivation," the crucial point remains just this:
If we had started out by describing the world without our ordi-
nary criteria of unity, then there seems to be no inference of
any sort whatever that could have led to our adopting those
criteria.
My argument against the empiricist view has taken the rather
extreme form of denying not only that we do in fact arrive at
our criteria of unity by some kind of inference; I deny that we
could even in principle have arrived at these criteria in any such
manner. According to my version of the innateness hypothesis,
we are innately disposed to adopt the criteria of unity that we
have; but, apart from this specific and complicated disposition,
there is nothing about the world which could have taught us to
adopt our criteria of unity, since there are no considerations,
theoretical or practical, which mark off just those criteria as being
especially right or reasonable. It might seem initially tempting
to maintain an innateness view somewhat different from mine.
It might be maintained, contrary to my view, that our criteria
of unity are rationally derivable in principle, that they are the
peculiarly right criteria for us to have, and that precisely for
that reason we have evolved the innate propensity to operate
 A SENSE OF UNITY 255
19
with those criteria. But, if my arguments in this section against
the empiricist position have been successful, then they show that
our criteria of unity are not even derivable in principle.

V. Focusing on Objects
I have discussed three general approaches to our concept of
bodily unity. According to one approach, our criteria of unity
are essentially arbitrary conventions which could easily have been
otherwise. This idea seemed intuitively implausible, though I
have yet to present a definite argument against it. A second pos-
sibility was that these criteria are somehow derived from some
facts about the world, facts that one could describe without pre-
supposing the criteria. I tried to argue in the last section that
this view in untenable. So we are left with the third possibility,
which is that our criteria of unity are neither arbitrary conven-
tions nor learned, but are determined by our innate disposition
to experience the world in terms of just those criteria.
It should be understood that the issue posed by these alterna-
tives relates most directly to the bare foundations of our concept
of bodily unity. To believe in an innate sense of unity is to sup-
pose that our most basic and general criteria of unity result from
an innate disposition. This does not preclude the possibility that
these criteria may be enriched and elaborated in various ways,
indeed in ways that are likely to include an element of conven-
tion as well as an element of practical and theoretical reasoning.
The conventionalist and the "empiricist" are not saying merely
that various nuances of our fully developed concept of unity may
result from convention or inference. They are denying an innate
status to any of our criteria of unity, even such seemingly funda-
mental criteria as cohesiveness and continuity. The position I
have been arguing for is that these most fundamental criteria
must be innate.

19. This seems in fact to be Kohler's position; see Kohler, Gestalt Psychology,
pp. 162-64. My own position might be compared to Chomsky's view that
"there is no a priori 'naturalness' to such a system [of innate grammatical
principles], any more than there is to the detailed structure of the visual
cortex." See Noam Chomsky, Language and Mind (Harcourt Brace Jovano-
vich, Inc., N. Y., 1968), p. 88. For further discussion of the rationality of our
identity criteria see above, Chapter 5, Section V.
256 THE CONCEPT OF IDENTITY

Earlier, when discussing the conventionalist position, I sug-

gested that one might try to clarify some of these issues by re-
flecting on how children are likely to experience unity before
the learning of language. In this connection one naturally thinks
of Jean Piaget's discussions of the child's development of the
object concept.20 According to Piaget's scheme, the child's con-
cept of an object invariably passes through a succession of stages.
The principle of transition from one stage to the next is charac-
terized by Piaget in a somewhat elusive manner. In some cases
the transition seems to be essentially nothing more than matu-
rational development, but the more fundamental idea seems to
be that each successive stage resolves with increasing success
various conceptual conflicts and tensions that arise at earlier
stages. We are invited to compare this process to the develop-
ment in theoretical science of increasingly more adequate ways
of thinking about the world. All of this would obviously repay
close examination. My concern at present, however, is less with
Piaget's account of these transitions, than with his depiction of
the very earliest stages. Would Piaget agree with my contention
that our experience is from the very start governed by our most
basic commonsense criteria of unity?
In the initial stages, according to Piaget, the child's orienta-
tion is essentially solipsistic; the infant does not initially appre-
ciate that objects can persist when they are not perceived. Piaget
bases this interpretation of the infant's experience on his (some-
what controversial) findings that infants at an early stage do not
engage in any "search behavior." If an infant at this stage "is
reaching for an object that is interesting to him and we sud-
denly put a screen between the object and him, he will act as if
the object not only has disappeared but also is no longer accessi-
ble."21 The infant, as here depicted, seems to treat the object in
rather the way in which we would treat an after-image, as some-
thing whose esse is percipi.
Now the most striking feature of this account, from my pres-
ent standpoint, is the way that Piaget unabashedly describes the
20. Jean Piaget, The Construction of Reality in the Child (Balantine Books,
N. Y., 1954), ch. i; Genetic Epistemology (Columbia University Press, N. Y.,
1970), pp. 43-44, 52-57; On the Development of Memory and Identity (Clark
University Press, Worcester, Mass., 1968), pp. 17-37.
21. Piaget, Genetic Epiitemology, p. 43.
 A SENSE OF UNITY 257

infant as reaching for an object, as being interested in the ob-

ject, and as, apparently, noticing the object's disappearance from
his field of view. In fact, Piaget consistently describes infants in
very early pre-linguistic stages as focusing on objects and follow-
ing them as they move. These descriptions seem perfectly nat-
ural, even inevitable. But it is important to see that these de-
scriptions imply that the infant's experience is directed toward
units approximating to our ordinary things, units that he can
focus upon, reach for, follow with his eyes, lose sight of, and so
on. In the initial stages, at least according to Piaget, these units
are treated as having a status akin to after-images (or akin to the
philosopher's sense data), and, like after-images, their unity seems
to be defined in terms of such familiar criteria as collusiveness
and continuity. Hence Piaget apparently would agree that our
experience is at the very earliest stages—certainly before the ac-
quisition of language—governed by our most basic commonsense
criteria of unity. Indeed it may not even be completely clear
that the child's drift away from solipsism, as characterized by
Piaget, deserves to be counted as a development in the child's con-
cept (definition) of unity, rather than merely an alteration in the
child's beliefs about the unperceived persistence of the units he
has picked out.
It may be somewhat incautious of me to try to relate Piaget's
highly complex thesis to the present discussion. I can, however,
more confidently cite the recent work of T. G. R. Bower, in
which a modified version of Piaget's developmental scheme is
expounded and impressively argued. The point that presently
concerns me is Bower's depiction of the infant's experience in
early pre-linguistic stages.
The evidence that infants do segregate their environments into units
is clear. A large number of studies on the eye-fixation behavior of
infants has shown that infants will fix on the external contours of
objects in their visual field. If the objects are moved, the infants will
track them. If after moving together, the contours of an object break
and begin to move independently, very young infants will display
massive surprise. This indicates that the common motion (common
fate) has specified for them a single unit. 22

22. T. G. R. Bower, Development in Infancy (W. H. Freeman and Co.,

San Francisco, Cal., 1974), p. 102.
2^8 THE CONCEPT OF IDENTITY

The condition that Bower calls "common motion" (or "com-

mon fate") is essentially what I earlier called "dynamic cohe-
siveness." As I pointed out, this condition presupposes various
criteria of unity through time, such as qualitative and spatio-
temporal continuity. So Bower is in effect saying that the in-
fant's experience is pre-linguistically organized around units that
are defined in terms of such fundamental commonsense criteria
as cohesiveness and continuity. 23
To speculate about the experience of infants may seem a
rather dubious undertaking for a philosopher. But J would sub-
mit that the facts I am here rehearsing are, for the most part,
so commonplace, and seem so central to our intuitive grasp of
what human nature consists in, as almost to invite the designa-
tion "transcendental." Try to imagine a person whose eyes and
hands do not fixate upon objects in an essentially ordinary way,
but whose attention meanders about without ever settling on
(what we ordinarily regard as) a single object. Can we imagine
what it would be like to initiate such a creature into our ordi-
nary thought-world? Where could we begin?
I am now taking back my earlier tentative concession to
Quine that perhaps the child's quality space is essentially all
that is required to explain how the child can learn the body-
minded way of talking. Something else that is required is the
child's instinct to focus on objects in the ordinary way. This in-
stinct is just another form of what I have been calling our
"innate sense of unity." In order for the child to learn the
ordinary way of talking, he must already be focusing on objects
in a manner that exhibits his disposition to adopt our basic cri-
teria of unity.
The easiest case to reflect upon is one in which the infant is
tracking a moving object. Here the characteristic alterations in
the infant's eyes, face, and body vividly display the effort to keep
an object in focus. For the infant, as indeed for us, a moving
object evidently stands out as something to be focused upon.
23. For the purposes of the present discussion I am deliberately leaving it
quite vague just what our "fundamental criteria" are, except to suggest that
they would undoubtedly involve some appeal to continuity and cohesiveness.
I attempted a more thorough presentation of these fundamental (or basic)
criteria in Chapter 3 above, and especially Section VI. (For some weird com-
plications, however, see Bower, Development in Infancy, pp. 189-98.)
 A SENSE OF UNITY 259

When a moving object passes through our field of view, we, even
as adults, experience the unmistakable impulse to fix our gaze
on it and follow it as it moves. Our various purposes, expecta-
tions, concerns—our whole "mental set"—will eventually deter-
mine where we look, and for how long. But the primitive power
of an object, especially a moving object, to fix our attention is
unmistakable.
A point that I especially want to stress, in the light of Quine's
approach to this topic, is that our quality space (our sense of
similarity) cannot account for our disposition to focus on objects
in the way we do. The infant who is tracking a moving object
is not merely registering passively some complex similarities be-
tween the presented scene and various other space-time portions
of the world which, as he later learns, are called "moving bod-
ies." He is exhibiting the quite irreducible instinct to direct his
attention in a distinctive way, by correlating the position of his
eyes (or hands) with the position of the object he is tracking. It
is, I am inclined to say, strictly a logically contingent fact that
the focus of the infant's (and our) attention tends (however
briefly) to follow the path of an object, rather than to meander
through space-time in any number of other imaginable ways.
However, as I remarked a moment ago, this fact seems com-
pletely central to our way of experiencing the world.
What it means for us to focus (our attention) on an object can
be explained partially by reference to the manifest correlations
between the movements of our eyes and hands and the move-
ments of the objects in our surroundings. At a deeper, and some-
what more nebulous, level our focusing propensities can be seen
as providing us with a general epistemic orientation toward the
world, with a general schema for learning. Consider, for example,
how readily we learn about the shape of an object, though an
object's shape is equivalent to a complicated fact about the inter-
relations between the object's parts. Compare this with the rela-
tively slower and more arduous process of learning about the
interrelations between things that do not add up to an object,
as ordinarily conceived. Or consider how readily we discern the
various changes (e.g., the patterns of movement) of a single ob-
ject, though these changes are equivalent to complicated inter-
relations between the object's temporal stages. It is generally far
more difficult for us to size up the interrelations between object-
260 THE CONCEPT OF IDENTITY

Fig. a

stages that do not add up to what we ordinarily regard as a single

persisting object. We can say, in general, that the qualities and
internal constitution of a unitary object, as ordinarily conceived,
are far more readily discerned than the qualities and internal
constitution of any other kind of space-time portion. (In be-
havioral terms we can say that we are far more likely to "re-
spond" to the former properties as "stimulus," than to the lat-
ter.) Our heightened readiness to focus upon, and hence to learn
about, an object, rather than any other kind of space-time por-
tion, is evidently central to what it means for us to regard a por-
tion of the world as a unitary object.
Some of the points that I have just been discussing can be illus-
trated by reference to Figures 2 and 3 (which are adapted from
Kohler). The reader will readily discern that there is a central
portion of Figure 2 shaped like the letter H. Let me call that
portion "a." On the other hand, one is not apt to notice that,
embedded within the center of Figure 3, there is also an H-

Fig-3
 A SENSE OF UNITY 261

shaped portion. I will call that portion of Figure 3 "ft." The

evident fact, then, is that we readily tend to focus on a, but not
on/3.
This fact about our focusing propensity, to recast an earlier
point, cannot be accounted for by reference to our sense of sim-
ilarity. What is peculiar about /? is not that we fail to see it as
similar to this or that, but rather that, in a sense, we fail to see
it, period. Once /? is pointed out to us, however, we readily see
it as being similar to various things (if nothing else, at least to
corresponding portions of replicas of Figure 3).
This distinction between « and /3 can be expressed in terms of
our disposition to learn. If presented with Figure 2, we would be
apt to learn various things about a, for example, that it is shaped
like an H, or that it contains three line segments, or that it is
not round. By contrast, we are not likely to learn anything at
all about /? when we are presented with Figure 3. (In behavioral
terms, the idea would be that «'s properties are far more likely
than /?'s to elicit a response from us.)
Now a space-time portion of the world which does not corre-
spond to what we ordinarily count as a unitary object can be
compared to ft: we do not readily focus on (or learn about) such
a portion of the world. One might almost say that, primitively,
a unit is just this: a portion of reality that we naturally focus on.
Or, to put this from another angle, one might almost say that
the root impulse behind our body-minded ontology is to exclude
those portions of reality which "we do not see," in the peculiar
sense in which we do not see /?•
I have been suggesting that there is a close connection be-
tween the following two ideas: (i) "experiencing (and thinking
about) bodily unity in an essentially ordinary way" and (a) "ex-
hibiting essentially ordinary focusing and tracking behavior in
the presence of bodies." I am uncertain just how close we ought
to say this connection is.2* (This seems of a piece with the un-
24. If the connection is sufficiently close then, it may be suggested, at least
an indirect or derivative pragmatic justification of our criteria of unity may
after all be possible, for our focusing and tracking behavior (it will be said)
surely serves our practical concerns. I do not rule out the possibility of devel-
oping such an argument, but I would stress two requirements for the argu-
ment to be convincing: first, some alternative patterns of focusing and track-
26a THE CONCEPT OF IDENTITY

certainty one feels about the closeness of the connection between,

for example, anger and anger-behavior.) Even most modestly
construed, the connection seems amply to warrant the conclu-
sion that before the acquisition of language very young infants
experience bodily unity in a manner akin to adults. Most mod-
estly construed, the connection is explanatory: We explain peo-
ple's focusing and tracking behavior by reference to their ex-
perience of unity. Put in these terms, my point about infants is
this: A seemingly plausible (indeed a seemingly compelling) ex-
planation of the infant's focusing and tracking behavior, and the
similarity of that behavior to our own, is that the infant ex-
periences the world as broken up into units in essentially the
way that we do. Someone who wants to reject this explanation
certainly has the burden of suggesting an alternative.

