Philosophical Studies 39 (1981) 61-78. (Received 4 March, 1980)



Perhaps the best-known attack on foundations in recent years is Sellars' criticism of 'the myth of the given'. Although there are several senses of the given, I will argue that the foundations theory need not accept the doctrine in Sellars' sense. This might seem quite serious for Sellars' account, but actually I think he could admit the point without serious damage. A more important problem is that his general argument against foundations (of which rejection of the given is a part) is inconclusive and indeed implausible at the crucial point. Hence, so far as I can make out, he has not refuted the foundations theory. On the positive side, I will defend Sellars' answer to the objection that a non-foundationalist theory generates an unacceptable regress.


Sellars' most prominent use of the phrase 'the given' is for the claim that we possess a faculty of knowledge that "presupposes no learning, no forming of associations, no setting up of stimulus-response connections" (SPR, 131; also 157-159).1 In its most well-known form, the doctrine holds that we have an innate skill to be aware of immediate experiences or mental states. The doctrine concedes that we have to learn that certain types of experiences are signs of certain objects; e.g. we have to learn that a bulgy red impression of a certain shape indicates the presence of a tomato. It is also admitted that we have to learn how to report the findings of our senses since reporting presupposes language and such skills are obviously learned. However, as Sellars understands it, the doctrine cannot admit that our ability to know our immediate experiences themselves is also acquired. This might be called the classic version of the doctrine of the given.

One might think that the culprit here is some account of sense data or even the more modest claim that knowledge of sense experience is basic to knowledge, but Sellars thinks this is mistaken. He thinks that direct realists often hold the corresponding doctrine about nearby physical objects, i.e. they hold that our ability to gain knowledge of our surroundings (as opposed to just our experiences) is unlearned (SPR, 88). For Sellars, this is just as much the doctrine of givenness as the more traditional claims that such knowledge is learned but knowledge of one's experiences is not. In other words, Sellars takes the essential claim to be that we are innately able without some prior conditioning to have either sort of knowledge. He thinks this doctrine is a myth in either form and holds instead that our ability to monitor our immediate experiences as well as nearby physical objects is acquired.

Two points should be noted about this conception. First, the issue concerns cognitive skills and not just the ability to report the findings of some innate faculty. Everyone admits that our reporting skills are acquired since these involve language. Sellars, however, goes further and holds not only that we have to learn to report our findings, but also that we have to acquire the ability to have any findings in the first place. Second, givenness is independent of the question whether there is innate knowledge. The doctrine of the given claims only that we have the innate ability to recognize sorts when we experience them without prior conditioning. This does not imply that we have knowledge before we have any experience as the doctrine of innate knowledge requires. Whether there is innate knowledge is thus another issue.

Sellars' alternative is a complex theory that takes the development of the ability to be a multi-staged process. First, the child learns to monitor nearby physical objects by being taught to respond to them with simple sentences such as 'This is red' and 'That's a ball' (SPR, 90, 167). In time, he becomes an efficient meter of such objects in the sense that his spontaneously uttering such sentence tokens is a reliable indicator of the presence of the objects. Sellars does not think, however, that the child can yet be said to believe or know, say, that a nearby object is red, since at this stage he is simply reacting to objects somewhat in the way in which a thermometer records or meters the temperature without having any beliefs (SPR, 167; also 162). Nor does he think the statement can be warranted for the child, since statements can be warranted only for those who have the relevant concepts. We might say that the child believes or has concpeis, but, according to Sellars, these claims are only true when taken in an analogical sense: the child's behavior warrants us in saying that he believes but only in an extended sense since 'believes' is not used in the same sense in which we use it when we apply it to adult humans. It is only later that the child acquires concepts and the ability to have beliefs about his environment and hence that the statements become warranted for him.

Sellars would admit, I think, that it is difficult to determine when we acquire concepts or even to provide a sufficient condition for when we have them. He does think, however, that two necessary conditions for entering the belief stage are, first, the possession of a battery of linguistic abilities including 'word-word' connections as well as 'word-thing' connections, and, second, the awareness that one has these abilities, i.e. the awareness that one is an adequate meter of his environment (SPR, 90, 168). These abilities develop piecemeal but remain at the level of association without belief until, at a certain stage, the child becomes aware of the fact that his spontaneous utterances are reliable indicators of truth and hence comes to see that he has an entire range of disparate abilities.

