Personal Identity

views updated


One of the commonest of daily experiences is that of recognizing our friends. A less common, though still fairly familiar, experience is the decision that a certain person is or is not the person he claims to be. The problem of personal identity is that of clarifying the principles behind these indispensable processes of reidentification. To reidentify someone is to say or imply that in spite of a lapse of time and the changes it may have wrought, the person before us now is the same as the person we knew before. When are we justified in saying such a thing, and when are we not?

The Basic Problems

Some philosophers have said that we are never justified, because sameness and change are, in themselves, incompatible. They have argued that it is almost paradoxical to say that something has changed and yet is still the same. There is nothing special about the case of persons in this connection, except, of course, that we might, as persons ourselves, be expected to be more concerned about this case or to have access to some of the facts needed to deal with it. One set of such facts is the private set of thoughts, feelings, and images that each of us has, and such philosophers as David Hume have emphasized how constant and rapid are the changes in them with which our identity has to contend. The problem generated by this alleged paradox will be referred to as the problem of the unity of a person through change or, more briefly, as the problem of unity.

Most discussions of personal identity, however, have taken it for granted that sameness and change are, at least, often compatible and have concentrated on the conditions under which reidentification of persons can take place. What enables us to say, in spite of the changes wrought by time, that person A, before us now, is the person B whom we formerly knew and that person C, also before us now, is not?

The problem of the conditions for reidentifying persons should be distinguished from the problem of individuating persons. To individuate among a class of beings is to pick out one from another; to reidentify a member of a class of beings is to recognize him as the same as someone known at an earlier time. It is, of course, unlikely that these two notions can be kept separate, since, on the one hand, one has to be able to pick out a being from among his contemporaries before one is able to identify him with a past member of his class (which, in turn, had to be picked out) and since, on the other hand, it is hard to see how a being that exists in our world of time and change can be picked out, at least in the deeper sense of being recognized, without being picked out as a being with a certain history. It is not accidental that the word identify can sometimes mean the one procedure and sometimes the other. This article will be concerned directly only with the problem of the reidentification of persons, which will be called the problem of criteria.

It has had two main competing answers. One is that the criterion of the identity of a person is the identity of the body that he hasthat it is either a necessary or a sufficient condition of saying correctly that this person before us is Smith that the body this person before us has is the body that Smith had. The other answer is that the criterion of the identity of a person is the set of memories he hasthat it is either a necessary or a sufficient condition of saying correctly that this person before us is Smith that he should have memories of doing Smith's actions or of having Smith's experiences.

It is clear that in practice we settle problems of identification in both ways. But we can still ask of each one whether it is necessary or sufficient; we can ask whether each is independent of the other; and we can ask whether one is more fundamental than the other. It is in connection with these questions that we find what are usually called puzzle cases. These are stories, sometimes true but usually imaginary, which are thought to contain prima facie conflicts between the two criteria. In deciding how the conflict is to be resolved, it is thought that we show the order of priority of the two criteria. For instance, there are the cases of ostensible "bodily transfer," like that of the cobbler and the prince mentioned by John Locke. In this story what physically seems to be a cobbler wakes one morning with all the apparent memories of a prince, with no knowledge of shoe mending, and with disgust at his present sordid surroundings. We might make the story harder by imagining that at the same time what looks like the prince wakes up in the royal palace with cobbler memories. In a story like this, persons seem to recall actions and events associated with a body other than the one they now have. Should we say that they are the persons their supposed memories suggest they are or the persons they physically seem to be? To decide this entails deciding on the relative importance of the two criteria of identity.

Related Issues

The two problems I have distinguished are bound to and do overlap in the literature. The difficulty and importance of the question of personal identity, however, are greatly increased by the fact that it lies at the point of intersection of several major lines of philosophical inquiry.

influence of dualism

The problem of personal identity has traditionally been raised in a dualist context. Those who have discussed it have been greatly influenced by the picture of a person as composed of two entitiesbody and mindwhich are only contingently related to each other. This has restricted the problem of unity so that it has become the problem of how one can be justified in attributing unity to the mind. This looks much harder than the problem in its more comprehensive form, since the thoughts, feelings, and images a person has are far less stable than is his body and since it is, to say the least, not easy to find what Hume calls "the bond that unites" them. Failing to find it, a philosopher may resort to a doctrine of spiritual substance and say that within each person there is some central component that preserves his identity because it never changes as his thoughts and feelings do; the philosopher must then decide whether this component can be detected by introspection or is unknowable. If he rejects this doctrine, as Hume did, he may give way to complete skepticism about identity.


The second issue with which the problem is involved is the relation between the knowledge a person has of himself and the knowledge that others have of him. There are a great many facts about a person that others can learn, it is often said, only by inference but to which he himself seems to have direct and privileged access. The usual examples are facts about his present thoughts, feelings, and intentions. But it looks as though something similar may be true about the past. Although others may have to ascertain whether I am a certain previously known person or did a certain past action by reference to external records or to my observable appearance, I seem to know this directly, in memory. This bears on the puzzle stories. It seems absurd, if we imagine ourselves as one of the participants in these tales, to suggest that someone else might know better than we who we are. If this is really absurd, the puzzles have to be settled in favor of memory; if it is not, we have to explain our natural tendency to want to settle them this way.


A third connected issue is the possibility of survival. If the unity of a person is necessarily connected with the continuance of his body through time, then it is logically impossible for a person to survive the death of his body. If bodily identity is a necessary criterion of personal identity, then even if it could be shown that some nonphysical characteristics of a person continued after his bodily death, the person himself would not have been shown to have survived any more than (to use Antony Flew's example) he would have been if it had been possible to preserve his appendix in a bottle. On the other hand, if bodily identity is not a necessary criterion of personal identity, perhaps bodily death is merely one major event in a person's history and not the end of him. And if the fundamental criterion of identity is memory, it would seem to follow that a person might be known, at least to himself, to have survived death because he continued to have memories in his disembodied state.

moral considerations

The concept of a person has moral connections. Problems of reidentification arise in practice largely when we have to decide questions of right or responsibility, such as right to inheritances or responsibility for crimes. Identity is a necessary though not a sufficient condition of someone's being accorded rights or being made to shoulder penalties. This applies in the afterlife, too. Only if beings who exist after our death can be identified with us can they rightly be held heirs to our merit or blame. A theory of personal identity must take this fact properly into account.

The "Self"

One result of these wider connections has been an unfortunate technical restriction on the language in which personal identity has come to be discussed. It has been referred to as the problem of the self. This word is sometimes used to mean the whole series of a person's inner mental states and sometimes, more restrictedly, the spiritual substance to which the philosopher says they belong. The use of the word self, however, has the effect of confining the question to the unity of the mind and of preventing the answer from relying on the temporal persistence of the body. This has made the unity problem seem intractable, especially when the fleetingness of mental images, feelings, and the like is contrasted with the temporal persistence their owner needs in order even to engage in the relatively lengthy processes of dreaming, reasoning, or scrutinizing the external world. This article therefore avoids a terminology that has ruled out one line of solution ab initio by making it impossible to endow the owner of mental processes with physical characteristics.