VI. Conclusion
My argument against the empiricist position consisted in main-
taining that there is nothing about the world which could have
taught us to adopt our ordinary criteria of unity, had we started
out without those criteria. My argument against the convention-
alist view was basically to suggest that children experience unity
in an essentially ordinary way before they acquire language. Per-
haps this latter argument also works to some extent against the
empiricist view, for it may seem immediately implausible to
suppose that very young infants have already arrived at their
experience of unity by way of learning.
Our concept of bodily unity, or at least the basic core of that
concept, is rooted in our primitive, pre-conventional experience
of unity. And it seems that only our innate constitution can
plausibly account for the specific and complicated conditions
that a portion of the world has to satisfy-if it is to be experienced
primitively as a unit. As far as defining what enters into this ele-
ment of innateness, one point I have repeatedly stressed is that

ing behavior must be described, and second, it must be carefully explained

why our ordinary focusing and tracking behavior is more expedient than
any of the alternatives.
 A SENSE OF UNITY 263

our quality space, our sense of similarity, cannot explain why we

experience unity in the way we do; nor can it account for our
correlative focusing and tracking behavior. The conclusion sug-
gested by this whole train of argument is that an innate sense
of unity is a quite irreducible feature of our experience of the
world.
 9
Natural Kinds
and Natural Units

I. Kinds and Units

A NATURAL kind, as this notion has been employed in some
recent literature, is a class of things that it seems in some sense
natural to bring together under a general concept. As Quinton
puts this: "[T]he members of some of the collections of things
which it is formally possible to construct must have a natural
affinity for each other, there must be some collections of things
which it is natural to class together in contrast to other collec-
tions whose association is arbitrary."1 Otherwise it seems that we
could never even learn how to use a general term, for we would
have no basis to extrapolate from old cases to new cases.
Just what to make of this "natural affinity" between the mem-
bers of a natural kind is one aspect of the traditional problem
of universals. If we permit ourselves to talk about properties,
then we can say that a natural kind is a class of (all and only)
things sharing some property or, perhaps better, some "genuine"
property. I assume that both the class of tigers and the class of
lions would be regarded as natural kinds, but that none of the
following are natural kinds: the class of things which are either
tigers or lions; the class of things which are not tigers; and the
class of things which are not lions. We do not regard the latter
classes as natural kinds because they do not intuitively strike us
as being defined by "genuine" properties. Evidently our judg-
i. Quinton, The Nature of Things, p. afla. Quinton talks here about natural
classes rather than natural kinds, but I doubt that this has any significance.

264
 NATURAL KINDS AND NATURAL UNITS 265

ments about which classes are natural kinds are closely correlated
to our judgments about which terms signify genuine properties.
Indeed the only immediate advantage of talking about "natural
kinds" rather than "genuine properties" is that the former no-
tion might be acceptable to a nominalist who eschews properties
while admitting classes. Such a nominalist might be content to
characterize a natural kind as a class of things that are in some
sense "sufficiently similar" to each other. I shall return to that
characterization shortly.
My use of the notion "natural kind" may depart somewhat
from the use associated with Kripke and Putnam. In the writings
of these philosophers the notion of a natural kind is often associ-
ated with a distinctive form of semantic analysis of such terms as
"lion" and "tiger," which are said to refer rigidly to underlying
structures as opposed to superficially observable qualities.2 But
I want to stipulate that this association is not to be taken as
definitive of what a natural kind is. One can hold that such terms
as "spherical object," "red object," and "painful sensation" de-
note natural kinds, in the sense that concerns me, even if it is
assumed that the application of these terms depends wholly on
superficially observable qualities. In this sense, natural kinds, and
the properties which define them, may be either manifest or hid-
den, and they may be referred to in any number of ways. A simi-
lar point is brought out by Quine, who distinguishes between
natural kinds of two varieties, which he calls "intuitive kinds"
and "theoretical kinds."3 As an illustration of this difference he
mentions color classifications: the class of red things constitutes
a natural kind from an intuitive common sense standpoint but
perhaps not from a more theoretical standpoint. In what follows
I will be talking about "natural kinds" in Quine's broad sense,
and indeed my emphasis will be on kinds at a rudimentary in-
tuitive level.
The notion of a natural kind is surely problematical. I want,
however, to take this notion pretty much for granted in the
present discussion, and to explore certain possible points of con-
nection between it and a seemingly kindred notion. Our basic
2. Kripke, Naming and Necessity; Putnam, "Is Semantics Possible?," reprinted
in Mind, Language and Reality.
3. Quine, "Natural Kinds," in Ontological Relativity and Other Essays, p.
i 3 iff.
266 THE CONCEPT OF IDENTITY

form of proposition, according to an enduring tradition, is one

in which the subject-part picks out some concrete thing, and the
predicate-part classifies that thing as being of one kind or another.
If there is with respect to the predicate function a distinction be-
tween natural and artificial classifications, one might hope to
find a parallel distinction with respect to the subject function.
And looking in that direction one readily sees the possibility for
drawing such a distinction. A concrete thing, such as a tree or a
cat, is comprised of the contents of a particular portion of space
and time. Evidently, however, not just any arbitrary space-time
portion of reality constitutes what we would naturally regard as
a thing. Suppose that a cat is lying a few feet away from a tree.
We would not naturally regard the cat together with the tree as
making up some unitary thing, a thing composed of the cat and
the tree. Nor less would we naturally regard some early stage of
the cat's history together with some later stage of the tree's history
as comprising the history of some unitary thing, a thing which
was first a cat and then a tree. A cat, it appears, is a natural
unit, as is a tree; but the result of combining in thought a cat,
or some stage of a cat, with a tree, or some stage of a tree, is not
a natural unit.
A natural unit, then, is a concrete portion of reality which, at
some level of common sense or science, we treat as a unitary
thing. Paralleling the distinction between natural kinds and
artificial classes we have the distinction between natural units
and artificial portions of reality. As before we can distinguish
further between units at a relatively nonspecialized commonsense
level and more theoretical units; and, again, my concern will be
primarily with the former.
The notion of a natural unit, as I intend to employ it, encom-
passes only such items as philosophers call "concrete" or "sub-
stantial." Concrete things have, I assume, a special and central
connection to the subject-role of our basic subject-predicate
proposition. And the point is that apparently not every portion
of reality which could in principle be treated as a concrete thing
is naturally so treated. It is indeed true that some philosophers,
notably Quine, have recommended metaphysical systems in which
any space-time portion of reality, however discontinuous or non-
cohesive, is to be treated as a concrete thing on a par with an
 NATURAL KINDS AND NATURAL UNITS 267
4
ordinary body, In these metaphysical systems the sum of a cat
and a tree is a concrete thing, as is the sum of a cat-stage and a
tree-stage. Even these philosophers would agree, however, that
at the level of common sense, as well as at various scientific-
theoretic levels, not just any arbitrary portion of reality is treated
as a concrete unit. Relative to these levels, at any rate, we have a
distinction between natural units and artificial portions of reality.

II. Kinds and Similarity Classes

This discussion, as I have already said, presupposes the notion
of a natural kind and aims primarily to connect this notion to
that of a natural unit. There are, however, several essential as-
sumptions about natural kinds which I must make explicit be-
fore proceeding.
a. The naturalness of a "natural kind" seems to convey two
somewhat different ideas. On the one hand there is the idea that
we find it natural to regard certain things as forming kinds. On
the other hand there is the idea that certain things form kinds in
the natural order, apart from how these things relate to our
human purposes or attitudes or interests. I think that both of
these ideas are generally implied by philosophers who employ
the notion of a natural kind. A natural kind is thus a class of
things that strike us as forming a kind independently of our
human activities.
The antithesis of a natural kind is an artificial class, the mem-
bers of which strike us as not just "going together" in the natural
order of things, but rather as being "put together" by us. One
sort of example, mentioned earlier, is classes defined on the basis
of essentially disjunctive or negative descriptions. A rather dif-
ferent sort of example of an artificial class is one defined on the
basis of a description that implicitly or explicitly refers to our
human activities. Thus the term "article of clothing" would not,
I assume, be said to define a natural kind, on the grounds that
articles of clothing have nothing distinctively in common but
their relationship to some human activity. Many other examples
are much more difficult to assess. Should we say that the class of
4. Quine, Word and Object, p. 171; "Worlds Away," Journal of Philosophy,
73. 22 (>97 6 )-
268 THE CONCEPT OF IDENTITY

tables is a natural kind, on the grounds that even apart from our
human activities tables are similar to each other (e.g., in shape
and size)? This does seem somewhat plausible, at least in the
"weak" sense of natural kind which I will explain in a moment.
So membership in a natural kind must not depend on how a
thing relates to our human activities. But should we in fact make
the much stronger remark that membership in a natural kind
must not depend on any of a thing's external relationships?
Something like this may often be tacitly assumed (at least in
the choice of examples), though the motivation for such an
assumption is far from clear. In any case this question need not
be resolved for the limited purposes of the present discussion.
What is essential to note, however, is that a natural kind cer-
tainly can be defined on the basis of internal relationships, that
is, on the basis of a description of the sorts of parts a thing has
and how these parts are interrelated. Thus "molecule of HaO"
denotes a natural kind, membership in which depends on
whether a given molecule is made up of certain kinds of atoms
related in certain ways. Membership in a natural kind will very
typically depend upon a thing's internal structure. Indeed it
seems to me reasonable to suppose that any description of an
internal structure will define a natural kind, so long as the de-
scription does not depend on essentially disjunctive or negative
properties. Thus the class of things which are partly red and
partly green ought to count as a natural kind (at least at a rudi-
mentary level).
b. This leads directly to another point. Certainly we must not
assume that in order for a term to define a natural kind the term
must be in any sense logically simple. The most typical natural
kinds are based on terms of some relatively high degree of logical
complexity; terms such as "molecule of H2O," "tiger," and
"spherical object" are evidently in no sense simple or indefinable.
A consideration of such examples suggests indeed that any
conjunctive term, built up out of terms which define natural
kinds, itself defines a natural kind. Thus if "(bit of) H2O" and
"solid" define natural kinds, so does "solid bit of H2O" (i.e.,
"ice"). And if "cubical" also defines a natural kind, then so does
"solid cubical bit of H2O" (i.e., "cube of ice"). There seems to be
no intuitive reason to place any general limitation on such con-
structions.
 NATURAL KINDS AND NATURAL UNITS 269

That a natural kind may have even a high degree of logical

complexity follows directly from one of our earlier characteriza-
tions of what a natural kind is, viz. a class of things which are
sufficiently similar to each other. Obviously if cubical things con-
stitute a similarity class, and bits of H2O constitute a similarity
class, then without question cubical bits of H2O also constitute
a similarity class.
So complex descriptions of internal structure as well as com-
plex conjunctive descriptions can typically define natural kinds.
On the other hand, two forms of logical complexity which typi-
cally cannot define natural kinds are, as mentioned several times
earlier, disjunction and negation. This too follows immediately
from the characterization of a natural kind as a similarity class.
For we cannot say that things which are either tigers or lions
form a class in virtue of being sufficiently similar to each other,
nor that things which are not lions form such a class.
c. There is, however, a difficulty in defining what we mean by
a similarity class, or a class formed on the basis of the similarity
of its members. A tempting definition is this: S is a similarity class
if anything outside of S is less similar to something inside S than
any pair of things inside S are similar to each other. (That is, if
x is not a member of S, then there is a y such that y is a member
of S and such that for any z and w which are members of S, x is
less similar to y than z is to w.) Can we now simply equate the
notion of a natural kind with this notion of a similarity class?
Quine argues against this on the grounds that we would then
have to say that disjunctions of overlapping conjunctions define
natural kinds.5 Consider, for example, the class of things which
are either red and spherical, or red and solid, or spherical and
solid. This will qualify as a similarity class on the previous defini-
tion, because any pair of members of the class must at least be
similar with respect to one of the three properties, whereas any-
thing outside the class will lack even this degree of similarity to
some members. But such a class, Quine insists, should not count
as a natural kind.
I shall not, however, follow Quine in this point of terminology,
but shall stipulate instead that any similarity class does qualify
as a natural kind. This stipulation accords, I think, with one

5. Quine, "Natural Kinds," pp. 119-23.

270 THE CONCEPT OF IDENTITY

tendency in current philosophical usage, though the termino-

logical point as such has of course no great importance. (In fact
the reader who wishes may substitute "similarity class" for "nat-
ural kind" in all that follows.) I do not question Quine's intui-
tion that there is an important difference between the similarity
class last mentioned, which was denned on the basis of a dis-
junction of conjunctions, and a full-blown natural kind such
as, perhaps, the class of spherical objects. We might mark this
difference by distinguishing weak natural kinds from strong
natural kinds/' It has often been noted that Wittgenstein's notion
of a "family resemblance" is closely related to a disjunction of
overlapping conjunctions. 7 My usage will allow us to say that a
natural kind (of the "weak" variety) can be based on a family
resemblance. This seems correct at least insofar as it is obviously
in some sense natural to bring under a general concept objects
which are related by a family resemblance.