Sellars makes the same claims about sensations, although he thinks that our ability to believe basic statements is derivative from our ability to monitor nearby physical objects.2 Specifically, he claims that the concepts we apply to sensations derive from the more 'public' concepts we employ in thinking about our environment. The important point, however, is that, unlike' Descartes, he thinks that we have to learn how to characterize our experiences and hence denies that this ability is, instinctive (although our capacity to acquire the ability itself is). Furthermore, he does not think that we can have beliefs about sensations or that such beliefs can have authority for us, as he puts it, until we know that we are reliable meters of them. For Sellars, there is still a sense in which the Cartesian emphasis on immediacy is correct: once we acquire the requisite skill, we are able to have spontaneous and thus non-inferential beliefs about experiences whereas others can only, know what our experiences are inferentially on the basis of our words and general behavior (SPR, 194-195). But this privacy, according to Sellars, is not the function of a special innate faculty but rather of the fact that the experiences in question are ours. The Cartesian error is to take this immediacy to be the result of a mental eye which, like the eye itself, operates instinctively rather than as a result of spetial training.

How does this relate to the foundations theory? Since the situation is rather complicated, let us proceed in steps. The first step concerns Sellars' conception of the foundations theory. The theory holds that some knowledge must be non-inferential (or basic, as I will call it), but this is not its most distinctive feature. It must also hold that this knowledge presupposes "no knowledge of other matter of fact, whether particular or general" (SPR, 164). Thus the fact that one accepts non-inferential knowledge or even insists on it does not make him a foundationalist.

The distinction Sellars has in mind may be brought out by considering the direct realist version of the theory. Such a realist holds that statements about tables and chairs in my vicinity are not inferred from some more basic statements such as sense-data statements, but rather are non-inferential. As it stands, this theory is open to attack. The skeptic will argue that we can know statements about physical objects only if we also know statements about the observation conditions, e.g. that the lighting is adequate and that I am a normal observer. These background statements are presuppositions in Sellars' sense since (presumably) we must know that they are true in order to know statements about nearby objects as the direct realist claims. The skeptic, however, finds it easy to attack the validity of such knowledge, since the only way we have of checking on the background statements is by further statements about physical objects. But these, in turn, are warranted only if we know that the conditions are normal -- the result is a circle or a regress, neither of which seems defensible.

What the direct realist must hold if he wishes to accept the foundations theory (and to avoid skepticism) is that we need not know the background statements in order to know, say, that the table is brown. That is, he must hold that these statements are not presuppositions of the observational knowledge he takes as the foundation. The point is that the mere fact that one accepts non-inferential knowledge does not commit him to a foundations theory. He is committed to such a theory only if he holds in addition that this knowledge does not have any presuppositions.3

Sellars does not say so explicitly, but it is clear that the only sort of presupposed knowledge the foundationalist rules out is empirical knowledge; a priori presuppositions would be perfectly acceptable. The reason is that the theory takes all empirical knowledge of general truths to rest on basic knowledge as a foundation. Basic knowledge cannot then in turn presuppose more general empirical knowledge, for this would involve the theory in the same circle to which the skeptic appeals, namely, general empirical knowledge resting on basic knowledge which in turn presupposes general empirical knowledge (SPR, 164). Indeed, the foundations theory was originally formulated to avoid just this circle. There is nothing in the theory, however, to rule out a prior presuppositions to basic knowledge. Since a priori knowledge does not rest on basic empirical knowledge but rather on what is knowable by reason alone, no circle is involved in accepting them as presuppositions to empirical knowledge.

The second step in Sellars' argument is the principle to which I have already alluded, namely, that a person can have basic beliefs only if he is aware that his spontaneous judgments about statements of the same type are reliable. Since knowledge implies belief, this principle implies that one can have basic knowledge about a certain kind of statement only if he also knows that he is a reliable meter of statements of that kind. In other words, Sellars' general account of the conditions under which one can have basic beliefs implies that basic or non-inferential knowledge presupposes knowledge of at least one general truth, i.e. that most tokenings of sentences of the same type are true.

Putting these two strands together: the foundationalist cannot argue that we can know empirically that we have this ability to monitor our experiences (or nearby objects); as we have just seen, this would commit him to the very circle he wishes to avoid. He must hold that we can know a priori that we have it and, I think, Sellars believes the only way to defend this claim is to argue that having the ability is a matter of human nature, i.e. part of our innate equipment as members of the species. That is, he thinks the foundationalist is forced to treat the presupposition as a general condition we can know to obtain simply from the concept of being human. This, however, commits the foundationlist to the myth of the given, i.e. the claim that all such skills are innate.

