The worth of A. D. Woozley's well-known book, Theory of Knowledge, is seriously compromised by an error which runs through the entire work. The error is of such a fundamental nature, and of such import for epistemology, that his very statement of the nature of epistemology presupposes it.

Mr. Woozley says that the main question which epistemology is designed to answer is the question, "What is present to my mind when I think? " (p.11). Now "a man is thinking", he says, "whenever he is conscious of anything..., in short, whenever his mind is not a blank.... Therefore the original question, "What is present to my mind when I think?" can now be seen to be asking what are the objects of consciousness, using consciousness in its widest sense " (p. 13). He goes on to say, "The Theory of Knowledge, then, is that branch of philosophy which has for its study the nature of cognition and its objects" (p.14). Although this last quotation suggests that Woozley is distinguishing between the nature of cognition (or consciousness), on the one hand, and the objects of cognition, on the other hand, he in fact continues throughout the book to say and imply that there is no difference, at least for the purposes of epistemology, between discussing the nature of cognition and discussing its objects. For example, he says on page 38, at the opening of his two chapters on memory, "What, then, does remembering consist in? More specifically, what is the immediate object of an act of remembering?..." Thus, to ask what the immediate object of memory is is passed off as simply a "more specific " way of asking what memory is. Then, too, in the last half of the first chapter of his book there is a discussion of the question, "What are we aware of in sense perception? " Chapter V is wholly devoted to the question, "What are we aware of in knowing and believing? (p. 102); and the remaining part of the book is given to discussions of truth, and of the knowledge/belief distinction, in terms of "Propositions", which, allegedly, we are aware of in knowing and believing. So it is not unjust to say that Woozley holds that what is important for epistemology about the nature of cognition is amply covered by talking about what the cognition is cognition of, i.e. by talking about the "objects" of cognition. He therefore takes the description of these "objects" as the epistemological problem. <514>

Within philosophical circles what Woozley says here may be regarded as commonplace and not to be questioned. After all, the entire "new way of ideas" was a discussion, alleged to be about the nature and limitations of "cognition", but carried out almost entirely in terms of cognition's "objects". But that Woozley's manner of describing the epistemologist's task—whether commonplace or not—is unfortunate, and in what way it is so, may be seen by considering, that the question of epistemology, as Woozley has stated it, can be taken as a serious question only if there is no obviously true answer to it, and consequently only if what many persons take to be the obviously true answer to it is in fact not obviously true. And many persons, including some epistemologists, have thought that the obviously true answer to Woozley's question is that the objects of cognition are such things as trees, numbers, events of yesterday and today and tomorrow, colours, and persons. But the epistemological question as Woozley sees it presupposes that this answer, though it might prove out (it does not for him), is not obvious. Why is Woozley willing to accept this presupposition?

The old explanation of why people have doubts about what I shall call the realist answer to Woozley's question is that they are disturbed by such things as dreams, perceptual errors and false beliefs. This, while true, is not very helpful by itself, since it does not lay bare the conviction which makes error and falsity disturbing in that peculiar way that casts doubt upon the realist's answer: the conviction, namely, that a relation obtains between a cognition and what it is a cognition of. This mistaken conviction is an all-pervasive feature of Woozley's book. And, as I have said, it shows up right in his formulation of the epistemological question. For did he not have this conviction, he would not be bothered by error and falsity into wondering whether or not we cognize trees, numbers, events, and so on; and he consequently would not regard the question, "What are the objects of cognition?", as the open and urgent question which he plainly takes it to be. But how does the conviction that a relation obtains between cognition and what it is of have the effect indicated?

Error and falsity are disturbing to one who holds that a relation obtains between cognition or consciousness and what it is of, because they threaten to push him into admitting instantiation of a two-term relation by one term alone, which would be like admitting that someone is a father although there is no one of whom he is the father. For example, the relation theory of perceiving gets along fairly well as long as perception remains veridical. But if, stumbling across the desert, blind with thirst <515> and deranged by heat, I suddenly see an oasis, when in fact there is no oasis, how is my seeing to be understood, given that it is a relation ? We cannot admit, on this view, that there is nothing which I see. And if I see something, and if it cannot be an oasis, since there is no oasis, what can it be? Thus we get what Woozley calls the epistemological question. Given the relational analysis of seeing, we have a puzzle on our hands.