By far the most important classical discussions are those of Locke and Hume, and it is therefore useful to begin consideration of the problem of personal identity by reference to their attempts to solve it.


incompleteness of the concept of identity

Locke began his discussion of identity in Chapter 27 of the Essay concerning Human Understanding by pointing out a vital fact that others, including Hume, have since neglected. The concept of identity has to be joined to some substantive notion like that of a tree or a person in order to have any use at all. What makes us say that a given entity is the same depends on what sort of entity it is. This implies an answer to the unity probleman entity of any sort can remain the same throughout its changes provided that the changes that take place in it are characteristic of entities of that sort and are allowed for in their concept. Over the years a tree can double its size and remain the same tree since this sort of change is characteristic of trees and is allowed for in the concept of a tree. It cannot, however, sprout wings and fly or burn to ashes and still remain a tree, for changes of these kinds are not allowed for. This being so, no hidden substance is necessary for the retention of its identity since there is no need of the unchanging character that this is said to provide. The same is true, presumably, of persons, and all that seems to remain is the much harder question of what changes are allowed for in this conceptthe problem of the criteria of identity. Locke characteristically failed, however, to follow through the implications of his own insight. Although he saw the inutility of the concept of substance, he still retained it and led himself into some confusions.

These confusions are partly engendered by his apparent assumption that is it possible to find one single criterion of identity for each sort of being. Our concepts are not as tidy as this. When the assumption is brought to bear on the very untidy concept of a person, the result is a distortion of the concept's logical character. This takes the form of a supposed distinction between "person" and "man."

"man" and "person"

A man, according to Locke, is a certain sort of living organism whose identity depends on its biological organization. On the other hand, he defined a person as "a thinking intelligent being, that has reason and reflection, and can consider itself as itself, the same thinking thing, in different times and places; which it does only by that consciousness which is inseparable from thinking and essential to it." Further, "as far as this consciousness can be extended backwards to any past action or thought, so far reaches the identity of that person." To sever the two notions in this way is a radical departure from ordinary usage, in which the two words are often interchangeable. Locke admitted this, without, however, seeing that the admission conceded that his account must be inaccurate as a description of the two "ideas." Of course, there is a point in the division; behind it lies the recognition that there are two criteria of identity for persons. This Locke tried to accommodate to his belief that for each sort of entity there is one criterion only, by arguing that there are two distinct concepts, each of which has its own unique criterion, rather than one concept with two criteria. But Locke was not trying merely to be tidy; more important is the motive supplied in his claim that "person" is what he called a "forensic" term. A person is a morally responsible agent. It is clear that to establish by physical evidence that the man before us in the dock is the one who did the deed is not sufficient to show that he should suffer the penalty (though it is surely sufficient to show that no one else should, unless he instigated or compelled the deed). Locke wanted to mark this fact by a special restriction on the notion of a person, so that to state that someone is the same person who did the deed is to imply accountability without room for more (or much more) dispute. He thought it obvious that what makes people accountable for their actions is their ability to recognize them as their own. This seems to mean two things: first, an awareness of what one is doing when one is doing it and, second, an ability to remember having done it. Hence, he said that the criterion for the identity of persons, as distinct from men, is consciousness, a concept intended to embrace both awareness and memory. The fact that the same man is before us does not mean that the same person is, since the man may not be conscious of having done the deed in question and if the man is not conscious of having done it, then the person did not do it. Here Locke brought in the puzzle cases:

Should the soul of a prince, carrying with it the consciousness of the prince's past life, enter and inform the body of a cobbler, as soon as deserted by his own soul, everyone sees he would be the same person with the prince, accountable only for the prince's actions. Had I the same consciousness that I saw the ark and Noah's flood, as that I saw an overflowing of the Thames last winter, I could no more doubt that I who write this now, that saw the Thames overflowed last winter, and that viewed the flood at the general deluge, was the same self than that I who write this am the same myself now whilst I write that I was yesterday.

Locke was misconstruing the facts to which he draws our attention. Even granting that only persons are accountable, persons are still men (for men are accountable). We may be morally right in making the memory of crimes a condition for punishment, but memory does not thereby become the sole criterion of identity, for physical presence at the crime is also a condition of responsibility for it. Both the criteria are used together, and the most Locke has shown is that the satisfaction of only one is not, for moral purposes, enough; he has not shown that each serves a different concept. One is tempted to sever them only because of the puzzle stories. These, however, do not represent the conditions under which our concepts have been evolved but, rather, imaginary new conditions that might force us into the decision to change them. As things now stand, we have one complex concept, represented variously by words like "person," "man," or "human being" and embedded in the specific notions of cobbler, prince, beggar, or thief. This concept has two complementary criteria of identity. If we allow ourselves to be forced to say that there are two concepts, each with one criterion, we are saying that our criteria here and now allow us to hold that the memory of a crime, even without physical presence, is enough to establish responsibility for it.

There is a possible Lockean reply to this. It is to say that when a person remembers his deeds but clearly does not have the body that performed those deeds, the deeds can nevertheless still be his because he may have done them in a previous body and have inherited another since. The same person will then no longer be the same man. This cannot be evaluated until we have considered the puzzle cases at some length. For the present let us turn to Locke's attempt to make memory the single necessary and sufficient criterion of personal identity. If this attempt is successful, his treatment of the puzzles is made highly plausible; if not, it becomes highly suspect.

identity as memory

That there is a big difficulty in the problem of identity as memory was clear to Joseph Butler and has recently been very skillfully argued by Antony Flew. Locke wished to say that Smith is the same person who did or witnessed X if, and only if, Smith has the memory of doing or witnessing X. But this is unclear. The verb remember and its cognates have a strong and a weak sense. In the strong sense, to say that someone remembers something is to imply the correctness of his recollection (at least in all but minor details). To say in this idiom that someone's recollection is erroneous is to say that he does not really remember, but only seems to. In the weak sense, to say someone remembers something is merely to say that he sincerely claims to remember it (in the strong sense). In the weak sense, memories can be mistaken. Now, it is clear that even though we do pay special attention to what people claim to remember when settling questions of identity, the fact that someone claims to remember doing or witnessing something does not show that he did it or witnessed it. Even though sincere, he might be mistaken. Thus, to say that Smith is the same person if he has the memory of X must, it seems, mean that he has to remember X in the strong sense of "remember." But here a twofold difficulty arises.

How are we to decide between a genuine and an apparent memory in any given case? The candidate's inner conviction is unreliable. We seem to have to resort to more than the memory claims themselves. And the critical evidence would seem to have to be evidence of the person's physical presence at the scenes he describes. This suggests that the memory criterion is not self-sufficient, as Locke says it is, for in order to know that it is satisfied on a given occasion, we seem to have to use the bodily criterion first.

Apart from this it is much too stringent to restrict personal identity to cases where a person can actually recall his past actions or experiences. People forget. Therefore, we must alter our wording. Smith, we have to say, is the same person who did or witnessed X if, and only if, he could remember it. But what does "could" mean here? Taken in a practical sense, it seems too strong, for this would imply that if Smith did do or witness X, there is some actual set of procedures that, if we applied them, would enable him to recall it. But this may not be so; even psychoanalysts fail. If, on the other hand, "could" is not given this sort of sense, it is hard to see what its use here contributes, unless it is merely another way of saying that Smith is the one who did or witnessed X if, and only if, he is the person to whom the application of procedures designed to induce recollection is appropriate. Unfortunately, this is either straightforwardly untrue (since before we discovered who did or saw X, it would be appropriate to apply such procedures to all likely candidates, not just to Smith) or merely a concealed way of saying that Smith is actually the person who did it, so that no one other than he could remember it. Thus, the concept of memory seems, in this argument, to presuppose that of personal identity, rather than the reverse.

