I do not myself believe that the analysis of knowledge can be effected entirely by means of purely external observation, such as behaviourists employ. I shall discuss this question in later lectures. In the present lecture I shall attempt the analysis of memory-knowledge, both as an introduction to the problem of knowledge in general, and because memory, in some form, is presupposed in almost all other knowledge. Sensation, we decided, is not a form of knowledge. It might, however, have been expected that we should begin our discussion of knowledge with PERCEPTION, i.e. with that integral experience of things in the environment, out of which sensation is extracted by psychological analysis. What is called perception differs from sensation by the fact that the sensational ingredients bring up habitual associates—images and expectations of their usual correlates—all of which are subjectively indistinguishable from the sensation. The FACT of past experience is essential in producing this filling-out of sensation, but not the RECOLLECTION of past experience. The non-sensational elements in perception can be wholly explained as the result of habit, produced by frequent correlations. Perception, according to our definition in Lecture VII, is no more a form of knowledge than sensation is, except in so far as it involves expectations. The purely psychological problems which it raises are not very difficult, though they have sometimes been rendered artificially obscure by unwillingness to admit the fallibility of the non-sensational elements of perception. On the other hand, memory raises many difficult and very important problems, which it is necessary to consider at the first possible moment.
One reason for treating memory at this early stage is that it seems to be involved in the fact that images are recognized as "copies" of past sensible experience. In the preceding lecture I alluded to Hume's principle "that all our simple ideas in their first appearance are derived from simple impressions, which are correspondent to them, and which they exactly represent." Whether or not this principle is liable to exceptions, everyone would agree that is has a broad measure of truth, though the word "exactly" might seem an overstatement, and it might seem more correct to say that ideas APPROXIMATELY represent impressions. Such modifications of Hume's principle, however, do not affect the problem which I wish to present for your consideration, namely: Why do we believe that images are, sometimes or always, approximately or exactly, copies of sensations? What sort of evidence is there? And what sort of evidence is logically possible? The difficulty of this question arises through the fact that the sensation which an image is supposed to copy is in the past when the image exists, and can therefore only be known by memory, while, on the other hand, memory of past sensations seems only possible by means of present images. How, then, are we to find any way of comparing the present image and the past sensation? The problem is just as acute if we say that images differ from their prototypes as if we say that they resemble them; it is the very possibility of comparison that is hard to understand.* We think we can know that they are alike or different, but we cannot bring them together in one experience and compare them. To deal with this problem, we must have a theory of memory. In this way the whole status of images as "copies" is bound up with the analysis of memory.
* How, for example, can we obtain such knowledge as the
following: "If we look at, say, a red nose and perceive it,
and after a little while ekphore, its memory-image, we note
immediately how unlike, in its likeness, this memory-image
is to the original perception" (A. Wohlgemuth, "On the
Feelings and their Neural Correlate with an Examination of
the Nature of Pain," "Journal of Psychology," vol. viii,
part iv, June, 1917).
In investigating memory-beliefs, there are certain points which must be borne in mind. In the first place, everything constituting a memory-belief is happening now, not in that past time to which the belief is said to refer. It is not logically necessary to the existence of a memory-belief that the event remembered should have occurred, or even that the past should have existed at all. There is no logical impossibility in the hypothesis that the world sprang into being five minutes ago, exactly as it then was, with a population that "remembered" a wholly unreal past. There is no logically necessary connection between events at different times; therefore nothing that is happening now or will happen in the future can disprove the hypothesis that the world began five minutes ago. Hence the occurrences which are CALLED knowledge of the past are logically independent of the past; they are wholly analysable into present contents, which might, theoretically, be just what they are even if no past had existed.
I am not suggesting that the non-existence of the past should be entertained as a serious hypothesis. Like all sceptical hypotheses, it is logically tenable, but uninteresting. All that I am doing is to use its logical tenability as a help in the analysis of what occurs when we remember.
In the second place, images without beliefs are insufficient to constitute memory; and habits are still more insufficient. The behaviourist, who attempts to make psychology a record of behaviour, has to trust his memory in making the record. "Habit" is a concept involving the occurrence of similar events at different times; if the behaviourist feels confident that there is such a phenomenon as habit, that can only be because he trusts his memory, when it assures him that there have been other times. And the same applies to images. If we are to know as it is supposed we do—that images are "copies," accurate or inaccurate, of past events, something more than the mere occurrence of images must go to constitute this knowledge. For their mere occurrence, by itself, would not suggest any connection with anything that had happened before.
Can we constitute memory out of images together with suitable beliefs? We may take it that memory-images, when they occur in true memory, are (a) known to be copies, (b) sometimes known to be imperfect copies (cf. footnote on previous page). How is it possible to know that a memory-image is an imperfect copy, without having a more accurate copy by which to replace it? This would SEEM to suggest that we have a way of knowing the past which is independent of images, by means of which we can criticize image-memories. But I do not think such an inference is warranted.
What results, formally, from our knowledge of the past through images of which we recognize the inaccuracy, is that such images must have two characteristics by which we can arrange them in two series, of which one corresponds to the more or less remote period in the past to which they refer, and the other to our greater or less confidence in their accuracy. We will take the second of these points first.
Our confidence or lack of confidence in the accuracy of a memory-image must, in fundamental cases, be based upon a characteristic of the image itself, since we cannot evoke the past bodily and compare it with the present image. It might be suggested that vagueness is the required characteristic, but I do not think this is the case. We sometimes have images that are by no means peculiarly vague, which yet we do not trust—for example, under the influence of fatigue we may see a friend's face vividly and clearly, but horribly distorted. In such a case we distrust our image in spite of its being unusually clear. I think the characteristic by which we distinguish the images we trust is the feeling of FAMILIARITY that accompanies them. Some images, like some sensations, feel very familiar, while others feel strange. Familiarity is a feeling capable of degrees. In an image of a well-known face, for example, some parts may feel more familiar than others; when this happens, we have more belief in the accuracy of the familiar parts than in that of the unfamiliar parts. I think it is by this means that we become critical of images, not by some imageless memory with which we compare them. I shall return to the consideration of familiarity shortly.
I come now to the other characteristic which memory-images must have in order to account for our knowledge of the past. They must have some characteristic which makes us regard them as referring to more or less remote portions of the past. That is to say if we suppose that A is the event remembered, B the remembering, and t the interval of time between A and B, there must be some characteristic of B which is capable of degrees, and which, in accurately dated memories, varies as t varies. It may increase as t increases, or diminish as t increases. The question which of these occurs is not of any importance for the theoretic serviceability of the characteristic in question.