A Tale of Two Distinctions

G.E. Moore and Bertrand Russell on Memory and Ways of Knowing
Charles W. Johnson
22 October 2002

This essay is © 2001 by Charles Johnson, and is freed for credited reprinting and derivative works (but not, of course, for plagiarism) under the terms of the Creative Commons Attribution-ShareAlike 2.0 copyleft license. If you have any questions or comments about the essay, feel free to contact the author.

They are deeply immersed in illusions and dream images; their eye glides only over the surface of things and sees forms; their feeling nowhere leads to truth, but contents itself with the reception of stimuli, playing, as it were, a game of blindman’s bluff on the backs of things.

—Nietzsche, On Truth and Lie in the Extra-Moral Sense

I. Introduction

In Some Main Problems of Philosophy, G.E. Moore famously holds up an envelope and directs his audience to inspect it. In inquiring with them as to what happened, Moore launches into one of the most hotly debated issues in the past century of Analytic philosophy: the nature of apprehension, our cognitive relationships to the external objects amongst which we live, and our ability to make judgments about the world in virtue of what we see. In his effort to complete the story about the envelope, Moore comes to the topic of just what seeing is and what sort of cognitive relationships we bear to the things we see. In so doing, he turns to the issue of the knowledge of things, or apprehension, the cognitive relationship by which I am able to pick out a thing in the world and speak of it. In order to complete his story about the envelope, Moore must make a distinction between direct and indirect apprehension of things, which allows him to explain how we speak of those things which are not immediately picked out by their presence in our awareness. In considering similar problems in the similarly titled The Problems of Philosophy, Bertrand Russell develops a distinction that is in many ways similar—the famous distinction between knowledge by acquaintance and knowledge by description. Although Russell and Moore’s distinctions address a common problem and seem in almost every respect to be similar, a tension lurks beneath the surface between the two accounts. The tension arises from the different standpoints from which Moore and Russell attack the problem of knowledge of things, and it is cast in the starkest relief by consideration of their diverging views on the problem of memory and our awareness of the past. Careful consideration will show that, although neither Moore nor Russell’s approach to our awareness of objects is likely to be fruitful, the considerations that such comparisons raise highlight a crucial issue for any doctrine of apprehension, and Russell’s account will prove superior to Moore’s in expressing what it is to speak of a thing as it was. (I.1)

II. Moore: Direct and Indirect Apprehension

Moore’s distinction between direct and indirect apprehension arises in his effort to complete the story about the envelope. Both Moore and Russell have thus far agreed that when I beheld the envelope, a part of what happened to me was that I saw certain sense-data—a particular whitish patch of colour, of a certain size and shape (46). In the course of developing this part of the story, Moore is greatly worried by the apparently mutually exclusive properties borne by the sense-data observed by each of us. From different perspectives, it seems as though the envelope appears with many different patterns of color, perhaps even with different sizes and different shapes (from different distances and different angles). It seems, then, that the sense-data impressed upon me cannot be identical with the one envelope that we all see. Nor can they be identical with the other, mutually exclusive sense-data seen by the others around me. Moore—in the midst of a great deal of caviling—lays down three principles, described as the accepted view (because of their supposed widespread acceptance by philosophers) which he proposes as the most convincing account of how all of this is: (II.1)

  1. Esse is percipi: absolutely no part of the sense-data, which I ever apprehend, exists at all except at the moment I am apprehending it. (M. 40) (II.2.1)
  2. Privacy: no two of us ever apprehend exactly the same sense-datum. They would allow that we might, perhaps, apprehend sense-data exactly alike; but they would say that even though exactly alike—the same in quality—they cannot ever be numerically the same. (M. 41) (II.2.2)
  3. Dislocation: none of the sense-data apprehended by any one person can ever be situated either in the same place with, or at any distance in any direction from, those apprehended by any other person. (M. 42) (II.2.3)