These arguments show that Locke was mistaken in trying to define personal identity in terms of memory because such a definition is necessarily circular. In at least this sense Butler was correct when he said that memory presupposes, and does not constitute, personal identity. Some philosophers have gone on to say that memory is not a criterion of identity for persons at all, since, they say, we cannot know whether someone's apparent memories are real without knowing by physical means that he is the person who was involved in the events he recalls. But this, it will be argued later, is also a self-defeating move. For the present it can be seen that Locke was undoubtedly wrong in holding that memory could be the sole criterion of identity for persons.

spiritual substance

A great deal of the argument of Locke's chapter is designed to reconcile his preference for memory with his doctrine of spiritual substance. The doctrine of spiritual substance is inherited from his view that some doctrine of substance is necessary to account for the fact that the qualities of an object cohere. This is presumably intended to account for their exhibiting a permanent ownership through time, as well as their belonging together in one region of space. Yet Locke denied that we have any knowledge of what substance is like, since our knowledge is restricted to the qualities of things. In the case of persons the doctrine is one of spiritual rather than material substance (whatever the difference between two unknowns may be). But it is clear that nothing whose character is totally unknown can be detectable by the senses or by introspection, so that the doctrine of substance, as Locke held it, cannot provide any answer to the problem of criteria. No one could be said to be applying a concept on the basis of facts to which he has no access. An intractable problem now arises. Granting for the moment that memory is the sole criterion of identity, what is the relation between this fact and that of the existence of the underlying substance? Is it not possible that the application of the memory criterion might lead us to ascribe identity when this was not metaphysically backed by the continuance of one substance? If this should happen, would we have made a mistake?

The most straightforward answer is the paradoxical one of saying that the memory criterion is merely a guide for making identity judgments and that their ultimate metaphysical justification must forever elude uswhich would mean that we could never be more than roughly sure we were punishing the right people for crimes. Locke sought to soften this by two devices. One was to sever "substance" and "person" in the same way that he severed "man" and "person" and to insist that only persons are bearers of responsibility, the concept of substance being obscure and irrelevant. The difficulty with this is that it leaves the doctrine of substance without any connection to those entities whose unity it was supposed to explain. The other device was to say that it is the "more probable opinion" that the consciousness that makes for personal identity is "annexed to" one immaterial substance rather than a plurality and to found the faith in its not being otherwise on the goodness of the Deity "who, as far as the happiness or misery of any of his sensible creatures is concerned in it, will not, by a fatal error of theirs, transfer from one to another that consciousness which draws reward or punishment with it."

But these are no more than devices and have to be used only if we represent the identity of persons as composed of one kind of fact yet recognized through another. For Locke himself, in his early comments on the varying criteria of identity for objects of different kinds, has provided us with a demonstration of the total inutility of the doctrines of substance. We do not need them to account for our ascriptions of identity through change; these rest upon our noticing characteristic patterns of sequence in things. But these patterns do not just supply the criteria for ascribing continuance. They are also the reasons for our doing so at all. In other words, the answer to the unity question lies in the same facts that yield the answer to the criteria question. The invention of substance was intended to explain a practice whose explanation Locke had himself provided in another way. That he did not draw the moral and altogether abandon this invention may in part be the result of his having inherited it from others and in part the result of the incompleteness of his account of the criteria of personal identity.

In Locke, then, we find: one answer to the unity problem in terms of substance and another in terms of the objects' characteristic patterns of change, which renders the first answer unnecessary; a clear recognition of the connection between problems of practical identification and moral responsibility, which is exaggerated to the point of caricature by the separation of the concepts of a person and a man; an unambiguous claim for the priority of the memory criterion of identity for persons, which seems on superficial examination to lead to circularities; and an introduction of the puzzle cases to force a decision in favor of the last claim. With the lessons of Locke's insights and errors behind us, we turn to Hume.


In Hume's famous section on personal identity (Treatise of Human Nature, Book I, Part IV, Sec. 6), we find a treatment of the topic that is, as would be expected, more polished and consistent than that of Locke. But since it is also radically defective, its very tidiness makes it less fertile. It has had a baffling effect on generations of readers because of Hume's ability to destroy metaphysical palliative solutions to problems without uncovering the confusions that give rise to them. This, in turn, issues in a paralyzing skepticism that rendered Hume even less capable than Locke of reaching a clear understanding of the conceptual structure he examines.

Hume began by attacking the spiritual-substance solution to the problem of unity, as it appears in the claim that there is a unique and simple "self" that each person is able to detect within himself. He argued with effective simplicity that he was unable to detect it in himself. He was accordingly forced to conclude that the belief in personal identity, since it lacks this justification, is erroneous. People are "nothing but a bundle or collection of different perceptions" in a constant state of changefor perceptions are all that Hume could detect in himself. In this situation all that a philosopher can do is examine how it is that men (himself included) "suppose ourselves possessed of an invariable and uninterrupted existence through the whole course of our lives." This psychological objective Hume tried to attain by uncovering a basic conceptual confusion that he claimed we all fall into. We fail, he said, to distinguish properly between two thingsthe "idea of an object, that remains invariable and uninterrupted thro' a supposed variation of time" (which is the prototype of identity) and the "idea of several different objects existing in succession, and connected together by a close relation" (which is as good an example as any other of diversity). We confuse these two ideas because of the mental laziness that makes us content with their superficial similarity. Strictly ("to an accurate view"), change destroys identity, but we are easily beguiled into overlooking that change has occurred. Once launched upon this convenient path of error, the mind is led further and further along it by certain recurring factsit is easier for us to overlook than notice gradual changes, changes that are characteristic of certain objects, and changes that occur according to certain smooth and regular patterns, and so we choose to overlook them. Everyone is prone to this error, which therefore acquires the dubious sanction of custom. Sooner or later, however, philosophers arrive on the scene and notice the recurrent paradox in which men have thus involved their thinking. They see both that we do ascribe identity to changing things and that we have no apparent ground for doing this. The result is that since they cannot find such a ground, they invent one. Hence, the metaphysical fancies of substance and the self. But these are hollow solutions; there is no discernible bond uniting a person, though there are sufficient interrelationships between his thoughts, feelings, and memories to explain why we erroneously ascribe unity to him. Hume had no consolation to offer us in this alleged predicament other than his usual one: Even though philosophical constructions cannot justify custom, philosophical criticism cannot dislodge it. For philosophical reasonings have power only in the study, not at the backgammon table.

sameness and change

Given the premise that Hume shared with the philosophers of substance, his conclusions follow only too clearly. This premise is that there is indeed a paradox in ascribing both change and identity to the same subject, since to ascribe change is to deny that we have the same subject. To agree to this is to deny that there can be any genuine solution to the problem of unity and to show that even a substance solution is at best a palliativeand a misleading one. But this is a very odd premise to concede without a battle. It has the extreme, language-destroying consequence that no predicates that cannot be simultaneously ascribed to one subject can be ascribed to a subject at two different times. If it is a mere matter of custom that we violate this principle, at least the custom is indispensable. Surely, much argument is needed to show that the custom is paradoxical and the principle necessary. And there is very little argument in Hume to this effect. His account of the fundamental confusion he claimed to have detected is made plausible only by its vagueness. It looks reasonable to say there is a contrast between one continuing object and a succession of related objects, but this is so only if "object" is tacitly replaced by the same noun in each case. There is a contrast between one continuing note and a succession of related notes (and who would confuse one with the other?) but not between one continuing tune and a succession of related notes. It is by means of the second sort of arrangement, not the first, that we incorporate change into our language. In order to understand the unreality of the contrast that Hume was foisting upon himself, one has merely to recall Locke's principle that "same" is an incomplete term that functions only in conjunction with substantives. There are some conjunctions that would yield the contrast"same note" and "succession of different notes" is obviously one. It is equally obvious that "same tune" and "succession of different notes" is not one. Thus, Hume was wrong to look for the source of the contrast, when it does exist, in the concepts of identity and diversity considered alone. The concepts do not operate alone and yield his conflict only in those cases where they are joined with the right substantives. In most cases it does not exist, because most substantive concepts (including that of a person) are designed to incorporate changes.