It is important to see just what is and what is not compatible with the foundations theory in Sellars' account. He seems in his early writings at least to take the nativist thesis, i.e. the myth of the given, to be essential to the theory, but it is not essential.4 Perhaps some versions must hold this, but there is no reason a more liberal foundationalist has to. He could hold with Sellars that we have to learn how to monitor our experiences yet still hold that basic statements can be self-warranted for us given that we have the requisite abilities. Again with Sellars, he could take the mark of having the abilities to be the fact that our spontaneous beliefs about our experiences are usually true. What he cannot hold is Sellars' further thesis, that a person has basic knowledge only if, in addition to being a good meter of his mental states (or environment, if he is also a direct realist), he also has empirical evidence for believing this is the case. In other words, the foundationalist cannot accept Sellars' claim that the following is true:

(S) An individual can know that a basic statement is true only if he knows empirically that he is an adequate meter of statements of that type.
The reason turns on the foundationalist's conception of empirical knowledge. As we saw, he holds that non-basic empirical statements must be justified in terms of basic statements, i.e. statements one knows non-inferentially to be true. Basic, statements are thus the source of all empirical justification and, since the theory wishes to avoid a circle and a regress, cannot presuppose any other empirical knowledge. If he accepted S, however, he would be committed to just this, for he would then have to hold that non-inferential knowledge has an empirical presupposition. He would thus have to accept a regress of high-level statements resting on low-level which in turn rest on further high-level statements and so on; or he would have to accept a circle of low-level statements presupposing the more genera] statements they justify. The central issue is thus S and not the myth of the given itself. We must ask what defense Sellars can give of his principle. In the next section I will examine his claim that S does not lead to an objectionable regress and in section III his general argument for S.


As we just saw, the regress problem is that, according to the empiricist, general empirical knowledge implies basic knowledge and, since S requires that basic knowledge implies empirical knowledge of at least one general truth, Sellars seems to require general knowledge resting on basic knowledge which in turn rests on general knowledge, and so on. That is, S generates a regress of knowledge resting on knowledge without end.

Sellars thinks this objection holds only if we hold that observational knowledge must precede general knowledge (SPR, 169), i.e. only if we accept:

(1) A person can know a general truth at t only if he has basic knowledge before t.
This principle together with S does generate a regress and a vicious one at that because it implies that empirical knowledge cannot get started: there is no time at which one could have either basic or general knowledge since, given S and (1), each type presupposes the prior possession of each. The response of most empiricists would be to reject S, but Sellars thinks we should reject (1) instead. Furthermore, he does not think that this compromises empiricism. According to him, all the empiricist has to hold is:
(2) A person can know a general truth only if he has basic knowledge.
Unlike (1), this principle does not require that basic knowledge precede any general knowledge. It leaves open the possibility of acquiring basic and general knowledge at the same time rather than in sequence or 'stepping-stone' style (as Aune puts it).5 As a result, it does not lead to a regress. What it commits us to instead is that we can have basic knowledge only if, at the time we acquire it, we also come to know the general truth that we are reliable monitors of that sort of experience.

It should be noted that this applies to the acquisition of concepts as as well as. to basic knowledge. As we have seen, the child who responds to the environment like a thermometer without being aware of his abilities does not, on Sellars' view, have any beliefs or concepts except in some analogical sense. The child must become aware of his ability to respond correctly (as well as meet other conditions) before he has concepts, beliefs or knowledge, all of which go together. Sellars thus denies that we can have concepts before having basic knowledge just as he denies that we can have basic knowledge before having general knowledge. Furthermore, since Sellars holds that concepts are not isolated, but rather are implicitly defined by their relations to other concepts, we must have "a whole battery of concepts" as well as other cognitive, abilities (SPR, 148; also 90). If we apply this to Sellars' account of knowledge, the picture that emerges is this: at first the child learns how to respond to various objects and experiences by uttering the appropriate tokens: 'Mama', 'Bottle' or some similar terms. This state is pre-cognitive, however: even when these habitual responses become firmly ingrained, the child as yet has no concepts, beliefs or knowledge. At some point, after enough habits have been formed, the child will becorne aware of his abilities, have concepts, and begin to have beliefs and knowledge. According to Sellars, since these abilities occur all at once, the child will pass from the pre-cognitive to the cognitive state in something like a quantum leap instead of by small incrimental steps as traditional theories of concept formation hold (SPR, 6).