Now, however, we are in something of a pickle. At the beginning of the exercise we had wanted to say that we all saw the same envelope. But if all we are aware of is the sense-data that present themselves to us, then there is nothing common to all of us that we can bring before our minds and say that we saw the same thing. The sense-data before our minds, on the accepted view, are different from one another just because they are before our minds, rather than a single mind. And thus, having for the moment accepted the philosophical view that all the sense-data seen by any one of us are seen by that person alone, we are committed to arguing that if we do in fact all see the same envelope, this seeing of the envelope cannot possibly consist merely in our seeing of those sense-data (M. 46). (II.3)

In order to resolve the tension between philosophical reflection and common-sense intuitions, we must now distinguish (at least) two different ways for a thing to be before the mind. With the accepted view of sense-data in hand, one way of knowing things will be the apprehension of sense-data that has already been explored. This relationship Moore dubs direct apprehension, and characterizes in terms of apprehension of sense-data under the accepted view, i.e., that which happens when you actually see any colour, when you actually hear any sound, when you actually feel the so-called sensation of heat … etc., etc. (M. 46). Moore later also allows that we have direct apprehension of propositions, but his basic characterization of direct apprehension is carefully tailored to accompany sense-data under the accepted view. In any case it will seem as though anything which counts as the object of a direct apprehension, except for propositions, will be something sense-datummy and subject to the conditions of the accepted view. (II.4)

However, our ability to bring one and the same envelope before all our minds and say, We all see the same envelope, there must be another way of knowing things, another mode in which something can be brought before the mind. This relationship Moore characterizes as the relation which you have to a thing when you do directly apprehend some proposition about it, but do not directly apprehend the thing itself (74), and he gives it the name indirect apprehension. Whereas both forms of apprehension places us in some cognitive relation (M. 78) which enables us to know truths about the thing that we apprehend, direct apprehension is the picking out of data that are directly confronting the mind in their experiential richness, whereas indirect apprehension is a much barer relation, only obtaining in virtue of having the ability to make judgments about the thing indirectly apprehended. Direct apprehension will be phenomenologically basic, whereas indirect apprehension will depend upon bringing various propositions before the mind. (II.5)

With this distinction in hand, the story about the envelope can now be completed. We all see the same envelope before us—but see here is being used in a sense other than the direct visual sensations that we have of its color, shape, position, and so on—for the objects of these sensations are, on the accepted view, essentially private. Rather, our seeing of the same material envelope consists, partly in directly apprehending certain sense-data, but in addition to this, being endowed with a cognitive relationship that can allow direct apprehension of a proposition connecting those sense-data to a material object, that is, knowing, besides and at the same time, that there exists something other than these sense-data (M. 51). Picking out the material object for all of us to talk about requires our perceptual faculty to convey the double existence that Hume so vigorously denied, presenting to us both a private sense-datum (directly apprehended) and also the ability to speak of a something quite other than these sense-data (M. 51). (II.6)

III. Russell: Knowledge by Acquaintance and by Description

Russell’s parallel distinction between knowledge by acquaintance and knowledge by description also takes its cue from Russell’s efforts to work out a story about the perception of office supplies. Like Moore, he is confronted with the question of how to string together many people’s radically private sense-data and connect them to the common envelope (or table) that is being observed. However, Russell leaves much of the ontological story about sense-data open, and he does not commit himself to the accepted view of sense-data in its entirety—in particular, he does not seem to take any particular stance on the esse-is-percipi thesis (1). The worry driving Russell in making his distinction is primarily a concern about how the sentences regarding a material envelope can be made meaningful. (III.1)

Russell frames this worry in the familiar terms of knowing what you’re talking about: it is scarcely conceivable that we can make a judgement or entertain a supposition without knowing what it is that we are judging or supposing about (R. 58). Our ability to talk about things (such as the material envelope) which are not brought before our minds in an unmediated confrontation requires us to have some cognitive relationship with the thing of which we intend to speak. But how can we account for the ability to bring things before the mind beyond the range of that which we primitively know? (III.2)