III. Is the Class of Units a Kind?

Let us now consider some possible connections between natural
kinds and natural units. One question that might be asked is
whether the class of natural units is itself a natural kind. This
question is surely obscure and difficult, but I think it may be
worthwhile trying to address it.
The question might be compared with asking whether the
class of beautiful objects constitutes a natural kind. This would
amount to asking whether there is, apart from our human atti-
tudes and purposes, some point of similarity between all and
only those objects which we regard as beautiful. In the same
vein we are now asking whether there is, apart from our attitudes
and purposes, some point of similarity between all and only
those portions of reality which we naturally treat as concrete
units. What we are seeking, it might be said, is an elucidation of
6. For degrees ot natural kinds see Anthony Quinton, "Properties and Classes,"
Proceedings of the Aristotelian Society, 58 (1957—58), pp. 47-48.
7. See, e.g., Renford Bambrough, "Universals and Family Resemblances," re-
printed in George Pitcher, ed., Wittgenstein (Anchor Books, New York,
1966), p. 189. In order for us to have a similarity class, or a family resemblance,
we must have a disjunction of extensively overlapping conjunctions, roughly
a term which applies to a thing only if it satisfies most o£ some list of
conditions.
 NATURAL KINDS AND NATURAL UNITS 271

the relationship between our similarity intuitions and our unity

intuitions.
As stipulated earlier, a natural unit must be concrete. Does
this amount to saying that, at least for common sense, the class
of natural units coincides with the category of a body, so that
all and only bodies are natural units? The answer to this ques-
tion would hinge on several nebulous points. First of all, one
would have to decide whether persons are bodies, since persons
are surely to be regarded as natural units. Furthermore, one
would have to clarify the notion of concreteness or substantiality
before one could assess the substantive status of various items,
for example, particular sounds and flashes, marks on bodies
(such as a figure drawn on paper), and segregated clusters of
bodies (such as a flock of birds). Even if we conceived of the
class of natural units as encompassing, besides bodies (and per-
sons), these other sundry items, it might still be possible to
maintain that all of these things have something in common
which marks them off from other portions of reality that are not
natural units. However, the issue in these terms seems fairly
intractable.
Let me in fact simplify our question significantly by focusing
exclusively on bodies as natural units. Bodies (together perhaps
with persons, if these are to be kept distinct) seem to be, in any
case, our most basic paradigms of concrete units; and it may
even be arguable that the class of natural units can properly be
regarded as forming a natural kind if and only if the class of
bodies can be so regarded. The simplified question, then, is
whether the class of bodies is a natural kind. Should we say that
there is some point of similarity that marks bodies off from all
other portions of reality?
An affirmative answer to this question is indicated by some of
Quine's remarks about the child's acquisition of the concept of
a body. Quine suggests that there are some "very general body-
unifying considerations that preserve the identity of each dog,
each rabbit, each apple, each bundle, in short each body." These
general body-unifying considerations, he says, constitute a
"similarity-basis for the child's ostensive learning of the general
term 'body' itself, or 'thing,' to take the likelier word."8

8. Quine, The Roots of Reference, p. 56.

272 THE CONCEPT OF IDENTITY

To know how to apply the term "body" requires that one

should know how to judge of the unity of a body; one must
know when to count a space-time portion of reality as constitut-
ing a single self-same body. Quine's remarks imply that if a child
learns how to apply the term "body" (or "thing") to a fair assort-
ment of bodies then he can go on to apply the term correctly to
new cases, basing himself on the similarity between bodily unity
in the new cases and bodily unity in the old cases. A body, ac-
cording to Quine, is a distinctive kind of space-time portion,
one whose spatial and temporal parts are structured in some
distinctive way. This is tantamount to saying, at least in the
rough terms of the present discussion, that the class of bodies
is a natural kind.
Wiggins's sortal theory of identity, on the other hand, would
seem to-imply the opposite, that the class of bodies is not a
natural kind. Wiggins holds that a judgment about bodily unity
must always be mediated by some "substance-sortal" concept,
where a substance-sortal is some relatively specific noun such as
"cat," "tree," or "car." Wiggins's view implies that there are no
"general body-unifying considerations" of the sort that Quine
refers to, that there is no "similarity basis" which could permit
one to extrapolate from the unity conditions of one body to
that of another, where these bodies fall under different substance-
sortals. This is why Wiggins repeatedly insists that a term like
"body" (or "thing" or "object") is only a "dummy sortal" which
cannot operate independently of the specific nouns subordinate
to it.9 One might indeed say that for Wiggins the term "body"
is in effect equivalent to some long disjunction of the form
"either a tree, or a cat, or an apple, or ... ," where each different
substance-sortal enters as a disjunct. The term "body," so re-
garded, could not be said to denote a natural kind.
As regards this apparent controversy between Quine and
Wiggins my view would be that Quine is right at one level and
Wiggins at another. Quine is surely correct in supposing that
there are some body-unifying considerations which can operate
independently of our various sortal differentiations. An example
I have discussed previously is that of an Eskimo who has never
before seen (or heard of) a tree, and who is now presented with

9. Wiggins, Identity and Spatio-Temporal Continuity, pp. 29, 33, 35.

 NATURAL KINDS AND NATURAL UNITS 27g
10
a tree for the first time. Though he lacks the general concept
of a tree there can be no doubt, I think, that the Eskimo is im-
mediately in the position to make such judgments as the follow-
ing: "There is a tall, queerly shaped, mostly brown body (thing)
over there, with a cylindrically shaped central portion, out of
which there is jutting at various angles a number of twisted
parts of various shapes and sizes, at the very ends of which are
some green things." This judgment reveals the Eskimo's basic
grasp of the tree's spatial extent, of the tree's unity through
space. And other judgments would no doubt reveal the Eskimo's
basic grasp of the tree's temporal extent, the tree's identity
through time.
What we can appreciate when we reflect upon this sort of
example is the extent to which our judgments of unity need not
depend upon any sortals: we can often base a judgment of unity
on body-unifying conditions that are not linked to any sortals.
Perhaps this sort of example does not conclusively show that the
sortal-neutral body-unifying conditions are similar from one
case to another. For it might be possible to suppose that when
the Eskimo confronts the tree, he just finds it natural or reason-
able to apply some wholly new criteria of unity, criteria essen-
tially different from any that he has previously employed in
other cases. Certainly, however, the far more plausible explana-
tion of the Eskimo's judgments is that he bases himself on some
analogy, on some point of similarity, between the unity of the
tree and other cases of bodily unity which he has encountered.
So it seems that Quine is right in saying that there are some
general body-unifying considerations that are the same for bodies
of all sorts. And presumably we could, with some care, even
formulate more or less what those considerations are. Quine
indeed makes the rough suggestion that a body's unity is deter-
mined by its "synchronic continuity" and its "diachronic con-
tinuity of displacement and deformation."11 This suggestion
needs a great deal of elaboration, but surely it points in the
right direction.12
10. See above, pp. 75-78.
n. Quine, The Roots of Reference, p. 54.
12. A more elaborate sortal-neutral formulation might be in terms of the
"basic rule" and the "principle of articulation"; see above, Chapter 3, Sec-
tions II and VI.
274 THE CONCEPT OF IDENTITY

So there are, as Quine suggests, some general body-unifying

considerations which can operate independently of sortals. It
seems, however, that Quine may overrate the scope of such con-
siderations. For the fact is (and this is Wiggins's point) that cases
can arise where one's ignorance of the applicable sortal (one's
ignorance of "what the thing is") will give rise to mistaken
judgments. Our Eskimo, for example, who has no idea what a
tree is, could not be expected fully to share our sortal-relativized
view of just what kinds of changes the object he picked out can
suffer without ceasing to exist.13 Quine's account, one might say,
answers to our most basic, our most unsophisticated, concept of
bodily unity. This is the concept we would apply when explor-
ing some wholly new terrain and, like the Eskimo in the pre-
vious example, have no sortals to apply. But this most basic con-
cept of bodily unity is eventually refined by our sortals, and the
sortals, when they are available, can make a difference to our
identity judgments. Wiggins is right therefore in denying that
our most sophisticated judgments of bodily unity can always be
based on some general similarities between bodies coming under
different sortals.
So we can conclude, perhaps, that whereas our most rudimen-
tary concept of bodily unity does denote a natural kind (of at
least the "weak" variety), to the extent that the concept is com-
plicated by sortal differentiations it no longer does denote a
natural kind.

IV. Kinds and Individuation

Whether or not we judge the class of natural units (which, for
simplicity, I continue to identify with the class of bodies) as a
natural kind, it seems clear that many natural units belong to
some natural kind or other. If x is a lion then x belongs to the
natural kind of lions, and if x is a tree then x belongs to the
natural kind of trees, and so on. This point might suggest the
possibility of explaining the notion of a natural unit in terms
of the notion of a natural kind. Could we not say, perhaps, that
a natural unit is just a portion of reality wich belongs to some
natural kind? This suggestion may seem plausible, for if a por-

13. Cf. above, Chapter 3, Section III, on the "limitations of the basic rule."
 NATURAL KINDS AND NATURAL UNITS 275

tion of reality does not belong to any natural kind, and hence
cannot be classified in any natural way, this might explain why
we do not regard that portion of reality as a unit.
There are two questions that we can raise about this sugges-
tion. First, is it true that every natural unit belongs to a natural
kind? And, second, is it true that every artificial portion of
reality fails to belong to a natural kind? The answer to the first
question would be trivially affirmative if we could assume that
the class of natural units is itself a natural kind. But given the
more nebulous position arrived at in the last section, it may not
be immediately obvious what the answer to the first question is.
I shall not pursue this, however, for I think that the more im-
portant and revealing issues arise in connection with the second
question.
Consider, as an example, a discontinuous portion of reality
which is comprised of two trees that are standing next to each
other (surrounded, let us say, by other trees). This portion of
reality, which I will call a "tree-cwm-tree," is evidently not a
natural unit. According to the above suggestion the reason why
the tree-cum-tree is not a natural unit is that it fails to belong to
any natural kind. But what of the classes denoted by the terms
"(mostly) brown" and "(mostly) wooden"? Presumably these
classes are natural kinds, and it seems that the tree-cwm-tree
belongs to both of them.
It might be answered that the terms "brown" and "wooden"
must be understood as equivalent to "brown body" and "wooden
body," so that the tree-cwm-tree, which is not a body, fails to
qualify under these terms. But what about the terms "brown
portion of reality" and "wooden portion of reality"? Obviously
the tree-cwm-tree qualifies under these terms, and it seems that
these terms denote natural kinds. For it seems that we ought to
say that all brown portions of reality have some property in
common, as do all wooden portions of reality.
It would do no good to suggest that the class of brown portions
of reality, and the class of wooden portions of reality, are not
natural kinds precisely because some of their members (e.g., the
tree-cwm-tree) are not natural units. If one simply stipulates, as
part of the definition of "natural kind," that the members of
natural kinds must all be natural units, then one has obviously
given up the hope of explaining what a natural unit is in terms
276 THE CONCEPT OF IDENTITY

of the independently understood notion of a natural kind. The

only legitimate procedure here is to retain our earlier intuitive
characterization of a natural kind, as a class whose members are
in some sense sufficiently similar to each other, and see whether
on that basis one can explain what a natural unit is. It seems,
at this point, that one cannot.
It might still seem possible to attempt a slightly different
approach. A natural unit is a portion of reality which we find
it natural to pick out as a distinct and unitary thing. Now it
seems that the most fundamental way to pick something out is
by an expression of the form "that F," e.g., "that tree." In order
for a term F to function properly in the expression "that F,"
the term must be individualive, where we might say, for present
purposes, that a term is individuative if and only if it typically
happens that exactly one instance of the term is present in a
given region of space. Hence "tree" is individuative, since there
might typically be exactly one tree in the observable region, or
one tree in the direction that someone is pointing. This is why
the expression "that tree" can often serve to pick out a tree.
Clearly such terms as "brown portion of reality" and "wooden
portion of reality" are not individuative. You could never pick
something out by way of the expression "that brown portion of
reality," or "that wooden portion of reality," because wherever
there is a brown or wooden portion of reality there must be an
indefinite number of such portions of reality.
Let it be granted, it might now be said, that the terms "brown
portion of reality" and "wooden portion of reality" do denote
natural kinds. It follows from this that a tree-cwm-tree does
belong to some natural kinds. It does not follow, however, that
the tree-c?<m-tree belongs to any individuative natural kind
(where a natural kind is individuative if it is denoted by some
individuative term). Perhaps the reason why the tree-CMr«-tree
does not qualify as a natural unit is that it does not belong to
any individuative natural kind, and therefore cannot be picked
out in any natural way.
The present suggestion, then, is that a natural unit is a portion
of reality that belongs to some individuative natural kind. But
this suggestion fails too, I think. For consider the term "tree-
ewm-tree" (or, equivalently, "portion of reality that is comprised
of two trees.") This term is quite definitely individuative; you
 NATURAL KINDS AND NATURAL UNITS 277

can just as easily point out a particular tree-cwrn-tree by way of

the expression "that tree-cwm-tree," as you can point out a tree
by way of the expression "that tree." Consequently the class of
tree-cum-trees is an individuative natural kind, if, that is, it is a
natural kind at all. But why would it not be a natural kind?
If trees are (sufficiently) similar to each other, then it seems that
tree-cMw-trees ought to be (sufficiently) similar to each other. (To
make this even clearer, we might stipulate that the trees in a
tree-cwm-tree must be similarly situated with respect to each
other, say a few feet away from each other.) So it seems that a
tree-cwm-tree, which is decidedly not a natural unit, does after
all belong to an individuative natural kind. And, in general, it
seems that, contrary to the above suggestion, various artificial
portions of reality will belong to individuative natural kinds.
Let me try to bring this point out in a slightly different way.
We can classify not only unitary things but pairs of things, and
it seems plain that some of these classifications will constitute
natural kinds (i.e., similarity classes) and some will not. Suppose
that S is the class of all pairs of trees (or, if one wants, pairs of
trees a few feet away from each other), and S' is the class con-
taining pairs of trees and pairs of tigers. It seems plain that S
is a natural kind (of pairs), whereas S' is an artificial class. Sup-
pose it is now asked why we do not treat the pairs in S as consti-
tuting unitary things. The answer cannot be that we are pre-
vented from doing so because these would-be units could not
then be classified as belonging to any individuative natural kind.
On the contrary, if we did treat the members of S as constituting
units then S would itself provide us with the required classifica-
tion.
Though I have been carrying on this discussion from an ele-
mentary commonsense standpoint, I think that these points
would remain essentially intact even if we shifted to a somewhat
more theoretical level. From a biologist's standpoint there is, I
take it, no such thing as a tree-cwm-tree. We might be tempted
to explain this fact by saying that, for biology, there are no
kinds, or no individuative kinds, to which a tree-ct<m-tree might
belong. But this explanation is doubtful since it is not clear why
the class of tree-cum-trees could not itself qualify as a biological
kind. If any tree has the same underlying structure as any
other tree, would it not follow that any tree-ewm-tree has the
278 THE CONCEPT OF IDENTITY

same underlying structure as any other tree-cwrn-tree? (This

would surely seem to follow if a thing's underlying structure is
a function of its atomic structure.) And, as before, it would not
help to answer that the class of tree-cwm-trees is not a biological
kind precisely because the members of this class are not biologi-
cal units. For we would then be explaining kinds in terms of
units, rather than the other way around, which was what we
wanted. Taking kinds simply as similarity classes, the point is
that there is apparently no way to explain the biologist's sense
of unity by reference to his sense of biological similarity.
It might be said that the biologist eschews tree-cwm-trees be-
cause biological theory is simplified by not acknowledging (not
quantifying over) any such units. Actually I think it is quite
unclear that the biologist is here motivated by intuitions about
theoretical simplicity, rather than straight-out intuitions about
unity. Perhaps, though, this is not a critical question, since one
might say in general that our "unity intuitions" form one impor-
tant aspect of our overall "simplicity intuitions." Be this as it
may, the important point for my present purposes is that there
seems to be no reason to regard the biologist's units as dependent
upon his kinds, rather than vice versa.
What I want to conclude from this discussion is that the notion
of a natural unit cannot be explained or reduced in terms of
the notion of a natural kind. And I take it as obvious that a
reduction in the opposite direction is impossible. That we find
it natural to regard only certain portions of reality as units, is one
fact about us. That we find it natural to regard these units as
classifiable, as similar to each other, in only certain ways, is
another fact about us. These are two fundamental features of our
thought, neither of which can be explained in terms of the other.

