The word "memory" is highly ambiguous, even when it is not being used in admittedly paradoxical and uncommon senses, as when people talk of "racial" or "ancestral" memory. I call such uses of the word paradoxical because even those persons who hold that in performing an instinctive action we are "remembering" similar actions which were performed deliberately by our remote ancestors would have to admit that, in the ordinary sense of "remembering", we certainly do not remember the actions or thoughts of our ancestors. Even apart from these odd senses of "memory" it is quite certain that the word covers a number of very different acts. We talk of remembering a set of nonsense-syllables; of remembering a poem; of remembering a proposition in Euclid, though we have forgotten the words in which it was expressed when we originally learnt it; of remembering past events, and of remembering people, places, and things. To remember a set of nonsense-syllables is merely to have acquired the power of repeating them at will; and remembering, in this sense, seems to be no more an act of cognition than is the act of riding a bicycle or of swimming. To remember a proposition of Euclid is no doubt to perform a genuine act of cognition; and the same is true of remembering events, persons, and places. But the first kind of act has an abstract and timeless object; whilst the second has a concrete particular object which exists in time. Presumably then the memory of propositions is something quite different from the memory of mere sentences, on the one hand, and from the memory of events, persons, and places, on the other. This of course is quite compatible with the view that there may be intimate relations of causal dependence between the various kinds of memory, and that there may be something common and peculiar to them all in virtue of which they are all called " memory ".

    It seems plain that there is one and only one kind of memory which can plausibly be regarded as closely analogous to perception; and this is the memory of particular events, places, persons, or things. Let us call this "Perceptual Memory". My main object in this chapter is to discuss perceptual memory, to compare it with perception, and to consider some of the epistemological problems to which it gives rise. At the end of the chapter I shall say something about the other senses in which the word "memory" is used, and shall consider the mutual relations and the common features (if any) of all kinds of memory.

Memory-Powers and Memory-Acts.

    I must begin by pointing out an ambiguity which applies equally to all kinds of memory and does not apply to perception. If a man said to me: "Do you see Jones?" and I answered: "Yes", I should be lying unless I were actually at the time subject of a visual situation whose objective constituent I took to be an appearance of Jones. But, if he said to me: "Do you remember Jones?" or "Do you remember Euclid i. 47?" or "Do you remember the first line of the Aeneid?" I might quite truly answer "Yes" even though I were not at the time performing any memory-act at all. So long as I believe that I could remember these things if I tried I should be justified in saying that I do remember them. If, on the other hand, he had said to me: "Are you remembering Jones?" I should not be justified in saying: "Yes" unless I were at the time actually the subject of a memory-situation with Jones as its epistemological object.

    The point may be put shortly as follows. "To remember"is an ambiguous word, which covers both an act and a power. When I say that I remember so-and-so I may be referring either to the power or to a particular present exercise of the power. When I say that I am remembering so-and-so I am understood to be referring to a particular present exercise of the power, and not merely to the power itself. We do not use words like "seeing" and "hearing" in this ambiguous way. Whether I say that I see so-and-so or that I am seeing so-and-so I am understood to be referring to a present act of perception, and not to a mere power of perceiving. Thus, in discussing memory and trying to compare it with perception, I must be understood to be talking about particular acts of remembering so-and-so, and not (unless I specially say so) about the general power of remembering so-and-so at will. We must distinguish then between "Memory-acts" and "Memory-powers", and we shall be talking about the former unless we explicitly say that we are talking about the latter.

Perceptual Memory.

    It will be admitted by everyone that such phrases as "I remember having my hair cut last week", "I remember the tie which my friend wore yesterday", " I remember the feeling which I had when I last went to the dentist", and "I remember hearing Mr Russell lecture ", all stand for familiar cognitive situations which do arise from time to time. We will call them "Perceptual Memory-Situations", or in the present section simply "Memory-Situations". It will be noticed that the four examples which I have given differ from each other in the nature of the object which we profess to be remembering. In the first we profess to be remembering an event of the physical kind. In the second we profess to be remembering a certain physical thing. In the third we profess to be remembering a past feeling. And in the fourth we profess to be remembering a past perceptual situation. Let us first consider the relation between memories of events and memories of things.

Memories of Events and of Things.

    We have a like distinction in the case of perception. We talk of seeing a flash of lightning and of hearing a clap of thunder; and we also talk of seeing a cloud and hearing a bell. In the one case we claim to be perceiving a physical event, in the other a physical thing. If we reflect we see that the two kinds of perceptual situation are very closely connected. When we say that we are perceiving a certain physical event, as distinct from merely having an auditory or visual sensation, we mean that we regard the sensum which we are sensing as either a part or an appearance of a part of the history of a certain physical thing. On the other hand, when we say that we are perceiving a certain physical thing, we are really perceiving a physical event or a series of them and regarding them as parts of the history of this physical thing. All perceptual situations refer beyond themselves to physical things; if we confine ourselves to saying that we perceive a certain physical event we simply leave the further reference rather more vague than when we say that we perceive a certain physical thing. Now the same is true of perceptual memory. I say that I remember the late Master of Trinity, and I say that I remember dining with him. But, on the one hand, I remember him only in so far as I remember events in which he was concerned. And, on the other hand, when I remember any physical event I, ipso facto, remember to some extent the thing in which I believe this event to have happened. Thus I think we may say that to remember a thing or person simply means to remember certain past events and to regard them as incidents in the history of that thing or person. We can, and very often do, remember things which still exist and which we are now perceiving. I both perceive and remember the chairs in my room, in so far as I perceive certain present events and remember certain past events and regard them as so many successive phases in the history of my chairs. So I think that the fundamental point to be considered in dealing with perceptual memory is the memory of events. Memory of things depends on this, and no principles are involved in passing from the memory of events to the memory of things which are not equally involved in the passage from the perception of events to the perception of things. We may add that the perception of things, in so far as it involves the belief that the event which I am now perceiving is the present phase of the history of a certain enduring thing, is inextricably bound up with memory of things and therefore with memory of events,

Repetition and Perceptual Memory.

    It has sometimes been held that a criterion by which we can distinguish Perceptual Memory from mere Habit Memory, such as is involved in repeating a poem by heart, is that Habit Memory depends on repetition whilst Perceptual Memory from the nature of the case cannot do so. This of course would at best be only a distinction between the conditions under which the two kinds of memory-power are acquired; it would not be a distinction between the two kinds of memory-act. But it seems to me to be a rather inaccurate statement; and I think that those who have made it have failed to distinguish between perceptual memory of events and perceptual memory of things. It will be worth while to clear up this point before going further. In the first place, repetition is not essential, though it is helpful, for the establishment of a habit-memory-power. A man, like Lord Macaulay, with a very quick and retentive verbal memory, may be able to repeat sentences or sets of nonsense-syllables which he has met with only once. Secondly, it is obvious that our power of remembering a person, place, or thing is in some ways improved by repeatedly perceiving the object in question. It will, therefore, be well to clear up this question of the relation between repetition and the acquirement of memory-powers.

1. It is of course perfectly true that the power of remembering a certain definite event cannot be due to repetition, and cannot be improved by repetition. For a definite event with a definite date cannot be repeatedly perceived, although other events very much like it may be perceived at various times.
2. But perceptual memory-power which is concerned with people or things is improved by repetition in two ways.
1. It is improved in content by repetition with variation. I have said that my memory of a thing consists of my memories of various events all of which I regard as so many slices of the history of this thing. Now a thing shows different sides of its nature by being placed in different kinds of situation. I can remember a thing or person better (in the sense that I can know more facts about it by memory) in proportion to the number of events of different kinds in which I have perceived it to be concerned. And the same thing can be concerned in a large number of events of very different kinds only on a number of successive occasions. Hence the power of adequately remembering a thing or person does need repetition for its establishment. But the repetition which is needful for this purpose is quite different in kind from that which is helpful in establishing a power of habit-memory. In the former case repetition is important only as a necessary condition for variation. In the latter case what is wanted is pure repetition with as little variation as possible.
2. I think that it is also true that pure repetition without variation plays a part in the establishment and improvement of the power of remembering persons and things; for I think that an element of habit-memory is involved in perceptual memory To have an accurate memory of a person or thing it is useful, if not essential, to be able to call up an accurate image. Now the power to call up an image is just a habit, like the acquired power to repeat a sentence which one has learnt by heart. And to establish this power repetition with as little variation as possible is helpful. The perceptual memory-act does not indeed consist in calling up the image; but the power to do so is an important condition of an accurate perceptual memory. As this power is established and improved by bare repetition, such repetition is so far helpful in establishing a perceptual memory-power.

Nature of Memory-Objects and of Perception Objects.

    If the reader will refer back to the examples which I gave of perceptual memory-situations he will see that we claim to remember, not only physical things and events, but also feelings and perceptual situations. We claim to remember, not only our friend's tie and Mr Russell's lecture, but also seeing the tie, hearing the lecture, and feeling toothache. Now we do not claim to perceive anything but physical events and objects. I do not think that we ought to exaggerate this difference between the possible objects of memory I and the possible objects of perception. If we confine the name "perception" to sense-perception, it is true that the objects of memory are less restricted than those of perception. But perhaps this restriction is unwarranted. We certainly seem to have some kind of intuitive knowledge of contemporary feelings and perceptual situations, and it is possible that we ought to regard this as a form of perception. I shall deal with this question in the next chapter. In the meanwhile we must recognise that, whilst the objects of memory are certainly less restricted than those of sense-perception, they may coincide with those of perception in the wider sense.

