From "Personal Identity", ed. John Perry (Berkeley: University of California Press, 1975)
Persons, unlike other things, make statements about their own pasts, and can be said to know these statements to be true. This fact would be of little importance, as far as the problem of personal identity is concerned, if these statements were always grounded in the ways in which people's statements about the past histories of things other than themselves are grounded. But while our statements about our own pasts are sometimes based on diaries, photographs, fingerprints, and the like, normally they are not. Normally they are based on our own memories, and the way in which one's memory provides one with knowledge concerning one's own past is quite unlike the way in which it provides one with knowledge concerning the past history of another person or thing. It is largely for this reason, I believe, that in addition to whatever problems there are about the notion of identity in general there has always been felt to be a special problem about personal identity. It is, for example, the way in which one knows one's own past that has led some philosophers to hold that personal identity is the only real identity that we have any knowledge of, the identity we ascribe to ships and stones being only, as Thomas Reid expressed it, "something which, for convenience of speech, we call identity."1 What I wish to do in this paper is to consider how the concept of memory and the concept of personal identity are related. In particular, I want to consider the view that memory provides a criterion of personal identity, or, as H. P. Grice expressed it some years ago, that "the self is a logical construction and is to be defined in terms of memory."2
1. Clearly the concepts of memory and personal identity are not logically independent. As has often been pointed out, it is a logical truth that, if a person remembers a past event, then he, the very person who remembers, must have been a witness to that event. It is partly this logical truth that has led some philosophers to hold that personal identity can be wholly or partially defined in terms of memory. And this view may seem to be supported by the fact that we sometimes use, as grounds for saying that a person was present when an event occurred, the fact that he apparently remembers the event, i.e., is able to give a correct and detailed account of it and does not appear to have anything other than his own memory on the basis of which he could know of it.
But it does not seem, offhand, that these considerations force us to accept this view. For it might be held that while there is a logical relationship between the concepts of memory and personal identity, this is because the former is definable or analyzable in terms of the latter and not vice versa. The assertion that a person A remembers an event X can plausibly be analyzed as meaning (1) that A now has knowledge of X, (2) that A's knowledge is not grounded inductively or based on the testimony of other persons, and (3) that A witnessed X when it occurred. To know with certainty that A remembers X, it might be held, we would have to know all three of these conditions were satisfied, and we could know that (3) is satisfied only if we had a criterion of personal identity by which we could judge that A, the person who now has knowledge of X, is identical with one of the persons who witnessed X. Obviously our criterion of identity here could not be the fact that A remembers X, for we could know this fact only if we had already established that such an identity holds.3
The view just described, I think, must be the view of any philosopher who thinks that the identity of a human body is the sole criterion of personal identity. And this view seems compatible with the fact that sometimes, when we do not have independent grounds for saying that a person witnessed an event, we accept his being able to describe the event as evidence that he was a witness to it. For it might be held that in such cases we are reasoning inductively. We have, it might be said, found out empirically (using bodily identity as our criterion of personal identity) that when someone claims to remember a past event it is generally the case that such an event did occur and that he was a witness to it. On this view it is an inductively established correlation, and not any logical relationship between memory and personal identity, that justifies us in using the memory claims of persons as evidence for identity judgments about them.
2. On the view just described the criteria of personal identity are simply the criteria of bodily identity (i.e., I suppose, spatiotemporal continuity). But it is often argued that bodily identity is not even a necessary condition of personal identity, let alone a sufficient condition, and the same arguments have been alleged to show that memory is a criterion of personal identity. We must now consider some of these arguments.
Considerable attention has been paid, in discussions of personal identity, to so-called "puzzle cases," ostensible cases of what I will call "bodily transfer." It has been argued that if certain imaginable events were to occur we would be obliged to say, or at least would have good grounds for saying, that someone had changed bodies, i.e., had come to have a body that is numerically different from the body that had been his in the past. Locke, it may be recalled, thought it conceivable that the soul of a prince might "enter and inform" the body of a cobbler, "carrying with it the consciousness of the prince's past life," and said that if this happened the cobbler would become "the same person with the prince, accountable only for the prince's actions."4 And it is certainly imaginable that a cobbler, living somewhere in the Bronx, might awake some morning and show great surprise at the appearance of his body, that he might claim to find his surroundings, and the persons who claim to know him, totally unfamiliar, that he might exhibit a detailed knowledge of the past life of Prince Philip, reporting the Prince's actions as his own, and that he might, in his subsequent behavior, exhibit all of the mannerisms, interests, and personality and character traits that Prince Philip had displayed in the past. Let us imagine this happening immediately after the death of the man now known as Prince Philip.
