Wilfrid Sellars Edited in Hypertext by Andrew Chrucky, 1995. Note: This paper was first presented as the University of London Special Lectures on Philosophy for 1955-56, delivered on March 1, 8, and 15, 1956, under the title "The Myth of the Given: Three Lectures on Empiricism and the Philosophy of Mind."
Reproduced with the permission of the University of Minnesota Press from: Wilfrid Sellars, "Empiricism and the Philosophy of Mind," in Herbert Feigl and Michael Scriven, eds., Minnesota Studies in the Philosophy of Science, Volume I: The Foundations of Science and the Concepts of Psychology and Psychoanalysis (University of Minnesota Press, 1956), pp. 253-329. When the essay was reprinted in Wilfrid Sellars, Science, Perception and Reality (London: Routledge & Kegan Paul, 1963), Sellars added a few notes which have been incorporated into the present version. I should also add that in the reprint British spelling was used. Republished as a separate book, Empiricism and the Philosophy of Mind, with an introduction by Richard Rorty and a study guide by Robert Brandom (Cambridge, Mass. : Harvard University Press, 1997). Republished in Willem deVries and Timm Triplett, Knowledge, Mind, and the Given: A Reading of Sellars' "Empiricism and the Philosophy of Mind" (Indianapolis, IN: Hackett Pub., 2000). French translation Empirisme et philosophie de l'esprit (Combas (France): l'Eclat, 1992). I. AN AMBIGUITY IN SENSE-DATUM THEORIES
I PRESUME that no philosopher who has attacked the philosophical idea of givenness or, to use the Hegelian term, immediacy has intended to deny that there is a difference between inferring that something is the case and, for example, seeing it to be the case. If the term "given" referred merely to what is observed as being observed, or, perhaps, to a proper subset of the things we are said to determine by observation, the existence of "data" would be as noncontroversial as the existence of philosophical perplexities. But, of course, this just is not so. The phrase "the given" as a piece of professional -- epistemological -- shoptalk carries a substantial theoretical commitment, and one can deny that there are "data" or that anything is, in this sense, "given" without flying in the face of reason.
Many things have been said to be "given": sense contents, material objects, universals, propositions, real connections, first principles, even givenness itself. And there is, indeed, a certain way of construing the situations which philosophers analyze in these terms which can be said to be the framework of givenness. This framework has been a common feature of most of the major systems of philosophy, including, to use a Kantian turn of phrase, both "dogmatic rationalism" and "skeptical empiricism". It has, indeed, been so pervasive that few, if any, philosophers have been altogether free of it; certainly not Kant, and, I would argue, not even Hegel, that great foe of "immediacy". Often what is attacked under its name are only specific varieties of "given." Intuited first principles and synthetic necessary connections were the first to come under attack. And many who today attack "the whole idea of givenness" -- and they are an increasing number -- are really only attacking sense data. For they transfer to other items, say physical objects or relations of appearing, the characteristic features of the "given." If, however, I begin my argument with an attack on sense-datum theories, it is only as a first step in a general critique of the entire framework of givenness.
2. Sense-datum theories characteristically distinguish between an act of awareness and, for example, the color patch which is its object. The act is usually called sensing. Classical exponents of the theory have often characterized these acts as "phenomenologically simple" and "not further analyzable." But other sense-datum theorists -- some of them with an equal claim to be considered "classical exponents" -- have held that sensing is analyzable. And if some philosophers seem to have thought that if sensing is analyzable, then it cannot be an act, this has by no means been the general opinion. There are, indeed, deeper roots for the doubt that sensing (if there is such a thing) is an act, roots which can be traced to one of two lines of thought tangled together in classical sense-datum theory. For the moment, however, I shall simply assume that however complex (or simple) the fact that x is sensed may be, it has the form, whatever exactly it may be, by virtue of which for x to be sensed is for it to be the object of an act.
Being a sense datum, or sensum, is a relational property of the item that is sensed. To refer to an item which is sensed in a way which does not entail that it is sensed, it is necessary to use some other locution. Sensibile has the disadvantage that it implies that sensed items could exist without being sensed, and this is a matter of controversy among sense-datum theorists. Sense content is, perhaps, as neutral a term as any.
There appear to be varieties of sensing, referred to by some as visual sensing, tactual sensing, etc., and by others as directly seeing, directly hearing, etc. But it is not clear whether these are species of sensing in any full-blooded sense, or whether "x is visually sensed" amounts to no more than "x is a color patch which is sensed," "x is directly heard" than "x is a sound which is sensed" and so on. In the latter case, being a visual sensing or a direct hearing would be a relational property of an act of sensing, just as being a sense datum is a relational property of a sense content.