There is, of course, one sense of the words same and identical in which sameness and change are incompatible. This is the sense of "same" in which, if applied to two distinct things, it means "alike" and, if applied to one thing at different times, it means "unaltered." This we might call the comparative sense of the word. It is to be distinguished from the numerical sense, in which two things said to be the same are said not to be two, but one. Clearly, one thing cannot be said to be both changed and the same if the comparative sense is intended, but this is not the sense we intend when we wonder whether we are entitled to consider someone the same throughout changes. Once this is noted, we can easily see that there is no need to assume that "to an accurate view" an object has to be the same in the comparative sense to remain the same in the numerical sense. If this is missed, a sense of paradox will be only too easy to sustain.

On the other hand, our concepts do not allow all kinds of change indiscriminately. How much is allowed depends on the concept in question. A man can change in more ways before he is destroyed than a chair can. To know what alterations are and are not allowed is to know, among other things, what the criteria of identity are for the class of entities grouped under the concept in question. These matters may not always be easy to settle precisely. We may not be in a position to say whether we have the same things on certain occasions. When the roof is removed, does the house still exist, or are we left with something else? If the walls are torn down and rebuilt, do we have the same house or another? Sometimes the only answer at such a point is a decision on the scope of the concept. But for general purposes usage over the years has provided us with rough and ready conventions that (this is a truism) language-users know.

Hume was aware of this fact, but the logic of his position forced him to misrepresent it. Instead of presenting us with some general indications of the sorts of change that tend to be allowed under concepts (changes that are gradual, small, functionally absorbable into the whole, and so on), he claimed to present us with the factors which, in his view, beguile us most regularly into the habit of ignoring the changes taking place in objects right under our noses. But these factors (which do not at all conceal the changing character of our world from us) are the same ones that appear without this disguise in a correct account of the situation. It is from a detailed knowledge of the very facts he outlined that we derive the criteria for those very identity judgments that he declared to be always unjustified. This is not the first or last time a philosopher has drawn our attention to facts supposed to support one theory when they in fact support another.

Similar considerations apply to what Hume said about the creation of substance doctrines. It is probably true that philosophers have invented these in order to answer the unity problem, and it is, of course, a merit in Hume that he saw that there is no independent evidence for the truth of such doctrines. But he did not see that the primary objection to them is not that they cannot be shown to be true, but that they are unnecessary. They are invoked to soften a paradox that does not exist. There is no contradiction between saying a thing or person has changed and remains that same thing or person if the changes are characteristic of that sort of thing or of persons. If there is no paradox here, there is no need of any metaphysical postulate to conceal it. If Hume had seen this, he would not have tried to render more palatable the skepticism to which he was led by rejecting the doctrine of substance, for such skepticism could arise only if the doctrine were thought to be both false and necessary. But it is only false. The substantialists do not vindicate the ordinary language-user, and Hume does not convict him. Both have misdescribed what he is doing.


In the specific comments that Hume made about the identity of persons, he was clearly working, as was Locke, in the restricted framework in which "person" means "mind." Only thus can we read his statement that people are nothing but bundles of perceptions. The restriction makes him exaggerate for skeptical purposes the discontinuity he claimed to have discovered in the life histories of personsa discontinuity that does not exist if we include the history of each person's body as well as that of his mind.

But this error hides a deeper one. There is a curious unreality about Hume's discussion of whether we can observe any real bond between the perceptions of a person. This question cannot, of course, be raised unless we can already distinguish between one person and another. Hume, that is to say, was asking whether there is any uniting bond among those perceptions that belong to one person. But why should this question puzzle him if he can already distinguish between those perceptions that belong to one person and those that belong to another? It is at least likely that those features of persons that enable us to distinguish one from another (to individuate) at any one time should also enable us to reidentify people after lapses of time. Yet these features are, and have to be, largely physical ones. For each of us can have (or perceive) only his own perceptions, and without the recognition of the bodies of others, there would be no question of the ownership of perceptions other than one's own ever arising (or, therefore, of the ownership of one's own). In asking his question, Hume was assuming that the perceptions persons are alleged to consist of are somehow known to be in parallel strings, so that the only question remaining is what unites those perceptions that belong on any one string. But if, as he saw, there is no clear psychical factor uniting them, it might still be true that whatever determines their belonging to a particular string also serves to join them together along it. And this, after all, is part of what the body does. His puzzle arises in the form that baffled him only if we first differentiate persons from one another on the basis of their bodies and then, forgetting that we have done it this way, look for some substitute for this principle among the contents of the mind. The principle that the question throws into doubt has to be assumed for the question to be raised.

In Hume, therefore, we find a dismissal of metaphysical construction and an awareness of the general characteristics of the complex facts out of which we forge our criteria of identity. These, however, are rendered completely sterile by the skeptical use to which Hume had to put them. The skepticism is, in turn, the result of a rationalistic oversimplification of the notion of identity that prevented Hume from discovering the muddle at the heart of the unity puzzle and of the dualistic framework of thought within which he worked.

Some Interim Conclusions

We can now draw some conclusions from this investigation of the two main classical discussions of self-identity. The first is that the problem of the unity of persons is a spurious problem that rests upon two errors concerning the idea of identity. One of the errors is the failure to take enough note of the distinction between comparative and numerical identity. The other is the failure to note that the concept of numerical identity works in harness with substantive class concepts that provide those who know how to employ them with rules for making correct identity judgments on entities within their classes.

The second conclusion is that the concept of spiritual substance is not only unverifiable (as Hume saw) but also unnecessary (as Locke saw and Hume did not).

The third conclusion is that the unity problem has acquired a specious appearance of difficulty because of a tacit restriction placed by philosophers on the concept of a person. Since only the psychical components of the person are considered, a picture of change and discontinuity is conjured up that makes the fictitious contrast between identity and change seem even more alarming.

This leads naturally into the fourth conclusionthat it is salutary to remind ourselves that our actual concept of a person is of a psychophysical being. Hence, talk of the criteria of identity for purely psychical beings is not talk of the concept of a person that we actually have. How far they would differ has yet to be decided, but we must at least begin by asking what the actual criteria for embodied human beings are. Here we must bear in mind the apparent circularity of the view that memory is the sole criterion for the identity of human beings. The examination of Locke suggested that in order to apply it some covert reference to the identity of the body has to be made. We must first examine this suggestion with some care.

We shall begin by trying to clarify further the notion of a criterion. It will then be argued that the bodily criterion of identity is in certain important ways more fundamental than the memory criterion in present discourse, although memory is still properly called a criterion in spite of Locke's failure. It could not, however, be the sole criterion. We shall finally consider the puzzles and argue that although they present us with some difficult conceptual decisions, they would not necessitate a change of convention in favor of memory, although this is a possible response to them. An attempt will be made to show that the response, if made, is innocuous, so that the puzzles are devoid of the wide implications philosophers have thought them to have.


Thus far, two things have been meant in calling bodily identity and memory criteria of personal identity. One is that it is by reference to one or the other of these facts about people that questions of identity are usually settled. The other is that practical knowledge of how to settle these questions in these ways is a necessary part of having the concept of a person. More needs to be said than this.