What we say about such cases is clearly relevant to the question whether memory is a criterion of personal identity. If the above case inclines us to say that bodily transfer is possible, this is largely because the cobbler is imagined to be able to describe in detail, thereby giving evidence of being able to remember, the past life of Prince Philip. That this so much inclines us to admit the possibility of bodily transfer, whether or not we do admit it, seems to be grounds for saying that bodily identity is not our sole criterion of personal identity, and that memory, and perhaps also sameness of personality, has a place among our criteria.
Many philosophers have held that personal identity and bodily identity are logically quite distinct. This view is implied by the Cartesian conception of the mind (or soul) as a substance distinct from the body, and it also seems to be implied by the view of Locke, that it is "same consciousness" that "makes" the same person, and by the views of those philosophers, such as Hume and (at one time) Russell, who have held that the persistence of a person through time consists simply in the occurrence of a series of mental events ("perceptions," "experiences") that are bound together by a non-physical relationship of "co-personality" (perhaps the relation "being the memory of"). in short, it is implied by any view according to which the identity of a person is essentially the identity of a mind, and according to which a mind (whether regarded as a Cartesian "spiritual substance" or a Humeian "bundle" of mental events) is something logically distinct from a human body. To hold such a view is to admit the possibility of bodily transfer, and it is partly the prevalence of such views that accounts for the attention that philosophers have paid to "puzzle cases" such as the one I have described. But it is hardly plausible to suppose that those who have held such views have come to hold them because they have been persuaded by such cases that bodily transfer is possible. For even if it is admitted that such cases would be cases of bodily transfer, it by no means follows that personal identity and bodily identity are logically independent. It does not follow that bodily transfer could become the rule rather than the exception, and it certainly does not follow that a person could exist without having a body at all. Indeed, the view that bodily transfer is possible is quite compatible with a completely behavioristic view concerning the nature of mind and a completely materialistic conception of the nature of a person. After all, in the case I have imagined it is bodily and behavioral facts (the behavior of the cobbler and the past behavior of Prince Philip) that incline one to say that a bodily transfer has occurred.
So while such cases provide some grounds for thinking that memory is among the criteria of personal identity, we must look further if we wish to account for the plausibility of the view that the criteria of personal identity are "mental" or "psychological," one version of which being the view that memory is, to the exclusion of bodily identity, the sole criterion of personal identity. But we need not look much further; all that we have to do, in fact, is to describe such cases in the first person rather than in the third person. For it is when one considers the way in which one knows, or seems to know, one's own identity that it becomes plausible to regard personal identity as something logically independent of bodily identity. One does not have to observe, or (it seems) know anything about, the present state of one's body in order to make past tense statements about oneself on the basis of memory. But such statements imply the persistence of a person through time, and it is natural to regard them as expressing knowledge of one's own identity, knowledge that a "present self" (that to which the word "I" refers) is identical with a "past self" (the person who did such and such in the past). One is inclined to suppose that the real criteria of personal identity must be criteria that one uses in making statements about one's own identity. And since it appears that one can make such statements, and know them to be true, without first knowing the facts that would justify an assertion about the identity of one's body, the conclusion would seem to be that bodily identity cannot be a criterion of personal identity. The real criteria of personal identity, it seems, cannot be bodily or behavioral criteria of any sort, but must be criteria that one can know to be satisfied in one's own case without knowing anything about one's body. For similar reasons one is inclined to reject the view that the notion of memory is definable or analyzable in terms of the notion of personal identity. For when one says that one remembers a past event it is surely not the case that one has first established that one is the same as someone who witnessed the event, and then concluded, on the basis of this fact and others, that one remembers the event. That one remembers an event seems, from one's own point of view, a brute, unanalyzable fact. But if there is a logical relationship between the concepts of memory and personal identity, and if the former is not definable or analyzable in terms of the latter, what seems to follow is that the latter is somehow definable in terms of the former, and that memory provides the criterion of personal identity.
3. Whether or not memory is a criterion of personal identity, it is not the criterion. As I will argue later, it cannot be the sole criterion that we use in making identity statements about other persons. And while it is true that one does not use bodily identity as a criterion of personal identity when one says on the basis of memory that one did something in the past, this is not because one uses something else as a criterion, but is rather because one uses no criterion at all.