V. The Basis of Kinds and Units

We might now ask how these features of our thought originate.
How do we arrive at our intuitions about kinds and units? With
respect to kinds Quine has in several places sketched an account
which seems plausible. 14 The sketch, with some minor modifica-

14. Quinc, "Natural Kinds," p. 123(1.; The Roots of Reference, p. igff.; Word
and Object, pp. 83-84.
 NATURAL KINDS AND NATURAL UNITS 279

tions of my own, goes something like this. We start out with an

innate quality space, an innate sense of similarity, relative to
which certain classifications seem natural and others not (or,
more likely, certain classifications seem more natural than
others). It is easy to conceive that, of those classes which could
be deemed natural kinds relative to the innate starting point,
some emerge as more important than others. These would be the
kinds that figure most prominently in our commonsense practical
principles and low-level laws of nature. One might speculate
that there is some rough correlation between the importance of
a kind and the likelihood of our having a single word in our
language that denotes it. The commonsense kinds are eventually
reshaped by scientific posits of deeper structures and properties,
yielding theoretic kinds which may diverge rather sharply from
the commonsense kinds.
I would suggest that a roughly comparable account ought to
be given of the genesis of our natural units. Studies by develop-
mental psychologists, such as Piaget and Bower, suggest that the
neonate is innately disposed to single out only certain portions
of reality as units.15 The conditions defining these most primi-
tive units probably include some of those same basic body-
unifying conditions that, as maintained earlier, one would appeal
to in the absence of any relevant sortals. If Piaget is right the
child's most primitive notion of unity undergoes a succession of
transitions to increasingly more sophisticated notions. These
transitions can be seen as motivated in part by the aim of
achieving a simpler and more precise system of units, and this
same aim yields at scientific levels theoretic units such as atoms
and electrons.
Both our units and our kinds emerge, according to these
speculations, from instinctive roots, and evolve through various
levels of common sense and science. There seems, however, to be
a crucial difference between the two cases. It seems to be a
matter of a priori necessity that any language-learner should
start out with a sense of similarity, that is, with a propensity to
classify in certain ways rather than others. Otherwise there could
apparently be no way to learn the use of a word, for there would
15. Cf. above, Chapter 8; Piaget, The Construction of Reality in the Child,
ch. i; On the Development o/ Memory and Identity, pp. 17-37; and Bower,
Development in Infancy, p. loaff. and ch. 7.
28o THE CONCEPT OF IDENTITY

be no basis for extrapolating from observed cases of the word's

application to new cases. It does not seem equally clear that, as
a matter of a priori necessity, a language-learner, or even a
learner of our language, must start out with a propensity to
pick out only certain portions of reality as units. We can appar-
ently imagine someone who starts out by regarding any portion
of reality as equally a unit, and who then classifies these units
in various ways. One such classification might indeed distinguish
between units that are bodies and units that are not bodies. The
final step to ordinary language would be to repudiate the non-
bodies as units, thus arriving at our commonsense ontology of
bodies. This would seem to be one conceivable route to common
sense, though it is quite certainly not the route actually taken
by ordinary human children.
It appears, then, that natural kinds are necessarily implicit in
language-learning in a way that natural units are not. There is
a closely related point. As mentioned before, many philosophers,
including Quine, have espoused metaphysical systems in which
all portions of reality count equally as units. As Quine puts it,
a physical object is simply "the material content of any portion
of space-time, however scattered and discontinuous." Hence:
"There is a physical object part of which is a momentary stage
of a silver dollar now in my pocket and the rest of which is a
temporal segment of the Eiffel Tower through its third decade."16
This seems as much as to say that, according to Quine, at the
highest level of philosophical theorizing the distinction between
natural units and artificial portions of reality simply falls away.
Could there be a philosophical system in which, in a com-
parable way, the distinction between natural kinds and artificial
classes falls away? In at least one sense this seems evidently im-
possible. Any philosophical system must begin with some stock
of primitive terms (i.e., terms which are not defined within the
system), relative to which things are classified in certain ways

16. Quine, "Worlds Away," p. 859. Has my talk throughout this whole dis-
cussion of "space-time portions of reality" committed me all along to Quine's
position? Not really; for such talk can be taken innocuously, as merely an
expository device amenable to various interpretations in terms of set-
theoretical constructions out of bodies, places, times, etc. But Quine's view
is that any space-time portion has the same metaphysical status, the same
status of concrete unity, as a body.
 NATURAL KINDS AND NATURAL UNITS 281

rather than others. If a class of things is denoted by a primitive

term, or conjunction of such terms, then it might be said to
constitute a kind within the system; whereas artificial classes are
denoted only by disjunctive or negative constructions. This char-
acterization is very rough, but does it not at least suffice to show
that the core distinction between natural kinds and artificial
classes must remain intact?
These considerations may seem to suggest that the notion of
a natural kind is more fundamental than that of a natural unit.
For it seems that natural kinds are necessarily indispensable
both to language-learning and to metaphysics, in a way that
natural units are not.
But let us reflect more closely on these points. The fact that
any philosophical system must contain some primitive vocabu-
lary does not entail that this vocabulary must include any such
term as "kind" or "similarity class" or even "similarity." That
is, the fact that we classify things in certain ways does not entail
that we must go on to classify our ways of classifying in terms
of the distinction between kinds and artificial classes. No such
distinction need be acknowledged.
Of course we can simply define a natural kind as a class which
corresponds to some primitive term or conjunction of such terms.
But this would not establish that there is any objective basis
for distinguishing between natural kinds and artificial classes. A
philosopher might maintain that this distinction is wholly de-
pendent on our classificatory apparatus, and that there is nothing
in the objective world corresponding to the distinction. Let me
call such a philosopher an Extreme Nominalist. This philosopher
will accept the existence of classes (and perhaps even the exist-
ence of properties), but he does not accept any objective distinc-
tion between natural kinds and artificial classes (or between
"genuine properties" and "mere constructions"). Since a natural
kind is (or is at least closely related to) a similarity class, the
position of Extreme Nominalism can also be expressed by saying
that there is no objective similarity relation. I think it is fairly
clear that Nelson Goodman is an Extreme Nominalist, and per-
haps Quine is too.17
17. See Nelson Goodman, "Seven Strictures on Similarity" in Problems and
Projects (The Bobbs-Merrill Company, Inc., Indianapolis and New York,
1972), especially pp. 444-46. In "Natural Kinds," Quine maintains that it is
282 THE CONCEPT OF IDENTITY

It is important not to jump to the conclusion that the Extreme

Nominalist must adopt an idealistic or subjectivistic attitude
about facts in general. His position might rather be this: "There
is an objective fact of the matter as to whether something is a
lion, just as there is an objective fact of the matter as to whether
something is either-a-lion-or-a-tree. But there is no objective fact
of the matter as to whether the class of lions, or the class of
lions-or-trees, constitutes a kind. That just depends upon our
classificatory practices."
The essential argument for Extreme Nominalism derives from
the fact that our judgments about kinds, or about similarity, are
often highly nebulous, uncertain, and contingent upon our inter-
ests.18 I personally doubt that these considerations can ever really
convince us that there is no objective sense in which, say, the
class of grue things is artificial. 19 However, my main concern now
is not to assess Extreme Nominalism, but rather to see how this
position bears on the comparisons I was making earlier between
kinds and units.
Let me return to the question I raised earlier: Could there be
a philosophical position with respect to kinds that parallels
Quine's position with respect to units? It may now seem that
there could be such a position, and that Extreme Nominalism
is it. But the parallel is not really exact, as I now want to explain.
It is necessary to be clear in what sense Quine's position in-
volves a rejection of the ordinary distinction between units and
artificial portions of reality. Certainly Quirie does riot deny that
there is an objective basis to the ordinary distinction. On the
contrary, he even tries to explain what that objective basis is:
From the ordinary standpoint we regard as a unit only a portion
of reality which satisfies certain continuity (etc.) conditions,
whereas a portion of reality which does not satisfy these condi-
tions is not regarded as a unit. Quine maintains, however, that
a metaphysically more perspicuous standpoint would treat all

a "mark of the maturity of a branch of science that it no longer needs an

irreducible notion of similarity and kind" (p. 138).
18. Sec Goodman, "Seven Strictures on Similarity."
19. See Nelson Goodman, Fact, Fiction, and Forecast, grd ed. (The Bobbs-
Merrill Co. Inc., Indianapolis and New York, 1973), p. 74, where the predicate
"grue" is denned as applying to all things examined before a given time just
in case they are green but to other things just in case they are blue.
 NATURAL KINDS AND NATURAL UNITS 283

portions of reality equally as units; for why should we meta-

physically despise those portions of reality which do not satisfy
the continuity (etc.) conditions? We can treat all portions of
reality as units, as referents of concrete nouns and as bearers of
qualities and spatiotemporal relations, and still distinguish be-
tween those portions of reality which satisfy the continuity (etc.)
conditions and those which do not. We can reject the ordinary
distinction between units and artificial portions of reality while
still holding on to the objective facts that underlie that distinc-
tion.
This position is almost the exact reverse of what the Extreme
Nominalist is saying about kinds. Whereas Quine does not ques-
tion the objective basis of the ordinary notion of a unit, the
Extreme Nominalist does question the objective basis of the ordi-
nary notion of a kind. Furthermore, whereas Quine wishes to
dispense with the distinction between units and artificial por-
tions of reality it is not clear that the Extreme Nominalist wishes
to dispense with the distinction between kinds and artificial
classes. To regard the distinction as subjective (or relative to our
interests or language) is not necessarily to dispense with it.20
Indeed it seems that the distinction between kinds and arti-
ficial classes is indispensable for language-learning, as I noted
earlier. And this of course is only one aspect of the more general
point that the distinction is indispensable for inductive general-
ization. In fact both Quine and Goodman stress the indispen-
sability of the distinction in these respects, even while in the
same breath disparaging the distinction's objectivity.
The notion of a kind, or similarity is ... disreputable. Yet some such
notion, some similarity sense, was seen to be crucial to all learning, and
central in particular to the processes of inductive generalization and
prediction which is the very life of science.21
[W]e must recognize that similarity is relative and variable, as unde-
pendable as indispensable.22
20. The Extreme Nominalist might even be able to say "Lions constitute a
kind apart from their relationship to our human activities" in the sense
"Lions strike us as constituting a kind apart from their relationship to our
human activities." Compare with an ethical subjectivist who says "The pain
of animals is bad apart from its relationship to us."
21. Quine, "Natural Kinds," p. 133.
22. Goodman, "Seven Strictures on Similarity," p. 444.
384 THE CONCEPT OF IDENTITY

The Extreme Nominalist is saying that, though our ordinary

similarity intuitions do not correspond to any objective facts,
they are nevertheless indispensable to our thought. On the other
hand Quine's view with respect to units is that our ordinary
unity intuitions do correspond to some objective facts, but we
can more perspicuously describe these facts without our ordinary
concept of unity.
If there could be a position with respect to units that more
accurately parallels the Extreme Nominalist's position with re-
spect to kinds, it would have at least to entail that there is no
objective basis for the ordinary notion of a unit. The only
semblance of an argument that I can think of for such a position
might run along the following lines. To the extent that the
unity of a thing can be said to depend upon the sortals that
apply to the thing, the unity of a table depends on the fact that
the thing is a table, depends, that is, on how the thing relates
to our human interests and purposes. This may seem to imply
a sense in which the unity of the table is not fully an objective
matter. Whatever the merits of this argument, however, it obvi-
ously casts no aspersion on the objective unity of such non-
artifacts as lions, tigers, and trees. So at least if we reserved the
notion "natural unit" for nonartifacts the objectivity of the no-
tion could apparently not be called into question.23
And let us ask, finally, what could be the accurate counterpart
with respect to kinds of Quine's view with respect to units. This
position would have to hold that our ordinary judgments about
kinds, or similarity, are objectively based, but that they can be
more perspicuously rendered in some different terms. I cannot
readily imagine what this position could amount to.
Let me conclude this chapter by considering again the question:
Which is more fundamental, the contrast between kinds and
artificial classes or the contrast between units and artificial por-
tions of reality? A case might be made out in either direction. It
might be held that the former contrast is more fundamental

23. Notice that even if the class of natural units is not a natural kind, this
in itself casts no aspersion on the objectivity of unity. The class which con-
tains tigers and trees is not a natural kind but it is a wholly objective matter
which things belong to the class. In the same way it might be an objective
matter which portions of reality are natural units even though the class of
natural units is not a kind.
 NATURAL KINDS AND NATURAL UNITS 285

because it is more obviously indispensable to thought; or it

might be held that the latter contrast is more fundamental be-
cause its objective status is less disputable. My own opinion is
that the contrast between kinds and artificial classes is more
fundamental. For I think, contrary to the Extreme Nominalist,
that this contrast is objectively based, as is indeed the contrast
between units and artificial portions of reality; but the former
contrast seems indispensable to thought in a way that the latter
contrast does not.24

24. The contention that thought must necessarily contain some contrast be-
tween kinds and artificial classes (some standard of similarity) docs not entail
that thought must necessarily contain our contrast between kinds and arti-
ficial classes (our standard of similarity). Only the stronger claim could sup-
port the judgment that our natural kinds are (in terms of the distinctions
of Chapter 6) metaphysically more basic than our artificial classes. The posi-
tion I am here advancing would obviously not support the judgment that
our natural units are metaphysically more basic than our artificial portions
of reality. (Cf. ftn. 9 of Chapter 6.)
 IO

Constraints on Self-Identity

THE ORDINARY distinction between "me" and "not-me,"

between that which does and that which does not lie within the
boundaries of a single self, seems at least on first reflection com-
pletely inevitable. It is difficult even to understand the sugges-
tion that this distinction might be arbitrary, or that it might
legitimately be redrawn in some other way. Here, if anywhere, a
"conventionalist" attitude is likely to strike us as intuitively
incredible.
Even here, however, our intuitions can be seriously challenged;
it is possible to argue that there is in fact nothing which con-
strains us to think of the identity of the self in the way we do.
One rather direct way to broach this issue is actually to try to
imagine what it would be like for people to operate with a con-
ception of the self radically different from ours, and then to ask
what, if anything, would be wrong with such a conception. In
the first section of this chapter I will construct one such alien
conception as an illustration. 1 Of course many other illustrations
i. The illustration is somewhat akin to that discussed by Sydney Shoemaker,
in "The Loose and Popular and the Strict and Philosophical Senses of Iden-
tity," in Care and Grimm, eds., Perception and Personal Identity, p. nyff.
The illustration could be generated to apply to the case of bodily identity,
and as such it might serve to elucidate the issue of conventionalism with
respect to bodily identity which was discussed in Chapter 8. There are indeed
close parallels between the discussion of Chapter 8 and that of the present
chapter, and a connection between the two discussions will be drawn later in
Section V of this chapter. But it seems advisable to address the case of per-
sonal identity separately, because of the quite special—and especially strong
—intuitions that we have about this case.