Sellars' view here resembles Wittgenstein's claim that "understanding a sentence means understanding a language"6. Wittgenstein takes this 'roughly'; taken literally, it is a very extreme view, for it implies that we can understand nothing until we understand everything. Do we have to attribute this view to Sellars? I think not. He could hold that concepts have 'content' only by virtue of their relations to other concepts without holding that all the concepts of a language are interconnected (see his note on SPR, 148).7 On this view, concepts could be grouped according to types and their content characterized by their relations to other concepts of the same type. Perhaps one could not have the concept of red until he also had other concepts or the concept of a triangle until he had other geometrical concepts as well, but there would be no reason he could not have color concepts without having geometrical ones. This modified holism is still quite vague. Does one have to have all the concepts of the type to have any? How are we to individuate concept types? But it is probably no less vague than the traditional view that the mind acquires concepts piecemeal by 'seeing' them and a great deal more plausible than the kind of holism Wittgenstein suggests.

Even if we admit this, the theory is still apt to seem implausible. How, after all, can we lack concepts at one moment, then all at once in the next acquire a whole set of them as well as come to believe a basic proposition and become aware of the fact that our basic belief is more than likely true? Limiting what we have to know does not make the acquisition of numerous mental abilities at the same time any less plausible, since the notion of acquiring more than one at any given time seems implausible in itself. From this perspective, it isn't the scope of the 'battery' that is problematic, but the idea that concepts come in batteries at all.

Actually, Sellars' claim is not so implausible, since it turns out on examination that the foundations theory is also committed to the 'battery' thesis. Usually the foundationalist holds that concepts arise out of sensation somewhat as follows. Objects in the environment stimulate the child's sense organs. After repeated sensations, he comes to recognize recurrent features in them and acquires concepts, e.g. he comes to notice the similarity between the ball and rattle and forms the concept of red. Later the child learns the word for the concept and starts to acquire language. Put in this way, the theory is highly problematic. Noticing or recognizing a color is an act of awareness; it is cognitive and cannot occur unless one has the appropriate concept. If I recognize that this and that are the same color, it follows that I have the concept of the color. Thus I cannot recognize the color one moment and come to have the concept of it in the next.8 Nor can we explain concepts on the basis of sensations alone. Since the theory holds that we can have sensations without concepts, i.e. before we notice their recurrent features, it takes sensations to be non-cognitive stimulations of the sense organs. Repeating a sensation then will not bring about a concept unless the child recognizes similarities among the sensations. Yet such recognition is only possible for one who has the concept. The result is a circle: recognition of a common feature requires a concept even though concepts are supposedly derived from such recognitions.9

The problem here derives from an ambiguity in the use of 'sensation'. In one sense, any stimulation counts as a sensation whether it is an awareness or not. In another sense, however, sensations are cognitive and always involve awareness, but can occur only if one has concepts. If we fail to make this distinction, it is easy to think that repeated stimulation forces concepts on us. Actually, we are not having sensations in the same sense throughout this process. First we are stimulated non-cognitively in a certain way; later we notice the similarity and have a sensory awareness. This is a completely different sort of thing, however, despite the fact that we can call it a sensation. For this awareness cannot occur unless we have a concept and thus cannot be brought in to explain how it is that we have a concept. Once we distinguish these two senses, the theory becomes less attractive. We cannot appeal to sensations in the first sense since they cannot yield concepts by themselves; nor can we appeal to them in the second sense to explain concepts, since sensations in this sense presuppose concepts.

Sellars seems to consider this a decisive objection to the traditional theory, but I think this is hasty. The theory can hold that concepts arise when we recognize some quality shared by stimulations (which are non-cognitive). Since recognizing is cognitive, it is only at this time that we come to have a sensation in the cognitive sense. Before that, we were just being stimulated in a certain manner. Awarenesses and concepts thus occur together rather than in sequence. This account makes the theory less explanatory since noticing a quality implies having a concept. But this does not make the account impossible. It can still hold that having a sensory awareness and a concept. are cotemporaneous or, in other words, that these mental states occur in batteries. The point is that, if the theory takes this line and I do not see how it can avoid it, the foundationalist cannot object to the acquisition of mental abilities in batteries since he is as committed to it as Sellars.