Suppose that I make the statement "I broke the front window yesterday." If this statement is based on a criterion of personal identity it must be the case that I know that someone broke the front window yesterday, and that I have found out, by use of my criterion, that that person was myself. And my statement must be based, at least in part, on what I know about that person as he was at the time at which he broke the window. Let us suppose that my own memory is my only source of knowledge concerning the past event in question, for that is the sort of case that we are interested in. Then my statement must be a conclusion from what I remember about the person who broke the window yesterday, and perhaps from other facts as well (facts about my "present self"), and my criterion of identity must be what justifies me in drawing this conclusion from these facts. Presumably, if I had remembered different facts about that person I would have drawn a different conclusion, namely that he was not myself. It should be noted that, if all of this were so, then, strictly speaking, it would be incorrect for me to say "I remember that I broke the front window yesterday." For if my statement "I broke the front window yesterday" expresses a conclusion from what I remember it is not itself a memory statement, i.e., is not simply a description or report of what I actually remember. We must distinguish statements that are "based" on memory simply in the sense of being memory statements from those that are "based" on memory in the sense of being conclusions drawn from remembered facts.s If one thinks that one cannot make a first person past tense statement except on the basis of a criterion of identity, one must accept the consequence that no such statement can be a memory statement. In the case at hand, if my statement is grounded on a criterion of identity then what I actually remember cannot be that I broke the window yesterday, but must be that someone of such and such a description broke the window, the assertion that it was myself being a conclusion from what I remember about the person.
Now it is a logical truth, as I have already said, that if a person remembers a past event then he, that same person, must have been a witness to the event, i.e., must have been present when it occurred and in a position to know of its occurrence. So if I remember someone breaking the front window yesterday it follows that I was present at the time. And since, if I remember this, I am entitled to say "I remember someone breaking the front window yesterday," I am also entitled to say "I was present yesterday when the front window was broken." But this last statement is a first person past tense statement, so let us see whether it can be grounded on any criterion of personal identity. Clearly it cannot be. It is not, as it would have to be if based on a criterion of identity, a conclusion from what I know about someone who existed in the past. What I know about the past, in the case we are considering, is what I remember, but this statement is not a conclusion from what I remember at all; it is a conclusion from the fact that I remember something, not from any of the facts that I remember.
But if I can know that I was present when an action was done without using a criterion of identity, why can't I know in this way that I did the action? Is it that I must employ a criterion in order to know which of the persons present was myself? In that case, presumably, I would not need to employ my criterion if I remembered that only one person was present, for that person would obviously have to be myself. But the trouble is that he would have to be myself no matter what I remembered about him. i.e., even if the remembered facts were such that I would have to conclude, in accordance with my criterion, that he was not myself. If I had a criterion of identity that I could use in such cases, it seems to me, it would be possible for me to remember someone doing a certain action, discover by the use of my criterion that he was not myself, and then find, by consulting my memory of the event, that he was the only person present when the action was done. And clearly this is not possible.
It is sometimes suggested that one is able to identify a remembered "past self' as one's own self by the fact that one is able to remember the private thoughts, feelings, sensations, etc., of that self. There does seem to be a sense in which my own thoughts and feelings are the only ones that I can remember. Certainly they are the only ones that I can remember having. But it is a mistake to conclude from this that memory is used as a first person criterion of personal identity. The sentence "I remember having a headache yesterday" does not differ in meaning from the sentence "I remember my having a headache yesterday." But if what I remember when I remember a past headache is my having a headache, or that I had a headache, my statement "I had a headache" is a memory statement, not a conclusion from what I remember, and cannot be grounded on any criterion of identity. If, however, what I remember is that someone had a headache, or that a headache occurred, it is clear that the remembered facts provide no grounds for the conclusion that I had a headache. Nor can we say, as some have said, that the relation "being the memory of" is the relation of "co-personality" between mental events, and that I know that a past sensation was mine because I have established that one of my present mental states is a memory of it and therefore co-personal with it. For, contrary to what Hume and others seem to have supposed, in the sort of case we are considering it makes no sense to speak of comparing one's present memory with a past sensation and finding that the one is the memory of (on Hume's theory, that it resembles) the other. One could make such a comparison only if one knew of the past sensation on some grounds other than one's memory of it, and our concern here is with cases in which one's memory is one's only source of knowledge concerning the past events in question. In such a case, comparing a past sensation with one's memory of it could only be comparing one's memory with itself—and comparing something with itself (if that means anything) is certainly not a way of discovering whether two events are related in a certain way. One can raise the question whether two events are related in a particular way (in any given way) only if one knows of the occurrence of both events. And if one knows of one of the events on the basis of memory, one must, in inquiring whether it is related in some way to the other event, be relying on one's memory of it, and clearly cannot be raising any question as to whether one does remember it (or whether one of one's present mental states is a memory of it). Indeed, if one's knowledge of a past sensation is memory knowledge it is misleading to say that one knows that one remembers a particular past sensation. It makes sense to speak of knowing that one remembers a particular event (knowing of an event that one remembers it) only where it would also make sense to speak of knowing of that event that one does not remember it (as is the case if one's knowledge of an event is based on something other than, or in addition to, one's memory). When I say that I have a headache I am not mentioning some particular headache and reporting, as a fact that I know about it, that it is experienced by me; likewise, when I say that I remember a headache I am not, in most cases, saying of some particular headache that I remember it. Normally I can identify a past sensation only as one that I remember (or, as I should prefer to say, one that I remember having). And when this is so there cannot arise any question concerning the ownership of the sensation, and there is no room for the employment of criteria of ownership or criteria of personal identity.