286
 CONSTRAINTS ON SELF-IDENTITY 287

are possible; mine perhaps has the virtue of being logically

simple and easily generalizable. This alien conception is designed
to strike us intuitively as utterly bizarre and crazy; and the philo-
sophical problem will then be to explain what these intuitions
amount to. In subsequent sections of the chapter I will discuss
various ways of responding to this problem.

I. A Strange Identity Concept

It will be useful to develop the fantasy of this alien conception
as involving two stages. Suppose that there is a community of
people who, to begin with, speak ordinary English. At the first
stage we imagine that these people introduce a set of linguistic
conventions, which I will tentatively characterize, subject to
further discussion, as altering their descriptions of the identity
of the self. We might say that as a result of these linguistic con-
ventions they wind up with a new language. At the first stage
we imagine that English remains their primary language; per-
haps they speak the new language only on special occasions. At
the second and more critical stage of the fantasy we will try to
imagine that English is forgotten and only the new language is
used. Let me indicate how this new language works.
Suppose that A and B are two people who come into physical
contact with each other (say, they shake hands). Then in the
new language the term "person" will denote neither A nor B,
as ordinarily conceived, but will denote instead two individuals
A' and B' who stand to A and B in the following sort of way.
The history of A' will contain all the stages of A's history during
periods when A is not touching another person, together with
the stage of B's history during the period when A and B touch;
correlatively, the history of B' will contain all the stages of B's
history during periods when B is not touching another person,
together with the stage of A's history during the period when A
and B touch.
More generally, suppose that A is a person who throughout
his entire life-history (as ordinarily conceived) makes physical
contact with only the people B, C, D, . . . etc. Then in the new
language the term "person" will denote an individual whose
whole life-history consists of all the stages of A during periods
when A is not touching another person, together with all the
288 THE CONCEPT OF IDENTITY

stages of B during periods when A is touching B, together with

all the stages of C during periods when A is touching C, together
with all the stages of D during periods when A is touching D, . . .
etc. Crudely put, the idea is that in the new language it is a rule
of "personal" identity that when two "people" come into physical
contact, each takes over the physical and mental characteristics
of the other, and then when they cease to be in contact they
again exchange their physical and mental characteristics. (What
happens if someone is simultaneously in contact with two or
more people? Let us stipulate that this will not qualify as contact
in the sense relevant to the rule of identity.) We imagine cor-
relative changes in the use of personal pronouns, and in nouns
subordinate to "person." For example, in the new language it
would be correct to say: "A person must use the word 'I' to refer
to that person."
Let us consider a few examples of this language in operation.
We might call the language "Contacti." Suppose that a man,
who is not touching anybody, approaches a woman, who is not
touching anybody, and they embrace. Let us imagine that the
man is standing to the left of the woman. We can picture the
situation as in Figure i. In the diagram the full arrows repre-
sent our ordinary identity relations, whereas the broken arrows
represent the Contacti identity relations. We imagine that at
1 P.M. neither the man nor the woman is touching anybody; at
2 P.M. they are touching each other; and at 3 P.M. they are again
separate.
This situation could be described in Contacti as follows: "At
2 P.M. a man embraces a woman. At the moment of embrace the
first person, who was previously a man standing to the left,
becomes a woman standing to the right; whereas the second
person, who was previously a woman standing to the right, be-
comes a man standing to the left." When they cease to embrace
the description in Contacti would continue: "The first person
now reverts to being a man, whereas the second person reverts
to being a woman."
Since it is a rule of Contacti that "a person must use the word
'I' to refer to that person," during the embrace the man could
say in Contacti, "A few moments ago I (who am now a man) was
a woman." The woman, in turn, could say, "A few moments
ago I was a man." After the embrace (e.g., at 3 P.M.) the man
 CONSTRAINTS ON SELF-IDENTITY 289

Fig. 1

could say, "A while ago I was a woman embracing a man, and
before that I was a man standing alone." And the woman could
say after the embrace, "A while ago I was a man embracing a
woman, and before that I was a woman standing alone."
Let us suppose, further, that prior to the embrace the man, but
not the woman, was feeling disconsolate. Then during the em-
brace the man could say in Contacti, "This woman whom I am
embracing was feeling disconsolate a few moments ago (when
she was a man), and I can well remember what that feeling was
like." She, on the other hand, would say, "I was feeling dis-
consolate a few moments ago, or so I'm told." But if she were
asked, "Can you remember your feeling of disconsolation?" she
290 THE CONCEPT OF IDENTITY

would reply, "Of course not! How could one conceivably re-
member while touching somebody what one felt when not
touching anybody?" We imagine, in short, that their use of
"remember" (as well as various other psychological verbs) is made
to accommodate their criteria of "personal" identity. 2 To remem-
ber an event is, roughly, to be able to make a judgment that
stands in a distinctive kind of causal relationship to the event.
Now we are certainly not to imagine that speaking Contacti
alters any such causal relationships. The point is rather that the
peculiar concept of "personal" identity in Contacti requires that
the causal chains constitutive of memory be regarded as typically
traversing the boundaries between one "person" and another.
There are many details and niceties here that I skip over. I
think, however, that for present purposes the general idea of
Contacti ought to be clear enough.
We are imagining that in the first stage the speakers of Contacti
still regard English as their primary language. We can even
imagine that they speak English silently under their breaths,
and then translate into the appropriate Contacti sentences. In
the second stage, however, English is completely forgotten. Let
us imagine indeed that a whole new generation of speakers is
raised on Contacti. These people have never even heard of
English, and we will assume that they do not at any level (con-
scious or unconscious) derive their Contacti sentences from
English; they "think in Contacti," as we might put it.
Now here is the difficulty. The speakers of Contacti apparently
describe the identity of the self in an abnormal way. We would
like to say (if we are not conventionalists) that they are wrong
and we are right. But how can they be wrong when their state-
ments are logically equivalent to ours? Suppose that E is a true
English statement, and that C is the crazy-seeming counterpart
of E. Eor example, E might be the statement "I am now touch-
ing somebody (while neither of us touches anybody else), and I
felt disconsolate a few moment ago (when not touching any-
body)," while C is the statement "I am now touching somebody
(while neither of us touches anybody else), and that person felt
disconsolate a few moments ago (when not touching anybody)."
a. For the sort of revision in the concept of memory that would be required
in Contacti, see Shoemaker's notion of "quasi memory," in "Persons and
Their Pasts."
 CONSTRAINTS ON SELF-IDENTITY 2Q1

Since E is true and C is logically equivalent to E, then C must

also be true. This is completely obvious at the first stage, since
at that stage someone who utters C will be silently translating
from E. But if the point holds at the first stage it must apparently
also hold at the second stage. For the advent of the second stage
does not alter the truth conditions of any statements in Contacti.
These statements remain logically equivalent to their English
counterparts.3
The first stage need not puzzle us. At this stage they still think
in English; that is, they still think about the identity of the self
in the normal way. At the first stage Contacti can be dismissed
as merely a strange "code" of some sort. (We might even imagine
that Contacti was introduced during a prolonged war in order
to confound the enemy.) Our puzzle arises, however, at the sec-
ond stage. At this stage they are not thinking about the identity
of the self in the normal (English) way. Yet their statements
remain logically equivalent to ours. How, then, can they be
wrong when we are right? How can a change in language convert
truth into falsehood (or sanity into craziness)?
Someone might object to my characterizing the Contacti
speakers, at the second stage, as conceiving of the identity of
the self differently from the way we do. Why, it might be asked,
should we say that they conceive of the self at all? Since their
use of "person," and correlative terms, is so very different from
ours, perhaps we should not associate these uses with one an-
other; perhaps we should not say that they have any concept
(like our concept) of a person.
But this point is not crucial. It does not matter whether we
say that they have a different concept of a person or no concept

3. Note that I take it for granted as part of the fantasized example that the
Contacti statements are indeed logically equivalent to (have the same truth
conditions as) the English statements that would typically be uttered in the
same contexts. I make no pretense of offering some general (e.g., behavioral)
test which would enable us to verity such an assumption in practice, or to
exclude other possibilities (such as that the Contacti speakers are deluded
about the causal relations between people who touch each other). Of course
my characterization of Contacti will invite various questions about "radical
translation." Some of these questions will be implicitly dealt with in what
follows. But my general assumption is that it makes sense to characterize a
language as containing sentences with certain specified truth conditions, and
terms with certain specified denotations.
2Q2 THE CONCEPT OF IDENTITY

of a person. In either case they evidently do not have our concept

of a person; they evidently do not conceive of the self as we
do, whether they can be said to conceive of the self at all. And
this threatens us with the conventionalist conclusion that it is
arbitrary for us to have our concept of a person, that it is
arbitrary for us to describe the identity of the self as we do.
There is one feature of the Contacti example that I have so far
left open. I have said nothing about their behavior and atti-
tudes. This feature of the example might be filled out in several
ways, but, in order to prepare for a later point, I want at present
to fill it out as follows. I want to imagine that their behavior
and attitudes will seem completely normal from our (English
speaking) point of view. Suppose, for example, that a certain
Contacti speaker is about to shake hands with someone who is
obviously limping with intense pain. Then the first person will
say, "When I shake hands with him I will have great pain," but
that judgment would not trouble him in the least. If we ask
him why the judgment does not trouble him he would reply,
"Because (obviously!), though I will have great pain, the person
whom I will be touching will have no pain." In this sense, their
behavior and attitudes, apart from their speech, remain just
like ours.
If we complete the fantasy in this fashion then we increase
the dissimilarity between their talk of "persons" and ours, since
their "person"-talk and ours would not be linked up to behavior
and attitudes in a comparable way. This would reinforce the
suggestion that perhaps we should not regard them as having
any concept of a person at all, that perhaps we should not asso-
ciate their use of "person" with ours. But, as I said a moment
ago, that point is not crucial, since, in any case, they certainly
do not conceive of persons as we do.
Contacti is evidently just one of an indefinite number of lan-
guages that can be constructed along similar lines. It seems that
we can imagine any number and variety of conceptual schemes
that dispense with our ordinary concept of the self, conceptual
schemes that combine in a single individual the successive stages
(or even the simultaneous states) of what we normally regard as
different persons. Any such conceptual scheme might contain
statements that are logically equivalent to our ordinary state-
 CONSTRAINTS ON SELF-IDENTITY 2Q3

nients about the self. The decisive problem is to explain how it

would be wrong to adopt such a conceptual scheme.

II. Metaphysical Constraints

Certainly our first, and perhaps inexorable, impulse is to insist
that Contacti must be ruled out on metaphysical grounds. To
speak Contacti, it will be said, is simply to misclescribe the ulti-
mate nature of the self.
This response as it stands, however, appears to miss the point.
For let the ultimate nature of the self be whatever you like. Our
problem still remains to explain how it could be wrong to shift
from one statement about the self to another logically equiva-
lent statement. Whatever might be the nature of the self it
remains clear that insofar as our ordinary statements of self-
identity are true, so must be the Contacti statements. So it is
not immediately obvious how our problem can even be addressed
by appealing to some facts about the nature of the self.
Nor should it be suggested that Contacti runs afoul of some
explanatory requirements. For we can explain in Contacti pre-
cisely what we can explain in English. If Ej explains E2, these
being English statements, then Ct explains C2, where these are
the Contacti equivalents of E x and E2. We do not lose our ex-
planations just by putting them into different words.
A suggestion that seems more seriously tempting is that the
second stage of the Contacti fantasy ought to be dismissed as
logically or metaphysically impossible. Though we can indeed
describe (and explain) the facts in terms of Contacti it is, accord-
ing to the present suggestion, impossible that Contacti should
be our primary language. This is because our use of the Con-
tacti concept of personal identity necessarily depends upon our
prior use of the ordinary concept.
But should we concede this alleged necessity? Let us reflect
that according to many philosophers our ordinary concept of
personal identity can be explained somewhat as follows: A suc-
cession of person-stages constitutes a persisting person if the
succession is R-interrelated, where the relationship R consists
of some combination of physical, psychological, and causal con-
tinuity. R, it should be noted, is not itself the relationship of
 THE c
294 ONCEpT OF
IDENTITY

identity, since R relates different person-stages. But we can say,

perhaps, that the ordinary concept of the identity of a person
is based upon R (that personal identity is, perhaps, a "logical
construction" out of R).
Along these same lines it might be suggested that in Contact!
a succession of person-stages constitutes (what they call) a per-
sisting person if the succession is R'-interrelated, where R' in-
volves both the continuity relation R and various facts about
human contact.4 So the Contacti concept of personal identity
is based upon R'.
How could it now follow that our employment of the Con-
tacti concept of personal identity must necessarily depend upon
our prior employment of the ordinary concept? It could be main-
tained, with at least a fair degree of plausibility, that anyone
who has the concept of R' must necessarily already have the con-
cept of R. (Even this may not be completely clear; but I will
assume it for the sake of the argument.) It would be a non
sequitur, however, to conclude from this that anyone who has
the Contacti concept of personal identity (which is based on R')
must necessarily already have the ordinary concept (which is
based on R). Why should it be impossible for someone to pro-
ceed directly from R and R' to the Contacti concept of personal
identity, without taking any detour through the ordinary con-
cept? Why, that is, should it be impossible for someone to pro-
ceed directly from the thought of certain kinds of continuity
and contact relations to the Contacti conception of personal
identity?
The issue I am here raising is related to a certain traditional
question about the nature of the self. Derek Parfit has expressed
this question as follows: "Does personal identity just consist in
bodily and psychological continuity, or is it a further fact, in-

4. Roughly: The person-stage x-at-i, is R' to the person-stage y-at-( 2 if either

(a) x-at-f, is not exclusively in contact with anyone, and y-a.t-t« is not exclu-
sively in contact with anyone, and x-at-« a is R to y-al-t2, or (b) x-at-tt is
exclusively in contact with z-at-ij, and y-at-J 2 is exclusively in contact with
w-at-t,,, and z-at-J, is R to w-at-< 2 , or (c) x-at-t, is exclusively in contact with
%-at-t,, and y-at-i 2 is not exclusively in contact with anyone, and z-at-^ is R
to y-at-tj, or (d) x-at-< 4 is not exclusively in contact with anyone, and y-3.l-tz
is exclusively in contact with w-at-t2, and x-at-^ is R to w-at-t2.
 CONSTRAINTS ON SELF-IDENTITY 295
5
dependent of the facts about these continuities?" In Parfit's
terminology the "Complex View" holds that personal identity
consists in just continuity, whereas the "Simple View" holds that
there is some further fact, independent of continuity. 6 Now I
think it is tempting to suppose that there is a close connection
between this issue of the Complex View versus the Simple View
and the problem of explaining why our ordinary concept of the
self is peculiarly right. One might be inclined to reason as
follows:
"If the Complex View is correct, then a person reduces to a
succession or bundle of momentary stages related by continuity.
Then it does seem that we might have used the word 'person'
to denote some other kind of succession, which is what they do
in Contacti. But if the Simple View is correct, then the identity
of a person is ultimate and unanalyzable, so that the only way to
talk about persons is the way we do."
This line of reasoning may seem convincing, but I question
whether it is really cogent. Consider the following point. If we
accept the Simple View then we regard personal identity as con-
sisting in something other than continuity. There is, however,
nothing in this idea which prevents us from regarding a person
as a succession of stages that are related to each other in some
way. We can even give a name to this relationship, say "person
kinship."7 We can then say, on the Simple View, that a person
is a succession of stages related by person kinship, where person
kinship is independent of continuity. Hence it seems that even
the Simple View would allow us to think of a person as a "mere
bundle" (and would therefore allow us to think of collecting
together other kinds of bundles).
It seems, in other words, that both the Simple View and the
Complex View can accommodate the general formula: A suc-
cession of person-stages constitutes a single person if the succession