    It is one of the characteristics of sense-perception that the objects which we claim to perceive are public and neutral. When we remember physical things and events memory claims to reveal public and neutral objects to us. A number of people can remember the same man or the same flash of lightning. When we remember feelings or perceptual situations memory claims only to reveal private and personal objects to us. It might be argued that there is something more private and personal about the object of a memory even of a physical thing or event than there is about the object of a perceptual situation. We can remember only the things and events which we have perceived, and only those phases of the past history of things which fell under our notice. I think that this does impose a restriction on the range of memory as compared with that of perception; but I do not think that it introduces any special privacy into the objects of memory I can remember only those things and events which I have perceived, and it happens to be true that I perceive many things and events which I cannot remember. But it is also true that I can perceive only those things and events which produce sensations in me, and that many things do in fact produce sensations in me without giving rise to perceptions. This restriction in the range of perception does not make the perceived objects essentially private in character; similarly, the further restriction in the range of memory does not make the remembered objects essentially private in character. In each case the class of objects perceived or remembered is determined by factors which are personal to the experient; but each of the individual members of the class may still be of such a nature that a number of experients could perceive or remember it

    It is only on one very special theory about memory that it could be maintained that there is something essentially private about the objects even of memories of physical events and things. It might be suggested that, when we say that we remember a certain physical thing or event, what we primarily remember is a past perceptual situation of which we were the subject and the event was the epistemological object. E.g., it might be held that, when I say that I remember a certain lecture being given by Mr Russell, I primarily remember the perceptual situation of myself hearing Mr Russell speaking; and that my belief that Mr Russell did speak is inferred from my memory of hearing him speak. On this view that what we primarily remember is perceptual situations with ourselves as subjects, and that our memory-beliefs about physical events and objects are secondary and derivative, it would be true to say that the primary epistemological objects of all perceptual memories are private and personal in a way in which the epistemological objects of perceptual situations are not. A milder view would be that, although my memory-judgments about physical events and things are not derived from a more primitive memory of myself perceiving these objects, yet, in fact, the former kind of memory-situation does not and cannot exist without the latter. On this milder view the total epistemological object of any memory situation would be complex; and, since one part of it would always be a past perceptual situation with myself as subject, it would be true that there is something private and personal in the epistemological object of every memory-situation even when there is also something public and neutral. I shall have to discuss these two views a little later, when I try to analyse the typical perceptual memory-situa:ion.

    In the meanwhile there is one other remark to be made before leaving this part of the subject. We shall find that it is necessary to distinguish between the objective constituent of a memory-situation and its epistemological object, just as we have had to do in the case of perceptual situations. Now it is arguable that the objective constituents of memory situations are private and existentially dependent on the body, if not on the mind, of the experient. For the objective constituents of memory-situations are generally believed to be images, and images are generally believed to be private in this sense. This, however, would not make any sharp distinction between memory-situations and perceptual situations. For, on the sensum theory, the objective constituents of perceptual situations are sensa; and there are reasons for thinking that sensa are, in some degree, dependent on the body and perhaps on the mind of the percipient.

    We may sum up the discussion as follows:

1. The objective constituents of all memory-situations may very well be private, in the sense that they are existentially or qualitatively dependent on the body or the mind of the subject. But much the same may be said of the objective constituents of perceptual situations.
2. The epistemological objects of some memory-situations (viz., of those in which we claim to remember a feeling or a perceptual situation) are undoubtedly private and personal.
3. The epistemological objects of other memory-situations (viz., of those in which we claim to remember a physical thing or event), are, on the face of them, as public and neutral as those of perceptual situations. It is true that the range of physical events and things which anyone can remember is limited by his past perceptions; but it is equally true that the range of things and events which he can perceive is limited by his sensations. And in neither case does this render the objects themselves private or personal in their essence.
4. There is a certain theory about memory according to which all that is primarily and strictly remembered is past perceptual situations of which the experiment was subject. On this theory it would be true that the primary epistemological objects of all memory-situations are private and personal to the experient.
5. Even if this extreme view be not taken it remains possible that all memory-situations have a complex epistemological object, one part of which is a past perceptual situation. If this be so there will be something personal and private in the complete epistemological object of every memory-situation, although there will also be something which is public and neutral in the epistemological objects of memories of physical things and events.

General Analysis of a Memory-Situation.

    I propose now to analyse a typical perceptual memory-situation, having a physical thing or event for its epistemological object. I shall begin with a rough general analysis, in which I shall try to bring out the apparent analogies between such a situation and a typical perceptual situation. I shall then go into greater detail about certain special points where the two kinds of situation seem to differ fundamentally from each other. Let us compare the two situations which are expressed respectively by the two phrases: "I am remembering the tie which my friend wore yesterday" and "I am seeing a certain penny."

    (1) Both are plainly situations with epistemological objects, and in both cases the epistemological object is known as soon as we understand the phrase which expresses the situation. In both cases the fact that the situation has such and such an epistemological object is wholly independent of the question whether there is an ontological object which accurately corresponds with the former. Here I must make much the same warning about the use of words as I did when discussing perception. There are memory-situations which we have good reason to think delusive; e.g., George IV used to say that he remembered leading a charge at the Battle of Waterloo, and there is every reason to believe that he was never within a hundred miles of the battle. Now in common speech we should be inclined to say that he did not "really remember" the event in question, just as we are inclined to say that the drunkard does not "really see" pink rats. But in both cases this is to mix up psychological, epistemological, and ontological considerations in a way which is most detrimental to philosophical discussion. Assuming that the First Gentleman in Europe was correctly describing his state of mind, he was subject of a situation which has just as good a right to be called a memory-situation as a veridical memory of the Duke of Wellington on the same subject-matter. It was a situation with a certain epistemological object, a certain kind of objective constituent, a certain kind of subjective constituent, and a certain characteristic kind of reference. There is nothing to distinguish it from what we should unhesitatingly call a memory-situation in the case of the Duke of Wellington, except that there probably is an ontological object accurately corresponding to the Duke's situation and that there almost certainly is not an ontological object corresponding to the King's. I therefore propose to call all such situations "memory-situations", regardless of whether there is or is not reason to think them delusive.

    (2) In both cases, and for similar reasons, it is impossible to hold that the ontological object, even if there be one which accurately corresponds to the epistemological object, is literally and bodily a constituent of the situation. The most that we could plausibly hold in either case is that each situation contains an objective constituent which is literally a slice of the history of an ontological object that corresponds accurately to the epistemological object of the situation. The objective constituent of the perceptual situation in our example is a certain patch which looks brown and elliptical. To make the analogy as close as possible we will suppose that, when I am remembering my friend's tie, a visual image of it is before my mind. And we will take this visual image to be the objective constituent of the memory situation. Now the strongest claim that the perceptual situation can or does make is that this patch which looks brown and elliptical is literally a contemporary slice of the history of a certain enduring physical object, viz., the penny. And the strongest claim that the memory-situation could make with any plausibility is that the visual image is literally a past slice of the history of a certain enduring physical object, viz., my friend's tie. Whether it is any part of the memory-situation to make even this claim we shall have to discuss later. But it certainly does not and cannot claim more than this.

    (3) An essential feature in both kinds of situation is a belief which refers beyond the situation and its constituents. This belief may not be explicitly formulated; but we are ready to act in accordance with it if occasion arises, and we are surprised if the results are such as would conflict with this belief. It is an essential factor in the perceptual situation that we believe that the penny now exists and that it is now manifesting certain aspects of itself to us. It is an essential factor in the memory-situation that we believe that the tie has existed and that a certain past phase of its history is now being manifested to us again. I leave the precise content of the memory-belief for more detailed discussion later.

    (4) In both cases we may say that the belief is

1. based upon the existence and character of the objective constituent;
2. refers beyond it to something which is not a constituent; but
3. is not reached by a process of deductive or inductive inference from the existence and nature of the objective constituent.

Memory-beliefs, like perceptual beliefs, not only are not reached by inference from the objective constituent of the situation but cannot be supported by such inference without logical circularity. When I remember the tie which my friend wore yesterday I do not first notice an image of a certain characteristic shape, colour, etc.; then recollect the general principle that the power to have an image always originates in a past perceptual experience whose objective constituent resembles the image; and then infer from these two premises that there must have existed a certain tie and that I must have seen it. And, if I did profess to reach my memory-judgments by an inference of this kind, the validity of my argument would be open to the following attack. How do I come to know the general principle that all images are copies of past sensa, which is an essential premise of the supposed inference? If we say that the general principle is established inductively, we must suppose that there are some cases in which we can remember a past sensum, compare it with a present image, and notice that the latter resembles the former. Now these instances will be useless for establishing the general principle unless, in these cases at any rate, we can remember the past sensum without using the general principle and making an inference from it. It would therefore be impossible to establish the general principle inductively unless there be some non-inferential memory-judgments about past sensa. The only way of avoiding this objection would be to take the desperate step of saying that we know a priori that every image must be a copy of a past sensum, and that the power to call up an image must have originated through the sensing of this sensum. It seems to me quite plain that this principle is not a priori and I do not know that anyone has ever asserted that it is so. Even Mr Hume, who had a conviction on the subject which seems to me to be quite unintelligible on his own principles, cannot have regarded the proposition as a priori; since he recognises and discusses the possibility of exceptions to it in the case of images of certain shades of colour.