3. Now if we bear in mind that the point of the epistemological category of the given is, presumably, to explicate the idea that empirical knowledge rests on a 'foundation' of non-inferential knowledge of matter of fact, we may well experience a feeling of surprise on noting that according to sense-datum theorists, it is particulars that are sensed. For what is known even in non-inferential knowledge, is facts rather than particulars, items of the form something's being thus-and-so or something's standing in a certain relation to something else. It would seem, then, that the sensing of sense contents cannot constitute knowledge, inferential or non-inferential; and if so, we may well ask, what light does the concept of a sense datum throw on the 'foundations of empirical knowledge?' The sense-datum theorist, it would seem, must choose between saying:
It is particulars which are sensed. Sensing is not knowing. The existence of sense data does not logically imply the existence of knowledge.
Sensing is a form of knowing. It is facts rather than particulars which are sensed.
On alternative (a) the fact that a sense content was sensed would be a non-epistemic fact about the sense content. Yet it would be hasty to conclude that this alternative precludes any logical connection between the sensing of sense contents and the possession of non-inferential knowledge. For even if the sensing of sense contents did not logically imply the existence of non-inferential knowledge, the converse might well be true. Thus, the non-inferential knowledge of particular matter of fact might logically imply the existence of sense data (for example, seeing that a certain physical object is red might logically imply sensing a red sense content) even though the sensing of a red sense content were not itself a cognitive fact and did not imply the possession of non-inferential knowledge.
On the second alternative, (b), the sensing of sense contents would logically imply the existence of non-inferential knowledge for the simple reason that it would be this knowledge. But, once again, it would be facts rather than particulars which are sensed.
4. Now it might seem that when confronted by this choice, the sense-datum theorist seeks to have his cake and eat it. For he characteristically insists both that sensing is a knowing and that it is particulars which are sensed. Yet his position is by no means as hopeless as this formulation suggests. For the 'having' and the 'eating' can be combined without logical nonsense provided that he uses the word know and, correspondingly, the word given in two senses. He must say something like the following:
The non-inferential knowing on which our world picture rests is the knowing that certain items, e.g. red sense contents, are of a certain character, e.g. red. When such a fact is non-inferentially known about a sense content, I will say that the sense content is sensed as being, e.g. red. I will then say that a sense content is sensed (full stop) if it is sensed as being of a certain character, e.g. red. Finally, I will say of a sense content that it is known if it is sensed (full stop), to emphasize that sensing is a cognitive or epistemic fact.
Notice that, given these stipulations, it is logically necessary that if a sense content be sensed, it be sensed as being of a certain character, and that if it be sensed as being of a certain character, the fact that it is of this character be non-inferentially known. Notice also that the being sensed of a sense content would be knowledge only in a stipulated sense of know. To say of a sense content -- a color patch, for example -- that it was 'known' would be to say that some fact about it was non-inferentially known, e.g. that it was red. This stipulated use of know would, however, receive aid and comfort from the fact there is, in ordinary usage, a sense of know in which it is followed by a noun or descriptive phrase which refers to a particular, thus
Do you know John? Do you know the President?
Because these questions are equivalent to "Are you acquainted with John?" and "Are you acquainted with the President?" the phrase "knowledge by acquaintance" recommends itself as a useful metaphor for this stipulated sense of know and, like other useful metaphors, has congealed into a technical term.
5. We have seen that the fact that a sense content is a datum (if, indeed, there are such facts) will logically imply that someone has non-inferential knowledge only if to say that a sense content is given is contextually defined in terms of non-inferential knowledge of a fact about this sense content. If this is not clearly realized or held in mind, sense-datum theorists may come to think of the givenness of sense contents as the basic or primitive concept of the sense-datum framework, and thus sever the logical connection between sense data and non-inferential knowledge to which the classical form of the theory is committed. This brings us face to face with the fact that in spite of the above considerations, many if not most sense-datum theorists have thought of the givenness of sense contents as the basic notion of the sense-datum framework. What, then, of the logical connection in the direction sensing sense contents-->having non-inferential knowledge? Clearly it is severed by those who think of sensing as a unique and unanalyzable act. Those, on the other hand, who conceive of sensing as an analyzable fact, while they have prima facie severed this connection (by taking the sensing of sense contents to be the basic concept of the sense-datum framework), will nevertheless, in a sense, have maintained it, if the result they get by analzing x is a red sense datum turns out to be the same as the result they get when they analyze x is non-inferentially known to be red. The entailment which was thrown out the front door would have sneaked in by the back.