Similarly with belief. Given that it is a two-term relation, we can say that the terms are my consciousness and a fact, so long as my beliefs are true beliefs about the present state of the world. But when my belief is false, or about the past or future, there seems to be no fact which can serve as the other term of the alleged belief relation. Columbus is not now discovering America, so how can I now, in my belief (or, for that matter, in my doubt or supposition) that he discovered America, be related to him-discovering-America ? And if I am not related to that, and if I am related to something, what am I related to in that true belief? On the other hand, if my belief is false—e.g. I believe that Columbus sailed to India—that also seems to exclude what it is about from being a fact; so what is the object of a false belief? And there is Woozley's question again.

Anyone who doubts that Woozley does suppose that cognition involves a relation between subject and object need not, to settle his doubt, rely upon my interpretation of why he puts the epistemological question as he does. By doing little more than turning the pages of Theory of Knowledge one discovers how he repeatedly uses this supposition to show that certain things are not objects of cognition. This supposition is used for that purpose on page 20, with reference to perception, on page 39, with reference to memory, and on page 102, with reference to belief. It also shows up on many other pages, but it is never questioned. So I shall regard it as established that Woozley does assume, without question, the relational analysis of cognition, and that this assumption determines, not only his overall conception of epistemology, but also the details of many of his arguments. What I have to do now is give some reason for supposing that this analysis is wrong and then suggest an alternative analysis.

The first task of one who accepts the relational analysis is to round up terms to serve as "objects" for the cognizing relations. There has been very little variation during the last few centuries as to what these "objects" are to be called and how they are to be regarded. "Idea" has been, over the long run, the undoubted favourite as a name for these objects. But "impression" and <516> "image" have also come in for heavy use. Hume's generic name for object was simply "perception". Except in Hume's case, the more complex cognitions, such as belief or doubt, have been regarded as having for objects some combination of "ideas". The so-called "concept" has also had a complex and interesting career as an object. "Sense data for perceiving, images for memory and imagination, and propositions for judgment, belief, doubt, and supposition", has been the full answer of many leading thinkers of this century to the quest for terms, with Moore and Russell leading the way.

Now here is one objection to the acceptability of such objects. It has been suggested in various writers, I believe, but is seldom adequately appreciated. In fact, the objection goes, no one ever is aware of such things as on the relational analysis we are said to be aware of, and not even in the case of illusion. If in the desert I have the experience described as perceiving an oasis, what I perceive, so far as anyone can tell, is moist and is frequented by wild beasts. (If I do not perceive that, why call my perception erroneous?) Yet it is certain that not even the most ardent enthusiast of the relational analysis, of sense-data, of images, and so forth, has ever held that what he says I see is either moist, or frequented by wild beasts, or has many other of the innumerable qualities and relations which that has which my perception must be a perception of in order to be erroneous. And in all cases, what one is cognizing turns out, on the type of view in question, to be something very different from what one thought he was thinking of. "I thought I was thinking about the discovery of America", one might say, "but I find now that, whatever else I may be doing, I am really thinking about, or am directly aware of, something very different from that—a proposition." (Many ways in which my alleged object differs from the discovery of America can easily be stated on the basis of what object-epistemologists have said about propositions. Similar remarks hold for memory and imagination and their alleged "images". It is never explained how people could be so mistaken about what is the object of their cognition.) It is not too hard to see, therefore, how one may get the impression that he has never met any of these "objects". But then the relational analysis must be false.

What always seems to go completely unrealized is that, even given the existence of such "objects", substantially the same problem as they were introduced to solve yet stands, with only a slight modification. I will illustrate what I mean by reference to Woozley"s views. Woozley clearly does reject the view that the <517> object of belief is a fact. He then considers the view that belief's object is a thing called a proposition, which is somehow other than the belief, but may be related to it. This type of proposition he calls a " substantial proposition" (p. 116). He rejects this view" because it requires us to accept entities of very doubtful character . . ." (p.121). Now I want to state another reason for rejecting this view, and any relational analysis of thought which might depend upon it, because I think some important philosophers have held it, and because I suspect that the view which Woozley does finally accept is not different enough from this one to prevent what I shall say about it from applying to his view as well. The reason is this: If the aboutness of a false belief calls for an object, then no less does the aboutness of a false proposition. It is just as absurd to speak of a proposition being about nothing as it is to speak of a belief or judgment being about nothing. So if a false belief needs some object other than a fact to be about, so does a false proposition. Hence, one of the main problems which propositions have been introduced to solve arises about the proposition itself; and that is no solution at all. The cognizing "relation" might just as well hang loose from the false belief as from the false proposition.