    Of course I am not maintaining that particular memory-judgments, like particular perceptual judgments, may not be supported or refuted by argument. They certainly can be. But the arguments always presuppose other memory-judgments which must simply be accepted on their own merits. If I claim to remember that my friend was wearing a tie of such and such a kind yesterday, and I find that he and other people who saw him agree with me, my memory-judgment will be supported. If I find that they disagree with me and agree with each other, my memory-judgment will be rendered improbable. But, in using this test, I presuppose the validity of their memory-judgments. Again, when I claim to remember a certain event, I may test my judgment by inferring what events would be likely to follow such an event as I claim to be remembering. If I find that I can remember and perceive these consequences, my memory-judgment will be supported by inference. If I find that I remember and perceive events which are incompatible with these, my memory-judgment will be made improbable. But, even when I test the memory-judgment by present perception and not by memory, I presuppose the general validity of my memory-judgments. For I start by inferring that I shall be likely to perceive so-and-so if the event which I claim to remember really happened. And, if the chain of inference be of any length, my guarantee for the conclusion is my memory that the earlier stages of the argument satisfied me. In exactly the same way we may support or refute particular perceptual judgments by argument; but these arguments always presuppose the general validity of perceptual judgments and the validity of certain particular perceptual judgments made by myself or others.

    (5) We see then that memory-judgments, like perceptual judgments, are "direct" or "immediate", in the sense that they are not in fact reached by inference from the nature and existence of the objective constituent of the situation, and that any such inference would be logically circular. But there was another sense in which we said that perceptual situations were direct and immediate. We contrasted the two statements: "I am hearing the bell" and "I am thinking about the bell"; and we said that the perceptual situation claimed to bring us into more direct cognitive contact with objects than the thought-situation. This we expressed by calling perceptual situations "intuitive" and thought-situations "discursive". Of course perceptual judgments (like all judgments, whether reached by inference or not) are "about" their subjects. But, in the perceptual situation, we claim to be also in peculiarly direct contact with the subject which the perceptual judgment is about. Now, very closely connected with this is the fact that the perceptual judgment is "sensuous". I think that the intuitive character of the perceptual situation lies in the fact that we are directly acquainted with the objective constituent, and that this constituent is regarded as being literally a part of the physical object which we are said to be perceiving. Can we say that the typical perceptual memory-situation is intuitive and sensuous?

    (i) Ordinary language suggests that the memory-situation is intuitive and not discursive. We say "I remember my friend's tie", just as we say "I see my friend's tie". In neither case does the verbal expression for the situation contain a preposition like "about" before the phrase which stands for the epistemological object.
(ii) The memory-situation often does contain as its objective constituent a visual or auditory image; and an image is obviously very much like a sensum. Nevertheless, it is not clear to me that the memory-situation is so obviously intuitive and sensuous as the perceptual situation. The connexion between the image and the remembered object seems much looser than the connexion between the sensum and the perceived object. In some cases the objective-constituent seems to be merely images of words; and in that case we cannot claim to be in direct contact with a past slice of the history of an object. And, even when the image is visual and is held to resemble a past phase of the remembered object, it is not clear to me that we claim that it is literally a part of the past history of the object. Thus we come here to a point at which the analogy between perceptual situations and perceptual memory-situations begins to fail. In the next section I propose to consider in more detail certain points, such as this, in which there seem to be essential differences between the two.

More detailed Discussion of certain Points.

    Under this heading I shall consider two closely connected points, viz,

1. the precise content of the memory-belief; and
2. the nature of the objective constituent and its connexion with the epistemological object.

When I speak of the "content" of a memory-belief or a perceptual belief I mean the propositions which are believed in the typical memory or perceptual situation. Now, in comparing the content of a memory-belief with that of a perceptual belief, there are two points to be considered.

1. Does the total content of one include some proposition to which nothing corresponds in the total content of the other? and
2. How far are the parts which may fairly be said to correspond analogous to each other?

The latter question forms a transition to the question of the nature of the objective constituent of a memory-situation and its connexion with the epistemological object.

    (i, a) When I am the subject of a perceptual situation I believe that such and such an event is happening and that it is part of the history of a certain physical object which still exists. But, so long as I merely perceive and do not begin to reflect on my perception, I do not believe anything about myself and my cognitive situations. When I am the subject of a perceptual memory-situation I believe that such and such an event has happened, and that it was part of the history of a certain physical object which may or may not still exist. This part of the content of the memory-belief is evidently analogous to the content of the perceptual belief. But is it the whole of the content of the memory-belief? When I remember an event do I not also always believe that I have perceived it as well as that it has happened? If so, there is an essential part of the content of every memory-belief which refers to myself and my past perceptual situations; and there is nothing analogous to this in the content of the perceptual belief.

    There is, I think, no doubt that in most memory-situations I judge, not merely that "This happened before", but also that "I have perceived this before"; and that neither of these judgments is inferred from the other or from anything else. And I suppose that I could hardly judge that "I have perceived this before" without, ipso facto, judging that "This has happened before." The only question then is this: Are there any memory-situations in which I judge that "This happened before" and do not judge, in the same intuitive and non-inferential way, that "I have perceived this before"? I find this question very difficult to decide, for the following reasons.

    In the first place, every one believes that, under normal circumstances, we cannot have direct and noninferential knowledge of a past event unless we have in fact at some time perceived it. If we ask what can be the evidence for this universal belief, I think we shall find that it has two sources.

1. It is quite certain that, in the vast majority of cases in which I remember an event, I do also remember perceiving it. We might induce from this that, in all cases in which I remember an event, I must have perceived it; though, in a small minority, there are special circumstances which prevent me from remembering my perception of it.
2. We have a great difficulty in conceiving the causal mechanism by which we could have direct and non-inferential knowledge of a past event which we had never perceived. When we perceive anything we have a sensation, and this involves a characteristic change in our bodies. We think of this change as leaving a persistent "trace" in our bodies or minds or both, and we think of this trace as carrying with it the permanent possibility of intuitively apprehending the past event. But, if the event did not affect our bodies or our minds when it happened, we find it hard to conceive that it should be causally possible to have subsequent direct knowledge of the event. For my own part I think that there is decent evidence that, in certain abnormal cases, people do have direct knowledge of certain past events which they never perceived in their present bodily life. But the evidence is not known to most people; the phenomenon, if genuine at all, is very rare; and those who are acquainted with the alleged facts would admit that it is dangerous to put too much weight on them. I think then that we may start by accepting the following propositions, one about a matter of fact and the other about the usage of words.
1. It is almost universally held that we cannot have direct knowledge of a past event unless we have in fact perceived it, whether we remember doing so or not. And
2. if cases could be produced in which there was direct knowledge of unperceived past events, we should refuse to call such knowledge "memory". It is part of the meaning of the word "memory" that a perceptual memory-situation shall in fact be due to a past perceptual situation of the same subject. The question that remains is: Do situations ever arise in which a past event, which has in fact been perceived by us, is remembered whilst our perception of it is not remembered? And would such a situation be called a "memory-situation"?

    I think it is certain that situations sometimes arise which it would be natural to describe as follows: "I remember that man's face, though I do not remember seeing it before." No doubt we should immediately add: "I must have seen it before." But this word " must" is the mark of an inferential belief, not of a direct intuitive one; and no doubt the inference is made from the generally accepted premise that I cannot remember anything unless I have at some time perceived it. Now, if this phraseology is to be taken at its face value, such situations are called "memory-situations", and the content of the memory-belief does not include the proposition: "I have seen this before." But I am not at all sure that this is the right interpretation. Memory-beliefs may be of various degrees of determinateness in at least three respects. In the first place, we may have a more or less determinate memory-belief about the qualities of the remembered object. We may, e.g, remember that it was brick-red, or only that it had some shade or other of red, or only that it was either red or green. Secondly, we may have a more or less determinate memory-belief about the spatial relations and contemporary context of the remembered object. E.g., we may remember that it was to the right of the fireplace in a certain room, or only that it was somewhere in this room, or only that it was somewhere in some room of a certain house, and so on. Lastly, we may have a more or less determinate memory of the temporal relations and the earlier and later context of the remembered object. We may remember that it happened just before dinner ml Monday last, or that it happened some time last Monday, Or only that it happened some time last week. Now it seems to me quite likely that, when we say that we remember a certain physical event or thing and that we do not remember perceiving it, we really mean only that our memory of our past perceptual situation is extremely indeterminate in these three respects whilst our memory of the event or thing is relatively determinate.

    On this view the total belief which accompanies a memory-situation may always include a genuine memory-belief about our own past perception; but in certain cases the latter may be so extremely indeterminate that we say that we remember only the physical event and not the past perception of it. We must, however, recognise the following argument which favours the opposite view. Since we all believe strongly that nothing can be remembered unless it has been perceived by us, we shall almost inevitably infer when we remember an event that we must have perceived it. And we may very well confuse this natural and immediate inference with a genuine memory-belief; and thus think that the proposition: "I have perceived this" was part of the content of the original memory-belief, when really it is a reflective and inferential addition. In view of these two considerations which point in opposite directions I do not feel able to make up my mind on the question; and I must content myself with saying that the majority of memory-situations do include a noninferential belief that the remembered object has been perceived by us; and that this belief may be of any degree of indeterminateness. I think, however, that the discussion is enough to refute a view which was mentioned as possible earlier in the chapter; viz., that what we primarily remember is always our own past perceptual situations, and that our knowledge of the past physical events which we are said to "remember" is derived from our memory of the perceptual situations of which these events were the epistemological objects. It is difficult to see how highly determinate beliefs about a past physical event could be derived from a knowledge of the situation in which it was perceived which is so indeterminate that many people deny its existence.