This is even clearer if we accept the further thesis that cognition cannot occur without propositional knowledge, i.e. knowledge that something is the case. For then one cannot hold that it is possible to have a sensation first and later knowledge that a proposition is true; in the case of one's first sensation, both would have to occur at the same time. The result is that the minimum number of mental events that could occur in the first moment of awareness would be three: the awareness, acquiring the concept and coming to know that one was appeared to in a certain way. Most foundations theories do not accept this further thesis about the priority of propositional knowledge, but hold instead that acquaintance can occur alone. However, they must still admit that sensations and the acquisition of concepts occur together rather than in sequence, as is commonly held.


The fact that Sellars' theory is coherent does not mean that we ought to accept it. We must thus ask what reasons there are for thinking it is true, specifically, his ground for S, the principle that one can have basic knowledge of a certain type only if he knows that he is an adequate meter of statements of that type. Clearly, we are justified in taking what other people say about their experiences as true only if we are warranted in believing that they are adequate meters. In such cases they act as witnesses to events we are not party to. We also need other assumptions -- that they are sincere and know how to communicate their beliefs in language. What is not clear is whether those having the experiences themselves have to know these things as well in order for the statements to be warranted for them. A foundationalist who accepts the liberal theory sketched at the end of section I will deny that the subject enjoying the experience is in the same position as us. His reason would be that having the skill without knowing that he has it is sufficient for the statement to be warranted. We must thus ask what reason Sellars has for denying this asymmetry and thus for holding that what applies to an independent party applies equally to the person undergoing the experience.

So far as I can make out, Sellars' argument for this and thus for S is that an individual who is able to read off his inner experiences without knowing that he has this ability is not yet a human agent in some full sense, but is more like a thermometer which simply registers temperatures without any awareness of its capacity to do so. According to Sellars, such an individual lacks the feature that is most distinctive of a human agent -- the awareness of himself and his abilities -- and hence is not yet a person in the full sense of the term. As he puts it, the difference between man and his precursors involved "a jump to a level of awareness which is irreducibly new, a jump which was the coming into being of man" (SPR, 6). The argument for this is by no means clear in Sellars, but I think what he has in mind is that man is by nature a social being capable of forming communities and transmitting culture (SPR, 16-17; 39-40). As Sellars interprets it, this means that to be a human agent, one must be capable of teaching concepts to others since this is a prerequisite of passing along culture to others. If we consider only concepts about immediate experience, however, this seems to imply that a human agent must be aware of his ability, say, to monitor color experiences and this in turn implies that he is aware that his success rate is better than chance, i.e. it implies that S is true.10 Indeed, I think Sellars would go further than this and hold that one cannot have concepts of color experiences unless he can teach others to have them and hence unless he is aware of his ability. (This is the source of his view that animals only have concepts in an analogical sense -- they do not have them in the full sense since they cannot teach other members of their species to monitor their environment.) Sellars' argument then is that S is something of a meaning postulate for being human. If we take it as true that

  1. only persons can have knowledge,
  2. to be a person one must be aware of his cognitive abilities, and
  3. givenness is a myth and thus that we cannot know a priori what our abiltities are, we can deduce that S is true.
If we assume that a given individual has basic knowledge of a certain type statement, it follows by (1) that he is a person and by (2) that he knows that he is a reliable meter of that type of statement. Since (3) rules out the possibility that this further knowledge is a priori, it follows that it must be empirical. Thus S: one can have basic knowledge only if he knows empirically that he is a reliable meter of that sort of statement.

One clear problem with this argument is (2). viz. Sellars' strictures on human agency or personhood. Why should we agree that a person must be aware of certain abilities in addition to merely having them? The main difficulty, however, is with (1). Even if we agree that persons must be aware of their cognitive skills, it is still an open question whether statements can be warranted only for persons characterized in this strict way. In other words, we could deny that basic knowledge presupposes knowledge that one has the appropriate ability by denying that only persons (who by definition must be aware of their abilities) can have knowledge. In support of this, we could argue that aninials other than humans can have beliefs and even knowledge despite the fact that they are not 'persons' since, so far as we can tell, they are not also aware of their abilities.