It is interesting to note, in this connection, that those who, in the classical period of sense-datum theories, say from Moore's "Refutation of Idealism" until about 1938, analyzed or sketched an analysis of sensing, did so in non-epistemic terms. Typically it was held that for a sense content to be sensed is for it to be an element in a certain kind of relational array of sense contents, where the relations which constitute the array are such relations as spatiotemporal juxtaposition (or overlapping), constant conjunction, mnemic causation -- even real connection and belonging to a self. There is, however, one class of terms which is conspicuous by its absence, namely cognitive terms. For these, like the 'sensing' which was under analysis, were taken to belong to a higher level of complexity.
Now the idea that epistemic facts can be analyzed without remainder -- even "in principle " -- into non-epistemic facts, whether phenomenological or behavioral, public or private, with no matter how lavish a sprinkling of subjunctives and hypotheticals is, I believe, a radical mistake -- a mistake of a piece with the so-called "naturalistic fallacy" in ethics. I shall not, however, press this point for the moment, though it will be a central theme in a later stage of my argument. What I do want to stress is that whether classical sense-datum philosophers have conceived of the givenness of sense contents as analyzable in non-epistemic terms, or as constituted by acts which are somehow both irreducible and knowings, they have without exception taken them to be fundamental in another sense.
6. For they have taken givenness to be a fact which presupposes no learning, no forming of associations, no setting up of stimulus-response connections. In short, they have tended to equate sensing sense contents with being conscious, as a person who has been hit on the head is not conscious, whereas a new-born babe, alive and kicking, is conscious. They would admit, of course, that the ability to know that a person, namely oneself, is now, at a certain time, feeling a pain, is acquired and does presuppose a (complicated) process of concept formation. But, they would insist, to suppose that the simple ability to feel a pain or see a color, in short, to sense sense contents, is acquired and involves a process of concept formation, would be very odd indeed.
But if a sense-datum philosopher takes the ability to sense sense contents to be unacquired, he is clearly precluded from offering an analysis of x senses a sense content which presupposes acquired abilities. It follows that he could analyze x senses red sense content s as x non-inferentially knows that s is red only if he is prepared to admit that the ability to have such non-inferential knowledge as that, for example, a red sense content is red, is itself unacquired. And this brings us face to face with the fact that most empirically minded philosophers are strongly inclined to think all classificatory consciousness, all knowledge that something is thus-and-so, or, in logicians' jargon, all subsumption of particulars under universals, involves learning, concept formation, even the use of symbols. It is clear from the above analysis, therefore, that classical sense-datum theories -- I emphasize the adjective, for there are other, 'heterodox,' sense-datum theories to be taken into account -- are confronted by an inconsistent triad made up of the following three propositions:
A. x senses red sense content s entails x non-inferentially knows that s is red.
B. The ability to sense sense contents is unacquired.
C. The ability to know facts of the form x is ø is acquired.
A and B together entail not-C; B and C entail not-A; A and C entail not-B.
Once the classical sense-datum theorist faces up to the fact that A, B, and C do form an inconsistent triad, which of them will he choose to abandon?
He can abandon A, in which case the sensing of sense contents becomes a noncognitive fact -- a noncognitive fact, to be sure which may be a necessary condition, even a logically necessary condition, of non-inferential knowledge, but a fact, nevertheless, which cannot constitute this knowledge.
He can abandon B, in which case he must pay the price of cutting off the concept of a sense datum from its connection with our ordinary talk about sensations, feelings, afterimages, tickles and itches, etc., which are usually thought by sense-datum theorists to be its common sense counterparts.
But to abandon C is to do violence to the predominantly nominalistic proclivities of the empiricist tradition.
7. It certainly begins to look as though the classical concept of a sense datum were a mongrel resulting from a crossbreeding of two ideas:
The idea that there are certain inner episodes -- e.g. sensations of red or C# which can occur to human beings (and brutes) without any prior process of learning or concept formation; and without which it would in some sense be impossible to see, for example, that the facing surface of a physical object is red and triangular, or hear that a certain physical sound is C#.
The idea that there are certain inner episodes which are non-inferential knowings that certain items are, for example, red or C#; and that these episodes are the necessary conditions of empirical knowledge as providing the evidence for all other empirical propositions.
