You are here

Series I

Lecture 4: Belief and Evidence (I)

The Evidence of Perception, Memory and Self-Consciousness

We should all agree that a person can only believe reasonably when he has evidence for the propositions believed. Moreover, our evidence for a proposition p which we believe must be stronger than our evidence (if any) against that proposition, if our belief is to be reasonable. Nor is this all. Belief admits of degrees. And if we are to believe reasonably, the degree of our belief must be no greater than our evidence justifies. For instance, it would be unreasonable to believe a proposition with complete conviction if our evidence, though good as far as it goes, falls short of being conclusive, though it would still be reasonable to believe with a good deal of confidence.

But unfortunately the notion of ‘evidence’ is full of difficulties. For one thing, the meaning of the word in modern English is quite different from the one its etymology suggests. The Latin word ‘evidentia’ literally means ‘evidentness’. In the French word ‘evidence’ and the German ‘evidenz’ the etymological meaning is still retained. But in modern English it only survives in the term ‘self-evident’. A self-evident proposition is one which is ‘evident of itself’. You have only to consider the proposition itself and then it is evident to you that it is true. We notice however that here, in talking of ‘self-evidence’, we are concerned with the evidence of a proposition and not with the evidence for it. It does make sense to speak of the evidentness of a proposition, but it makes none at all to speak of the evidentness for it. And in modern English (apart from this one exceptional case), ‘evidence’ does always mean ‘evidence for…’. The evidence for a proposition consists of those considerations which support that proposition or confer some degree of probability upon it, great or little. The evidence for a proposition may be strong or weak, whereas evidentness is a matter of all or nothing. Again, there may be evidence against a proposition, consisting of considerations which decrease the probability of that proposition; whereas the contrast between ‘for’ and ‘against’, or between favourable and unfavourable, does not apply to evidentness at all.

Presumably the word ‘evidence’ came to have this meaning in modern English, because we may acquire information which makes a proposition evident or obvious though it would not have been evident or obvious otherwise. The testimony of the witnesses makes it evident or obvious that the prisoner is guilty. But though this was presumably the first step in the divergence between the modern English sense of the word ‘evidence’ and its etymological sense of ‘evidentness’, there was more to come. ‘Evidence’ eventually came to mean not just considerations which make a proposition evident or obvious, but any considerations which make it in any degree probable.

Another point of terminology may be briefly mentioned. We speak of ‘reasons’ for believing something, and also of ‘evidence’ for believing it. What is the relation between the two? The answer is, I think, that the notion of ‘reasons for’ is wider. We may have reasons for doubting something, or for being surprised at it, or for suspending judgement about it. These three—doubting, being surprised, suspending judgement—are of course, fairly closely related to believing. But we may also have reasons for hoping or fearing, for being anxious, for liking someone or disliking him, or even for being angry with him. Again, we are expected to have reasons for the advice which we offer to someone, or the recommendations we make. Most important of all, we may have reasons for doing something, or for deciding to do it.

In this discussion, however, we are concerned only with reasons for believing. In some of the other cases just mentioned, when we speak of ‘reasons for’ such and such, it would be quite inappropriate to substitute the word ‘evidence’ instead. ‘What evidence have you for being anxious about John's health?’ would be a strange question. And ‘what evidence have you for deciding to take the 4.15 train?’ would hardly be intelligible. But in the special case of believing, the two questions ‘what is your evidence for…’ and ‘what are your reasons for…’ amount to pretty much the same thing. Perhaps ‘reasons’ here are primarily concerned with the mental attitude of believing (What reason have you for taking that particular attitude, rather than another, e.g. doubting or disbelieving?) whereas ‘evidence’ is primarily concerned with the proposition believed (what is the evidence for the proposition that this is the road to Aylesbury?). But when it is belief that we are talking about, there is at any rate a very close relation between the two questions. An answer to either would also serve as a satisfactory answer to the other.1

I think, then, that in this discussion we may safely confine ourselves to questions about evidence, though anyone who wishes to may translate what I am going to say into the terminology of ‘reasons’.

The Evidence ‘There Is’ and the Evidence X ‘Has’

Now it would often be said that the evidence for a proposition p consists of some relevant fact or set of facts which ‘support’ that proposition, or increase its probability; and that the evidence against p consists of some relevant fact or set of facts which ‘weaken’ that proposition or decrease its probability.