In order to get out of this pickle, Russell develops his distinction between knowledge by acquaintance and knowledge by description—a distinction which will, initially at least, track many of the same things as Moore’s distinction between direct and indirect apprehension. We have acquaintance, Russell writes, with anything of which we are directly aware, without the intermediary of any process of inference or knowledge of truths (R. 46). In acquaintance the thing that we know presents itself, as it were, nakedly before the mind, placing us in a cognitive relationship to, for example, the color of the envelope such that I know the colour perfectly and completely when I see it, and no further knowledge of it itself is even theoretically possible (R. 47). Acquaintance, like direct apprehension, is an essentially simple relationship between the knower and the thing known such that the thing is indubitably identified and available for judgments, without any need to refer outside of the experience of acquaintance itself. (III.3)

So far, so good. But Russell has already examined the phenomenology of seeing an object before us, and he has confessed that he failed to find in it any acquaintance with material objects such as the envelope. If he is to speak of the one envelope that we all say we see, then he will need to distinguish a second way of knowing things, which does not require this primitive acquaintance with the thing known. Russell finds this in the notion of knowledge by description. Anything which does not disclose itself to the mind in its full richness, as objects of acquaintance do, must be known to us indirectly, through a description. On Russell’s account, like Moore’s, our ability to speak of things to which we do not have a direct cognitive relation requires us to pick them out in virtue of our ability to make statements about them. For Russell, this ability is explained in virtue of picking out the object by a description phrased in terms of other things. Thus, (III.4)

There is no state of mind in which we are directly aware of the table [qua material object]; all our knowledge of the table is really knowledge of truths, and the actual thing which is the table is not, strictly speaking, known to us at all. We know a description, and we know that there is just one object to which this description applies, though the object is not directly known to us. (R. 47-48) (III.5)

The table itself does not show up on our cognitive charts, but we can use our knowledge of other things to, as it were, triangulate to the table itself and speak of it, even though it is not directly before the mind. (III.6)

If we are, then, to pick out objects outside of the range of acquaintance by speaking of them in terms of other objects which we have cognitively picked out, it soon becomes clear that all of the terms we use in the description must ultimately be reduced to the names of things with which we are acquainted. On Russell’s view, only objects of acquaintance present themselves as already picked out; therefore, only objects of acquaintance can provide a foundation from which to pick out objects beyond the range of immediate experience. He must introduce the principle of acquaintance as the basic requirement for any meaningful cognitive relationship involving description. Thus Russell: The fundamental principle in the analysis of propositions concerning descriptions is this: Every proposition which we can understand must be composed wholly of constituents with which we are acquainted (R. 58). (III.7)

With this distinction in hand, Russell can now survey the field of human experience to see what sort of things might count as objects of acquaintance, and what sorts of things will have to be triangulated by a description. Sense-data have already been taken as paradigmatic examples of objects of acquaintance, but this is not enough for Russell. In order to connect objects of immediate acquaintance with things outside of our acquaintance, we clearly cannot be acquainted only with self-contained particulars. Otherwise there could be no descriptions that identify anything other than a laundry list of objects of acquaintance. Russell, therefore, adds acquaintance with universals (according to what he describes as a roughly Platonic theory of their nature and self-disclosure) alongside our acquaintance with sense-data. With these two objects of acquaintance in hand, along with a rich set of logical operators, he can now begin to form descriptions that pick out objects outside of the range of acquaintance, as when he speaks of a material envelope or table as the physical object which causes such-and-such sense-data. Such a description can now be broken down into an existential quantification (logical operator) for a thing which is (1) a physical object (universal), and (2) causes [universal] such-and-such sense-data [various particulars]. By combining universals and particulars with which we are acquainted, we can reason outward beyond the range of immediate acquaintance and thus pass beyond the limits of our private experience (59). As long as a thing to be spoken of can be completely described in terms of other things which are directly before the mind, we can say that (in a certain sense) we know it, and we will be able to speak of it even though our native faculties do not allow us to have it disclose itself directly to us. (III.8)