    (i, b and ii) We have now to consider how far the memory-belief that "This physical event happened and formed part of the history of a certain physical object" is analogous to the perceptual belief that "This physical event is happening and is a contemporary part of the history of a certain physical object." It is hardly possible to discuss this question apart from that of the nature of the objective constituent and its relation to the epistemological object. I shall therefore take (i, b) and (ii) together. There are, as usual, two questions, one purely psychological and the other epistemological. The purely psychological question is: "What does the memory-situation claim to be the connexion between its objective constituent and its epistemological object?" The epistemological question is: "What view about the connexion between the two can be justified in face of all the known facts?" 1 shall defer the seeond question to the next section.

    I will begin by taking a case where the memory involves an imitative image. I will suppose that I am remembering my friend's tie, in his absence; and that I have an imitative visual image of a tie. This makes the analogy to sense-perception as close as possible; so that any differences that we may discover will be fundamental. Let us then contrast this situation with that of seeing a penny, and judging that it is brown and round. I will suppose that I judge that my friend's tie was red. Now I think that the most notable difference between the two cases is the following. When I am simply seeing a penny and making perceptual judgments about it, and am not philosophising, I draw no distinction between the objective constituent of the situation and the surface of the penny. I am acquainted with a certain particular (the objective constituent); I regard it as part of the surface of the penny, and I regard the qualities which seem to be sensuously manifested by it to me as, ipso facto, qualities of the penny. It is true that I find on careful inspection and reflection that the qualities which I have been ascribing to the penny are not exactly those that are sensuously manifested to me in this situation, and that there are strong reasons for refusing to identify the objective constituent with the surface of the penny. But these views are reached only by careful and critical reflection, and they exist only while we are reflecting. They vanish at once when we again begin to perceive, and the naively realistic view is reinstated as if it had never been questioned. Now it seems to me that, when I remember my friend's tie by means of an imitative image, I do not regard the image as literally a part of the surface of the tie which existed yesterday, and I do not regard the qualities which seem to be sensuously manifested to me in the image as, ipso facto, qualities formerly possessed by the tie. So far as I can judge, the perceptual situation definitely identifies its objective constituent with a contemporary part of the perceived object, whilst the memory-situation neither identifies the image with a past part of the object nor definitely distinguishes the two. The memory-judgment most certainly is not: "This, which I am now acquainted with, is a part of the tie as it was." But it also is not: "This, which I am now acquainted with, is numerically different from but qualitatively similar to a part of the tie as it was." It seems to me that both these suggested judgments are reflective theories about the memory-situation, and not judgments which form an essential factor in the memorysituation itself. While we are actually living through the memory-situation, and not philosophising about it, the belief that we have is vaguer than either of these suggested beliefs. If we want to put into words we must use some such formula as follows: "There is some peculiarly intimate relation between this, which I am now acquainted with, and a certain part of the tie as it was." And this statement must be taken neither to assert nor to deny that the two may be numerically identical.

    The difference which I have been trying to indicate between the memory-situation and the perceptual situation may be expressed as follows. Naive Realism is not merely a theory about perception; it is the explicit formulation of the belief which forms an essential part of the perceptual situation as such. But Naive Realism is merely a theory about memory, just as the Sensum Theory is a theory about perception. All that the memory-situation itself claims is that somehow the image enables us to have an intuitive non-inferential knowledge of the occurrence of a certain past event and of some of its qualities and relations. That it enables us to do this because it is numerically identical with the past event is neither asserted nor denied in the memorysituation itself. It is perhaps worth while to make the following remark at this stage. The facts which make it difficult or impossible to accept on reflexion the naively realistic claims of perception are by no means obvious. In order to recognise them we have to inspect our sensa with special care; to compare notes with others; to know something of the physiology of the nervous system, and of physical optics; and so on. It is therefore plausible to suppose that, if we start with an innate tendency towards naively realistic perceptual beliefs, the tendency will be so strengthened by habit in childhood that no arguments will eradicate it in practice later on. But the difficulties about a naively realistic view of memory are glaring, and need no special knowledge or careful inspection to reveal them; as we shall see in the next section. Hence, even if we started with a tendency to identify the memory-image with the remembered event (which there is no reason to think that we do), it seems doubtful whether it would survive the continual assaults of the objections which would arise almost automatically even in the least reflective mind.

    I have so far taken a case in which the objective constituent of the memory-situation is an imitative image, and have argued that even here it is no part of the memory-belief to claim that the image is numerically identical with a certain slice of the past history of the remembered object. I wish now to consider cases in which the memory-situation does not contain an imitative image as a constituent. It seems certain that there are such cases; and I think that a discussion of them will throw light on the part played by the imitative image in those cases where it is present. Even if it be granted that all perceptual memory-situations contain images of some kind as objective constituents and that these images play an analogous part to that which is played by sensa in perceptual situations, it must be granted that the connexion between the objective constituent and the epistemological object is much looser in the memory-situation than in the perceptual situation. I will now try to illustrate this point.

    Suppose that someone says to me: "Was the tie that your friend wore yesterday, red?" I may answer at once: "I remember quite distinctly that it was not red, but I cannot remember what its colour was." I will call this a "negative memory-situation ". Now there is nothing in perception which is strictly analogous to this. Of course, if my friend is wearing a green tie and I am looking at it and am not colour-blind, I can say: "I see plainly that his tie is not red." But I say that it is not red, and that I see that it is not, because I do see that it is green. I should never say that I see that it is not red unless I also saw that it had such and such another colour. The perceptual denial of a certain determinate colour depends upon the perceptual recognition of the presence of another determinate colour. But the memory-denial of a certain colour, so far from depending on the memory-recognition of another determinate colour, may precede the latter and may exist though the latter never supervenes on it. Now, in the negative memory-situation at least, the presence of an imitative image seems quite unnecessary. It may happen of course that, when the word "red" is mentioned by the questioner to me, I have a red image; but, so long as I understand the meaning of the word "red" in any way, the negative memory-situation may arise even though there is no red image. And in many cases I am quite sure that I understand the meaning of the word and that the memory-situation does arise in the absence of a red image. Again, it may of course happen that I have an image which does in fact resemble my friend's tie in colour; though it is evident that I do not recognise the fact at the time. But it is quite certain that the presence of such an image is not essential to the negative memory-situation, and that the latter does in fact quite often arise in the absence of the former.

    When I try to analyse a negative memory-situation as carefully as I can the essential point about it seems to be the following. In some way or other a certain determinate characteristic is presented to me for consideration. It may be presented by an imitative image, or by actually hearing and understanding the word which stands for it without using an imitative image, or by calling up for myself an image of the sound or the appearance of this word. The method of presentation seems to be absolutely unimportant so long as it succeeds in making me think of the characteristic in question. The other factor is that I then have a peculiar feeling which can only be described by the phrase: "This doesn't fit the object". Of course the feeling is one thing and the judgment is another; but this is the kind of judgment which we consider to express and to be justified by this kind of feeling. Lastly, the belief that the characteristic does not fit the object is not based on a comparison with the object and its remembered characteristics. For, if it were, the negative memory-situation, like the negative perceptual judgment, could not exist apart from the corresponding positive situation; whereas the former certainly can and does exist apart from the latter. The nearest analogy that I can give is that of trying a lock in the dark with a number of keys that do not fit it.

    Let us now consider a positive memory-situation. Suppose that I go on trying various suggested colours, red, blue, yellow, etc., and that they all fail to fit. At last, perhaps, I try green and I have a new and unique kind of feeling which I should express by the statement: "This fits the object". I then say that I remember that the tie was green. This feeling, which is naturally expressed by the judgment: "It fits the object", and is regarded as justifying that judgment, is the characteristic mark of a positive memory-situation, And up to this point no imitative image has been needed. There is no more need for the alternative which does fit to be presented for consideration by an imitative image than for the various alternatives which did not fit to be presented in this way. But at this point an imitative image very often does supervene; and then I think we are said, not merely to remember things about the object, but in the strictest sense to remember the object. (Of course, in a looser sense, we are said to remember an object provided we remember anything about it.)

    In the examples which I have just been giving we are supposed to be trying to remember something, and to succeed after failures. In other cases it seems to me that the imitative image comes first. It floats up; we notice certain characteristics in it which are felt to fit a certain past object, and others which are felt not to fit it. And the two processes may happen alternately. I may begin by merely remembering things about an object; then I may have an imitative image of it; and finally I may read off from the image further characteristics which are felt to fit the object, and so may remember further things about it. The essential point is the felt fitting or non-fitting of suggested characteristics; the way in which these characteristics are presented for our consideration is of minor importance.

    I think that I can now state more clearly what seem to me to be the essential points of difference between perceptual situations and even perceptual memory-situations.