4. If, as I have argued, one does not use criteria of identity in making statements about one's own past on the basis of memory, the criteria of personal identity must be third person criteria. And if memory were the sole criterion of personal identity it would have to be the sole criterion that we use in making identity statements about persons other than ourselves. It is easily shown, however, that if we did not have some criterion other than memory that we could use in making statements of personal identity we could not use what others remember, or claim to remember, as evidence of any sort (criteriological or otherwise) for identity statements about them.
To begin with, if the word "remember" is to have any meaning it must be possible to establish whether someone is using it using it correctly. If some of the utterances that persons make are to count as memory claims, and therefore as evidence of what they remember or seem to remember, it must be possible to establish what a person means by the words he utters. But establishing what a person means by a term, or whether he is using it correctly, involves observing his use of it in various circumstances and over a period of time. This, of course, involves being able to know that it was one and the same person who uttered a given word on two different occasions, and to be able to know this one must have a criterion of identity. What could this criterion be if not bodily identity? It could not be any "psychological" criterion (such as memory or sameness of personality), for the use of such criteria (if criteria they are) involves accepting what a person says as indicating what his psychological state is (e.g., that he seems to remember doing a certain thing), and one could not do this if one were trying to establish what he means by, or whether he understands, the expressions he is using. In some circumstances, at least, bodily identity must be a criterion of personal identity.
Moreover, memory claims can be mistaken, and there must, accordingly, be such a thing as checking on the truth of a memory claim, i.e., establishing whether a person remembers something without taking his word for it that he does. And this, if he claims to have done a certain thing in the past, would involve establishing whether he, the person who claims this, is the same as someone who did do such an action in the past. In establishing this we could not use memory as our criterion of personal identity, and it is difficult to see what we could use if not bodily identity. And if, in such cases, we could not use bodily identity (or something other than memory) as a criterion of identity, it would not be possible to establish whether someone understands the use of the term "remember," and that term could not so much as have a meaning. It is, I believe, a logical or conceptual truth, not a contingent truth, that memory beliefs, and therefore honest memory claims, are generally true.6 If someone frequently prefaced past tense statements with the words "I remember that," and these statements generally turned out to be false, this would be grounds for saying that he did not understand the use of these words. We would not think that we had succeeded in teaching a child the use of the word "remember" if he commonly said "I remember doing such and such" when he had not done the thing in question. Again, suppose that we had discovered a new people whose language we did not know, and that someone had proposed a way of translating their language that involved regarding a certain class of statements (or utterances) as memory statements. Clearly, if all or most of those statements turned out to be false if translated as proposed, there could be no reason for accepting that way of translating them as correct, and there would be every reason for rejecting it as mistaken. But if it is a conceptual truth that memory claims are generally true, establishing that someone understands the use of the term "remember" must surely involve establishing whether his memory claims (or what appear to be his memory claims) are true or false. And to be able to do this we must have something other than memory that we can use as a criterion of personal identity.