Sellars has a reply to this. He holds that when we ascribe beliefs and desires to animals

we invariably find ourselves qualifying these explanations in terms which would amount, in the case of a human subject, to the admission that he wasn't really thinking, believing, desiring, etc.11
For Sellars, animal behavior does not settle the question since at best it shows that animals have beliefs and concepts in some analogical or extended sense. Thus he does not think we can appeal to animal beliefs or knowledge to show that S is false. This reply is not adequate, however. If his only reason for thinking that there are no animal beliefs is that animals are not aware of their abilities while humans are, he has begged the question against the foundationalist. What he needs is some reason for denying warranted belief in animals other than that animals are not aware of their abilities, and, so far as I can I tell, he has no good reason for thinking this. The same comment applies to the other part of Sellars' argument -- the claim that animals lack concepts since they cannot teach them to other members of their species. At best this coheres with Sellars' general view of animals and their supposed beliefs, but there is no independent reason for holding that their inability to transmit concepts implies that they do not have any to transmit. As a result, Sellars' argument is inconclusive even if we go along with him about the conditions for being a person, i.e. even if we accept (2). He has not shown that one can know a basic statement only if he (or it) also knows that most of his spontaneous beliefs about his experiences are true.

There is another line of argument Sellars could take to support his position. He could argue that young children and animals do not have any concepts in the full sense since they either have insufficient or non-existent linguistic abilities. Sellars' views on the relation between language and thought are complex, but I do not think they are a help in this connection. He does not think that having a belief or a concept (which we may take as representative of all the intentional attitudes) can be analyzed into statements about linguistic ability (SPR, 170). He does think, however, that it is impossible to have beliefs unless you know a language. You have to learn to report on your beliefs as well as your sensations and, as we have seen, also be aware that your reports are reliable. As he inteprets it, this cannot take place unless you understand a language. Thus language is necessary to have concepts in the non-analogical sense (SPR, 188-189). This view coheres with his general theory about concepts and supposed animal beliefs, but it is of no help in settling the issue. The issue is whether animals can have concepts even though they (apparently) are not aware of their abilities. To hold that they do not since they do not have language (in the appropriately narrow sense at issue here) does not settle this question. It is just as much an open question whether they can have concepts without language as whether they can have them without being aware of their reporting skills. Nor will it do for Sellars to argue that animals cannot have very highly developed concepts without language. It seems clear that language is necessary in order to have sophisticated and extensive concepts. The issue is whether animals can have any concepts and beliefs even though they lack language, not whether they need it to have high-level concepts.

There is a point about animal concepts that confuses the issue. A dog who believes that his master is going to feed him has an analogical concept of master. Given Sellars' plausible view that concepts have content by virtue of their relations to other concepts, the dog could not have the full-blooded human concept of master; he cannot make the same connections to other concepts a human can make, although presumably he can make some of them (if he has any concepts at all). Thus, when we describe the dog's belief using the term 'master', we are not using it in the same sense we would use it to describe a human's belief, but rather in some analogical sense. But none of this implies that the dog doesn't have a concept in the same sense in which the human does: it only implies that he doesn't have the full concept of master we have. It may be that all or almost all animal concepts are analogical but it doesn't follow that animals only have concepts in an analogical sense: it is still an open question whether the dog has a concept in the same sense in which the human does. What Sellars needs to prove is not that animals have different concepts from humans or concepts that are only analogous to ours -- that goes without saying. He has to show that attributing any concepts at all to them is just a matter of charity since they do not have concepts in the sense in which we have them. As a result, the sense in which animal concepts are analogical is too weak to show what Sellars needs to show, namely, that they do not really have concepts at all.

I conclude then that Sellars has not made out his case for S and thus that he has not refuted the foundations theory.