5. Derek Parfit, "Personal Identity," in Jonathan Glover, ed., The Philosophy

of Mind (Oxford University Press, 1976), p. 162, ftn. 37; reprinted, with added
footnote, from Philosophical Review, 80, i (1971).
6. For this terminology, see Parfit's "On 'The Importance of Self-Identity,' "
journal of Philosophy, 68, 20 (1971).
7. Cf. Quine's use of "river kinship" and "water kinship" in From a Logical
Point of View, p. 66.
296 THE CONCEPT OF IDENTITY

is R-interrelated. The disagreement only concerns the nature of

R. But this disagreement has no obvious bearing on our question
about the viability of Contacti as a primary language. The
Contacti conception of personal identity is based on the relation-
ship R', where R' is defined in terms of R and contact. It seems
that even on the Simple View we are still left with the question
why it would be impossible for someone to proceed directly
from the thought of R and R' to the Contacti conception of
personal identity.
I am quite certain that many proponents of the Simple View
will want to insist that our ordinary concept of a persisting
person must necessarily precede any concept we form of person-
stages and their interrelations, or any such deviant concept of
personal identity as that of Contacti. But we need to understand
whether this is just another claim, in addition to the claim that
personal identity is not a matter of continuity. It is unclear
how these claims are related to each other.
I am therefore not confident about how to connect the Simple
View to the issue of the present chapter. In any case my own in-
clination is to favor the Complex View. And the Complex View
certainly does not seem to provide any explanation as to why a
language like Contacti must be ruled out as a primary language.
So the upshot of this argument seems to be that the preeminence
of our ordinary concept of self-identity lacks any clear logical
or metaphysical ground.

(Some philosophers might want to appeal to the traditional

distinction between "superficial grammatical form" and "true
logical form" to defend the claim that Contacti is necessarily
dependent on ordinary thought. The superficial grammatical
form of a Contacti identity statement, it will be suggested, dis-
torts the true logical or metaphysical form of the facts; whereas
the superficial form of an ordinary identity statement adequately
reflects the true logical form. This is why Contacti is necessarily
only a code, which must depend upon a language like English,
but could not possibly itself serve as a primary language.
One may question the principle, evidently assumed in this
suggestion, that our grasp of a statement whose superficial form
deviates from logical form must necessarily depend on our grasp
of another statement whose superficial form reflects logical form.
 CONSTRAINTS ON SELF-IDENTITY 2Q7

Actually, however, I suspect that this principle is only a tautol-

ogy, which in effect defines the technical distinction between
"superficial form" and "true logical form" in terms of the idea
that our grasp of one kind of statement necessarily depends on
our grasp of another. Be this as it may, my main response to the
above suggestion is already implicit in my earlier argument.
First of all, if we hold the Complex View, which is my inclination,
then there seems obviously to be no basis for regarding the super-
ficial form of our ordinary English identity statements as espe-
cially suggestive of the "true logical form"; we might then even
be led to say—in the spirit of Hume and others—that both our
English statements and Contact! statements distort logical form.
And second, even if we hold the Simple View—defined in terms
of the independence of identity and continuity—it is still not
clear what we should be led to say about the "true logical form"
of self-identity.)8

III. Pragmatic Constraints

It might now be suggested that we ought to look for a more
"pragmatic" approach to this issue. Even if I have raised doubts
as to whether there are any purely logical or metaphysical
reasons which constrain us to employ our ordinary concept of
personal identity, there are, it may be said, compelling practical
reasons to consider. According to this pragmatic approach our
ordinary concept of the self performs various essential roles in
structuring an individual's relationship to himself and to others.
And to dispense with, or radically to alter, these roles would
be, as a practical matter, unthinkable.9
Take Contacti, for example (our "pragmatist" might con-
tinue); such a language must lead to total havoc. For one of the
essential functions of our concept of the self is to enable us to

8. The flimsiness of our intuitions about "the logical or metaphysical form

of the facts," especially as regards such crucial examples as personal identity,
may generate scepticism about the general usefulness of this traditional notion.
Cf. Chapter 6, Sections IV and VI, and ftn. 9.
9. Two papers which suggest this pragmatic approach are: Terence Penelhum,
"The Importance of Self-Identity," Journal of Philosophy, 68, 20 (1971);
J. M. Shorter, "Personal Identity, Relationships, and Criteria," Proceedings
of the Aristotelian Society, 71 (1970-71).
2g8 THE CONCEPT OF IDENTITY

retain our separate identities in the course of complex social

interactions, including of course physical contact. In Contacti
this would be impossible. People who spoke that language would
feel impelled to touch each other, or not to touch each other, in
ways that are completely irrational or even socially harmful. For
example, if someone were in pain, no one would want to touch
him, including the doctors. On the other hand, everyone would
be anxious to touch the rich and successful (which is already
something of a problem). The whole idea is evidently insanely
unworkable.
It is all too easy to acquiesce to this kind of argument. Con-
sider, however, that I stipulated earlier that the Contacti speakers
would feel and act, apart from their speech, just as we do. I men-
tioned, in particular, that a Contacti speaker would not be at
all troubled by the thought he would express as "When I touch
him, I will be in pain." The previous argument seems to overlook
this stipulation.
To clarify some of the issues here, it might be useful to intro-
duce a certain distinction. When we fantasize about some lan-
guage such as Contacti there are two directions in which we
might try to develop the fantasy, which I will call the emotive
and nonemotive directions. If we develop the fantasy in the
emotive direction then we imagine that a given sentence (e.g.,
the sentence "When I touch him, I will be in pain") tends to be
linked up to the emotions and behavior of Contacti speakers in
much the same way that it is linked up to our emotions and
behavior. Hence, on the emotive interpretation, a Contacti
speaker would be troubled by the thought he would express as
"When I touch him, I will be in pain." Taking the fantasy in
the nonemotive direction, as I did earlier, we imagine that a
given sentence does not have the same emotional significance
for a Contacti speaker as it does for an English speaker. Rather,
on the nonemotive construal, a given sentence tends to have the
same emotional significance for a Contacti speaker as its English
equivalent has for an English speaker. On this construal a Con-
tacti speaker might be troubled by the thought he would express
as "When I touch him, he will be in pain," which is equivalent
to the English statement "When I touch him, I will be in pain."
(More strictly, the Contacti statement "I am not touching any-
body now and when he and I are touching only each other, he
 CONSTRAINTS ON SELF-IDENTITY 2Q9

will be in pain" is logically equivalent to the English statement

"I am not touching anybody now and when he and I are touch-
ing only each other, I will be in pain.")
To take another example, suppose that when I shake hands
with a certain person he squeezes my hand till it hurts. Escap-
ing from his grip I might say "You hurt me when we were shak-
ing hands"; and that fact would make me angry. If I spoke
Contacti, what I would say in that circumstance is "I hurt you
when we were shaking hands." So who would be angry at whom?
On the emotive construal, he would be angry at me. For it would
be he who could say "You hurt me when we were shaking
hands." On the nonemotive construal I am angry at him, just as
if I spoke English.
Taking the fantasy in the nonemotive direction, life goes on
in the normal way; only the words change. The pragmatist must
explain why this would be an impossibility. So far his remarks
seem only to apply to the emotive construal. My main concern,
on the other hand, is with the nonemotive construal.
If we tried to develop the Contacti fantasy in the emotive
direction, then we imagine people whose lives are in countless
ways bizarre and grotesque, from our ordinary point of view.
I am not in fact even confident that we can make the fantasy
fully intelligible in this direction. In any case, my primary ques-
tion is not "Why do we have to live the sorts of lives that we
do?" but rather "Given the sorts of lives that we live, why do we
have to talk (and think) about personal identity as we do?" The
pragmatist must explain why we could not speak a language
like Contacti and go on living as we do. Why should it make a
practical difference whether we utter certain sentences in English
or utter their logical equivalents in Contacti?
The inevitable suggestion will be that a language like Contacti
would have to be too cumbersome and complicated. Consider
one of our moral principles, such as "You ought to be punished
for a bad deed only if you yourself did the deed." This principle
obviously involves our ordinary concept of personal identity.
The fantasized people who speak Contacti, construed now non-
emotively, would also be committed to the content of this prin-
ciple, but would have to express the principle in terms of their
concept of personal identity. This might lead to a very compli-
cated formulation of the principle. And in general, it will be
gOO THE CONCEPT OF IDENTITY

said, all of their talk about persons would have to be too

complicated.
I think that this point needs to be approached with a good
deal of scepticism. What does it mean to say that the Contacti
statements are "too complicated"? Of course Contacti strikes us
as mind-boggling; that is, we have the greatest difficulty in even
grasping how the language operates. This fact is indeed the point
of departure of this whole discussion: we are trying to under-
stand why the language strikes us as mind-boggling. It is not
enough just to repeat this fact.
But perhaps it will be suggested that the Contacti statements
are too complicated in the obvious sense of being too wordy. It
does in fact seem that many of the things that we normally say
would require more words to be said in Contacti. But can this
point by itself bear any serious weight? It must not be supposed
that English is somehow ideally suited to minimize the number
of words that we utter. Obviously we could introduce various ab-
breviations into English which would further condense our state-
ments. And, by the same token, we could undoubtedly introduce
abbreviations into a language like Contacti which would make
the typical statements of that language relatively condensed.
Let me reinforce this last point by mentioning one abbrevia-
tory device that might be available in Contacti. Suppose that the
language contains a symmetrical relationship, say "C-partnership,"
denned as follows: "x and y are C-partners" means: Either x and
y are exclusively in contact with each other (i.e., each is touching
the other and no one else), or x = y and x is not exclusively in
contact with anyone. In other words, I am the C-partner of
anyone with whom I am exclusively in contact, and, as a de-
generate case, I am my own C-partner if I am not exclusively
in contact with anyone.
The reader can now verify that any English statement of the
form "There is a person x such that x is A at ^ and x is B at t2"
is equivalent to the Contacti statement "There are people x, y,
and z such that x and y are C-partners at i1; y and z are C-
partners at tv x is A at tit and z is B at f 2 ."
On the basis of this equivalence virtually any typical state-
ment of English can be rendered in fairly manageable Contacti.
For example, the English principle "You ought to be punished
 CONSTRAINTS ON SELF-IDENTITY 301

for a bad deed only if you yourself did the deed" comes out in
Contacti "You ought to be punished for a bad deed only if (the
person who is) your C-partner at the time of punishment once
had a C-partner who did the deed." Of course this latter formula-
tion remains utterly mind-boggling. And our philosophical prob-
lem is to understand why this is so. My point is that it would
surely be incorrect to answer that it is mind-boggling because it
contains too many words.
Perhaps it will next be suggested (still in a somewhat prag-
matic vein) that even if Contacti is not absurdly impractical it
is absurdly arbitrary. For we are imagining that Contacti is
just like English except for the different rule of personal iden-
tity. Only in this one case do contact relations play their distinc-
tive role. Is it not absurd for a language to contain a rule of
identity which diverges arbitrarily from the general pattern?
Of course this argument may merely encourage us to alter the
Contacti fantasy so that contact relations figure in other identity
rules as well. But even if we stick with the original fantasy the
argument has, I think, little force. Suppose that we confronted
a Contacti-speaker with this argument, in the hope of persuading
him that his way of talking is absurd. Surely we can imagine
him responding that, on the contrary, people are special and
therefore deserve an exceptional identity rule, indeed a rule
which highlights the special importance of human contact. For
the Contacti-speaker it might seem completely natural and in-
evitable that personal identity revolves around contact relations.
Consider, furthermore, that many languages contain various
"irregularities," e.g., of conjugation or pronunciation. These
languages thrive nonetheless. Indeed any student of a foreign
language must be struck by how the nuances of a language, even
(or perhaps especially) its "irregularities," can come to have a
distinctive sense of fittingness and elegance. Surely the same
might be said for Contacti.