5. The arguments of the last section may seem to give support to the view that bodily identity is, to the exclusion of memory, the sole criterion of personal identity. But this view seems to me to be mistaken. Bodily identity is certainly a criterion of personal identity, and if it were not, I have argued, nothing else could be so much as evidence of personal identity. But I do not think that it can be the sole criterion, and I think that there is an important sense in which memory, though certainly not the sole criterion, is one of the criteria.
Let us consider one consequence of the view that bodily identity is the sole criterion of personal identity. As I said in section 1, if this view were correct it would have to be the case that we are reasoning inductively when we use the fact that someone claims to remember something as grounds fora statement about his past. It would be a contingent fact, one that we have discovered empirically, that most memory claims are true, or that people generally remember what they claim to remember. This would, indeed, be nothing other than the fact that the memory claims that issue from the mouth of a certain body generally correspond to events in the past history of that same body. But I have argued that it is a logical fact, not a contingent fact, that memory claims are generally true. If this is so, inferences of the form "He claims to remember doing X, so he probably did X" are not simply inductive inferences, for they are warranted by a generalization that is logically rather than empirically true.7
Now let us return briefly to the case of the cobbler and the prince. If one is inclined to use the memory claims of the cobbler as grounds that he is (has become) the prince, the inference one is inclined to make is not of the form "He claims to remember doing X, so he probably did do X," but is of a more complex sort. Roughly, it is of the form "He claims to remember doing X, Y, and Z under such and such circumstances and at such and such times and places, and X, Y, and Z were done by someone under precisely those circumstances and at those times and places, so there is reason to believe that he is the person who did those actions." But it seems to me that if inferences of the first sort are not inductive, neither are inferences of the second sort. And I think that to say that inferences of the second sort are legitimate (as they certainly are, at least under certain circumstances), and that they are non-inductive, is tantamount to saying that memory is a criterion of personal identity.
It should be noted that if such inferences were merely inductive, and if bodily identity were the sole criterion of personal identity, it would be patently absurd to make such an inference in a case in which the body of the person making a memory claim is known not to be identical with the body of the person who did the action that he claims to remember. The absurdity would be that of asserting something to be true, or probably true, on the basis of indirect evidence, when one has direct and conclusive evidence that it is false. But in the imaginary case I have described, the claim that the cobbler is (has become) the prince does not, I think, strike us as having this sort of absurdity.
I have not attempted to say whether, if the events I have described were to occur, it would be correct to say that the cobbler had become the prince, and I do not know how this question could be settled. But this in itself seems to me significant. The fact that such cases so much as incline us to admit the possibility of bodily transfer, or leave us in doubt as to what to say, seems to me prima facie evidence that memory is a criterion of personal identity. It is not as if our doubts were due to ignorance of empirical facts that, if known, would settle the issue. Doubts of that sort are easily removed, for we need only add further details to the description of the case. But if, knowing all of the relevant facts, we are in doubt as to how we should answer a question of identity, this is surely an indication that the case is such that the question is not unambiguously decidable by our criterion of identity. This, in turn, suggests that there is a conflict of criteria. In the case at hand, our doubts are evidence that one criterion of personal identity, namely bodily identity, is in conflict with another, namely memory.
But now I must try to meet an objection. It might be argued that while the inference "He claims to remember doing X, so he probably did X" is not inductive, we are nevertheless reasoning inductively when we take what a person says as evidence for a statement about his past history. For what justifies us in taking the sounds that a person utters as expressing a memory claim? As was argued earlier, if a question arises as to whether a person understands the use of the word "remember," or is using it to mean what we mean by it, the question can be settled only by establishing, independently of what he says, whether the things that he claims (or apparently claims) to remember are things he actually did, endured, or witnessed in the past. If in a number of cases it turns out that the actions that he apparently claims to remember having done are actions that he actually did, this is evidence that he does understand the use of such words as "remember," and that his apparent memory claims are really memory claims and can generally be relied upon upon. Must it not be much the same sort of considerations, i.e., our having observed certain correlations between the sounds that people utter and what they have done in the past, that justifies our general reliance on people's memory claims, or rather our acceptance of people's utterances as memory claims? If so, it would seem that our use of people's memory claims as evidence for statements about their own pasts, including identity statements about them, is, in the end, inductively based. Though it is a logical fact that memory claims are generally true, what does this come to except the fact that if there did not exist correlations of the sort mentioned none of the utterance of persons would be memory claims? But the existence of such correlations is a contingent fact, and it is on this contingent fact, it might be argued, that inferences of the sort "He claims to remember doing X, so he probably did X" are ultimately based. As for the case of the cobbler and the prince, it might be argued that if what I said in section 4 is correct then the facts that I have imagined would be evidence, not that the cobbler had become the prince, but rather that his utterances were not memory claims at all, and that he did not understand the use of the term "remember."