But there are two objections to this formulation. If we are concerned to decide whether a particular person's belief is reasonable (or how reasonable it is, if reasonableness admits of degrees) what we must consider is not just what evidence there is for the proposition he believes, but the evidence which he has for that proposition. He may have evidence for it which others do not have, and equally he may fail to have evidence which others do have. Primitive people, we may suppose, believed pretty firmly that the sun is much smaller than the earth, and about the same size as the moon. Their belief was mistaken. But we must not conclude that it was therefore unreasonable. Neither telescopes nor trigonometry had been invented at that time. The evidence which we have about the size and position of the sun and the moon was not available. They only had the evidence of unaided sense-perception, and this (as far as it goes) really does give some support to the proposition which they believed.

Must the Evidence Consist of Known Facts?

Secondly, there is a difficulty about the word ‘fact’ (‘some relevant fact or set of facts which support the proposition believed’). It follows from what has just been said that if we are asking about the reasonableness of a particular person's belief, we must not just pay attention to the relevant facts which there are. Facts of which a man is completely ignorant, however relevant they may be, can have no bearing at all on the reasonableness of his belief. The only ones which do have a bearing on it are relevant facts which he himself is aware of.

And now we find that the word ‘fact’ gets us into difficulties. For when we say ‘It is a fact that so-and-so’, we are making a claim to knowledge. If I speak of the fact that to-day is Friday, I am claiming to know that to-day is Friday. Sometimes, no doubt, when I believe some proposition, my evidence really does consist of some relevant fact which I know. The primitive people I mentioned might fairly claim to know that the sun occupied a much smaller part of their visual field than the terrestial landscape did. But it is by no means obvious that when we believe reasonably our evidence must always consist of some relevant fact or facts which we know in any strict sense of the word ‘know’. On the contrary, it seems (at first sight at any rate) that our evidence for a proposition which we believe consists quite often in other propositions which we believe—believe in a sense in which belief is contrasted with or inferior to knowledge. My evidence for believing that the Conservatives will win the next election consists of a lot of propositions which I have read in the newspapers during the last month or two. I believe these propositions more or less firmly: but can I claim to know in any strict sense of the word that all of them are true, or even that most of them are? It does look as if my belief about the next election was supported by nothing better than a set of other beliefs.

At any rate, there is a problem here, and we prevent ourselves from discussing it if we begin by laying down a rule that a belief cannot be reasonable, unless the evidence for it consists of relevant facts known to the believer. Instead of speaking of ‘relevant facts’ it is better to use some more colourless phrase like ‘relevant considerations’; that leaves us free to enquire what kind of considerations they have to be. Might they just consist of other propositions which we already believe, but do not know to be true?

The suggestion that they might is strengthened when we consider the relations between belief and knowledge. As we have seen, we usually draw a distinction between belief and knowledge. It does not make sense to say ‘John knows that it is raining, but it isn't’: whereas it does make sense to say ‘He believes that it is raining, but it isn't’—even though he believes this with complete and unshakable conviction. Here there is a very sharp distinction between belief and knowledge. Any belief, no matter how firmly held, can be mistaken: the proposition believed may still be false, and it still may be false, even when the belief is reasonable. But to speak of ‘mistaken knowledge’ would be self-contradictory.

We have also seen, however, that some philosophers have suggested that ‘knowledge’ itself can be defined in terms of ‘belief’.2 But even so, the distinction between knowledge and belief is not abolished. For we still have to distinguish between a belief which amounts to knowledge and a belief which does not; or, if you like, between knowledge on the one hand and ‘mere’ belief on the other. And thus the question asked just now can still be raised: does the evidence for a reasonable belief have to consist of propositions known to be true, or can it consist (wholly or partly) of propositions which are ‘merely’ believed?

The suggestion that it can is supported when we consider why we need to have ‘mere’ beliefs at all. Why not be content with such knowledge as we can get? The answer is that unfortunately we can get so little. If we are to live at all, we must constantly make practical decisions without knowing what the result of our action is going to be. Let us consider an academic person who has to give a lecture in a distant university this evening. When he gets into the train to go there, he certainly does not have conclusive evidence that the train will get him there in time to give his lecture. He has to be content with forming the most reasonable belief that he can on the evidence available. It is inductive evidence: most trains reach their destination not more than half an hour later than the timetable says they will. On the strength of this, he can reasonably believe with some confidence (but not with complete conviction) that he will arrive in time to address his audience. But if he had demanded conclusive evidence for this proposition before deciding which train to catch, he would never have got there at all.