1. All perceptual denials are based upon perceptual affirmations of determinate characteristics which are incompatible with the characteristic which is denied. But there are independent memory-denials, which are not based in this way on corresponding memory-affirmations. This fact should lead us to suspect that there may be important differences between positive memory-situations and positive perceptual situations even where they seem most alike.
2. It is true that we say both that we perceive objects and tbat we perceive propositions about objects. I see a tie, and I see that the tie is green. But the latter is regarded as dependent on the former. When I say that I perceive a tie I do not seem to mean merely that I know various propositions by perception, such as "This is green", "This is long and thin"; that all these propositions have a common subject; and that the tie is known only as the common subject of all these perceptually known propositions. On the contrary I claim to be directly acquainted with a part of the tie; and the propositions which I claim to perceive about it seem to be "read off" from the object itself. The object (or, at any rate, a literal part of it) seems to be "given" bodily and the perceptual judgments profess to "analyse" it. Now, in spite of some appearances to the contrary I believe that the opposite is true of perceptual memory. I believe that what is primarily known by memory is propositions like "This was green", "This was long and thin", etc.; and that this is true both in positive and in negative memory-situations. Certain groups of such propositions are recognised to have a common subject; and the object is "remembered" only in so far as it is known as the common subject of such a group of remembered propositions.
3. In many cases an imitative image of the remembered object supervenes at this stage; and in some cases such an image comes first, and the characteristics which are asserted or denied in the memory-judgments are presented to our attention by it. It is in this last case that perceptual memory is most like perception; but even here it does not seem to me that we claim to be in direct cognitive contact with an actual fragment of the past history of the object and to be "reading off" the memory-propositions from this fragment.
4. The essential factor in the memory-situation is that peculiar feeling which seems to justify the judgment that a certain characteristic fits or fails to fit a certain past object. The characteristic need not be, and generally is not, presented to our attention by means of imitative images. And the object which the characteristic is felt to fit or fail to fit is not cognised by direct and sensuous acquaintance, as it seems to be in sense-perception, but is presented only to thought as the subject of such and such propositions.

    Thus, although perceptual memory agrees with sense-perception in the fact that the memory-judgment, like the perceptual judgment, is not inferential, I believe that it differs from sense-perception in that it is not strictly intuitive or sensuous. Some perceptual memory-situations certainly seem at first sight to have these latter characteristics; but it seems to me that careful inspection and comparison show that none of them really do so. And I think that the function of imitative images in perceptual memory has been greatly exaggerated. It is perhaps relevant to add that I am myself a strong visualiser, and that memory with me does in fact generally involve imitative imagery; so, if I am wrong on this point, my mistake is certainly not due [to] identifying a personal defect with a law of nature, as the Behaviourists do when they deny the existence of images.

Epistemological Questions about Perceptual Memory.

    So far I have attempted nothing but a descriptive analysis of perceptual memory-situations, and a careful statement of what they really do claim. I must now consider what claims would be justified on the part of the memory-situation. If I am right, the perceptual situation makes a stronger claim than we can admit to be justified. For every perceptual situation claims that its objective constituent is literally a part of the perceived physical object; and this claim is certainly false in some cases and extremely hard to maintain in any. Now it is of course possible that the memory-situation goes to the other extreme and is too modest in its claims. If I am right, it does not claim that its objective constituent is literally a part of the past event which it remembers; nevertheless this claim might be justified, and the New Realists may only be asking on behalf of the memory-situation what it is too modest to ask for itself. There are two reasons for wanting to be as realistic as possible about perception. One is that we are then running with the stream and defending what we all in fact believe except when we are philosophising. The other is that the rejection of the claim of the objective constituent of a perceptual situation to be literally a part of the perceived physical object makes it hard to explain how we can possibly know that there are physical objects and that perception gives us trustworthy information about them. Now there is not the first motive for a naively realistic view of memory; but there is no doubt the second motive. Unless the objective constituents of some memory-situations be literally identical with the events remembered, how, it might be asked, can we possibly know that there are past events and that memory gives us trustworthy information about them? What we have now to consider is whether any such claim could be admitted even if it were made.

    I will first try to show, as I did in the case of perception, that, even if the naively realistic view could be accepted, it would by itself go but a very small way in meeting the attacks of a sceptic. To be sure that there are past events or that there are physical objects it is not enough to be acquainted directly with what is in fact a past event or a literal part of a physical object; we need in some way to know that this is what we are acquainted with. Now, if there were no delusive memory-situations, or if it were found on careful inspection that those which are delusive differ internally from those which are not, it might be suggested that every memory-situation is accompanied by a kind of infallible revelation that its objective constituent is literally identical with some past event. But it is certain that there are totally delusive memory-situations, such as that of which George IV was subject. And it is certain that there is no inner difference which distinguishes them from memory-situations which are commonly believed to be veridical. Hence we must deny that there is an infallible revelation that we are in direct contact with the past in any case. And, in the absence of this, the mere fact (if it be a fact) that the objective constituents of some memory-situations are actually past events does not explain how we know that there are past events or how we know that we have trustworthy information through memory about some of them. We shall still have to rely entirely on the external tests of agreement or disagreement of one memory-judgment with others, and of agreement or disagreement between inferences from memory-judgments and present perceptions.

    The most that we can say then is that the claim which the extreme realists make for the memory-situation may sometimes be true, though it is certainly sometimes false. Can we go so far even as this?

    (1) I will first mention the only point, so far as I am aware, on which the memory-situation is in a stronger position than the perceptual situation in making such a claim. There is nothing in memory corresponding to the systematic difference in the apparent shapes and sizes of perceived objects when viewed by observers from different positions. Now it was this which made it so very hard to believe that the objective constituent of a visual situation can be literally a part of the surface of the perceived physical object. Nevertheless there often are positive discrepancies between different people who profess to be remembering the same event, and between successive memories of the same event by the same person. The mere fading of details as the event retreats further into the past presents no particular difficulty for a naively realistic view of memory; but the memory of details which are positively inconsistent with each other on different occasions when we profess to be remembering the same event does present a very serious difficulty.

    (2) I will now consider the special objections which might be made against a naively realistic theory of memory.

    (a) The first is a general metaphysical objection, which I believe to be baseless. It might be said that, when an event is past, it ceases to exist. Now, when I am remembering a past event, the memory-situation certainly exists and so does its objective constituent. Hence, it is said, the objective constituent of a memory-situation cannot be identical with the past event which is being remembered. This objection seems to me to be mistaken. It depends on a view of time and change which I am forced to reject. It appears to me that, once an event has happened, it exists eternally; all that happens henceforth to it is that, as more and more events occur and take their permanent place in the ever-lengthening temporal order of the universe, it retreats into the more and more distant past. If an event ceased to exist as soon as it ceased to be present it plainly could no longer stand in any relations to anything. But, when we say that it is past, we imply that it does stand in the relation of temporal precedence to the present; moreover, we say that one past event precedes a second past event and follows a third. All such statements would be nonsensical if events ceased to exist when they ceased to be present. It is perfectly true that certain objects which have existed (e.g., the town of Old Sarum) have ceased to persist. But this means only that after a certain time none of the events which happened were such as to continue the history of these particular objects; the earlier series of events which constitute the history of such objects are nevertheless a permanent part of the universe, considered as an existent which is extended in time. There is then no general metaphysical objection to a naively realistic view of memory. Past events are always "there" waiting to be remembered, and there is no a priori reason why they should not from time to time enter into such a relation with certain present events that they become objects of direct acquaintance. There is no a priori reason why a cognitive relation should not bridge a temporal gap, and connect a present mental event with a past event of any kind whatever.

    (b) The second difficulty is much more serious. Suppose that I remember the same event on several occasions, and that the objective constituent of each of these memory-situations is an imitative image. If the naively realistic theory of memory be right, this image is literally and numerically the same in all cases; it is the past event or a part of it; and so its date is that of the past event. On the other hand, the image certainly seems to be present on each occasion; and we should certainly judge that we were concerned with as many different and successive images as there are different and successive memory-situations, even though all these images were exactly alike in their qualities. Again, suppose that when I perceived the event I had a twinge of toothache; and that when I first remembered it I had no toothache, but a tickling sensation in my throat. I should certainly judge that the original event was contemporary with this twinge of toothache and preceded the tickling sensation. And I should certainly judge that the image was contemporary with the tickling sensation and not with the twinge of toothache. If the naively realistic view of memory be true it would seem that the same event can be both contemporary with a certain other event and can also succeed this event by a long interval.

    The present difficulty is evidently analogous to that which arises on the naively realistic theory of perception over mirror-images and "seeing double". There we seem to see an object in a place which is remote from its real position, or we seem to see several distinct though qualitatively similar objects in different places, whilst the theory requires that there shall be only one. Here we seem to be aware of a series of distinct images which are separated in time, whilst the theory requires that the objective constituents of all the memory-situations shall be numerically identical and shall have a certain one date in the past. In face of this situation the naively realistic theory of memory might take one of three courses.

1. It might suggest that the image only seems to be present, whereas it is really past; and that it only seems contemporary with the tickling sensation, whereas it really precedes it and is contemporary with the twinge of toothache. Or
2. it might suggest that we are using words in an ambiguous way. Perhaps when we say that the image is present and the original event is past we are using "present" in one sense and "past" in another. The statements may then be compatible with each other. And the same explanation may apply to the statements that the image is contemporary with a tickling sensation, whilst the original event is contemporary with a twinge of toothache and precedes the tickling sensation. E.g., the two statements: "Napoleon was greater than Og, King of Bashan" and: "Napoleon was less than Og, King of Bashan" may both be true if one refers to their relative heights and the other to their respective achievements. Or
3. it might suggest that we are not using words ambiguously, but that temporal relations are not dyadic, so that the minimum complete statement about the temporal relations of two events is of the form: "x is contemporary with y (or precedes, or follows y) with respect to z." In that case, when we say that the image is contemporary with the tickling sensation whilst the original event precedes the tickling sensation, both statements may be true even though the image is identical with the original event. For we may be using a different third term of reference in the two cases.