Given the outline of Russell’s distinction, we can see that there are very close parallels with the distinction made by Moore between direct and indirect apprehension—with acquaintance being roughly equivalent to direct apprehension and knowledge by description being roughly equivalent to indirect apprehension. However, Moore and Russell have approached their doctrines from different directions. Moore developed his distinction between direct and indirect apprehension in order to complete the story about the envelope in light of a doctrine about the ontological status of the objects of our direct experience. On his view, direct apprehension is a phenomenon tightly fitted to sense-data, and indirect apprehension will have to cover nearly everything else. Russell, on the other hand, develops his distinction in light of a principle about the meaningfulness of sentences, and his distinction has far fewer ties to a particular view on sense-data. Because of this, Russell can—and, ultimately, must—add much more to the roster of objects of acquaintance than simply sense-data. Acquaintance less narrowly constructed than direct apprehension, and and it will turn out that this difference in approach leads Russell and Moore’s apparently congenial distinctions come into conflict. (III.9)

IV. The Remembrance of Things Past

The most dramatic separation between Moore’s and Russell’s accounts can be seen when we turn to their accounts of memory. Returning to the envelope, Moore observes: (IV.1)

I look at the envelope again and I see the whitish colour. I turn my head away, and I no longer see it. But I remember that I did see it a moment ago. I know that I did see it. There is nothing that I know more certainly than this. Moreover I know that that whitish colour was: that there was such a thing in the Universe. (M. 49) (IV.2)

What is the nature of the presentation of the envelope in memory? One potential confusion must be mentioned in order to dismiss it immediately. Both Russell and Moore note that memory of an object is apt to be accompanied by an image of the object, and yet the image cannot be what constitutes memory (R. 114-115). This is shown clearly enough by the observations that (1) the image is in the present, whereas what is remembered is known to be in the past (R. 115), and (2) that if the image itself were memory, then I could not possibly know that the image which I now see was at all different from the colour which I saw a moment ago (M. 50), but in fact we can and do evaluate the images that accompany our memories and speak of whether or not they justly portray the thing remembered. (IV.3)

Remembering an object, then, is having that object, safe at home in the past, before my mind now, presented to me by the faculty of memory. But what is the nature of this presentation? Is it knowledge by acquaintance or knowledge by description? Is it direct apprehension or indirect apprehension? (IV.4)

Moore argues that memory is one way of having before the mind, which is not direct apprehension (M. 47)—indeed, in this passage he introduces it as a paradigm case of indirect apprehension. When you remember an object, Moore claims, you are no longer directly apprehending the coloured patch which you saw (M. 47). Moore draws an analogy between the relationship between present sense-data and the material object with which they are connected, on the one hand, and the relationship between the present image and the sense-data remembered, on the other. Just as direct apprehension of present sense-data somehow also elicits indirect apprehension of the material object, direct apprehension of the mental image also elicits indirect apprehension of the remembered object. When you turn your eyes away from the envelope and remember what you saw, there is—so to speak—a leftward shift in the ledger: (IV.5)

(IV.6)ImageSense-dataMaterial envelope
ThenN/ADirect app. (Sensation)Indirect app.
NowDirect app. (Imaging)Indirect app. (Memory)Indirect app.

Russell, on the other hand, introduces memory as a paradigm case of the extension [of knowledge by acquaintance] beyond sense-data (R. 48). On the account of memory he develops, the essence of memory is not constituted by the image, but by having immediately before the mind an object which is recognized as past (R. 115, emphasis added). Here the image is, at most, a psychological illustration that accompanies the essential acquaintance with past objects, and our relationship with these objects is no more strained, just as immediate and cognitively simple, as it was when we had seen them face to face. In light of further deliberations about problems of fallacious memory, Russell will eventually recognize that some memories (or memories so-called, at least) are examples of knowledge by description, but here as always they are rooted in the primary knowledge of things, which is the acquaintance—here the acquaintance conveyed in (primary) memories. (IV.7)

Moore and Russell are each driven to the conclusion that they adopt by the prior commitments that structured their respective distinctions. For Moore, the driving force is his ontological theses about the objects of direct apprehension. For Russell, it is his semantic and epistemic commitments in the analysis of knowledge by description. (IV.8)