We have seen that Sellars' argument for S contains assumptions that are very difficult to support. There is also reason to think that S itself is implausible. Although someone else might have to know that a given individual is a good meter in order to use his reports, cases can be cited to show that the person himself does not have to know this in order for his 'reports' to be warranted for him. Alvin Goldman offers such a case, although he does not direct it against Sellars.12 Imagine that Tom is a chicken sexer who is able to sort chicks into male and female with surprising accuracy long before the obvious marks of their sex are apparent to the rest of us. Tom, in other words, is a good meter of the sex of the chicks. Suppose, however, that even though we know this to be the case, we convince him that he is not a good chicken sexer by giving him phony evidence to show that he is a failure at his job. Perhaps we show him doctored records indicating that his performance in the recent past is no better than we would expect from an ordinary person sorting the chicks by chance. In such a case, he would not be warranted in believing that he is an adequate meter since, on the evidence available to him, his success rate is not better than chance. Suppose, however, that despite this he spontaneously believes that a certain chick we present to him is a female; he has a hunch or whatever it is he has about the chick even though he is no longer as confident as he once was that the chick is a female. Is this belief warranted for him even though it is not warranted for him that he is a good meter of the sex of chicks? It seems to me, as it does to Goldman, that it is. Even though he does not believe and, indeed, is not warranted in believing that he is a better-than-chance chicken sexer, his spontaneous belief that the present chick is female is still warranted for him, since, despite, his own doubts about his skill, he still has the ability to sort the chicks with surprising accuracy. In other words, he is still warranted in thinking that the present chick is female despite the fact that it is not warranted for him that most of his beliefs about the sex of chicks are true. If this is right, Sellars is mistaken in thinking that we must know that we have a certain skill in order for its use to yield warranted belief or knowledge. Here we not only have a case in which the person is unaware of his ability; he thinks that he does not have it. Yet his belief is still warranted for him.

One might argue that the example is inconclusive on the ground that we know that we have doctored the evidence; hence we take Tom's belief about the present chick to be warranted because we know that he still has the required skill even though he no longer believes that he has it. Actually, however, what we know and do not know as observers is independent of whether Tom's belief is warranted. Tom might have come to doubt his ability through some computer error that confused his past record with that of an actual incompetant. In that case, Tom would not be warranted in thinking that he is an adequate meter even though it is still warranted for him that the present chick is female and nobody knows that the computer has fouled up. Indeed the computer error might mislead us as it misleads Tom. But if we discovered the error later, this would not lead us to distrust Tom's judgments about the chicks during the period when we all believed that he had lost his skill -- these we would take to be warranted because he had the skill all along even though we were not warranted in thinking that he had it.

There is another reply to the example. Bonjour, who accepts Sellars' critique of foundations in outline, argues that unless a subject who believes that p has evidence that most of his spontaneous beliefs similar to p are true, he "has no reason at all" for believing that p is true: "From his perspective, it is an accident that the belief is true"13. Bonjour would apparently hold that Tom's belief about the present chick is not warranted for him under the conditions of the example even though Tom has the requisite ability since, from Tom's perspective, it is an accident that the belief is true. Bonjour's point is not convincing, however. In the first place, he would beg the question if he were to say that Tom "has no reason at all" for believing that the chick is female. Tom can give us no reason to trust his judgment, but this does not mean that his belief is not warranted. To assume that his belief is not justified because he cannot defend it is to assume at the start that the foundations theory is false. Presumably the only way he could defend his belief wouldbe by citing some evidence about his ability. And to assume that, unless he can do this, the belief is unwarranted is to assume without an argument that the theory is mistaken.

A second problem lies in Bonjour's use of the phrase 'from his perspective'. When applied to Tom, this means that Tom does not believe that it is warranted for him that the chick in front of him is a female; rather he believes that, if his spontaneous belief is true, it is simply an accident that it is since he has been convinced by the doctored evidence -- or the computer error. But this is not relevant. The question is not whether Tom believes that the statement is warranted, but whether it is in fact warranted despite his beliefs about it. If Tom is rational, he will not believe that the statement is warranted, but it is not necessary for him to believe this in order for the statement to be warranted for him. A person convinced of skepticism might not believe, for instance, that a certain class of statements is warranted (or he might even believe that they are unwarranted) but it does not follow from this that they are unwarranted. If the skeptic's claim about these statements is false, they would be warranted despite the belief or absence of belief about their being warranted. As a result, we must conclude that Bonjour's defense fails to carry conviction when examined closely.

The general conclusion, I think, is that Sellars' principle S is implausible and thus that his rejection of the foundations theory also is. Furthermore, as we saw in section III, he does not have a good argument for S that would override this implausibility. As a result, even if we reject the doctrine of the given as he argues we ought to, there is no reason to think that we must also reject the foundations theory in some more reasonable form.

State University of New York at Albany


1 References to Sellars will be given in the text using the following abbreviations: SPR for Science, Perception and Reality (London, 1963); SK for 'The structure of knowledge', in: Action, Knowledge, and Reality, ed. by Hector-Neri Castaneda (Indianapolis, Ind., 1975), pp. 295-346, and GEC for 'Givenness and explanatory coherence', Journal of Philosophy LXX (1973), 612-624.