The same thing holds if we talk about "images" instead of propositions. An image allegedly provides, sometimes, an object of cognition—in hallucination, memory, etc. Now an image is always an image of something. But suppose there is nothing of which it is the image. How then can it be an image at all, if it is the image of nothing? So the puzzle about the "object" of a dream, hallucination, or memory cannot be resolved successfully by introducing images. It will make just as good sense to speak of perceptions, memories, or dreams of nothing, as it will to speak of viewing an image which is an image of nothing.

A strongly analogous objection is to be made against the theory of belief or judgment which Woozley regards as correct. He does hold that there are propositions, and that there has to be such because of the nature of judgment (p. 120). He holds this because he equates the meaning of sentences with propositions, and holds that thinking or judging essentially involves sentences. He finds it "very hard to see how an animal having no power of speech could, for instance, wish that it could be moved from this field to the next in four days' time" (p. 120). That raises a tangle which need not concern us here. What is important is that Woozley holds there to be propositions and thinks that the serious objections against them will be obviated "if we can <518> produce an account by which propositions are mind-dependent although the elements of which they are composed are independent of mind. . "(p. 120). Now these propositions are to be present to the mind when we think or judge " (p.119), i.e. they are objects of belief; and, as he later says, "the unity of the objects of judgment is produced by the act or relation of judging itself" (p. 121). Hence, "a proposition is now constructed by the operation of judgment", and all that is required beyond the relating action of judging is "that the materials should previously exist, to be combined in whatever way is required " (p.121). He considers this view to be "substantially correct" (p.125).

So Woozley's view of judgment is that thinking that Columbus discovered America, for example, consists in bringing Columbus and America and discovering and the person doing the thinking together in that particular way in which a relation unites its terms. This particular relation must no doubt be regarded as of a unique kind. "Bringing together" certainly is only a spatial metaphor, which may be misleading. "Knitting", which both Bertrand Russell, the originator of this theory, and Woozley use (p. 124), is also a metaphor, one based upon uniting things by interlacing them. The relation of which Woozley speaks is certainly not reducible to relations of space, cause, etc. In a wholly unique fashion it ties things into what we might call a "mental" fact, e.g. the fact of Columbus being supposed by me to have discovered America at a certain time. In Russell's words, "What is called belief or judgment is nothing but this relation of believing or judging, which relates a mind to several things other than itself" (Problems of Philosophy, p. 126).

The only trouble is, the very thing which is an objection against the view that a fact is the object of belief is also an objection against this view of belief as a "multiple relation". Russell remarks, "Othello believes that Desdemona loves Cassio. We cannot say that this belief consists in a relation to a single object, 'Desdemona's love for Cassio', for if there were such an object, the belief would be true. There is in fact no such object, and therefore Othello cannot have any relation to such an object. Hence his belief cannot possibly consist in a relation to this object "(ibid. p.124). Here the line of reasoning is crystal clear and very familiar. Belief is a relation to something, and what does not exist does not stand in relations. But the same line of reasoning shows the multiple relation theory to be false. Columbus does not exist, and because he does not he cannot be related into a "proposition" of the type Woozley suggests.

I am aware that Russell has an escape clause. He held that <519> judgments such as mine about Columbus were not "really" about Columbus. They are general judgments and are (his attempted way out continues) about certain "propositional functions" or, more familiarly put, about certain qualities. All I can do with that move here is to suggest that it begs the very question here at issue. The theory of descriptions, the germ of which I have just stated, was promoted by the conviction that a proper name, if meaningful, must be related to that of which it is a name, and so will not be meaningful if there exists nothing to which the name has the required relation. Since "Columbus" is still meaningful, although Columbus does not exist, it is not a proper name, but a general name, the meaningfulness of which rests upon our acquaintance with qualities. But this only begs the question, at issue here, of whether or not thought, in any of its forms, is related to what it is of or about. For the analysis of names can proceed only by analysis of their use and understanding, which essentially involves thought. Russell simply presupposed a relational analysis of the understanding of proper names and of the sentences in which they occur.