There are two areas where the notion of a criterion has been of special concern in recent philosophy. One is the problem of the knowledge of the mental life of other persons. It has been said by some, following Ludwig Wittgenstein, that we can have this knowledge because people's behavior is able to supply us with criteria for saying correctly that they have certain mental states. The other is the problem of the relationship between judgments of fact and evaluative judgments. It has been said by J. O. Urmson, R. M. Hare, and others that certain facts about things or people serve as criteria for evaluating them as good or bad. In both these cases the relationship the word criterion names is thought to be tighter than an inductive one and yet looser than a deductive one. In this discussion the word will not be used in this sense, since the relationship between bodily identity and memory, on the one hand, and personal identity, on the other, seems to be closer than this; it seems, in fact, to be straightforwardly deductive. In the discussion of Locke we saw that saying someone remembers something in the strong sense entails that it forms a part of his life history. It is now claimed that if a person before us has the body that Smith used to have, it follows that he is Smith.

Two comments are necessary. First, this does not commit us to any view about how we know that the criteria are satisfied. To explain how we discover that this man really remembers or really has Smith's body, it might be necessary to use the notion of a criterion in some other, weaker senseto say, for example, that a certain accumulation of evidence left no more room for reasonable doubt on the matter. But this is another issue. Second, an objection has to be countered. It might be objected that if the relationship between memory or bodily identity and personal identity is deductive, then the criteria are sterile and unusable. For, the argument might go, if either of these facts entailed that this was the same person, we would have to know independently who it was before we could be sure the criterion was satisfied. (This is the objection mentioned in the case of Locke.) This is not a genuine difficulty, but it is instructive. The reason for introducing it can only be the doctrine that if one proposition, P, entails another, Q, then it is impossible to know P without first knowing Q. But this is only a dogma that has to be tested against the facts, which do not bear it out.

The difficulty can teach us, however, that the standard objection to Locke is too simple. Even though the fact that memory entails personal identity prevents us from defining one in terms of the other without a circle, it is still possible that we may sometimes know that a person remembers without having previously checked on his identity. If this were not so, then memory could not serve as a criterion, for it is an additional part of the notion of a criterion, as all philosophers have used the term, that it can be applied. I shall shortly argue that this knowledge is possible.

bodily identity

Some philosophers have said that the bodily criterion is not a criterion at all because there are some occasions in which we find human bodies that are not personsthat is, dead bodies or bodies that are biologically alive but incapable of exhibiting personality. But my thesis is that bodily identity is a sufficient criterion for reidentifying persons and by hypothesis these are not persons. If we are asking whether X before us, who is a person, is the same as Smith whom we once knew, who was a person, it is a sufficient condition of an affirmative answer to know that X 's body is Smith's body.

A more serious-looking argument against bodily identity comes from the puzzles. It might be said that when we use the bodily criterion, we are covertly assuming that there has not been any bodily transfer. This raises an important point of method: How are the puzzles to be treated? We shall treat them as cases of proposed conceptual innovation, as if those who invent them do so to make us imagine circumstances that would force us to change our conceptual habits and rely on one criterion alone, even though we now use two. I have argued that in using two criteria, we have not faced the sorts of problems the puzzles present. If this is right, then no proviso against them can be embodied in our present thinking, even covertly. (If anyone considers that such contingencies are already provided for, then what is said below about the puzzles can probably be transposed into the key needed to examine his view of what sort of provision we make.)

There are several ways in which the bodily criterion is more fundamental than the memory criterion. In the present thesis these statements should seem like truisms.

Although both criteria are sufficient, only bodily identity is necessary. "This is the person who fired the shot" is entailed equally by "This person has the body of the person who fired the shot" and "This person remembers firing the shot"; but although the third statement entails the first statement, it does not entail the second.

The bodily criterion is more extensive. It is a matter of chance that men remember the tracts of their lives that they do remember rather than those that they do not, and we can apply the memory criterion only when there are memories to use. But in a clear sense the bodily criterion can always be used, for the body is present whenever the person is.

The bodily criterion is more varied. There are more ways in which we can determine whether a person is physically the same as someone than there are ways of determining whether his recollections are genuine. There are blood tests, fingerprints, photographs, the testimony of witnesses, and much else. Of course, a candidate's memory claims can be used to support this evidence, just as physical evidence can be used to support memory claims. The resort to physical tests when the memory claims are in doubt, however, is much more nearly inevitable than the resort to memory claims when physical evidence is inconclusive, since there are so many ways of adding to the physical evidence and it is free from the nagging thought that there is more than one way of coming by information about the past.

These examples are enough to show that we should regard overconfident readings of the puzzle cases with some suspicion, since the normal order of priority between our criteria is not what these readings suggest that it is.


It has already been suggested that even though Locke was mistaken in thinking that he could define personal identity in terms of memory, it does not follow that he was wrong to think of memory as a genuine criterion of personal identity. It might be possible to know that someone remembered without first ascertaining in another way who he is. But if this is possible, it has to coexist with the fact that when men's memory claims are in doubt, decision hinges for the most part on physical tests.

One way of trying to relate these two is to say that when we accept a memory claim unchecked, as we often do, we are relying on an inductive connection between the memory claims of a person and the events he refers to. We have found, that is, that this man's memory claims are usually true or, perhaps, that most people's are usually true. We now accept his word on this basis. Sydney Shoemaker has argued persuasively that this is too simple. He has claimed that it is a logical truth that memory claims are usually true, not an inductive one. Following are his arguments: (1) If someone frequently said with sincerity that he remembered events that did not occur, we would be justified in concluding that he did not know how to use the word remember. (2) If a child learning the language were to behave in this way with the word remember or one of its cognates, we would tell him that he had not learned how to use it. (3) If we were translating an unknown language and were inclined to translate certain expressions in it as memory expressions, our decision whether to do so would have to hinge in part upon the truth or falsity of the statements beginning with those expressions; if they were generally false, we could not translate them in this way.

If these arguments are accepted, it should probably be added that in order to understand memory claims at all, we must be able to recognize cases of genuine memory, so that there must be some such cases and also that just as lies and false promises must be in the minority to succeed, so must insincere or mistaken memory claims. These arguments appear enough to refute any generalized skepticism about memory, unless the skeptic is prepared to deny that our language has those features on which these arguments dependthat its users are generally successful in communicating by means of it and that it is learned and not instinctive. We shall not investigate how far it is correct to regard something established by this sort of argument as a "necessary truth." Although the arguments do depend upon features of language that might be argued to be contingent ones, it is still clear that the conclusions are not straightforwardly inductive, and for this reason I shall allow the label to stick.

It is, then, a necessary truth that memory claims are usually true, from which it follows that they can usually be relied upon in practice. But this does not tell us whether any given memory claim is true. The situation here is, rather, that we are justified in accepting someone's memory claims unless there is some reason to doubt them. Only when there is such a reason do we need to check them. It is this that enables memory to serve as a criterion of identity.

But this is a far cry from Locke's theory that memory is the sole criterion. The very facts that show it to be a criterion at all show that it could not be the only one. We must be able to use the distinction between true and false memory claims (even to learn memory language), and this means we must have at our disposal a way or ways of checking the claims that are made. This implies, of course, that we must be able to discover whether the speaker was, indeed, present at that which he describes. Thus, the availability of the bodily criterion of identity is a necessary condition of our having made the distinction between genuine and false memories, even though it often must, from our previous arguments, be in order not to resort to it but to accept memory claims at their face value. Memory is thus a criterion of identity, but it is absurd to suggest it could be the only one, for without the ability to use another we would lack the ability to use it.

This bears out the view that the bodily criterion is more fundamental. There are arguments in Shoemaker, however, which suggest that just as the memory criterion depends on the bodily criterion in the way we have seen, a similar dependence exists the other way. There is a dependence the other way, but it is not a parallel one. The dependence is one found in all cognitive procedures. Unless people had memories, they could not know past facts. If they did not know past facts, they could not know past facts about themselves or other persons, for we have to depend on either our own recollections or those of other witnesses to learn about the past of a human body. At some point memory testimony has to be accepted without further question, and to accept someone's testimony is to accept that he was indeed a witness to some past event. This is true and supplies us with one more argument to show that memory claims must usually be correct, but it does not establish parity between the two criteria because it does not show that in dealing with a problem of reidentification, it is impossible in theory to dispense with the memory claims of the candidate himself. This is possible, however, and is one of the reasons for the greater importance of the bodily criterion.