    I think it is possible that when we say that an image is obviously present each time we remember a certain event we may only be justified in saying that it is presented each time, i.e., that it is an objective constituent of each situation and an object of acquaintance: Now "being presented" is certainly one of the tests that we use for "being present"; but it may not be an infallible test. Now of course, on the usual view of time, the same event cannot be present more than once; but there is no a priori reason why it might not be presented to acquaintance dozens of times, and it is of the essence of the naively realistic theory of memory to hold that this actually happens. Suppose then that, when we say that the image is present, we are justified only in saying that it is presented. If we think that it is present we shall infer that we must be dealing with a different image on each occurrence of the memory-situation. And yet really we may be dealing with a single entity which is presented many times but is present only once. I think it must be admitted that we have not the same direct and overpowering evidence that we are acquainted with a number of different images in a series of memory-situations with the same epistemological object, as we have for saying that we are acquainted with two distinct sensa when we "see double". In the latter case it is certain that no inference is involved; nothing is needed but inspection. In the former case I am not at all certain that the statement is guaranteed by inspection; I think that it may well rest on inference. And, as I have just pointed out, the premise of the inference might be derived from an uncritical jump from "presentedness" to "presentness." Similarly, when we say that the image is contemporary with the tickling sensation, we may only be justified in saying that both are presented together. And it may be that co-presentedness, though a test for co-presence, is not an infallible test. The original event might be presented (in perception) along with a twinge of toothache and apart from a tickling sensation, and it may be again presented (in memory) along with a tickling sensation and apart from a twinge of toothache. If we think that co-presentation is an infallible sign of co-presence we shall be forced to distinguish between the original event and the memory-image of it, or else to hold that simultaneity is a triadic relation. But, if we admit that the two relations are different, and that the former is not an infallible sign of the latter, we could hold that one and the same event is objective constituent of the original perceptual situation and of the subsequent memory-situation. This event is both co-present and co-presented with the twinge of toothache, whilst it is co-presented but not co-present with the tickling sensation. Thus I think that the naively realistic theory of memory could answer the present objection; provided it is allowed to distinguish between presentness and presentedness and between co-presence and copresentedness, to hold that the latter can occur without the former, and to hold that one and the same event can be presented at various times to the same mind. This last-point leads us to the third possible objection.

    (c) It is commonly held that the past cannot change. Would the naively realistic theory of memory be consistent with this doctrine? And, if it is not, ought we unhesitatingly to reject it? We must begin by distinguishing between pure qualities and relations. I think that every one would admit that an event cannot change in respect of any pure quality which it had when it happened. If it was, e.g., a red flash, it cannot cease to be red and become green. Again, I think it would be admitted that an event cannot change its relations, in the sense of ceasing to be related in a certain way to contemporary or earlier events and becoming related in different ways to them. But it seems to me that events can and do change, in the sense that they acquire additional relations through the occurrence of later events. This seems to me quite clear about temporal and causal relations. Queen Anne's death now precedes Queen Victoria's by so many years, and will do so for ever; but there was a time when Queen Anne's death preceded nothing. And, until Queen Victoria had died, Queen Anne's death stood in no relation whatever to the event which we now call "Queen Victoria's death". For there was then no such event; and an event cannot stand in any relation to a mere nonentity. Again, Queen Anne's death caused a feeling of annoyance in the Duke of Berwick when he heard of it; but it certainly did not stand in the causal relation to the Duke's feeling of annoyance until the Duke began to feel annoyed, which he did not until he heard of the death. There is then, in my opinion, no objection to holding that past events change, in the sense of acquiring relations to events which follow them. And this is the only kind of change in past events which the naively realistic theory of memory absolutely requires. It requires that the same event shall from time to time become a constituent of successive memory-situations. But the memory-situations are simply fresh events which happen after the remembered event is past; and whenever a past event is remembered it has simply acquired a relation to a certain later event, which it naturally could not do until that later event had happened. It is worth while to remark that even universals and other timeless entities can change in an analogous way to past events. The quality of redness is timeless; nevertheless it sometimes characterises one thing, sometimes another, and sometimes perhaps nothing at all. Again, it is sometimes thought of by me, sometimes by you, and sometimes perhaps by no one at all. Thus even the timeless may acquire certain additional relational properties through the happening of new events; and precisely the same is true of the past. It must be admitted then that the naively realistic theory of memory necessarily involves the proposition that past events change in certain respects; but it must also be admitted that there is no objection to the kind of change which it involves in past events. This brings us to the last objection which I propose to consider.

    (d) If the objective constituent of a memory-situation were found to differ in some of its pure qualities from the remembered event, it would seem to be impossible to identify the two. For the attempt to do so would involve the proposition that a past event had changed in respect to some of its pure qualities. And this is plainly impossible. I want to make quite clear what kind of qualitative difference between memory-image and remembered event would be fatal to the naively realistic theory of memory, and what kind would not, before I consider whether such differences are actually found.

1. A visual sensum must in fact have some perfectly determinate shade of colour; and so must a visual image. If the memory-image is to be incidental with the remembered sensum it is impossible that the determinate shade of the one should differ from that of the other. But it is one thing for a sensum or image to have such and such a determinate shade, and another thing for us to be able to judge that it has it. Our judgments about the determinate characteristics of an object may be of various degrees of determinateness, and they are no doubt never completely determinate. Now it might be that, when an event is presented in sensation, we are able to make more determinate judgments about its colour, for instance, than we can do when precisely the same event with precisely the same determinate shade of colour is presented in a memory-situation. If the only difference between the memory-situation and the corresponding perceptual situation is that the former permits of less determinate judgments about the same determinable characteristic than the latter, the memory-image and the remembered sensum may in fact be identical. But, if we can see that the image has a different determinate characteristic from the original sensum, the two cannot be identical; and the naively realistic theory of memory falls to the ground
2. Much the same remarks apply to differences of internal detail or external relation between the original sensum and the memory-image. Both must, in fact, be perfectly determinate in these respects. If the difference merely is that we can detect more detail in the perceptual situation than we can in the memory-situation, the image and the sensum may be identical, and the naively realistic theory of memory may be true. But, if we can detect details in the images which differ from and are inconsistent with those of the sensum, the two cannot be identical; and the theory falls to the ground. What are the actual facts?

    In the first place, there is a systematic and directly noticeable difference between the corresponding characteristics of sensa and of images. An auditory image never sounds exactly like an auditory sensum, and a visual image never looks exactly like a visual sensum. No doubt there are marginal cases where this systematic difference is hard to detect. A very faint sensum may be hard to distinguish by its intrinsic qualities from an image, and a very vivid image may be hard to distinguish by its intrinsic qualities from a sensum. But in general there is not the slightest difficulty in recognising that an image looks and sounds different from the sensum of which it is said to be a copy. Secondly, when I remember a thing or event by means of an imitative image, I can often say quite definitely that there are certain details in the image which are different from and inconsistent with corresponding details in the original. I may, e.g., call up an imitative image of my friend's head; and I may be able to say with complete conviction: "His hair is like that, but his nose is not of that shape." And, if I can often detect these positive differences of detail between the memory-image and the original, it is reasonable to suppose that they still more often exist when I cannot be sure that they do. Now it is impossible to believe that a past event actually undergoes a systematic change of intrinsic quality through lapse of time. And it is impossible to hold that a past event can undergo positive changes of internal detail through lapse of time. Hence we must either refuse to identify the memory-image with the remembered event; or we must hold that the image can seem to have characteristics which differ from and are inconsistent with those which it really does have; or that the characteristics which we detect when we are subjects of a memory-situation inhere in some different way from those which we detected when we were subjects of a perceptual situation. These three alternatives are analogous to the three which presented themselves when we tried to be naively realistic about perception.

    I do not think that the theory of two kinds of inherence will help us in defending a naively realistic view of memory. It seems quite clear that, if the characteristics which we seem to detect in images really inhere in them at all, they inhere in them in precisely the same way in which the characteristics which we seem to detect in sensa inhere in the latter. There does not seem to be the least reason to believe that the "is" in the two propositions: "This sensum is red" and "This image is green" stands for two different modes of inherence, such that the two statements would be compatible even though "This sensum" and "This image" were identical. A theory of triadic inherence for colours has a certain plausibility in dealing with perception, when we remember that the apparent colour of what we call "the same surface" varies according to the position and internal state of the percipient's body. But what is needed for the present purpose is that the colours, etc., which we detect when we are in a memory-situation shall inhere in some different way from the colours, etc., which we detect when we are in a perceptual situation. And we know of nothing that makes this suggestion plausible. I think then that we may rule out this line of defence for naive realism about memory. We are, therefore, reduced to saying that naive realism about memory is possible only on the supposition that memory images can seem to have characteristics and details which are other than and inconsistent with those which they really do have.

    We must note that this last statement needs a little further refinement. The essential point is that, when I remember something which I have perceived, the objective constituent of the memory-situation often seems to have characteristics and details which are, and are recognised at the time to be, other than and inconsistent with certain characteristics and details which the objective constituent of the perceptual situation seemed to have. I make this modification in case anyone accepts the theory that the objective constituent of a perceptual situation does not have the characteristics which it seems to have. Thus we may restate the position as follows. We can identify the memory image with the original sensum only on the supposition that one and the same event can seem to have one set of details and characteristics when it is the objective constituent of a perceptual situation and can seem to have another set of details and characteristics, partly inconsistent with the former, when it is the objective constituent of a memory-situation. Moreover, I must be able to know, with regard to certain determinate characteristics which the object now seems to have, that it did not seem to have these and did seem to have others when I perceived it. E.g., I may remember my friend's face, and I may remember that when I saw him his hair appeared to be bright yellow and his nose straight. Yet the image (which, on the present theory, is identical with the past sensum) may seem to have hair of a washy straw-colour and a crooked nose.