2 The necktie fable illustrates this, SPR, 142-146. Sellars holds that the concept of looking green presupposes that of being green which "involves the ability to tell what colours objects have by looking at them -- which, in turn, involves knowing in what circumstances to place an object if one wishes to ascertain its colour by looking at it", SPR, 146.

3 Sellars' view is thus foundational on the basis of Firth's well-known characterization in 'Coherence, certainty, and epistemic priority', Journal of Philosophy LXI (1964), 545-557. Firth holds that both the foundationalist and his opponent can accept the existence of non-inferential warrant-increasing properties'. Such properties increase the warrant of statements for a person at a given time. An example is the property of being a statement (a) I believe and (b) about a present sensation of mine. The foundationalist holds that any statement having such a property gains warrant regardless of whether the person also has evidence that statements having the property are likely to be true. The non-foundationalist requires that such a property is warrant-increasing only if this more general statement also has warrant. Thus he holds that the property is 'ultimately inferential' in a broader sense, 549-550. Like Sellars' account, Firth's non-foundationalist theory (or coherence theory, as Firth calls it) can accept non-inferential warrant-increasing properties but only if the covering statements about their reliability also have-warrant. In GEC, Sellars discusses Firth's defense of the foundationalist theory understood in these terms. Cornman takes a foundations theory to be one that holds (in effect) that there are self-justifying statements. i.e. any theory that accepts the existence of a non-inferential warrent-increasing property regardless of whether it also holds that it is 'ultimately inferential'. As he remarks, this makes Sellars a foundationalist. The Sellars-Firth account seems to me closer to the traditional understanding of foundations theory. For Cornman's view see James W. Cornman, 'Foundational versus nonfoundatibnal theories of empirical justification', in: Essays on Knowledge and Justification, ed. by George S. Pappas and Marshall Swain (Ithaca, 1978), pp. 229-252.

4 He seems to consider it as essential in 'Empiricism and the philosophy of mind', in SPR. The nativist claim is not prominent, however, in his discussion of Firth in GEC. Sellars still speaks of givenness in the latter, but he seems to mean by this only the claim that some statements have intrinsic warrant while not presupposing other knowledge and not that we have a primordial ability to recognize sorts.

5 Bruce Aune, Knowledge, Mind, and Nature (New York, 1967), pp. 100-101.

6 Ludwig Wittgenstein, The Blue and Brown Books (Oxford, 1964), p. 5.

7 I have put the point in terms of concepts here to avoid complications arising from the relation between language and thought; thus the obscure appeal to the content of a concept. The more usual way of expressing the point is to say that one can understand the meaning of any word in a language only if he understands the meanings of other words in it. This of course is trivially true in the case of defined terms, but the doctrine applies equally to primitive terms which are said to be defined implicitly by the sentences in which they occur, SPR, 302-306.

8 This is somewhat oversimplified. As Sellars remarks, the traditional theory posits "a level of cognition unmediated by concepts" (SK, 338). This 'direct apprehension' is supposed to stand between completely non-cognitive stimulations and conceptual awarenesses, since it is awareness without concepts. Such awareness is completely obscure, however, and ad hoc; the only reason we have for thinking it exists is that otherwise the traditional theory is false. I think William S. Robinson's discussion of the 'non-subsumptive use of concepts is an attempt to defend such non-conceptual awareness. See his 'The legend of the given', in Castaneda, Action, Knowledge, and Reality, pp. 102-103.

9 "For we now recognize that instead of coming to have a concept of something because we have noticed that sort of thing, to have the ability to notice a sort of thing is already to have the concept of that sort of thing, and cannot account for it," SPR, 176; original is in italics.

10 Jones, the mythical character who introduces the concept of sense impressions, has to teach our Rylean ancestors his theory of perception and also how to report their sensations. This implies that Jones has the ability and knows that he has it, SPR, 191-194.

11 'Chisholm-Sellars correspondence on itentionality', in: Minnesota Studies in the Philosophy of Science, ed. by Herbert Feigl et al. (Minneapolis, 1958), II, p. 527.

12 Alvin Goldman, 'Innate knowledge', in: Innate Ideas, ed. by Stephen P. Stich (Berkeley, 1975), pp. 114-115.

13 Lawrence Bonjour, 'Can empirical knowledge have a foundation?', American Philosophical Quarterly XV (1978), 8.