In spite of this many philosophers have accorded memory greater weight than the bodily criterion. This seems to be a result of what I shall call the "internality" of memory. In remembering, a person seems to have direct, rather than inferential, access to his own past, to know past facts about himself from the inside. This view of memory is reinforced by the fact that most people would admit to having quasi-perceptual experiences in the form of mental imagery when they remember. Most readers unhesitatingly follow the writers of bodily-transfer stories in assuming them to be intelligiblefor how could someone who had systematic recollections of this kind be proved wrong about his own identity by outsiders?

This attitude is not shaken as much as it should be by the fact that in ordinary unsystematic cases we frequently find that even the most vivid recollections are illusory. This is presumably because of the traditional picture of memory as some sort of introspective contemplation of imagery. But what brings memory into the public arena and enables us to use it as a criterion of identity is not this or that sort of private experience but the claims made as a result of it. Indeed, the memory claims of those who deny having memory images are as negotiable in common speech as those of the rest of us. If someone were to claim that he remembered an event and if we were able to determine that he had indeed witnessed it, could give us correct information about it, and could not have come by this information through later research or hearsay, there could be no doubt that he did remember it. The presence or absence, vividness or faintness, of his private images would be of no interest.

It is nevertheless characteristic that when people remember, they have images. If it were not, it is hard to see how the traditional picture of memory could have gained currency. It is true that memory claims are corrigible public claims to knowledge about the past and true that those who make them usually seem to have memory images. It is the first claim that explains why memory has the status it has as a criterion of personal identity. It is the second claim that helps us to understand why some have thought it more fundamental a criterion than it is. For although the subject's unique possession of his images does not confer immunity on the claims he makes, it may have much to do with the fact that he makes them. And it is easy to imagine cases where someone has such experiences and makes the memory claims that they characteristically engender only to find out afterward that these claims are unfounded. This is common enough. It is an easy extension of this to imagine situations in which the events described by such a person did in fact take place, but in the presence of a human body other than the one he has. We then have a typical philosopher's puzzle case. In such a situation characteristic image-laden experiences might take place, and the customary memory claims might be uttered, yet the contextual conditions surrounding correct memory claims would not exist. To allow in some such cases that the speaker really does remember is to change the meaning of this word, but the characteristic intimacy and feeling of conviction that such inner experiences engender might hide this fact from those imagining such examples.

Bodily Transfer

It is now time to look at the puzzles. There are, however, a great variety of these, and without deliberate restriction it is impossible to produce any example of the intricate conceptual decisions involved in them. We shall accordingly leave aside puzzle stories of persons who seem to vanish and reappear or who seem to be reincarnations of someone dead and keep to the case of apparent bodily transfer. What is said here is probably comparable to what could be said in these other cases.

Let us take a story in which the servants in a royal palace waken a person who looks as if he is the prince but who evinces complete bewilderment at his surroundings, utters memory claims befitting a cobbler, is astonished on looking into the mirror, and so on. At the same time a man who looks as if he is the cobbler produces princely reactions and memory claims and demands to be returned to the royal palace. What should we say?

B. A. O. Williams has pointed out that the puzzle cases are harder to state in detail than is usually thought. Are we really able to imagine a person with the cobbler's memories (which will include some acquired skills and personality traits) and the prince's body? I shall ignore this complication, though in fact it tends to support what I shall argue to be the best solution.

The first thing to notice about such a puzzle is that it is puzzling. We are torn two ways over it, as we would expect to be if we have two criteria in apparent conflict. On reflection, however, it is more puzzling because if what I have said above is correct, the bodily criterion is the more fundamental of the two, so that the priorities in present practice would lead one to expect that the puzzle should be settled in its favor. Yet those such as Locke, who invent these stories, take it for granted that our temptations are to settle it in favor of memory. And as far as their judgment of the temptations of most readers goes, they seem to be right. Any answer to the puzzle must take both sides of this paradox into account and try to reconcile them.

priority of bodily criterion

Let us first consider the recommendation that our cobbler-prince episode should make us abandon the bodily criterion in favor of the memory criterion.

Put in this bald way, the proposal is absurd. We have already seen reason to say that memory could not be the sole criterion for the identity of persons because using it requires the availability of another. But this, although true, is far too brusque a reaction to the puzzle, which could be used to argue a more modest proposalto weaken the bodily criterion in certain circumstances.

The advocate of bodily transfer could begin his case by making certain admissions and could then say that they do not destroy the case for it. The admissions would be these.

First, in order to set up any case at all, we have to have someone who now makes memory claims that fit a body other than the one he now has. This requires that he should be reidentifiable as the same throughout the period during which he utters the claims. The claims have to be systematic in the circumstances, so the period has to be considerable. For such reidentification the criterion of bodily identity would be necessary.

Second, in order to set up any case at all, we have to know that there was actually a person in the past about whose life these memory claims seem to be accurate reports and that all the claims fit the life history of the same person in the past, who was the person the claimant now says he is. This can be known only if in the past we were able to reidentify that person over the period of his life. This requires the past availability of the bodily criterion.

But when these admissions are made, the advocate of bodily transfer need go no further; he can hold his ground here and say that bodily transfer is still possible. If we had a case where the memory claims of the man who seemed to be the cobbler systematically fitted the past of the prince and vice versa, these claims could be checked up on in detail. And they would be found, ex hypothesi, to fit a past human body; the only difference from normal would be that the body that they fitted was not the body uttering them. Yet the past of the body uttering them would itself be taken care of by a systematic set of memory claims now uttered by that body which they did fit. In such circumstances it surely would be wholly natural to say that the two men had exchanged bodies.

In spite of much recent writing on the puzzles, there seems to be no satisfactory demonstration that the change in convention that would follow on our saying a transfer had occurred would lead to absurdities. It is therefore a possibility. If we make this decision, we would be forced to so weaken the bodily criterion that we were entitled to infer from its being the same body to its being the same person only if there were no (systematic) memory claims which pointed to its being another person. This would place the two criteria in a position of relative parity, for the memory criterion would hold in normal circumstances subject to bodily checks and the bodily criterion would hold except in those abnormal cases where there were detailed and systematic memory claims that conflicted with the normal reading of the bodily evidence.

Having allowed this, we must now emphasize two things. One is that other readings of these cases could be made, as will shortly be argued. The other is that even the adoption of the bodily-transfer reading of them does not have the exciting implications most have thought.

We have already seen that it lends no support to the view that memory either is or ever could be the sole criterion of identity for persons.

It also does nothing to support the suggestion that people could exist with no bodies at all or to give concrete meaning to the common picture of bodily transfer as someone's going out of one body into another.

Transfer cases, even if allowed, could only be exceptional. If they were not, we would have a world in which the procedures for applying memory concepts would be much more complex than they now are, and virtually impossible to learn. I do not think we could come to learn memory language if the basic use of the word remember were one in which it could refer not only to the past of the body uttering it but also to the past of another body (which, in turn, it could be allowed to "fit" only if it were certain that there were no other systematic memory claims to fit the same period available from that body itself). A concept as epistemologically fundamental as that of memory has to be more easily come by than it would be in this sort of world. But granted that it is simpler and has been learned in more straightforward ways, as at present, then it could be stretched to subsequently cover the exceptional cases.