    Now it is no doubt theoretically possible to hold that the sensum and the memory-image are numerically identical in spite of the inconsistency between their apparent determinate characteristics in the perceptual situation and in the memory-situation. But I fail to see what advantage accrues to the theory of memory from this supposed numerical identity. On any view I manage somehow to remember the original apparent characteristics by means of the different and incompatible apparent characteristics which are manifested in the memory-situation. If there is any mystery in this, I cannot see that it is in any way lessened by the supposition of a de facto numerical identity of image and sensum which is plainly contrary to all the appearances.

    I may now sum up what I have to say about the naively realistic theory of memory.

1. That the memory-image and the objective constituent of the original perceptual situation are numerically identical is a claim made, not by the memory-situation, but for it by certain philosophers. Hence this proposition has not the same strong antecedent claim on our belief which Naive Realism about perception undoubtedly does have.
2. The motive for a naively realistic theory of memory is undoubtedly the belief that, unless we are in direct cognitive contact with past events in memory, it is impossible to explain how we come to have the very notion of "pastness" or how we have trustworthy non-inferential beliefs about particular past events. Now this presupposes that the objective constituent of a memory-situation literally is past; and that we recognise its pastness in the memory-situation just as we recognise the redness of a sensum, which is in fact red, in a perceptual situation. Such a view cannot be maintained in face of totally delusive memory-situations. For in them the objective constituent manifests the very same characteristic which we took to be "pastness" in other situations. And here the objective constituent is not identical with a certain past event which we claim to be remembering, for there is no such event. Thus, even if the objective constituents of some memory-situations be in fact past events, it cannot be admitted that there is any infallible revelation of their pastness in the memory-situation. The objective constituent of a memory-situation does no doubt manifest a certain peculiar characteristic which we take as a sign of pastness. But the existence of totally delusive memory-situations, and their internal likeness to veridical ones, show that this manifested characteristic is not pastness and is not even an infallible sign of pastness. Thus naive realism about memory, even if it be sometimes true, fails altogether to solve the epistemological problem which gave rise to it.
3. The a priori objections to a naively realistic theory of memory, based on the nature of time and change, are invalid. Past events exist henceforth eternally; there is therefore no a priori objection to their being objective constituents of existent memory-situations. And, although the theory requires that past events shall be liable to a certain kind of change, this is not an objection. For it is a kind of change to which past events and even timeless entities may be subject without contradiction.
4. It is not a conclusive objection to the theory that the memory-image seems to be present each time the same event is remembered and that the memory-image seems to be contemporary with events which the remembered event precedes. There are various ways round this difficulty; and perhaps the simplest and most plausible is to draw a distinction between "presentness" (which cannot be repeated) and "presentedness" (which can), and to hold that the latter is a sign but not an infallible sign of the former.
5. The most serious difficulty is that we can recognise a general qualitative difference between any image and any sensum, and specific differences in detail and in determinate characteristics between a memory image and the remembered sensum. This can be reconciled with the naively realistic theory of memory only on the assumption that the same event can appear to have different and incompatible details and determinate characteristics according to whether it is the objective constituent of a perceptual situation or of a memory situation. Such an hypothesis is not impossible; but it entails the conclusion that either in the memory situation or in the perceptual situation (and possibly in both) an event appears to have determinate characteristics which are other than and incompatible with those which it really does have.

The conclusion seems to be that, if the naively realistic theory of memory had anything to recommend it, it would be possible to hold it provided we made odd enough supplementary assumptions. But there are, so far as I can see, no reasons direct or indirect for holding the theory.

    I must now raise the question which was raised by the naively realistic theory of memory, and which the latter failed to answer. If past events be never constituents of memory-situations, or if at any rate they never manifest the characteristic of pastness as sensa manifest colours, etc., how do we come to have the notion of "pastness" at all? It will be remembered that a similar question arose over the origin of the notion of a "physical object". Underlying such questions there is a particular theory about the origin of our knowledge of universal characteristics which is explicitly stated by Mr Hume and tacitly assumed by nearly every one else. The theory may be roughly stated as follows. If a universal characteristic be simple and unanalysable we can form a concept of it only by being acquainted with some particular which has or seems to have this characteristic. If the characteristic be complex and analysable, we may be able to form a concept of it without being acquainted with any particular which has or seems to have it; but we must have been acquainted with particulars which between them had or seemed to have all its simple constituents. E.g., red is a simple characteristic; and it seems obvious that we could not have had the notion of redness unless we had been subjects of certain cognitive situations whose objective constituents were or seemed to be red. The really important point is that the particulars in question should seem to have the characteristic. For it certainly would not suffice that they should have it without seeming to have it; whilst, so far as we can tell, it would suffice if they seemed to have it even though they really did not have it, if this be possible. Now the trouble is that pastness is certainly a simple characteristic, and that the peculiar characteristic which memory-images seem to have cannot be identified with pastness, for the reasons given above. At most this characteristic can be taken as a sign of pastness; but how can I know this unless in some cases I have found pastness and this other characteristic together? And how can I have done so if no instance of apparent pastness is ever presented to my acquaintance?

    If we accept Hume's principle the question seems insoluble. It is even more intractable than the similar question about the origin of the concept of "physical object". For it might be argued (though not, I think, successfully) that the concept of "physical object" is complex, and is constructed by us from simpler concepts which are abstracted from sense-experience. But I do not think that anyone could maintain this view about pastness. One solution would be to give up Hume's principle; which is what I have done over the notion of "physical object". I should then draw a distinction between "empirical" and "categorial" characteristics. I should call "red ", "hard", etc., "empirical characteristics", and I should be inclined to maintain Hume's principle about the origin of our concepts of these. I should count "physical object" "pastness" "causation", etc., as "categorial characteristics", and I should be inclined to deny Hume's principle about the origin of our concepts of the latter. I see nothing self-evident or sacrosanct about Hume's principle; it seems to work well for empirical characteristics, like colour, and to cause nothing but trouble over categorial characteristics, like cause or substance. The two kinds of characteristic are obviously extremely different, and there would be nothing in the least surprising in the fact (if it were a fact) that our concepts of the one arose in a quite different way from our concepts of the other. I am not assuming of course that categorial concepts are "innate", in the sense that we are born thinking of cause, substance, etc. So far as I can see each such concept arises only an the occasion of a certain specific kinds of experience, which can be analysed and described with fair accuracy. There are three stages in the development of these categorial concepts. At tbe first stage they exist only in the sense that, on the occasion of certain kinds of experience, we act as if we were recognising the presence of causation, of substance, etc. This stage is reached by all men and probably by the higher animals. A dog, in the situation described as "seeing a bone", treats his visual sensa as appearances of a permanent and present physical object; and, in the situation called "hearing the dinner-bell," he acts as if he believed there to be a causal cennexion between dinner-bells and food. At the second stage we make explicit judgments involving the categories in question; e.g.. "Quinine tastes bitter and gives me a headache." This stage is probably reached by all sane men, and probably not by any animals. Then there is a third stage at which we do not merely act as if we recognised the categories, and do not merely make particular judgments which involve the categories, but contemplate the categories as such and make reflective judgments about them. This stage is reached only by philosophers while philosophising.

    Let us now apply these general remarks to the particular problem under discussion. I suggest that the objective constituents of memory-situations are not in fact past and that they do not even seem to be past. But they do seem to have (and there is no reason to doubt that they actually do have) a certain peculiar characteristic which is not manifested by most images or most sensa. Let us call this "familiarity". Now we are so constituted that, when we are subjects of a cognitive situation whose objective constituent manifests the characteristic of familiarity, we inevitably apply the concept of pastness; and, if we make an explicit judgment, it takes the form: "There was an event which had such and such empirical characteristics." Familiarity is an empirical characteristic and pastness is a categorial characteristic; but the former "means" the latter to such beings as we are; and this "meaning" is primitive and unacquired, in the sense that it is not, like most meaning, due to the repeated manifestation of the two characteristics together. This is the only account that I can recommend of "how we come to have the notion of pastness at all". I owe the notion of "unacquired meaning" to Professor Stout; though I do not know in the least whether he would accept my exposition of it or the particular applications which I have made of the notion.

    If the reader cannot accept the above suggestion I have only one other to make, and I am not prepared to lay much stress on it. The suggestion is this. The specious present has a certain small temporal extension. Now it might be said that the earlier objective constituents of the specious present are actually past and that they manifest the characteristic of pastness. On this view pastness is an empirical characteristic which is manifested by part of the total objective constituent of a specious present, and we form the concept of "pastness" by abstraction in the same way in which we form the concept of "redness". We then apply this concept beyond the contents of the specious present, just as we apply the concept of redness to things that we have never seen. We might then suppose that the earlier parts of the total content of the specious present manifest both pastness and familiarity, so that familiarity has acquired for us the meaning of pastness. The objective constituent of the memory-situation manifests familiarity but not pastness; but we read pastness into it through the association which has been formed between these two characteristics in the case of the contents of the specious present. I do not think myself that this suggestion will work. Apart from any other difficulties the following strikes me as serious. If familiarity has come to represent pastness to us because the two are manifested together in the earlier part of the content of the specious present, I should expect the result of the association to be that, finding the memory-image to seem familiar, we should ascribe pastness to it. For in the specious present, on the theory under discussion, what seemed familiar itself seemed past. But we do not in fact regard the memory-image itself as past. The familiarity of the image makes us think of some event, other than the image, as past; and makes us say that this event had or had not such and such characteristics which are suggested to us by the image. I do not see that the present theory will account for this fact.