IV. Psychological Constraints

So where does this leave us? The preceding discussion suggests
that there may be neither metaphysical nor pragmatic constraints
which determine us to draw the boundaries of the self as we do.
302 THE CONCEPT OF IDENTITY

If this is so, are we then forced to embrace the conventionalist

position? This position would imply that our ordinary concept
of the self is an arbitrary convention, which could be replaced
by various other radically alien conceptions. According to the
conventionalist, a language like Contacti strikes us as impossible,
as mind-boggling, only because of the deeply entrenched habits
of our ordinary thought.
The most vividly implausible consequence of this position is
that we could in actual fact raise our children to speak a language
like Contacti, and they would be none the worse for it. Presum-
ably we (or at least some of us) could train ourselves to translate
very quickly from English into Contacti, and to do this silently
(i.e., under our breaths); so that, if we wished, we could speak
only Contacti in the presence of our children. What would
happen to these children? According to the conventionalist they
would grow up speaking Contacti as their primary language, and
their lives would be quite unaffected by this fact.
I know that many readers will share my own incredulity about
this outcome. Of course the intuitive incredibility of conven-
tionalism may not by itself constitute a decisive reason to reject
the position; but it will certainly prepare us to look favorably on
another alternative.
The alternative to conventionalism which I want to consider
consists in the following hypothesis: We conceive of the self as
we do because this is a basic kind of psychological necessity.
Another way to express this hypothesis is that it is a basic
part of human nature to conceive of the self as we do. Accord-
ing to this position our concept of the self is indeed not con-
strained by some metaphysical or pragmatic considerations; nor,
however, is the concept merely an arbitrary convention. Rather
it is a (more or less) specialized and irreducible fact about our
nature that we must think about the self as we do. Precisely
what this means is of course far from clear; and I do not pre-
tend that I will be able to clarify it fully. But in the remainder
of this chapter I will try to draw out some of the implications
of this idea.
We can, to begin with, associate the idea with some of Chom-
sky's views about "universal grammar." Chomsky holds that, as a
matter of empirical fact, there are certain features which must
 CONSTRAINTS ON SELF-IDENTITY 303

be present in any human (primary) language. These universal

features are said to be determined by some "innate properties
of the mind."10 What this means is that any human being is
innately disposed to speak a language containing the universal
features, though, as Chomsky stresses, both maturational processes
and environmental stimulations might have some role in the nat-
ural development of this disposition.11 Applying these views to
the present topic, we might speculate that, as a matter of em-
pirical fact, any human being is innately disposed to speak a
language in which the boundaries of the self are drawn in the
normal way. Chomsky holds that there is no a priori justifica-
tion for why certain linguistic features, rather than others, must
be universal; this is just a certain kind of contingent fact (per-
haps a very deep kind of contingent fact).12 In the same vein
we would hold on the present proposal that it is essentially a
contingent fact about our human nature that we must speak
about the self in the ordinary way.
This idea might be developed along the following lines. If
two statements are logically equivalent then we can say that
they have the same factual content. Statements having the same
factual content may nevertheless fail, in some sense, to be synon-
ymous (to have the same meaning). One important way in which
this can happen (and the only way that need presently concern
us) is that statements having the same factual content need not
even refer, via their singular and general terms, to the same
things. For example, a statement that refers to triangles may
have the same factual content as a statement that refers only to
the sides of triangles. Let us say that such statements have differ-
ent referential contents, though the same factual content. This
was the relationship that we found to obtain between English
and Contacti statements. For any English statement about the
self there is a Contacti statement with the same factual content.
But no Contacti statement has the same referential content as
any English statement about the self, for no Contacti statement
refers to a person as ordinarily conceived.
Suppose we ask what it is that fixes, that places a constraint
10. Chomsky, Language and Mind, p. 95.
11. Ibid., pp. 8ifE., 72—73.
12. Ibid., pp. 61-62, 88.
304 THE CONCEPT OF IDENTITY

on, the factual contents of our statements. One wants to answer:

the facts, reality, the world. That is to say, if we are going to
have any knowledge at all then we must make true statements
about the world, and this imposes a constraint on the factual
contents of our statements. But what can constrain the referen-
tial contents of our statements? Since for any true statement there
are an indefinite number of statements with the same factual
content, but differing only in referential content, it seems that
the demands of knowledge as such could not fix referential con-
tent. According to the present proposal it is an innate property
of the mind which fixes referential content. Or this is so at least
with respect to reference to the self.
This proposal should be contrasted with the view criticized
earlier in Section II of this chapter, according to which it is
logically or metaphysically impossible for people to speak a
language like Contacti as their primary language. The present
proposal implies instead that the relevant kind of impossibility
is natural or causal rather than logical or metaphysical. It is
logically conceivable that people should speak Contacti as their
primary language, and their view of reality would not therefore
be metaphysically less adequate than ours. But this is, for human
beings, psychologically impossible; Contacti is not the humanly
natural way to think. On the present position we see no meta-
physical significance in the difference between English and Con-
tacti, but we do see psychological significance in this difference.
The psychological significance of our ordinary way of referring
to the self lies most obviously, on the present proposal, in the
fact that we are innately disposed to refer to the self in just
this way. But we might speculate that this does not exhaust the
sense in which our ordinary mode of referring to the self is
psychologically significant. It seems plausible to suppose that
there are various psychological laws that relate our emotions
and behavior to our thought about the self. For example, there
seems to be a law roughly to the effect that a person will tend
to feel angry if he thinks that someone has injured him. Perhaps
there is another law roughly to the effect that a person will tend
to feel guilty if he thinks that he has injured someone. In terms
of the view being advanced, we might formulate these laws as
follows: A person will tend to feel angry, if he accepts as true a
 CONSTRAINTS ON SELF-IDENTITY 305

statement (or judgment) having the factual and referential con-

tent of the English statement "That person has injured me";
and a person will tend to feel guilty, if he accepts as true a state-
ment (or judgment) having the factual and referential content
of the English statement "I have injured that person."
Imagine, somewhat fancifully, that it is possible, by way of
drugs or brain surgery or whatever, to alter the "deep struc-
ture" of someone's language, so that he actually winds up speak-
ing (and thinking in) Contacti. Suppose that such a Contacti-
speaker is injured by someone during physical contact, say, while
shaking hands. After the contact is over, he could say in Contacti,
"I hurt that person when we were shaking hands (and now I
am not touching anybody)," which has the same factual content
as the English statement "That person hurt me when we were
shaking hands (and now I am not touching anybody)." But what
would his emotional response to this situation be? Given the
previous speculation, we would not assume that, just because
his Contacti statement has the same factual content as the anger-
inducing English statement, he would therefore feel anger. For
we would regard it as significant that his statement does not have
the same referential content as the anger-inducing English state-
ment. Nor could we expect him to feel guilty, for his Contacti
statement does not even have the same factual content as the
guilt-inducing English statement "I hurt that person." Since his
thought about the self is abnormal, we could not expect to find
any obvious correspondence between his responses and ours.
Perhaps all we could expect, given his extreme "identity con-
fusion," is that his responses will seem unintelligible to us.
The position I am advancing is evidently in line with some of
Chomsky's views. It is also more vaguely in line with a whole
strand of psychological literature which takes the development o£
a sense of self-identity to be subject to psychological law. To men-
tion one example, which seems fairly close to our philosophical
concerns, Margaret Mahler and her associates announce in one
study the aim "to learn how healthy children attain their sense
of 'individual entity' and identity."13 That question seems pretty
unmistakably related to the present discussion. So does Mahler's
13. Margaret S. Mahler, Fred Pine, and Anni Bergman, The Psychological
Birth of the Human Infant (Basic Books, Inc., N. Y., 1975), p. x.
306 THE CONCEPT OF IDENTITY

assumption that "the innate given is the drive toward individua-

tion."14 I take this philosophical-sounding remark to imply that
the "given" is our innate tendency to conceive of the self in the
normal way. ("Here is where justification comes to an end.")
In Mahler's account, we find that the innate drive toward in-
dividuation requires, for its normal development, various ma-
turational processes, and also various prototypical interactions
with the environment. Where, for one reason or another, the
drive toward individuation is stymied, the result is a "disturbed
sense of identity."15 In cases of "extreme disturbance" the out-
come is psychosis, in less extreme cases, various forms of neurosis.
I shall not here attempt to address the difficult question as to
what extent such terms as "psychosis" and "neurosis" are norma-
tive. I assume, however, that at least the notion of psychosis
contains an important descriptive element, which includes such
things as: being confused and incoherent, being unable to pro-
vide for one's physical well-being, and, typically, suffering in a
peculiarly terrible sort of way. This, on Mahler's account, ap-
pears to be the inevitable outcome for any human being whose
concept of the self is not at least essentially normal. (If we try
to imagine a human being who thinks in Contacti, then we are
certainly imagining someone with an "extremely disturbed"
sense of identity, and therefore someone who, on Mahler's as-
sumptions, would have to feel and behave like a psychotic.)
Mahler's kind of account supports the earlier conjecture that
there are essential linkups between our ordinary concept of
self-identity and our most rudimentary patterns of feeling and
behavior. It is not just that we are innately determined to think
about the self in a certain way, but that this way of thinking is
essentially tied to our sanity.
To summarize, according to the position I am now proposing,
human beings are, as a contingent fact of psychology, innately
disposed to develop an essentially normal concept of the self. I
have also suggested, in keeping with psychologists like Mahler,
that developing a normal self-concept is constitutive of human
sanity. So there is, after all, a deep constraint on our concept of
the self.

14. Ibid., p. 9.
15. Ibid., p. nff.
 CONSTRAINTS ON SELF-IDENTITY 307

V. The Sense of Self

The position outlined above is obviously highly tentative, and
would need to be significantly elaborated. Let me, in this final
section, indicate a few directions for further consideration.
a. I maintained previously in Chapter 8 that our most basic
criteria of bodily identity (e.g., continuity, cohesiveness, boun-
dary contrast) are innately determined. It is tempting to try to
draw a connection between that idea and the position I am now
advocating. One obvious difficulty in the way of drawing such a
connection is that our concept of bodily identity does not appear
to coincide exactly with our concept of personal identity. This
seems to be suggested by "brain-transfer" cases, and the like.
Still, it might be possible, for the purposes of drawing the con-
nection, to downplay such problematical cases and to emphasize
that at least normally our concept of personal identity does in-
deed coincide with our concept of bodily identity. It might seem,
therefore, that the innateness of our concept of bodily identity
(argued in the earlier chapter) could provide essentially all the
psychological resources required to explain the innate basis of
our concept of personal identity.
There are a number of difficulties here, but I want especially
to emphasize one of them. It seems that we can imagine an
"impersonal" language, in the sense of a language which con-
tains no personal pronouns. I want to imagine that, apart from
containing no personal pronouns, the general concept of a per-
son which operates in that language is exactly like ours. Someone
who speaks that language would refer to himself by way of
definite or demonstrative descriptions, such as "the (this) per-
son who is such-and-such," or by way of a proper name. But he
would never refer to himself by way of a first-person pronoun.
This "impersonal" way of thinking about oneself certainly
strikes us as significantly different from our normal way. Part
of the difference can be brought out by reference to the follow-
ing sort of example.16 Imagine that someone judges, "That per-
son whose trousers are on fire is in danger," but, because he does
not realize that he is looking in a mirror and seeing himself,

16. See Hector-Neri Castaneda, "He: A Study in the Logic of Self-Conscious-

ness," Ratio, 7 (1966), p. 141ff.
308 THE CONCEPT OF IDENTITY

he does not judge, "I am in danger." Then, we could not say,

"He believes that he himself is in danger"; nor would we expect
him to act accordingly. It seems that in order for someone to
ascribe a property to himself, in the fundamental sense of be-
lieving that he himself has that property, it is not in general
enough that he should conjoin a definite or demonstrative de-
scription of himself to that property. Nor, by much the same
argument, would it suffice for him to conjoin a proper name of
himself to the property. What is required, apparently, is that
he should conjoin the first-person pronoun to the property, that
he should judge, "I am such-and-such." It seems, therefore, that
the "impersonal" language would not even permit a person to
"think about himself" in the most fundamental sense.
Evidently much more would need to be said to show con-
vincingly that personal pronouns, in particular the first-person
pronoun, are indispensable to our ordinary thought.17 But if
this point is granted, then it follows that in order to conceive
of oneself in the normal way, it will not suffice that one should
have the ordinary general concept of the identity of a person.
Something else required is that one should employ a term or
concept having the special role of "I" (that is, roughly, the spe-
cial role of demonstratively picking out the person who employs
the term or concept). This seems to establish an essential gap
between the innateness theory being proposed in the present
chapter and the innateness theory of bodily identity I proposed
in Chapter 8. It seems the most we could possibly say, to bring
the two theories together, is that the innate disposition to con-
ceive of oneself in the normal way consists in the general dis-
position to adopt our ordinary criteria of bodily identity, to-
gether with the special disposition to employ a term or concept
with the role of "I." Even this might overstate the connection
between the two theories; so, for now, I shall have to leave this
connection open.
b. I hope it is clear that when I suggest that the normal way
of thinking about the self is "natural" or "innately determined"
I am talking only about our most primary and spontaneous level
17. For further discussion of related points, see Shoemaker's critique of the
"disguised description theory" in Self-Knowledge and Self-Identity, pp. 99-
106; sec also John Perry's account of "self locating knowledge" in "Fregc on
Demonstratives," Philosophical Review, 86 (1977), p. 492ff.
 CONSTRAINTS ON SELF-IDENTITY 309

of thought. What happens after that, at more inferential and

theoretical levels, is of course another story. Psychologists like
Mahler frequently employ such expressions as "a sense of iden-
tity," or "a feeling of self," or even "an experience of I" to
convey the immediacy and spontaneity of our judgments about
the self. This point can be brought out by considering the
familiar joke that many people have been certified for saying
some of the things that Descartes said. Descartes's position, on
at least one classic formulation, was that the word "I" designates
the soul. This position would have led Descartes to say such
things as "I am not here in this room. I am invisible and in-
tangible. No one can see me or touch me." The literature of
psychopathology is replete with just such statements.18 The dif-
ference between the pathological cases and Descartes's case is
that presumably Descartes's "feeling of self" was (more or less)
normal. That is, Descartes did not just find himself spontane-
ously thinking that no one could really see him or touch him,
which would be the pathological case, but rather he arrived
at his philosophical judgments by way of a complicated piece of
ratiocination. I imagine that often Descartes spontaneously
thought such things as "The queen has seen me. Let's hope she
doesn't want to touch me." While philosophizing, however, he
would censor these spontaneous judgments (these "feelings") for
the sake of what he regarded as a better theory.
There may be cases which are in a way the opposite of Des-
cartes's, i.e., cases in which someone's considered judgments
about the self are normal even though his sense of self is dis-
torted. Someone who suffers from such symptoms as "deperson-
alization" or "dissociation" may report the inclination to make
bi/arre judgments about his identity, but, if he is not psychotic,
he will "know the truth," which in this context may mean little
more than that he knows the ordinary way to talk. Perhaps it
would be more accurate to say that in these cases the patient's
sense of self is in conflict, in that the patient experiences a con-
flict between two spontaneous judgments, one normal and one
abnormal.
Obviously a great deal of further clarification is required here,