The conclusion is, therefore, that although the logical possibility of bodily transfer has to be admitted, the implications are small and the wisdom of this particular change in our conventions is not self-evident.

abandoning the memory criterion

We shall now consider the reverse suggestionthat in the face of such a puzzle we abandon the memory criterion and keep the bodily criterion.

It is not immediately obvious what could be meant by this. If it means that we should ignore the memory claims of candidates for reidentification, this is something we could do in any case; the point at issue is the status of those claims when they are considered. If it means that we should reject memory claims that clash with the bodily facts, then this is something we do already and no change in conventions is implied in it. It must mean that we disallow the inference from "He remembers X " to "X formed part of his life history." But the difficulty here is that in order even to gather the bodily facts, we need to learn about the pasts of others, we have to use either our own memories or those of witnesses, and checks on one set of memories, as we saw earlier, require reliance on other sets. So a change of convention here must allow for the continuance of this reliance.

It seems possible to allow for it in only one wayto continue to say that memory claims are generally correct accounts of past actions or events but to add that these actions or events may have formed part of the life of a person other than the one now making memory claims about them. People, in other words, would be allowed to recall events in the lives of others. Two comments may be made here.

For reasons that would parallel those in the previous section, it seems that cases where people did recall events in the lives of others would have to be rare.

Suppose that in spite of his protestations X was just admitted to be the prince because he has the prince's body. He now says, "But I remember mending the shoes last night." Suppose X finally gives in and concedes that he must be the prince although it is still agreed on all sides that the cobbler did mend the shoes last night. X cannot just say, "Oh, I really remember the cobbler's mending the shoes, not myself." This will not do because it fails to distinguish between the new, special case in which one person remembers the deeds of another without having done them (or even having been present) and the familiar case in which one person remembers another's deeds through having witnessed those deeds. It is the second case that would be conveyed by a sentence like "I remember the cobbler's mending the shoes." I am not sure how far this difficulty could be removed by verbal adjustment, but it is at the minimum an inconvenience under the new convention.

The conclusion is as before.

denial that one criterion is satisfied

There would thus seem to be two possible alternative conceptual changes that we could make, each of which would weaken a familiar inference and each of which would be awkward, though not demonstrably impossible. As a matter of fact, however, we already have at our disposal a much simpler device for dealing with such puzzles. Instead of pretending to abandon or to alter one criterion, we can refuse to allow that one of them is satisfied. This need not be thought of as merely a temporary device. If we were to come across odd examples of pieces of iron that did not obey the lodestone but seemed otherwise to satisfy tests for being iron, we could postpone conceptual change for some time by insisting either that the tests had not been properly administered or that it was not really a lodestone. Such moves would become irrational only if maintained in the face of repeated examples. It is hard to admit that the point of irrationality could ever be reached in the present case.

There are clearly only two moves of this sort here. We can deny that it is really the same body, on the grounds that the memory claims it utters fit another, or we can deny that it is really the case that the speaker remembers, because it was not the body before us that was present. Note that neither move involves denying a criterion as the term is being used here. It merely involves refusing to accept that one criterion is satisfied in those cases where accepting that both were satisfied would land us in direct contradiction. There seem to be insuperable obstacles in adopting the first move. For one thing, it would require us, in the case of human bodies, to adopt standards of reidentification that differ from those we accept in the case of all other physical objects. (And if we disregarded this and insisted on behavioral or memory criteria for the identity of human bodies, we would destroy the distinction between a human body and a person.) For another, we would find ourselves led straight into an absurdity. Note again that we are retaining the bodily criterion while making this move. If what is known to be spatiotemporally continuous with the prince's body utters cobblerlike memory claims and if for this reason we say that it is not really the prince's body, we are not able to go on to say that it is, instead, the cobbler's body; for, by hypothesis, it is not spatiotemporally continuous with the cobbler's body and is therefore not the same physical object as that body. Thus, it is nobody's body at all, which is absurd. Hence, we are not able to make the move of denying that it is the body it seems to be. But there is nothing to prevent us from making the other moveof saying that unless the bodily facts at least coincide with the memory claims a person utters, then these claims are false, however closely they fit the past of someone else. This would merely be the determined application to special cases of a procedure we now follow.

We could not, of course, stop there, for we would have to explain how the person came to forget his own past and have so much accurate information about another's. Heroic hypotheses of retrocognitive clairvoyance would have to be brought forward to deal with such strange things. Such hypotheses would have to explain how it was that a person could have information about someone else's past in a manner so phenomenologically similar to the way in which he normally remembered his own. But no greater heroism would be called for here than would be called for by accepting that one person could exchange bodies or memories with anotherfor the second idea would require much the same sort of hypothesis as the one I have mentioned, and the first would make it puzzling that people should remember their own pasts. Of course, each would introduce a difficult conceptual change.

puzzle cases becoming common

But would we not be forced into a conceptual change if such cases became common? For once, the complexities of our problem make it easier to deal with and enable us to give a negative answer. This can be understood from two sides. It has already been argued that either of the possible conceptual changes would require that the cases of bodily transfer or memory exchange be rare; otherwise, we would not have the memory concepts we do have. Yet in order even to state the problem, we must use memory concepts. From the other side, we have to remember that if we were to adopt the device recommended, then in cases like the one in our story we would say of the characters not that they remembered but that they "retrocognized." If such a convention were adopted, however, it would become the appropriate language for the persons to use in such situations. For what makes our problem is what makes the memory criterion possiblethe occurrence of memory claims. These are made in public memory language. If the public language changed so that the inappropriateness of a standardly worded memory claim for such circumstances became generally recognized, then the persons themselves, on discovering that the bodily facts did not fit, would not say that they remembered but that they "retrocognized." Thus, by the time the cases became common, they would cease to exist in the logically puzzling form, because they would cease to be heralded by claims to remember. Pieces of iron do not talk; people are different, and the very data of the puzzles would change if the cases occurred frequently.

primacy of memory

The solution has now to contend with the fact that we do feel a genuine compulsion to read these puzzle stories in some way that favors memory and to say that the claimant himself must know who he is better than others ever could. There are two reasons for this compulsion. One derives from the internality of memory, the other from psychophysical dualism.

On the internality of memory it is enough to repeat that although it is people's public memory claims that relate to decisions about their identity, such claims seem to be made for the most part when people have had characteristic image-laden experiences. Many philosophers consider these to be more closely related to the logic of remembering than they really are, and since the privacy of imagery places reports of it in an epistemologically privileged position, this privilege is erroneously thought to extend to memory claimsoverlooking the fact that memory claims are not reports of imagery. When a person imagines himself being involved in a puzzle story, he supplies himself with vivid and systematic imagery to occasion memory claims that do not fit his present body, and he forgets that the persistence and vividness of the memory could not override the impact of the public physical checks that are a necessary part of the conventions governing memory claims.

As for the theoretical dualism that lies behind so many arguments about personal identity, it has here been argued that however we read them, the puzzles do nothing to support dualism. But the investigation of them has been conditioned in many cases by dualist preconceptions.

Shoemaker correctly remarked that the concept of bodily transfer is compatible with a behavioristic view of the mind, for one might mean, when saying that the cobbler and prince had exchanged bodies, that in the case of each person his distinctive behavior patterns (including his memory claims and behavior) were to be found in a body other than the one in which they used to be found. This is true, but if this solution to the puzzles were urged upon us in conjunction with an overtly behaviorist view of personality, it seems plain that there would be no special obviousness in or compulsion toward this solution as opposed to the others, even though it would still be a possible one. The reason that we all feel some degree of compulsion toward accepting the bodily-transfer solution is that dualist preconceptions intrude themselves when we investigate the stories. It is taken for granted that we have an independently clear concept, with recognized criteria of identity, of a soul, spirit, or mind, which can be thought of as having a purely contingent relationship to the body, which it may abandon in favor of another body. (Locke's phraseology in introducing the puzzle is to the point: "Should the soul of a prince enter and inform the body of a cobbler".) The only available criterion for such a purely psychical being is presumably memory, but we have already seen that it cannot be self-sufficient in the way it would have to be for us to conceive such an entity independently. Yet this is necessary to justify otherwise vacuous talk about such an entity's entering one body, leaving another, and the rest. Anyone feeling impelled to read the puzzles in favor of memory is probably making covert use of this illegitimate picture.