    The question: "How do we come to have the concept of pastness?" is one question, and I have tried to answer it to the best of my ability. The question: "What right have we to believe that we have rightly applied the concept of pastness in any particular case?" is a different one. I have already said that we have no infallible revelation on the subject; that we can indeed test our memory-judgments by comparison and inference, but only on the assumption of the general trustworthiness of memory. I have only one thing to add. There are "secondary signs" of pastness; just as there are "secondary signs" of distance, such as the size, clearness, etc., of the visual sensum. By this I mean simply that there are certain empirical characteristics which are more often found in an image which seems familiar than in one which does not. These become empirically associated with familiarity; whilst familiarity is, in our view, non-empirically associated with the notion of pastness. Hence, by the ordinary process of "telescoping", these other empirical characteristics of images may come to stand for pastness.

Non-Perceptual Memory

    I shall end this chapter by considering very briefly some of the other senses in which the word "memory" is used.

    (1) It is a fundamental fact about living organisms that, when they have performed a certain set of movements several times, they tend to acquire a more or less permanent power of repeating these movements with greater or less accuracy from time to time when suitably stimulated. This general capacity of living matter is sometimes called "memory "; and it is in this sense and in it alone that heredity can plausibly be regarded as an extension of memory. It would be better to call this general capacity "retentiveness" or "perseverance".

    (2) We may acquire by practice the power of performing at will certain characteristic sets of bodily movements, such as those which are used in swimming. If we find that we can still swim when we get into the water after an interval, we should commonly say that we "remember how to swim" or "remember the movements of swimming". There is nothing cognitive about "memory", in this sense. To say that we remember how to swim is merely to state (a) that we can perform the proper movements after an interval, and (b) that we believe, or that the speaker who observes us believes, this to be due to our having performed them in the past. It would be better to call memory, in this sense, "retention of an acquired motor-capacity".

    (3) in precisely the same way we may acquire by practice the power of uttering or writing at will a certain set of noises or marks, as a parrot or a monkey might do. There is no essential difference between this and the last case, if the noises or marks are meaningless to us. What has really been acquired and retained is a certain motor-capacity in the throat and tongue or in the fingers. It is true that, in this case, an external observer would probably say, not only that we remember how to make certain movements in tongue, throat, or fingers, but also that we remember the original words. But this means no more than that the movements in question reproduce noises or marks which in fact resemble those which we had to imitate in acquiring the motor-capacity. I will call "memory", in this sense, "retention of an acquired speech-capacity", using "speech" in a wide sense to include the utterances of a parrot or the imitative scrawlings which a monkey might make. Here too "memory" is not a form of cognition; it is simply a kind of bodily action such as we perform when we find ourselves still able to swim after an interval.

    (4) One peculiar capacity which we may acquire by practice and retain is the power to call up an image which in fact resembles something that we have repeatedly seen or heard in the past. If we can do this, an external observer who knew of the fact would be inclined to say that we are remembering the thing which we have seen or heard in the past which the image in fact resembles. But this would not be an accurate use of the word "memory" unless the image seems familiar to us and leads us to make memory-judgments. Apart from this we have merely acquired and retained a peculiar kind of capacity; and "to remember", in this sense, is no more to perform a special kind of cognitive act than to swim is. The peculiarity of the present case is simply in the nature of the capacity which we have acquired. An image is not a series of movements, and to call up an image is not to move in a certain way. It is of course possible that the calling up of a certain kind of image is causally dependent on the occurrence of certain microscopic movements in the brain and nervous system, and that what we have primarily acquired is the capacity to initiate such microscopic movements at will. But this is purely hypothetical, and therefore it would be paradoxical and rash to count this capacity as a peculiar kind of motor-capacity. I will call "memory" in this sense "retention of an acquired capacity for imitative imagery".

    (5) The four kinds of " memory" which I have so far mentioned do not really deserve the name. In themselves they are modes of behaviour, and not modes of cognition. To call them "memory" is merely to state our belief that the capacity for behaving in these ways arose through our performing similar actions in the past, or through our having perceived something which resembled the image which we can now call up. Such acquired and retained modes of behaviour may be necessary conditions of genuine memory, but they are nothing more. But it is of course a fact that when I perform such actions as these a genuine memory-situation often does arise. The movements of swimming may seem familiar; so may the sounds which I utter, or the marks which I make, or the images which I call up. I see no reason why this should not happen even with parrots and monkeys. A further stage is that I may then make memory-judgments of various degrees of determinateness. I may judge that this has happened before, or I may definitely remember a certain occasion on which I have formerly swum. I see no reason to suppose that this stage is ever reached by animals. The first stage cannot properly be called "memory", though it approaches nearer to it than the four cases which we considered before. The second stage is definitely memory, i.e., a peculiar kind of cognition in which we seem to be in contact with a part of our own past history and with events which we then experienced. I think that the name "memory" is often applied by external observers to the first four cases because they unwittingly assume that what is in fact a repetition of a past mode of behaviour and is in fact causally dependent on past behaviour or past perception must be accompanied by a feeling of familiarity and a more or less determinate memory-judgment about the past. A very little careful introspection will suffice to show that this is a mistake.

    (6) There is one other important sense of "memory" to be considered. In dealing with perceptual memory we had occasion to consider a certain sense in which we "remember propositions". But there is another sense in which we are said to "remember propositions", and it is this which I want now to discuss. We often say that we remember propositions about historical characters, such as Julius Caesar, which we were taught at school. Again, I might well say that I remember Euclid's proposition I, 47, and Euclid's proof of it. Memory of propositions, in this sense, must be sharply distinguished from the memory of propositions which forms a part of perceptual memory. It must also be sharply distinguished from mere memory of sentences. We will take these two points in turn.

    (i) The propositions which I remember because I have once upon a time learned them may be about past events, but they need not be. Euclid I, 47 is about the timeless relations of certain abstract and timeless objects. Moreover, when the proposition which I remember in this way happens to be about a past event, I do not say that I remember the event because I remember the proposition about it. I certainly do remember that Caesar crossed the Rubicon, and I certainly do not remember the event which is described as "Caesar crossing the Rubicon". In perceptual memory the propositions remembered are always about past events; and, when we remember a proposition in this sense, we, ipso facto, remember perceptually the event which it is about.

    (ii) On the other hand, memory of propositions which we have been taught or have learnt for ourselves cannot be identified with a mere power to repeat the sentences in which these propositions were expressed when we learnt them, nor with such repetition accompanied by a feeling of familiarity in the words and by a perceptual memory-judgment. I remember Euclid I, 47 and his proof of it through having learnt it. But I certainly could not reproduce the words in my Euclid book; and I should recognise the proposition equally well if I now saw it stated for the first time in any foreign tongue that is known to me. Again, I might accurately reproduce a sentence without remembering a proposition This might happen if I did not understand the sentence e.g., if it were in Hebrew, a tongue which I do not understand though I can write the letters. Or it might happen even though I did understand the sentence when I learnt it, if I have now forgotten its meaning. Lastly, it might happen if I understood the sentence and could again understand it by giving enough attention to it, but I am now repeating it parrot-wise whilst thinking of other things. It is, therefore, impossible to identify memory of propositions with memory of the sentences in which they were originally expressed, for there is not even an invariable and reciprocal connexion between the two kinds of memory.

    I am said to remember a proposition which I have learnt provided I have acquired the power of contemplating it at will, and provided that, when I do contemplate it, it seems familiar to me. The first part of this definition might conceivably be fulfilled without the second. If it were, I do not think that the experient would himself say that he remembers the proposition; though an external observer would be very likely to say this of the experient. Now I think it likely that we cannot contemplate a proposition without some kind of concrete symbolism, though this may be to the last degree sketchy and vague without apparently interfering with our contemplation of the proposition. But, for the purpose of remembering the proposition, it is a matter of complete indifference what particular form the symbolism takes, and it is quite unnecessary for the present form of symbolism to resemble that which was used for expressing the proposition when we first met with it. No doubt the power to reproduce the original sentence at will is helpful as a means to enabling us to think of the proposition at will; but it is not essential, and it is sometimes positively harmful. Sometimes we have accepted a proposition in the past on authority or because of a process of reasoning which then satisfied us. Perhaps, if we were now to inspect and criticise the proposition, we should no longer accept it; the reasoning might not satisfy us now, and we might have lost our respect for the authority. But, unfortunately, we have acquired the power of reproducing the original sentence, in which the proposition was expressed when we first met it, at will. When we exercise this power we think that we are thinking of the proposition, and we remember that we have accepted it and that we had what seemed adequate grounds for doing so. But really we are not contemplating the proposition at all; we are just behaving like parrots or monkeys. Thus it comes about that intelligent grown men can honestly believe that they believe the most preposterous propositions in theology and politics, provided that these continue to be expressed in language that has been familiar to them since their childhood.

    I hope that I have succeeded in this chapter at least in showing how ridiculous it is to attempt to reduce memory to "language-habits". Such an attempt does not even seem to account for perceptual memory; and it fails to recognise the elementary distinction between remembering a sentence and remembering a proposition which one has learnt in the past. It is odd enough that the attempt should have been made; but it is far more odd tbat it should have been hailed as a wonderful step in psychology and as the last word in "advanced thinking".