Again, when he set out on his journey, he did not know that he would not be stricken by aphasia, on the way, or otherwise incapacitated, whether physically or mentally, before he got there. Such things do happen, and he had no conclusive evidence that one of them was not going to happen to him. But on the evidence available to him about the state of his health, he could reasonably believe with considerable confidence (but not with absolute conviction) that he was going to arrive there safe and sound.

One thing, then, which makes us believing beings is that we need beliefs for the guidance our actions and our practical decisions. Belief is a second best, but it is much better than nothing. But we need it in another way too. Knowledge is something which we value for its own sake. (Everyone has some curiosity, and what is curiosity but a desire to know?) But here again, our trouble is that very often knowledge in any strict sense of the term is not available, especially, but not only, when the knowledge we should like to have concerns something rather remote in space or time. If we were to say ‘we must have knowledge or nothing’ then—very often—we should have to be content with nothing. We should have to remain in a state of suspended judgement, a state of complete ‘agnosticism’ about many of the questions which interest us (for example what were the motives of the Emperor Constantine when he adopted Christianity?). But though we have not got conclusive evidence which would enable us to know the answer, we may still have evidence which supports one answer rather than another. Thus we may have evidence which makes it probable, though not certain, that Constantine's motives were at least partly political ones, and improbable, though not certainly false, that they were purely religious. And here again reasonable belief is a second best; but still it is very much better than nothing, though knowledge would be better still if only we could have it.

Belief and Inference

Now let us consider what we do with our beliefs when once we have got them. The mathematical philosopher F. P. Ramsey suggested that when we come to believe a proposition, we ‘add it to our stock of premisses’. I think he has put his finger on a very important function which our beliefs have. They enable us to draw inferences. Indeed, believing a proposition seems to consist at least partly in a tendency to draw inferences from the proposition believed. If someone claimed to believe that to-day is early closing day and yet set out on a shopping expedition this afternoon, we should doubt whether he did really believe what he claims to believe. For if he really does believe that it is early closing day, he must surely be capable of drawing the very simple inference ‘The shops are shut this afternoon’.

As I have said already, we need beliefs for the guidance of our actions and our practical decisions. This is another way of saying that we draw practical inferences from the propositions we believe, or use them (when relevant) as premisses in our practical reasoning. But we draw theoretical inferences from them too. If one likes to put it so, we use them for the guidance of our thoughts as well as our actions. If I believe that the motives for Constantine's conversion were at least partly political, I shall think that this makes it likely that his motives for summoning the Council of Nicaea were partly political also. In other words, when we believe a proposition p, we do use that proposition as evidence to support other propositions. The inference we draw from the proposition p takes the form ‘p, so probably also q’. Indeed, this is one of the most important uses we have for our beliefs, once we have got them. When we come to believe a proposition p we consider its implications—what follows from it either certainly or probably. If we are reasonable, the propositions which follow from p with certainty are believed as firmly as p itself, though not of course more firmly.

Let us take the case of suspecting (‘Suspecting is the name traditionally given to the lowest degree of belief). If I suspect that to-day is early closing day, I am entitled to suspect that the shops will be shut this afternoon: indeed I am logically committed to suspecting this, since the second proposition is logically entailed by the first. But I am not entitled to be sure, or even almost sure, that the shops will be closed this afternoon.

If however the inference I make from the proposition believed is what we call a probable inference, ‘p, so probably also q’, the degree of belief which I give to q must be lower than the degree of belief I give to p, if I am reasonable. For example, if I am absolutely sure that it is raining now, I am not entitled to be absolutely sure that it will still be raining in five minutes’ time (unless I have some other evidence for believing so). But I am entitled to believe this second proposition with considerable confidence.

It will be seen that when we consider the consequences of some proposition which we believe, our belief attitude spreads itself as it were, or extends itself, from that proposition to its consequences. But in a reasonable believer it suffers some degree of diminution on the way, if they are only probable consequences, supported but not logically entailed by the proposition originally believed.

Beliefs Supported by Other Beliefs

We may now return to the relation between belief and evidence. Our conclusion so far is this: on the face of it, it does not seem to be always true that when we believe reasonably our evidence consists of known facts (facts known to the believer). Quite often it seems to consist of other propositions which are themselves ‘merely’ believed.