What I have tried to show, mainly by talking about Woozley's book, is that the thesis that cognition is a relation is false. The reason for saying that it is false is that it implies that we are aware of certain things although we are not. Moreover, the existence of such things would present us with the same type of puzzle which they were supposed to rid us of, namely, two-term relations appearing to be instanced by one term alone. I submit that these are adequate reasons for rejecting what I call the relational analysis of cognition. But this leaves unexplained why this analysis is so attractive; and it certainly has been attractive and, consequently, influential. One will find it very instructive to read the significant epistemelogical writings, from Plato's Theaetetus on, with this analysis of cognition in mind, to see how far it influences the questions raised and the conclusions reached. So the issue here is much broader than whether or not Mr. Woozley is right.

There are at least two things which conspire to make the relational analysis so attractive. First, it is a plain and widely recognized fact—and one for which I here offer no analysis-that one cannot contemplate particular "cognitions" without having certain other things called to mind. Philosophers and psychologists have often remarked how one cannot reflect on his cognition of x without altering that cognition. The other side of the coin is that reflection on, or cognition of, the cognition of x involves <520> some sort of cognition of x. Thus, as I reflect upon and try to understand my own experience, all sorts of things other than my own thoughts are inescapably brought to mind—table, pen, a child's voice, etc. etc. (A few philosophers have attempted to hold a certain symmetry here, maintaining that cognition of x, in its turn, requires some sort of reflection on, or cognition of, the cognition of x. For example, Locke, Sartre, Hegel, Spinoza, Samuel Alexander.)

Now in order to show why this feature of cognition has led to it being treated as a relation, it is necessary to point out that, traditionally, two things have been said about "relations", one of which, because of the facts just considered, may also be said of cognition. These two things have been said, for example, by Aristotle (Categories, chap. 7), by Mill (System of Logic, chap. 2, sec. 7) and by Stebbing (Modern Introduction to Logic, chap. 10, sec. 2). First, it has been said that if you would understand the nature of a relation, or explain its nature, you must take into consideration, or refer to, something other than the things which have that relation, i.e. other than its referents. On this point there is a great similarity between cognitions and relations such as to the left of or father of. To understand a cognition, or to explain it, something other than a person cognizing must be considered.

But a second thing has, traditionally, been said about relations. In Mill's graphic language, "every other attribute of an object might, without any contradiction, be conceived still to exist if all objects besides that one were annihilated; but those of its attributes which are expressed by relative names would on that supposition be swept away". And Aristotle says that "...those things only are properly called relative in the case of which relation to an external object is a necessary condition to existence...". In short, in order for a thing to have a relation, such as to the left of, there must exist something other than itself; and the non-existence of anything else would be sufficient grounds for saying that that thing does not have the relation.

And now our question is this: Since the understanding of a "cognition" always requires that something other than the thought and thinker be considered, just as the understanding of a "relation" requires consideration of something other than relation and referent, should we not carry the parallel with relations on through, and refuse to allow that our thought, at least in error and falsity, was really about that thing which came to mind in reflection upon it, unless that things exists? Yet, if we do so, we must either hold that what came to mind was somehow a mistake, and that really our thought was not about it, but <521> about something else, or we must hold that the thought is not about anything at all. I have never heard of a case being made for the second alternative, and we have seen where the first one leads. Better to simply refuse to continue the parallel with relations, and to frankly grant that while thinking of x resembles being to the left of x in the one respect mentioned, it does not do so in the other.

Aristotle saw the dissimilarity of cognition and other things which he regarded as relations. He was bothered by the existence of objects of knowledge before and after being known (7b, 23-34). This seemed to him to constitute an exception to the claim that correlatives must exist simultaneously. But rather than question the advisability of describing cognition as a relation, he chose to water down that claim to something which is "for the most part true" (7b, 15).

The trouble with some words—of which "relation" is an outstanding case—is that they cover things which, for some common and important purposes, may be treated alike, but which differ in respects important in other contexts. If one has a word which, because it is like that, cannot be used in some contexts without obscuring the distinctions which in those contexts are crucial, and thus without leading people down blind alleys such as the chase after "objects", then the only help for the situation is to get a verbal distinction which reflects the real differences, and which holds them apart in reflection and discourse.