We reviewed earlier the three theses which Moore presents as the accepted view about sense-data, and noticed how Moore’s notion of direct apprehension is tailored to fit sense-data under the accepted view. It should come as no surprise, then, that the accepted view theses weigh heavily in Moore’s efforts to place memory in terms of the various ways of knowing a thing. The first of the three theses, the esse-is-percipi principle, holds a decisive weight here. Remember that on this principle, a sense-datum exists as long, and only as long as it is the object of direct apprehension. (IV.9)

As of yet, the thesis does not say anything about memory: whether I directly or indirectly apprehend something in memory, I apprehend what was then, not anything that is now. Nevertheless, when (1) is conjoined with the other two theses (privacy and dislocation), the doctrine that emerges does bear heavily on memory. Sense-data exist only as long as they are apprehended because they essentially dependent upon the act of apprehension. The relationship between the object of direct apprehension and the act of direct apprehension, in fact, is so tight that wherever there are two distinct acts of direct apprehension, there must be two numerically distinct sense-data to accompany them. And now, the pressure of the accepted view is brought to bear on memory. If, after turning away from the envelope, I directly apprehended the envelope’s sense-data in memory, then this would be a second, different act of direct apprehension. As Moore writes, the quality of the relationship changes such that (IV.10)

the relation which you now have to the image is obviously different from that which you have now to the sense-datum, which you saw but do not now see; while this relation which you now have with the image, is the same as that which you had to the sense-datum, just now when you actually saw it. (M. 47) (IV.11)

And since we have here two separate acts of apprehension, if both were acts of direct apprehension, rather than indirect, then that very fact would constitute two numerically distinct acts of direct apprehension, and thus two numerically distinct objects apprehended. You could not, then be thinking of the colour which you saw, and therefore having it before your mind in a sense (M. 47). Thus it must be that you are no longer directly apprehending it (M. 47), that you pick it out by entertaining propositions about it, but you do not have a direct apprehension of the thing that you remember. (IV.12)

For Russell, on the other hand, the looming issue in understanding memory is the principle of acquaintance. On his account, in order for me to have a thing before my mind, it must be picked out entirely by things with which I am acquainted. And so it is with judgments about the past just as much as judgments about the present. But here a worry arises: if the only vocabulary with which I can pick out a past thing is what is contained in the directory of objects of my acquaintance, then in order to describe anything before the present moment, I must have some acquaintance which allows me to speak of what has gone before, which is to say, some acquaintance with something in the past. Russell mentions his reasoning briefly twice: (IV.13)

This immediate knowledge by memory is the source of all our knowledge concerning the past: without it, there could be no knowledge of the past by inference, since we should never know that there was anything past to be inferred. (R. 49) (IV.14)

Thus the essence of memory is not constituted by the image, but by having immediately before the mind an object which is recognized as past. But for the fact of memory in this sense, we should not know that there ever was a past at all, nor should we be able to understand the word past, any more than a man born blind can understand the word light. (R. 115) (IV.15)

We can understand the arguments as a sort of elliptical transcendental argument, and reconstruct Russell’s reasoning more or less as follows: (IV.16)

  1. Knowledge of past things is possible. (IV.17.1)
  2. If knowledge of past things is possible, then either it is all knowledge by description, or else some memories convey knowledge by acquaintance. (IV.17.2)
  3. Either all knowledge of past things is knowledge by description, or else some memories convey knowledge by acquaintance. (IV.17.3)
  4. If all knowledge of past things is knowledge by description, then past things must be picked out entirely by from present or timeless things with which we are acquainted. (IV.17.4)
  5. Past things cannot be picked out entirely by present or timeless things. (IV.17.5)
  6. If all knowledge of past things is knowledge by description, then knowledge of past things is not possible. (IV.17.6)
  7. Not all knowledge of past things is knowledge by description. (IV.17.7)
  8. Therefore, some memories convey knowledge by acquaintance. (IV.17.8)