18. See, e.g., the discussion of the "unembodied self," in R. D. Laing, The
Divided Self (Penguin Books Ltd., Middlesex, England, 1965), p. 656:.
310 THE CONCEPT OF IDENTITY

from both a philosophical and a psychological standpoint. Cer-

tainly we need a more secure grasp of what is meant by a
"spontaneous judgment about the self," and the correlative no-
tion of a "sense (or feeling) of self." It may turn out, upon
deeper analysis, that there is no such thing as "the normal sense
of self," but rather a range of significantly different cases that
fluctuate around a central paradigm. Nevertheless I think it is
plausible to suppose that there are severe psychological con-
straints on such fluctuations, that there are severe limits on the
extent to which a person can, within the bounds of sanity, alter
his sense of self. And this leads directly to one final point.
c. Derek Parfit has argued that what matters is not personal
identity as such, but rather those continuity relations which (on
the Complex View) underlie identity. The importance of personal
identity is only derivative, stemming from the links between
identity and the inherently important continuity relations. Hence
if a situation should arise (e.g., the case of the person who
divides) in which our concept of identity has no definite appli-
cation, we can still express our reasonable concerns by refer-
ence to the continuity relations. 19
Let me say that someone has a "Parfitian attitude" insofar as
his judgments of concern are couched in terms of the continuity
relations, rather than in terms of personal identity. Someone
with this attitude might say "How terrible, tomorrow a person
who is continuous with my present state will be in pain!" rather
than "How terrible, tomorrow I will be in pain!"
One of Parfit's points is that we might be confronted with an
exceptional situation (e.g., the case of division) in which it would
be reasonable for us to adopt a Parfitian attitude. Perhaps I
can agree with this. What I want to question, however, is the
pervasive suggestion in Parfit's work that it would be possible,
perhaps even beneficial, for us to adopt the Parfitian attitude as
19. Parfit, "Personal Identity," and "On 'The Importance of Self-Identity' ";
see also "Lewis, Perry, and What Matters" in Amelie Rorty, ed., The Iden-
tities of Persons (University of California Press, Berkeley and Los Angeles,
Cal., 1976).
Parfit maintains that if a person divides into two people in such a manner as
to preserve all of the relevant continuities with each of the resultant people,
then this would not be in any sense that matters a case of ceasing to exist,
even though the original person could not be unambiguously identified with
either of the resultant people.
 CONSTRAINTS ON SELF-IDENTITY 311

our primary orientation. Here I am no longer picturing some-

one who on some rare occasion translates his ordinary identity-
related concerns in terms of the continuity relations, but rather
someone whose most spontaneous and unreflective judgments of
concern are typically couched in terms of the continuity relations.
The sort of hypothesis I have been considering implies that it is
psychologically impossible for a sane human being to have a
Parfitian attitude as his primary orientation. If one has a normal
sense of self then one's spontaneous judgments of concern must
at least typically be couched in terms of the ordinary concept of
self-identity. The Parfitian attitude can be at most an occasional
and sophisticated modulation of the more basic identity-related
orientation.
Let me emphasize that the issue here is not a priori. I am not
saying that it ought to be impossible for us to adopt a Parfitian
attitude as our primary orientation, or that this is a priori
inevitable. I am only saying that, given the sort of psychological
hypothesis I have been advancing, it is, as a matter of fact, im-
possible for human beings to adopt a Parfitian attitude at the
primary level.
My impression is that Parfit, and others who have taken up his
question about "what matters in identity," are not sufficiently
alive to the possibility that the way we are able to think and
feel about identity, at least at the most primary and spontaneous
level, may be severely restricted by psychological constraints
quite unrelated to the terms of philosophical justification. The
machinery of human sanity is complex and delicate. That may
not be an a priori fact, nor even a desirable fact; but it appears
to be a fact. And if my speculations here have been correct, an
essential element, or prerequisite, of sanity is that one should
spontaneously think of the self in a certain way, in essentially
the ordinary way.
This page intentionally left blank
 Index

Analysis of identity, 3-5, 103-5, Body-minded, 172-73, 23gff.

117, 123, 183-84, 201-2 Body-stages, 184-88
see also Identity criteria see also Object-stages
A priori, see Conceptual truth Boundary contrast, 90, io7ff., 307
and metaphysical necessity Bower, T. G. R., 257-58, 279
Articulation, 105-12, 114-15 Broad, C. D., 8
principle of, no, 27311 Brody, Alan, 2i6n
Artificial classes and natural kinds, Butler, Joseph, 161, 233
see Natural kinds
Atomism, 67, 121-23, 157
Carnap, Rudolph, 12in, i6on
Castaneda, Hector-Neri, 3070
Bambrough, Renford, 27011 Causality, see Continuity, causal
Basicness, 178-79, Chapter 6 Change:
passim as continuous, see Continuity
epistemological, 178-79, 182-84, condition of minimizing, 78ff.,
202-10 114—15, 181, 234; see also
metaphysical, 178-79, i82ff., Basic rule
2Ol-2n in which one thing turns into
test for, 187-88 another, 25-27, 51, 83, 148,
Basic rule, 8iff., 114-15, 165-66, 198; see also Existence, coming
2
73 n into and going out of
see also Change, condition of Change-minimizing conflict, 84ff.,
minimizing i n , 165
Bennett, Jonathan, ign Chisholm, R. M., 5gn, 13111, i32n,
Black, Max, 199 161, 233
"Body," 4, 107, 152, 219, 239, 242, Chomsky, Noam, 240, 241, 255n,
271, 272 302-3, 305
different senses of, 97-99 Compositional continuity, see
see also "Object"; "Thing" Continuity, compositional
313
314 INDEX

Compositional criterion, 24-25, Dynamic cohesiveriess, 88, io8ff.,

64-71, 81, 133, 134 237. 239. 250-51, 258. 3°7
see also Continuity, composi-
tional
Compositional stability, 213—16 Einstein, Albert, 217, 230, 232
Concept-dependence test, 187-88, Empiricism, 249-55, 262
Eskimo example, 75-79, 272-74
*93> '94 Events, 4, 5, 8, 14711, 152, 195
Conceptual truth and meta-
physical necessity, 212, 228ff. see also Object-stages
Concreteness, 141-42, 266-67, 271 Evidence of identity, 117, 120
Conflict: Existence, coming into and going
with respect to minimizing out of, 23-27, 47-56, 82-84,
change, see Change- 94-96
minimi/ing conflict
with respect to compositional Factual and referential content,
continuity, 68-71, 224-26 3° 3-5
with respect to self-identity, 309 Features, 103
Constitutive identity, see Identity, Fields, 217, 232, 252
"strict" vs. "constitutive" Figures, 260-61, 271
Contacti, 288ff. Focusing, 255ft.
Contiguity, spatiotemporal, 148-
49> '61
Continuity, Chapter i passim, Geach, P. T., 5811
184-85, Chapter 7 passim Genidentity, 121-23, 157-59
causal, 8n, 135, 218-22 Gettier examples, 190
compositional, 66-71, 216-18 Global contexts, see Continuity,
of global context, 199-201 of global context
qualitative, 7, 10-15, 99~ 1OO > God, 218-19
144, 238 Goodfield, June, i22n
spatial, 81, 52-53, 97ff., 147-48, Goodman, Nelson, 281-83
19811, 237, 239 Group-units, 6, 106, 207-8, 210,
spatiotemporal, 7, 15—21, 144— 245-46, 271
49, 196—201, 208—9, 238, 246 Grue, 282
temporal, 20-21, 67
Conventionalism, 162-63, 178, Histories of bodies, 186-88, 194,
244-49, 262, 286, 290, 292, 302 206
Correspondence rules, 130 Hobbes, Thomas, 68n
Count nouns, 42 Hume, David, 297
Criteria, see Identity criteria

Descartes, 123, 309 Identity:

Dispersiveness, 4off., 95, 113, 116, of bodies (objects), see Basic
165 rule; Continuity; Sortals;
and-, 44—47 Sortal rule; SQ principle
 INDEX 315

of indiscernibles, 199-200 James, William, 252-54

of matter, see Matter
of persons, see Personal identity Kant Immanuel, 11-12, 239-40,
optimal cases of, 211-16 252
"strict" vs. "constitutive," 5gff-, Koffka, K., 106-8, 11 in, 112
118 Kohler, Wolfgang, 106-8, 112,
ultimate cases of, see Persistence, 240, 244-51, 2550, 260
as "ultimate" Kripke, Saul, am, i36n, 228, 265
see also Analysis of identity;
Identity criteria; Perception
Laing, R. D., 3090
of identity; Persistence; Unity "Leibniz's Law," 60
Identity criteria, 3, 117-19, i23ff., Location, 7, 15, 79-82, 95, 115,
183-84, 203, 205-6, 209-10,
125, 129-31
236ff. see also Places; Continuity,
as compared with evidence of spatiotemporal
identity, see Evidence of Locke, John, 67, 132
identity
Logical constructions, 192, 201,
as observational, 119-23 281, 294
primary vs. supplementary, Logical form, 188, 193, 201, 296-97
24-85- 67-68, 133 Logical geography, 202
as sortal relative, see Sortal-
relativity
as theoretical, 121-23 Mahler, Margaret, 305-6, 309
see also Analysis of identity Mass, 134, 155
Identity-deviations, 26-30, 76, 84 Mass nouns, 42
Identity schemes, 22, 44, i4gff. Matter, 42, 92, Chapter 4 passim,
Immediacy, argument from, 203-5 i56ff., 217, 221, 224-25, 230,
Implicit inference, see 240, 252
Unconscious inference M-language, 1398.
Incars and outcars, 32-33, 35-36, Metaphysical necessity, see
Conceptual truth and meta-
76> 79
generalized, 152-55 physical necessity
Individuative terms, 276-77 Metaphysical structure of the facts,
Inductive generalization, 283 see Logical form
Infeld, Leopold, 217, 230, 232 Miracles, 218—19
Innateness, 107, 111-12, 153, 163, Movement:
172, 177-78 as continuous, see Continuity,
innate sense of self, 302-11 spatiotemporal
innate sense of similarity, as discontinuous, 134, 225-26
240-44, 278-80 see also Separate movability
innate sense of unity, Chapter 8
passim, 279-80, 307-8 Natural kinds, i36n, 177, 179,
linguistic, 255n, 302-3 20 in, Chapter 9 passim
Insanity, 306, 309-11 intuitive vs. theoretical, 265, 279
316 INDEX
Natural units, 177-79, 201-211, Persistence-free language, 139ff.
Chapter 9 passim see also M-language; Space-time
Nominalism, 179, 265, 281-85 language
Personal identity, 6, 138, 160, 162,
"Object," 4, 38, 41, 62, 92, 107, 171-72, 177-78, 201—an,
3~ 4> »33> HO. 239, 272
11 1 Chapter 10 passim
different senses of, 97—99 Complex and Simple View of,
see also "Body"; "Thing" 295-97
Objectivity, 142-44, 179-80, 204, Personal pronouns, 307—8
209, 281-85 Person-stages, 160, 293-96
Object-stages, 3-4 Phenomenalism, 142-43, 252
as concrete, 141-42 Phlogiston, 217
as momentary, 104, 139 Piaget, Jean, 256-57, 279
as ordered pairs, 9, 141 Places:
see also Body-stages; Temporal and absolute space, 197
parts as persisting, 145-47
Observational vs. theoretical, identity of, 2in, 145-47, 196-97
126-27, 132, 232-34 occupation of, 15—16
see also Natural kinds, intuitive on objects, 92
vs. theoretical; Perception of overlapping, i6ff., 145
identity; Theory Portions, 91-92, 124-25
Occam's razor, 195-96 of matter, 97, ngff.
see also Simplicity Pragmatic accounts:
of bodily identity, 167-73,
Parfit, Derek, i6on, 17 inn, 294-95, 252-54
310-11 of personal identity, 297-301
Parts and wholes, 35, 194 Predication, 63-64, 266
see also Portions; Temporal Priorities, see Basicness
parts Process-things, 152
Paul, G. A., i8gn Prolongation, principle of, 48-52,
Penelhum, Terence, 29711 101
Perception of identity, 67, 2O3ft., Putnam, Hilary, 13611, 177, 227-
219, 244-48 33, 2490, 265
Perry, John, 3o8n Properties, 6, 264-65, 275, 281
Persistence: see also Natural kinds
of bodies (objects), see Basic
rule; Continuity; Sortal; Qualitative continuity, see
Sortal rule; SQ principle Continuity, qualitative
of matter, see Matter Qualitative homogeneity, 99—100,
of persons, see Personal identity io8ff.
"real" vs. "fictitious," 156-62 Qualitative stability, 213-16
as "ultimate," 5, 119-23, 156, Quality space, see Innateness,
160-61, 295 innate sense of similarity
see also Identity; Unity Quine, W. V., 75, ii2n, 141-43,
 INDEX 31?
151, i6on, 168, 172-73, as applying in two senses,
238-39. 240-44. 248. 254. 259. 62-63
265, 266-67, 269-70, 271-74, definition of, 37-38
278, 280, 281-84, 29511 phase vs. substance, 53-56, 63-
Quinton, Anthony, 24, 101-2, 64, 94, 103
2ogn, 22211, 240, 264, 270n Sounds, 143-44
Space, see Continuity; Places, and
Randomness, 22 in absolute space
Reference, see Factual and Space-time language, i42ff., 151-
referential content; Rigid 53, 168-69, 173,247-48,
reference 25!~54
Reichenbach, Hans, i2i~22n, 16011 Space-time paths, 9
Reid, Thomas, 161, 233 Spatial continuity, see Continuity,
Relational properties, 80, 86, 88 spatial
Relativism, 155, 160, 162-63, J ?8 Spatiotemporal continuity, see
Rigid reference, 227ft., 265 Continuity, spatiotemporal
Russell, Bertrand, 7-8, i2n, Specious present, 205
117-18, 134-35 SQ principle, Chapter 7 passim
Star Trek example, 223, 224, 232
"Seeing as," 247 Stereotypes of identity, 177,
Self-completeness, 80 227-35
Self-identity, see Personal identity Strawson, P. F., 103, 143-44, 152,
Sense data, 142, i8gn, 209, 251, 257 18711, 2080
Separate movability, 86, io8ff. Strict identity, see Identity,
Shadows, 219 "strict" vs. "constitutive"
Shape, regularity of, io8ff. Substance, 5, 123, 2i7n, 240
Shoemaker, Sydney, 3n, 5811, 5911, Substantiality, 5
i23n, i3in, 161, 168, i 7 i n n , see also Concreteness
219-22, 231 n, 246-48, 286n, Substitutivity, 60
2gon, go8n
Shorter, J. M., ag7n Temporal continuity, see
Similarity classes, 269-70, 281 Continuity, temporal
see also Natural kinds Temporal order, 204
Simple Continuity Analysis, 8-10, Temporal parts, 100-102, 141-42,
15, 21, 22-33 157-58. 188-92
Simplicity, 125, 129—30, 137, 147, as ordered pairs, 192
158-60, 195-96, 201, 221, Temporal slices, 8
268-69, 278 Theory, 12 iff., 156-60, 166-67,
Solipsism, 143-44, 256-57 217, 221, 225, 230, 252,
Sorites, 215 277-78, 309
Sortal-relativity, 72ff., 272-74 Theseus, ship of, 68-71, 224-25
Sortal rule, 36 "Thing," 7-8, 239, 24.2, 271, 272
addendum to, 52-53, 83 different senses of, 97—99
Sortals, Chapter 2 passim, 272 see also "Body"; "Object"
318 INDEX
Toulman, Stephen, 1220 see also Innateness, innate sense
Translation, 248, 29111 of unity
Unity-making relationship, 4, 7-8,
Ultimate reality, 5, 182-83 104, 116, 122, 157-58, 160, 162
see also Logical form; Persist- Universal grammar, 302-3
ence, as "ultimate" Urmson, J. O., 155-56, 201 n
Unconscious inference, 2O5ff.
Unity: Verbal questions, 189-93
and similarity, 177, 240-44,
Chapter 9 passim Whitehead, A. N., io2n, i47n
criteria of, see Identity criteria Wiggins, David, 311, 37n, 45n, 53,
through space, 6, 97-112, 140, 55n, 58n, 59n, 6gn, 73, 272-74
237 Wisdom, fohn, 192
through time and space, 97-105, Wittgenstein, Ludwig, 155, 247,
111-12 270