But this involves us in an awkward problem. The problem is particularly awkward if we wish to define knowledge itself in terms of belief (‘believing a true proposition on conclusive evidence and with full conviction’), and it must be admitted that this definition does fit some sorts of knowledge quite well. If the evidence for a proposition p is to be conclusive, surely it must consist of other propositions which are themselves known to be true? So according to this definition of knowledge, these other propositions too must be believed on conclusive evidence. And the evidence for them, in their turn, must consist of still other propositions which are believed on conclusive evidence. Here we seem to have committed ourselves to a regress which has no discernible end—a regress of ‘evidence for our evidence for our evidence…’.

But the same sort of difficulty arises about beliefs which do not amount to knowledge. For instance, I do not even claim to know that John is away from home to-day; I only claim that it is reasonable for me to believe this with a considerable degree of confidence, the degree of belief which is traditionally called ‘opinion’. What evidence do I have for this belief? Well, I am fairly sure (but not quite) that he told me so when I saw him last Tuesday. My opinion is supported by nothing better than another opinion.

To put it metaphorically: we seem to be in a bog or a quagmire with no firm ground anywhere. Or, if one prefers another metaphor, we build up an elaborate structure of beliefs supported by other beliefs, and the whole thing just hangs in the air like a cloud. Is it anything better than a more or less coherent fiction? Some philosophers might be willing to accept this situation. They would say that if a set of propositions is sufficiently coherent, that in itself makes them true. Yet it seems obvious that a highly coherent system of propositions might still be wholly fictitious. (I recently read Professor Tolkien's series of novels called ‘The Lord of the Rings’. I was very much impressed by the coherence of this complicated story, and it occurred to me that if the Coherence Theory of Truth were correct the whole story would have to be true!)

Let us try to see whether we can solve the problem in another and more commonsensical way. We shall find, however, that we get into difficulties about two important sorts of evidence which we have for our beliefs, the evidence of memory and the evidence of testimony. In both cases the ‘coherence’ interpretation has very considerable plausibility.

What I shall try to show is that this regress of ‘evidence for our evidence’ does have a termination: so that in the long run (though not always immediately) the evidence we have for our beliefs does after all consist of known facts—that is, of facts known to the believer—if we are believing reasonably. But I shall only consider empirical beliefs, what Hume called beliefs concerning matters of fact, and shall not discuss a priori or logically necessary propositions. It is not clear to me whether the notion of believing (either reasonably or unreasonably) applies to a priori propositions at all, nor whether the notion of evidence applies to them, though the notion of proof certainly does. At any rate, when we speak of the evidence we have for believing something, the type of evidence we have in mind is always or nearly always empirical evidence. It may of course be maintained (and often has been) that some very simple mathematical and logical truths are self-evident. But as we have seen already, in the phrase ‘self-evidence’, ‘evidence’ means just evidentness. And in a discussion of belief the word ‘evidence’ has quite a different sense; it means evidence for… and that is something which is quite different from evidentness.

There seem to be four different ways in which the series of beliefs supported by other beliefs can be terminated. To put it differently, there are four different sorts of evidence which do not just consist in propositions which are themselves merely believed. They are: (1) the evidence of perception (2) the evidence of self-consciousness (3) the evidence of memory (4) the evidence of testimony. The fourth is the most puzzling, and moreover it is not wholly independent of the first and second. Testimony itself has to be perceived (e.g. heard or read) and it also has to be remembered if it is to be of any use to us. But we do normally think of it as a separate source of evidence for our beliefs, and it raises special difficulties of its own which do not arise about perception, self-consciousness or memory. So in the rest of this lecture I shall only consider the evidence of perception, of self-consciousness and of memory. The evidence of testimony needs a lecture to itself.3

The Evidence of Perception

We all think that some questions can be conclusively settled by means of sense-perception; and not only that they can be conclusively settled in that way, but that they very frequently are.

It is true that the epistemology of perception is a complex and highly controversial subject. The analysis of such a simple-looking statement as ‘I am now seeing a piece of paper’ is very difficult, and many strange opinions have been held about it. The distinction between ‘appears’ and ‘actually is’ is puzzling too: for example, on a foggy evening the sun often appears oval, but it is not actually oval. How can something appear to have a quality which it does not actually have, and in what sense exactly can we be said to be ‘experiencing’ this quality when the thing appears to us to have it? Or should we perhaps say that ‘actually is’ should itself be defined in terms of ‘appears’, and hold (as some philosophers have) that such a thing as this chair here is just a class or system of actual and possible appearances?