An important objection could now be raised. It might be said that even though much reflection has been infused with a dualist theory, this is a linguistic fact of life that philosophers must accept without complaint, for all language-users, not just philosophers, tend to be dualists. Thus, all language-users, if faced with the puzzles, would tend to opt for the memory solution. If so, how can a philosopher cavil at this solution? For what we should say is usually to be determined by a decision as to what we would say.

This raises the difficult general question of how to react to a misleading theory that has filtered into ordinary discourse. In the present case we could argue as follows. Philosophers such as Gilbert Ryle have exposed many errors and confusions in traditional dualism. But they have spoiled their own case by representing themselves as champions of the common man against the professional philosopher. It is easy enough to show that nonphilosophers are dualists, too. However, the common man is a dualist in the same sense in which the philosopher is onewhen he interprets his own thinking about mental qualities and conduct. What the antidualist arguments show is that laymen misconstrue in their interpretative moments the utterances and thoughts that they engage in in their day-to-day existence. (We could say that all of us are occasionally philosophers, when we think about our ordinary mental concepts, but most of us are bad philosophers because we misinterpret them.) These common theoretical misconstructions, though inconsistent with our daily use of such concepts, are usually harmless because of the merciful logical dispensation that allows us to make good sense with our concepts while talking nonsense about them. Occasionally, however, the prolonged continuance of the misguided theory can infect the practice. One such occasion is the present one, where the tacit appeal to the illegitimate concept of an independently identifiable psychical entity exerts a compulsion upon the reader of a puzzle story to interpret that story as a case of bodily transfer. Here it seems legitimate to replace bad theory by better and to argue against taking this solution for granted. The memory solution the dualist reading implies is at best one competitor among others, and one is led to think it is required only by our use of concepts on more normal occasions if one has misunderstood those occasions.


Of the two problems distinguished at the outset, this article has tried to show that the first, the unity problem, is spurious, since the paradox on which it rests is only apparent. The criteria problem admits of no such clear-cut solution, since it is clear on examination that both the bodily criterion and the memory criterion are ineluctable components of our concept of a person. The bodily criterion is more fundamental, but the memory criterion is, in its own way, indispensable because of the basic epistemological status of memory itself. This is one of the many facets of the irreducibly psychophysical nature of persons. One important result of this conclusion is that it is absurd to consider memory as the sole necessary or sufficient condition of identity. Thus, it would not even seem possible to construct a coherent concept of an independently identifiable bodiless person of whose identity memory would be the sole criterion. It would seem to follow that disembodied survival is logically absurd. It is impossible to decide here whether the doctrine of bodily resurrection fares better. Our examination shows that the puzzle stories can at most embody situations in which the relationships between the two criteria could be altered by conceptual decision. They could not embody situations in which either could be abandoned in favor of the other.

See also Butler, Joseph; Dualism in the Philosophy of Mind; Hare, Richard M.; Hume, David; Identity; Immortality; Locke, John; Memory; Persons; Reincarnation; Self; Self-Knowledge; Shoemaker, Sydney; Williams, Bernard.


The literature devoted explicitly to the problem of personal identity is fairly small, but the amount that is devoted to related questions is immense. These works were made direct use of in the article: John Locke, Essay concerning Human Understanding, edited by Campbell Fraser (Oxford: Clarendon Press, 1894), Book 2, Ch. 27; David Hume, A Treatise of Human Nature, edited by L. A. Selby-Bigge (Oxford: Clarendon Press, 1896), Book I, Part 4, Sec. 6; Joseph Butler, "Of Personal Identity," appendix to The Analogy of Religion, edited by W. E. Gladstone (Oxford, 1897), Vol. I, pp. 385ff.; Antony Flew, "Locke and the Problem of Personal Identity," Philosophy 26 (1951): 5368; Terence Penelhum, "Hume on Personal Identity," Philosophical Review 64 (1955): 571589, and "Personal Identity, Memory, and Survival," Journal of Philosophy, 56 (1959): 882903; B. A. O. Williams, "Personal Identity and Individuation," PAS 57 (19561957): 229ff.; Sydney Shoemaker, "Personal Identity and Memory," Journal of Philosophy 56 (1959): 868882.

The most important classical discussion that is not discussed in this article is Thomas Reid, Essays on the Intellectual Powers of Man (Edinburgh, 1785), Essay III, especially Chs. 4 and 6. Reid has some admirable criticisms of Locke, but he is too wedded to the concept of substance to see that Locke's departures from common conceptual practice are not remedied by the use of it.

Another classic discussion is Immanuel Kant, Critique of Pure Reason, translated by Norman Kemp Smith (London: Macmillan, 1929), pp. 341ff.

There are some recent books whose discussions repay close study. See C. D. Broad, The Mind and Its Place in Nature (London, 1937), Sec. E, pp. 553ff. For stimulating arguments in favor of the notion of a substantial self, see C. A. Campbell, On Selfhood and Godhood (London: Allen and Unwin, 1957). Risierei Frondizi, The Nature of the Self (New Haven, CT: Yale University Press, 1953), contains interesting discussions of Locke and Hume, but its positive discussion seems to be vitiated by the restrictions of the terminology in its title. P. A. Minkus, Philosophy of the Person (Oxford: Blackwell, 1960), is obscure to a degree but has the only extended discussion of Reid. See also Sydney Shoemaker, Self-Knowledge and Self-Identity (Ithaca, NY: Cornell University Press, 1963), and his entry "Memory" in this encyclopedia.

The following articles take positions that radically differ from the arguments of the present article. H. P. Grice, "Personal Identity," Mind 50 (1941): 330350, argues that the "self" is a logical construction consisting of experiences linked conceptually by memory. An authoritative presentation of the Kantian thesis that perceptual acts require a persisting subject or owner is found in H. J. Paton, "Self-Identity," in his In Defence of Reason (London, 1951). J. R. Jones, "The Self in Sensory Cognition," Mind 58 (1949): 4061, attempts to dispense with the notion of a subject of perceptual acts. This paper generated an exchange on the concept of the self between its author and Antony Flew; see Antony Flew, "Selves," Mind 58 (1949): 355358, and J. R. Jones, "Selves: A Reply to Mr. Flew," Mind 59 (1950): 233236. This article has not been able to deal with the detail of the arguments presented in these papers but would hold that each in its own way is handicapped by the restrictions placed on the discussion of personal identity by Hume and by the use of the terminology of the "self." On the perplexities surrounding the notion of a purely mental subject, rather than the psychophysical person, as the owner of mental acts and events, see Ch. 6 of Gilbert Ryle's The Concept of the Mind (London: Hutchinson, 1949), and Ch. 3 of P. F. Strawson's Individuals (London: Methuen, 1959). Both of these books have strongly influenced this article.

Other helpful recent treatments are C. B. Martin, Religious Belief (Ithaca, NY: Cornell University Press, 1959), Ch. 6, and A. M. Quinton, "The Soul," Journal of Philosophy 59 (1962): 393409.

Terence Penelhum (1967)

About this article

Personal Identity

Updated About content Print Article