And I think that once we are on the lookout for them, it is quite easy to see how these two ideas came to be blended together in traditional epistemology. The first idea clearly arises in the attempt to explain the facts of sense perception in scientific style. How does it happen that people can have the experience which they describe by saying "It is as though I were seeing a red and triangular physical object" when either there is no physical object there at all, or, if there is, it is neither red nor triangular? The explanation, roughly, posits that in every case in which a person has an experience of this kind, whether veridical or not, he has what is called a 'sensation' or 'impression' 'of a red triangle.' The core idea is that the proximate cause of such a sensation is only for the most part brought about by the presence in the neighborhood of the perceiver of a red and triangular physical object; and that while a baby, say, can have the 'sensation of a red triangle' without either seeing or seeming to see that the facing side of a physical object is red and triangular, there usually looks, to adults, to be a physical object with a red and triangular facing surface, when they are caused to have a 'sensation of a red triangle'; while without such a sensation, no such experience can be had.
I shall have a great deal more to say about this kind of 'explanation' of perceptual situations in the course of my argument. What I want to emphasize for the moment, however, is that, as far as the above formulation goes, there is no reason to suppose that having the sensation of a red triangle is a cognitive or epistemic fact. There is, of course, a temptation to assimilate "having a sensation of a red triangle" to "thinking of a celestial city" and to attribute to the former the epistemic character, the 'intentionality' of the latter. But this temptation could be resisted, and it could be held that having a sensation of a red triangle is a fact sui generis, neither epistemic nor physical, having its own logical grammar. Unfortunately, the idea that there are such things as sensations of red triangles -- in itself, as we shall see, quite legitimate, though not without its puzzles -- seems to fit the requirements of another, and less fortunate, line of thought so well that it has almost invariably been distorted to give the latter a reinforcement without which it would long ago have collapsed. This unfortunate, but familiar, line of thought runs as follows:
The seeing that the facing surface of a physical object is red and triangular is a veridical member of a class of experiences -- let us call them 'ostensible seeings' -- some of the members of which are non-veridical; and there is no inspectible hallmark which guarantees that any such experience is veridical. To suppose that the non-inferential knowledge on which our world picture rests consists of such ostensible seeings, hearings, etc., as happen to be veridical is to place empirical knowledge on too precarious a footing -- indeed, to open the door to skepticism by making a mockery of the word knowledge in the phrase "empirical knowledge."
Now it is, of course, possible to delimit subclasses of ostensible seeings, hearings, etc., which are progressively less precarious, i.e. more reliable, by specifying the circumstances in which they occur, and the vigilance of the perceiver. But the possibility that any given ostensible seeing, hearing, etc., is non-veridical can never be entirely eliminated. Therefore, given that the foundation of empirical knowledge cannot consist of the veridical members of a class not all the members of which are veridical, and from which the non-veridical members cannot be weeded out by 'inspection,' this foundation cannot consist of such items as seeing that the facing surface of a physical object is red and triangular.
Thus baldly put, scarcely anyone would accept this conclusion. Rather they would take the contrapositive of the argument, and reason that since the foundation of empirical knowledge is the non-inferential knowledge of such facts, it does consist of members of a class which contains non-veridical members. But before it is thus baldly put, it gets tangled up with the first line of thought. The idea springs to mind that sensations of red triangles have exactly the virtues which ostensible seeings of red triangular physical surfaces lack. To begin with, the grammatical similarity of 'sensation of a red triangle' to "thought of a celestial city" is interpreted to mean, or, better, gives rise to the presupposition, that sensations belong to the same general pigeonhole as thoughts -- in short, are cognitive facts. Then, it is noticed that sensations are ex hypothesi far more intimately related to mental processes than external physical objects. It would seem easier to "get at" a red triangle of which we are having a sensation, than to "get at" a red and triangular physical surface. But, above all, it is the fact that it doesn't make sense to speak of unveridical sensations which strikes these philosophers, though for it to strike them as it does, they must overlook the fact that if it makes sense to speak of an experience as veridical it must correspondingly make sense to speak of it as unveridical. Let me emphasize that not all sense-datum theorists -- even of the classical type -- have been guilty of all these confusions; nor are these all the confusions of which sense-datum theorists have been guilty. I shall have more to say on this topic later. But the confusions I have mentioned are central to the tradition, and will serve my present purpose. For the upshot of blending all these ingredients together is the idea that a sensation of a red triangle is the very paradigm of empirical knowledge. And I think that it can readily be seen that this ideal leads straight to the orthodox type of sense-datum theory and accounts for the perplexities which arise when one tries to think it through.