But in spite of these difficulties, we all do think that some questions can be conclusively settled by the evidence of sense-perception, for example the question ‘Are there any matches in this box?’ By looking inside, it is possible to know or be certain that there are some matches there, or that there are none. G. E. Moore was surely right in maintaining that we know many propositions of this sort to be true even though we do not know what the correct analysis of them is.

If so, it is perfectly proper to speak of observed facts, as we all do in practice, whatever philosophical theories we may hold. And this is one way (the most familiar way) in which the regress of beliefs supported by other beliefs comes to an end. This regress—the regress of evidence for our evidence, as I called it—is terminated sometimes by an observed fact, that is by a fact ascertained or discovered by means of sense-perception.

For example, suppose I have to go to Cambridge this afternoon, and I decide to take a bus to the railway station. My reason for this decision is that I believe I shall get very wet if I walk all the way. What is my evidence for believing this? It again is something which I believe but do not know. I believe that it will be raining at that time. And what is my evidence for that belief? This again might be something which I believe but do not know. I might have had the misfortune to be blind, or to be shut up all the morning in a room with no window; and then at half past twelve some kind person might have told me that it was raining steadily, and I might just have taken his word for it. But as it happens, I was not in this unfortunate situation. I was able to go out into the garden just before lunch and see for myself that it was raining steadily, to feel the raindrops on my hands and face, to look at quite a large area of the sky and see for myself that there was no break in the clouds there. This was enough to settle the question ‘What is the weather like here and now?’ It was an observed fact, a fact which I observed for myself at first hand, that at that place and time it was raining steadily and that there was no break in the clouds within my range of vision.

Some philosophers may maintain that this language I have used is too strong. They may tell me that even here I was still only believing, and had no right to use the word ‘fact’ (which implies a claim to knowledge). But perhaps this dispute is less important than it seems. Suppose we do say that I only believe that it is raining when I have the experience commonly described as ‘actually seeing the rain and actually feeling the raindrops’. Nevertheless, we have to admit that my evidence for this believed proposition is as strong as the evidence for any empirical proposition could be, and we cannot well conceive what it would be like to have better evidence for the proposition ‘It is raining’. To use an analogy suggested by Professor Ayer, so long as we agree that evidence of this kind gets ‘top marks’, it does not matter very much where we draw the line between knowledge and belief, or between the beliefs which amount to knowledge and the beliefs which do not. But in actual fact, we should all say that in the circumstances described a person does know that it is raining.

In our reaction against over-sceptical views of perception, we must not of course jump over to the opposite extreme and maintain that all propositions for which we have perceptual evidence are certainly true. There are illusions and hallucinations, and what we perceive is not always as it appears to be. There are perceptual mistakes, and there are several different kinds of them. But we have means of detecting them. Roughly speaking, we do it by finding that the expected consequences of our belief do not occur. For example, in a mirage there appears to be a pool of water some distance away. We go to the place, and instead of seeing the water more clearly and in greater detail (as we should expect to do if it were actually there) we see nothing but an expanse of sand or tarmac, and we cannot get the tactual experiences we expect to have when we put our hands in a liquid.

Again, quite apart from illusions and hallucinations, some perceptual experiences are better from the evidential point of view than others. A single glance is not always enough to settle the question which we wish to ask (for example, the question ‘Is that a swan over there?’) though it does provide us with some relevant evidence. Often we must move to another point of view. We must come closer, or look at the thing from another direction; or we must switch on the electric light, or put on spectacles. Sometimes we must touch as well as sight. In order to decide whether this is petrol or water, we may have to smell it or even taste it.

So, within the class of propositions for which we have perceptual evidence, we can still distinguish between (1) those which are known or certified and (2) those which are ‘only’ believed. In this sphere, as in others, there is plenty of room for mere belief which does not amount to knowledge, because our evidence is not conclusive, (although it is good evidence as far as it goes.) The important points for our present argument are these: (1) First, there are some propositions which are perceptually certified, conclusively established by perceptual evidence, and there are many others for which we have strong perceptual evidence although it is not conclusive. (2) Secondly, even when we have only relatively weak perceptual evidence, this is sufficient to bring the regress of ‘evidence for our evidence’ to an end. The series of beliefs supported by other beliefs is terminated when we get back to a belief supported by perceptual evidence. For this belief, though used to support other beliefs, is not itself supported by another belief. It is supported by something quite different, namely by an experience, for example an experience of seeing or touching or hearing.