To take the last point first, suppose that we were in doubt as to whether the cobbler really understood the words that he was using. Could we not satisfy ourselves that he did by observing his subsequent behavior, and by establishing (using bodily identity as our criterion of personal identity) that when he claims to have done an action that occurred after the alleged bodily transfer it is generally the case that he did do that action? When we are trying to establish whether a person understands the words he utters we must, I have argued, use bodily identity as a criterion of identity, but it does not follow from this that there cannot, in exceptional cases, be personal identity in the absence of bodily identity.
As for the rest of the objection, it is certainly true that unless there existed certain correlations between the sounds people utter and events in the past histories of those who utter them it would be impossible to have knowledge of the past that is based on the memory claims of other persons. These correlations are those that must exist if any of the utterances that people make are to be memory claims. But it cannot be the case, I believe, that we regard certain of the utterances of other persons as memory claims because we have established, inductively, that such correlations hold. To be sure, from the fact that a person utters the sounds that I would utter if making a certain memory claim it does not necessarily follow that he speaks the language that I speak and means by those sounds what I would mean by them. Under exceptional circumstances I might raise a question as to whether what sounds to me like a memory claim is really one, and such a question could be settled empirically, by observing the behavior of the person who made the claim. But except when we have definite grounds for supposing the contrary, we must, I believe, regard other persons as speaking a language, our own if the words sound familiar, without having any general empirical justification for doing so. Let us consider whether it would be possible for me to question whether there is anyone at all (other than myself) who speaks the language that I speak, and then to discover empirically, by observing correlations between the sounds people utter and their present and past behavior, that those around me do speak the language that I speak and that certain of their utterances are memory claims and can generally be relied upon. In carrying on such an investigation I would, of course, have to rely on my own memory. But one's memory can be mistaken. It is essential to the very notion of memory that there be a distinction between remembering something and merely seeming to remember something. And for there to be such a distinction there must be such a thing as checking up on one's own memory and finding that one does, or does not, remember what one seems to remember. As Wittgenstein pointed out,8 there are and must be circumstances in which we would accept other sorts of evidence concerning the past as more authoritative than our own memories. But an important--I think essential--check on one's own memory is the testimony of other persons. And this sort of check would not be available to me if I could not even regard the utterances of other persons as testimony until I had completed my investigation and established the required set of correlations. Unless there were some persons whose utterances I would be willing to accept as memory claims without having conducted such an investigation I would in effect be admitting no distinction between finding the correlations and merely seeming to have found them.
It is, I should like to say, part of the concept of a person that persons are capable of making memory statements about their own pasts. Since it is a conceptual truth that memory statements are generally true, it is a conceptual truth that persons are capable of knowing their own pasts in a special way, a way that does not involve the use of criteria of personal identity, and it is a conceptual truth (or a logical fact) that the memory claims that a person makes can be used by others as grounds for statements about the past history of that person. This, I think, is the kernel of truth that is embodied in the view that personal identity can be defined in terms of memory.
NOTES
1. Thomas Reid, Essays on the Intellectual Powers of Man, ed. by A. D. Woolly (London: Macmillan, 1941), p. 206.
2. H. P. Grice, "Personal Identity," Mind, (October, 1941), 340.
3. I use "remember" in its most common sense, in which "I remember that P" entails "P," and "I remember X occurring" entails "X occurred."
4. John Locke, An Essay Concerning Human Understanding, I, ed. by Fraser (Oxford: The Clarendon Press, 1894), 457.
5. Roughly speaking, a statement is a memory statement if (supposing it to be an honest assertion) it cannot be false unless the speaker has disremembered. A conclusion from what is remembered, on the other hand, can be false without there being a mistaken memory. E.g., I mistakenly identify the man I saw as John when in fact it was his identical twin.
6. The word "generally" is vague, but I doubt if this can be made much more precise. This statement should perhaps be qualified so as to apply only to memory beliefs concerning the recent past.
7. We can, of course, have inductive grounds for believing that one person's memory claims are exceptionally reliable and that another's are exceptionally unreliable.
8. Ludwig Wittgenstein, Philosophical Investigations (Oxford: Basil Blackwell, 1953), I, paras. 56 and 265.