Now the point here is that it is misleading to say that cognizing is a relation between a mind and its "object" or what it is thinking about. This is because of the wide presumption that the instancing of a relation requires the existence of all of the relation's "terms". It may not be best to use the word "relation" in that way. Nonetheless, some things commonly called relations (e.g. being to the left of) do require for their instancing the existence of all of their terms; and there is a strong presumption in the minds of most persons, when you say that such and such a "relation" obtains, that you are committed to the view that all of the things, to which one must refer in explaining that relation, exist. Only the existence, not the worth, of that presumption is in question here. It exists; and because we often do think of what does not exist, we should not call any of the specifications of cognizing or being conscious of a "relation". Rather, use the word "reference"; and say that in believing, doubting, imagining, perceiving, remembering, and the like, there is reference to things of many kinds, which may or may not exist. <522>

The second thing which renders the relational analysis of cognition so attractive is the very widespread conviction that in cognition there is interaction of mind, or thought, and its object. Philosophers have tended to divide over the question of which acts and which is acted upon. For example, sensationism (Hume, Mill) and idealism (Berkeley, Fichte) divide sharply on this issue. All of the action is in the object for the former, but it is ultimately all in thought for the latter. However, these are the extreme cases. Locke and Kant well represent a large group of philosophers who have put the main action in the object with one type of cognition—roughly speaking, sense perception, and the main action in the mind in other types of cognition—judgment and reasoning. But on one thing all of these philosophers agree, in so far as they admit a subject/object distinction at all: There is some type of interaction between thought and its object; and so cognition must resemble such relations as to the left of, at least to the extent that, just as there must be two things in order for one thing to be to the left of another, so there must be two things in order for one to cognize another. Interaction seems to presuppose something acting and something acted upon, and so the relational analysis of cognition once again appears unavoidable. But that involves us in the futile quest already discussed, and hence suggests that we should examine the view that seeing, believing, etc., are cases of interaction, in order to determine if we must accept it.

The view that thought and its object interact seems to rest upon the conviction that their interaction alone can guarantee their correlation. We need not debate whether this is so. Possibly some form of interaction does bring about that correlation between thought and the world which is present with veridical perception and true judgment. Yet all of that may be admitted without implying that interaction of thought and its object is the very thing which is asserted whenever it is said, for example, that I see this book. Hence, this admission does not force us into holding that seeing is interaction of the mind with what is seen. Leibniz was one philosopher who made it a cardinal point in his system that seeing this book was a very different thing from the book and my mind interacting (Discourse on Metaphysics, XXIX). Never denying that persons saw books, he did deny that minds and books interact, which would simply be absurd if seeing were interaction. But it is not absurd, though it may be false. Since it is not absurd, it is safe to say that cognizing is not interaction. That it is interaction cannot, therefore, be regarded as a good reason for regarding it as a relation. <523>

Over fifty years ago the American New Realists proclaimed the emancipation of metaphysics from epistemology". This at least meant a refusal to accept the understanding of knowledge as a doorway—much less as the doorway—to the understanding of the nature of "things". What both they and many other philosophers have failed to sufficiently appreciate, however, is the dependence of epistemology upon metaphysics. After all, cognizance is just as cold and hard a fact, is just as much of a reality, as is the growth of a tree, the chemical composition of water, or the motion of a planet. What is ordinarily expressed by a sentence such as "I see this book" or "I know that Columbus discovered America" admits of and requires categorial analysis—the exploration of its factors to determine the summum genus of each, as well as the higher species which fall under that genus—just as much as does what is ordinarily expressed by "I am near my desk" or "Mars and Saturn are now in opposition". Such analysis is the traditional business of metaphysics. To follow an old usage, metaphysics is "criticism of categories" and one of the things which it sometimes renders possible, by its elucidations of the ultimate distinctions between factors of the real, is the prevention and cure of "category mistakes". A category mistake consists in inferring, from one resemblance of distinguishable factors of the real, that those factors possess other (very abstract) similarities which they do not in fact possess. 

The "crucial error" to which my title refers consists in the acceptance of the view that cognizing is a relation. This is a categorial analysis of a certain familiar type of fact. I have argued that it is mistaken. It would be better to describe cognizing as a "referential quality" of persons; but a fully satisfactory categorial analysis of cognizing—the indication of its dissimilarities and similarities to other factors of the real—is yet to be worked out. Such an analysis would be only a part of the more extensive ontological task of discriminating and formulating the differences and similarities of the many kinds of things now indiscriminately called "relations".