The Evidence of Self-Consciousness

This phrase ‘an experience’ draws our attention to the second source of evidence I wish to consider—the second way in which the series of beliefs supported by other beliefs is brought to an end. This second sort of evidence is the evidence of self-consciousness. In the example I gave before about the rain, the experience of seeing and feeling the raindrops was my own. When someone has an experience of this sort, he need not of course attend to it. But if he does, he can know that he is having it. We can sometimes claim to have knowledge about ourselves, and here again the regress of beliefs supported by other beliefs is brought to an end. For example, I believe that I should make a hopelessly bad soldier or policeman or air-raid warden. What is my evidence? Well, I am a timid person. What is my evidence for believing this? My evidence is not just some other proposition which I believe. It is something I have noticed about myself. I have noticed on a great many occasions that I am easily frightened by persons, objects and situations which do not seem to frighten others at all. I feel fear in the presence of a large dog, or of a man who addresses me in a loud and angry tone of voice, or when I am driving a car on a main road on the day before Bank Holiday, or when I have to go to the dentist, or even sometimes when I have to deliver a lecture. My evidence for believing that I am a timid person is the evidence of self-consciousness, the frequent experiences of fear which I have noticed in myself on many different sorts of occasions.

In the present climate of philosophical opinion it is hardly necessary to point out that a person does need evidence for the statements he makes about himself, especially when they are statements about his own character. In these Behaviouristic days, we are in no danger of claiming that every human being (or every sane adult human being) has a vast store of infallible knowledge about his own mind. The danger nowadays is all the other way. We are more likely to assert that others know or can know far more about us than we know about ourselves, or even that each of us knows almost nothing about himself or his own mind, and that most of the statements he makes about himself express ‘mere’ beliefs of a highly questionable sort. So if you want to know the truth about yourself you had better ask someone else, and preferably a person who does not like you very much.

No doubt this is a sphere in which we are peculiarly prone to error. That is why the precept γνωθι σεαυτον (‘Know thyself’) was needed. Self-knowledge is often a difficult achievement, and it is difficult in several different ways. One difficulty is that many mental states and happenings are so fugitive. They come and go so quickly and do not ‘stay put’ to be examined. Sometimes, however, the difficulty is just the opposite. A persistent state of mild melancholy or depression which ‘colours’ all our thoughts and feelings for the whole day, or even for weeks on end, or even throughout the whole of our waking life, is easily overlooked. One may fail to notice it, because there is nothing to contrast it with. More important still, self-knowledge and especially knowledge of one's own character (one's conative and emotional dispositions) is often painful. This is because we are moral beings. Self-knowledge is hardly separable from self-judgement, and only too often the self-judgement has to be of a disapproving kind. It is not very surprising that self-knowledge is something which we tend to dislike, and that when it comes to us we often do our best to forget about it.

All the same, there is such a thing as self-consciousness. The idea that the only good evidence a person can get for belief about himself is the evidence of testimony is too absurd to be taken seriously, however true it is that external observers may sometimes have more correct beliefs about us than we ourselves have. I can sometimes notice that I am feeling frightened or tired or surprised, that I still feel resentment about an unkind remark made to me yesterday. I do not have to look in a mirror and observe my bodily behaviour in order to ascertain these facts about myself. I can sometimes notice that I am hearing or seeing or thinking, that I am imagining this or wondering about that. I notice these things for myself at first hand and do not need the confirmation of others. Self-consciousness is one source of evidence, and an indispensable source of evidence, for the beliefs which each person has about himself, although the testimony of others who observe his bodily behaviour is also relevant. You may think that this is a platitudinous conclusion. So it is, but platitudes are just the things which Philosophers are liable to forget. What is called ‘a firm grasp of the obvious’ is less common among learned men than one might suppose.

The Evidence of Memory

We may now turn to the third way in which the regress of beliefs supported by other beliefs may be brought to an end, namely by means of the evidence of memory. Since sceptical views about memory have been rather prevalent among philosophers, and some have doubted whether there can be memory-knowledge at all, I shall begin by considering the defects of memory.