The crucial step in establishing that memory must convey at least some acquaintance is the introduction of the principle of acquaintance at step 4, and then the introduction of a principle at step 5 which we may call the flatness of the present principle. Roughly speaking, on Russell’s account, my acquaintance with present things (sense-data, acts of consciousness) and timeless things (universals) leaves me with a flat perspective on time. There are no objects of acquaintance in either of these categories which give me cognitive license to speak of a duration of time beyond the present moment or the view sub specie aeternitatis. Unless I have acquaintance with past things as past, I have no way of spreading my judgments outward in time from the present moment. Flashbacks in the cinema of my awareness would be indistinguishable from the ordinary forward motion of the plot. (IV.17.9)

V. Different Strokes

It seems, then, that in spite of the initial similarities between Russell’s and Moore’s distinctions, there is a deep conflict between them. If we have knowledge of past things, then Moore’s accepted view with regard to objects of direct apprehension and Russell’s principle of acquaintance come into conflict with one another. Moore’s view requires an essentially ephemeral character for direct apprehension. Russell’s requires acquaintance with past things in the present. What, then, are we to do? (V.1)

Such questions may themselves be questionable. There are, after all, very good reasons to reject both Moore’s account of sense-data and Russell’s principle of acquaintance. Nevertheless, even if we reject the motivations for Russell and Moore’s solutions on memory, we still must come to some kind of peace with how it is we speak of things that we experienced in the past. How do we get a cognitive grip on such objects? Will a Moorean indirect apprehension do, or do we need Russellian knowledge by acquaintence to account for our knowledge of how things were? (V.2)

Moore, if he were prone to saying such things, might tell the Russellian that she has the phenomenology all wrong. When, immediately after directly apprehending some sense-datum, Moore writes, you remember that sense-datum, or remember that you did just now directly apprehend it, there is nothing more obvious than that you now stand in a different relationship to the object than you did when you were looking at it. It seems that you are, ex hypothesi, no longer directly apprehending the snese-datum in question (M. 74). Otherwise, you would still be seeing it, rather than remembering it. (V.3)

Such a response, however, will only work by begging the question. On Moore’s account, it is true that the only way you could still have direct apprehension of the sense-datum is by it continuing to present itself visually to you. A Russellian account, however, can just as easily account for the phenomenological change. Memory, for Russell, simply is a different faculty of acquaintance from the sense of sight. In this way, one is acquainted with the past sense-datum, and why should the phenomenal character of this relationship be like the phenomenal character of being acquainted with a present sense-datum? If Moore assumes that the change from sight to memory requires a change in the underlying cognitive relationship, then he has in fact merely ignored one of his own principles about direct apprehension: with each of the different sensory faculties, Moore argues, (V.4)

what I mean by direct apprehension, namely, the act of consciousness, is exactly the same in quality: that is to say, the actual seeing of a colour, considered as an act of consciousness, differs in no respect at all from the actual hearing of a sound, or the actual smelling of a smell. They differ only in respect of the fact, that whereas the one is the direct apprehension of one kind of sense-datum, the other is the direct apprehension of another kind. (M. 47) (V.5)

And so, with memory, the phenomenal difference can be understood not in terms of the difference between direct apprehension and indirect apprehension, but rather in terms of the difference between a past object of acquaintance and a present one. (V.6)

Russell, for his part, can renew his transcendental argument on behalf of acquaintance with the past, even without the demands of the principle of acquaintance. If the present is inferentially flat in the way that Russell argues that it is, then quite apart from Russell’s semantic worries, there will simply be no way to get an epistemic grip on the past unless I am able to stand towards past objects of awareness independently of the present objects around me. Without an independent epistemic grasp on the past, as Russell writes, we will not be able to infer anything about the past, because we will have no basis on which to suppose that there are past things to be inferred. Russell’s division of our epistemic and semantic lives into acquaintance and description is certainly not to be accepted uncritically, but examination of his distinction and comparison with Moore’s highlights the crucial need for a theory of apprehension which respects the timeliness of things in the world, and raises concerns for anyone explicating our relationship to the things of which we speak and think. (V.7)