There are two quite different ways in which a person's memory may be defective. Badness of memory is something like badness of character. There are sins of omission and there are sins of commission. I may fail to do the things I ought to do, or I may do the things I ought not to do; and very likely I have both these kinds of moral defect. There is a similar distinction between omission and commission where defects of memory are concerned. If you ask me where I was on this day of the month ten years ago, I cannot remember at all. Everyone has forgotten very many of his past experiences. And for some of our beliefs about the past—even though it was our own past—we have only the evidence of testimony, for example, what our parents or other relatives have told us about what we did in our childhood. Sometimes the testimony is in a way our own. I might be able to tell you what I was doing on this day of the month ten years ago if I could find my own diary for that year.

In such cases as these the defects of memory are defects of omission. It is important to mention then, but they do not raise any very difficult philosophical problem. The defects of memory which I compared to sins of commission are a different matter. For then it is not just a case of failing to remember but of mis-remembering, much as sins of commission are misdoings and not just failures to do something. And as reflection on our misdoings gives rise to pessimistic theories of human character, such as the doctrine that human nature is totally depraved, reflection on our mis-rememberings likewise gives rise to sceptical theories of memory. Memory is fallible. It sometimes turns out that what a man sincerely claims to remember did not in fact happen. Then how can we be sure that all our claims to remember are not equally mistaken? Whenever we claim to remember something, it is conceivable that we might be mis-remembering. In that case, how can we know anything about the past at all?

Now there is something wrong with this argument. It cannot even be stated unless we assume that some of our claims to remember are correct, that sometimes when we claim to remember we really are remembering and not mis-remembering. How do we know that memory claims are ever made at all? Because we remember making them ourselves and remember hearing others speak as if they were making them. And how do we know that some of these memory claims were incorrect? Because we are able, somehow, to find out what the facts about the past actually were. And in order to find out what they actually were, we ourselves must rely on memory at some point or other. For example, you claim to remember having lunch with Thomas in London last Thursday, but you must be mistaken. How do we know that you are mistaken? Because we all saw you having lunch in College in Oxford on that day. But we ourselves are relying on our own memories when we say this.

Let us now consider another way in which we might find out that our friend's memory claim is mistaken. Suppose he specifies the time. He claims to remember meeting Thomas in the restaurant in London at 1.5 p.m. Yet he was seen in Oxford at 1.6 p.m. on that day. Surely this is sufficient to show that his memory claim must be mistaken? It is, but only because we assume the validity of certain causal laws. It is not logically impossible that a person should travel from Oxford to London in one minute, but it is causally impossible, the laws of Nature being what they are. We must however ask ourselves what evidence there is for believing that the laws of Nature are of this particular sort. It is empirical evidence, the evidence of observations and experiments. And they are past observations and experiments. But if knowledge of the past is impossible (as the sceptic says it is) past observations and experiments cannot be evidence for anything. Nor will it do to say that we can rely on documentary evidence—books, articles in scientific journals, etc.

We do of course rely on documentary evidence, but here again it is reliance on what we have read, and not just on what we are at this present moment reading. And if it is not strictly correct to speak of the present as momentary, because it has a finite duration and is what is called ‘a specious present’, still the duration of the specious present is very brief indeed, only a very few seconds. There is another difficulty about documentary evidence which I will just mention. It would be absolutely useless to us, unless we were justified in assuming that ink-marks retain an approximately constant shape and approximately constant spatial relations to one another over long periods of time. But our evidence for this is again empirical (there is no a priori reason why they should not change their shapes every other minute or move about all over the page) and it again is the evidence of past experience.

I hope this is enough to show that we can have no ground for thinking that a particular memory claim is mistaken unless we assume that other memory claims are correct.

‘Saving the Appearances’

The pessimistic way of formulating our conclusion would be this. ‘Do what we will, we have to depend on the evidence of memory in the end, however sceptical we try to be. So we must just make the best of it. There is no prospect of getting anything better.’ If that is our view, we shall have to proceed in something like the manner recommended in the Coherence Theory of Truth. It will be a case of ‘saving the appearances’. We shall try to form a system of mutually supporting memory claims or memory propositions, in which as many of them as possible are retained and as few as possible rejected, though some, no doubt, will have to be rejected. In order to do this, we shall have to suppose that every memory claim has some degree of intrinsic weight or credibility. Unless we put in some proviso of this kind, our system of memory propositions, however coherent we make it, might be no more than a coherent fiction.

But once we have introduced this rather strange property of intrinsic weight or credibility, we can hardly fail to notice that it admits of degrees. Some memory claims have more of it than others. My claim to remember that I had several cups of tea at breakfast this morning has more ‘credit-worthiness’ than my claim to remember switching off the electric fire before I went to bed last Saturday night. Or again, I claim to remember (or, as we also say, I seem to remember) having seen this man before, when I was in California last winter; but the weight or creditworthiness of this claim is small, and I could easily be persuaded that it is mistaken.

So if we do proceed in this way, and try to construct a coherent system which will include as many memory claims as possible, we cannot follow the democratic principle of Bentham, that each of them is to count for one and none of them for more than one. Our aim is to ‘save the appearances’. But some of the appearances are more worthy of salvation than others, and we must save them first and be prepared to cast out the less worthy candidates, if there is a conflict between a more worthy one and a less.

But once we admit that this property of intrinsic credibility or credit-worthiness varies in degree, we may have to go farther. Might there not be cases where this property reaches a maximum? Might there not be some memory propositions which present themselves to us with a weight or credibility or credit-worthiness so great that no empirical proposition could possibly have more? On the face of it, there are such memory-propositions and we all of us are sure that there are.

For instance, could any possible adverse evidence induce me to reject the very clear recollection I now have that I have been sitting in a chair and writing for some time? Well, just conceivably it could. Unknown to me, a dose of lysergic acid might have been put into a cup of tea I had at breakfast, and for the past forty-five minutes I might have been in a state of hallucination. But even so, I very clearly recollect that I have for some time been having experiences which were as if I was sitting in a chair and writing, and I cannot conceive of any empirical evidence which would convince me of the contrary.

It seems likely that everyone has such memories pretty frequently, and can recognize them when he has them. Not all of them are memories of the very recent past, though these are the most obvious examples. They may be memories of something which happened many years ago, for example of some episode in a railway journey on one's first visit to the Continent. Such recollections of the distant past tend to be fragmentary and isolated. They usually come to us just as memories of ‘long ago’, and do not have a determinate date attached to them. To date them, we have to resort to causal inferences or to documentary evidence or oral testimony.

We should also notice that among these ‘unshakable’ recollections of our past experiences, recollections which no amount of adverse evidence would induce us to withdraw, there are some which have a general character. When put into words they take the form of statements which are in one way or another general statements: for example ‘I have gone to bed many times’, ‘I have often been to London’, ‘I have sometimes played cricket and have very seldom enjoyed it’.

It follows, if I am right, that the task which confronts us in our study of memory is not just one of ‘saving the appearances’, that is, our apparent or ostensible recollections, with due regard to the fact that some of them are ab initio more worthy of salvation than others. For some of them, and not so very few of them either, are improperly described as ‘apparent’ or ‘ostensible’. Mixed in among the appearances which have to be saved there are some realities. They are not just memory-claims but actual memories. Indeed, if this were not so, how could we talk of ostensible or apparent memories at all, or attach any meaning to the phrase ‘claim to remember’? What is this which we claim to be doing, or are ostensibly or apparently doing but perhaps not really? It is remembering. There happens to be such a thing. It is logically possible that there might not have been. But if there had been no remembering, there would have been no persons either, and indeed no minds or minded creatures at all. Leibniz' ‘mens momentantea seu carens recordatione’ would hardly deserve the name of mens.

So much for three sorts of evidence by which the series of beliefs supported by other beliefs is brought to an end: the evidence of perception, of self-consciousness and of memory. The problems we have considered in this chapter are at any rate familiar ones. They have been discussed by philosophers for many centuries. But that cannot be said about the fourth sort of evidence, to which we must now turn, the evidence of testimony. In practice, and in learned enquiries too, we do often rely upon it as a means of terminating the regress of beliefs supported by other beliefs. But to the best of my knowledge, epistemologists have had very little to say about it, and there are no ‘standard views’ which we might use as starting-points for our discussion. The next lecture, then, is bound to be difficult both for the lecturer and the audience.

  • 1.

    Cf. also the parallel between ‘What reasons are there for believing that p?’, and ‘What evidence is there for p?’: and likewise between ‘What reasons have you for believing that p?’ and ‘What evidence have you for p?’.

  • 2.

    Lecture 3, pp. 83–91.

  • 3.

    See Lecture 5, below.

From the book: