From Bruce Aune, Knowledge, Mind, and Nature: An Introduction to Theory of Knowledge and the Philosophy of Mind (Random House: New York, 1967).

Does Knowledge Have an Indubitable Foundation?

The position outlined at the end of the last chapter was built on the assumption that a man's sensory experience can provide a satisfactory foundation for the body of his empirical knowledge. Although this assumption has been a key tenet of traditional empiricism, most contemporary philosophers regard it as extremely questionable. In fact, it is now commonly believed that subjective experience, taken by itself, cannot possibly provide an adequate basis for any kind of knowledge at all. The main task of this chapter will be to examine the credentials of this current opinion.


Anyone who thinks it possible to develop his conception of himself and an external world solely by reference to his immediate experience is generally presupposing that the nature and interrelations of his sensory experiences can be classified and known independently of anything else. Although it will actually be shown only in later chapters that knowing and classifying are essentially linguistic and require some kind of conceptual scheme, I shall here assume that the presupposition can be expressed by the claim that one might possibly possess a primitive phenomenal language. Such a language or conceptual scheme may of course be used only in silent soliloquy. It will be considered primitive in the sense that its basic use will be to classify the phenomena of immediate experience.

The point in expressing the above presupposition by reference to a language is to come directly to terms with Wittgenstein's influential critique of private languages,1 which is commonly regarded as ruling out the possibility of a subjective basis for empirical knowledge. Since Wittgenstein's critique raises numerous questions of a highly controversial sort, I shall develop it dialectically as the chapter proceeds. As an introduction to the general strategy of his approach, the following remarks will be helpful.

If a man is to defend an assertion or proposition, he plainly needs some kind of argument. Arguments, however, need premises. Hence if there is to be a fundamental basis of empirical knowledge - something by which the truth of ordinary claims is to be defended - this basis must be propositional in character: it must be the kind of thing that can have a place in an argument. This being so, any adequate basis of knowledge will be radically different from raw experience. Pains, tickles, and itches are neither true nor false, and they cannot appear as premises in arguments. If, accordingly, the true basis of knowledge is regarded as phenomenal, it can at best consist of propositions about immediate experiences. Since these propositions are presumed to be true, they must themselves be capable of some kind of defense. If they cannot be defended, if no good reason can be given for supposing them true, then their claim to provide a secure basis of knowledge can be nothing but a sham.

Although traditional empiricists have notoriously regarded the truth of basic phenomenal premises as far too obvious to require explicit defense, this is exactly what Wittgenstein demanded. If the very foundation of our knowledge is to rest on truths about our sense experiences, then, considering the enormous stakes involved, we must surely have some guarantee that our knowledge does not rest on sand. We may indeed have no haunting doubts about the reliability of our beliefs concerning our immediate experiences, but this does not allow us, as philosophers, to assume without question that all such doubts are strictly unjustifiable, that they must be mad or wild. Admittedly, it is exceedingly difficult to imagine that we might chronically misidentify our own experiences. But this is merely a psychological matter: some people cannot imagine a lifeless universe. It will be of no use here to appeal to intuition, since the intuitions of one man may easily conflict with those of another. What is needed, plainly, is an argument, something showing just why the products of our awareness may justifiably be held reliable. Unfortunately, nothing but the flabbiest arguments have ever been given in defense of this basic empiricist idea. Empiricists have for the most part simply taken it for granted that our knowledge of our sense experiences is both noninferential and absolutely reliable. Yet taking things for granted is the mark of the dogmatist - not the free, critical spirit that the empiricist has taken himself to be.

Contrary to what one might expect, it happens that good reasons can actually be given for doubting the allegedly infallible character of immediate awareness, though these reasons were not advanced by Wittgenstein. Consider, first, the verbal behavior of hebephrenic schizophrenics. These people frequently utter what are graphically called "word salads"; they pour out chaotic jumbles of words, which often appear to be utterly unrelated. Verbally, at least, these people are totally confused. Is there any reason to think that their thoughts are less chaotic than their words? Evidently not. In fact even psychoanalysis would have to admit that, consciously at least, such patients are totally confused: their conscious thinking, if indeed it merits the name of thinking, seems to be every bit as flightful and disconnected as the words they utter. Yet if these patients really are intellectually deranged, thoroughly befuddled, there is obviously good reason to think that this derangement also extends to their awareness of their feelings.

It might be objected at this point that the kind of intellectual befuddlement just mentioned can occur only in psychotics, never in an ordinary perceiver. This may be granted, though it would appear that less dramatic forms of intellectual befuddlement could occasionally exist in almost anybody. The philosophically important fact has, nevertheless, been established. It is simply that serious intellectual confusion is a possibility, and that identifications of even feelings and mental images are not logically incapable of error. 1 myself have no reason, of course, to think that I am as befuddled as a hospitalized schizophrenic, and my confidence in the truth of what I say about my feelings is extremely high. But this confidence is something that requires some kind of justification; it cannot stand by itself as the basis of all my knowledge. If I cannot justify my conviction that I am not a walking whirly of confusion, I plainly cannot justify the idea that I have any knowledge at all, let alone a substantial stock of infallible knowledge about my momentary state of mind.

Actually the possibility of being mistaken about the character of one's momentary experience can be illustrated by reference to the behavior of perfectly sane adults. Consider, for instance, the following experiment.2 We have a man, noted for his integrity, who reports having extremely vivid imagery. His imagery is so vivid, he tells us, that he can generally read off from it all sorts of facts about objects he has recently seen. We present him with the following letter-square, and let him scrutinize it for a few seconds:

e m f
r z a
o w p

We then take the square away and ask him whether he has a clear image of it. He says that he does. We then ask him to read off the letters from left to right, starting from the top and working down. He does so, and makes no mistakes. We then ask him to read them off in the opposite direction. Suppose that, contrary to his likely behavior,3 he reads them off without hesitation, though he makes several mistakes. We have him do the same thing again - that is, read off the letters in both directions - and he gives the same answers. Without mentioning his error, we then ask him whether his image changed during the experiment. He says, "No; it remained the same throughout the experiment, vivid and sharp." In fact, he emphatically endorses all of the following claims, not even considering, so great is his confidence, that they might not be entirely consistent.

1. The image did not change during the experiment.
2. From left to right, top to bottom, the letters were: e, m, f, r, z, a, o, w, p.
3. From right to left, bottom to top, the letters were: p, w, o, r, a, s, f, m, e. (He was evidently wrong about the italicized letters.)

Now, if he has the image he claims to have, not all three of these assertions can be true of it; at least one must be mistaken. Whichever it is, we know that he has made a mistake about the character of his experience. Presented with this example, about the only thing the traditional empiricist can say, apart from questioning the honesty of the subject (which we have implicitly ruled out), is that the man's memory, as anyone will admit, is by no means infallible. We could, however, vary the experiment indefinitely, using either a very small square or even just a short sequence of letters. If we have more than one letter, error will always be theoretically possible: the statements the agent will make could always turn out to be inconsistent. To blame the error in every case on the weakness of memory seems ad hoc and theoretically desperate. Besides, what experimental sense could we give to the assertion that it is always memory that is at fault?

The admission that memory is intrinsically fallible is, however, extremely damaging to the idea that phenomenal identification could not possibly go wrong. There is plainly a sense in which memory is involved in all judgments of identification. To judge that a phenomenal occurrence has the property F is to assert that it belongs to the class of F's. But how could one know this infallibly, if one's memory is intrinsically fallible? - if one may well misremember the peculiarities, the distinguishing features, of F's generally? If it is replied that every assertion of the form "This is an F" is really an immediate matter, involving no reference whatever to other F's or to their distinguishing features, then the assertion evidentaly amounts to no more than "I shall call this 'F'." But if all phenomenal identification have this import, and only this import, then it would be impossible to establish any generalization, let alone infallible ones, relating different phenomenal items. Indeed, there would be no bona fide identifications at all; each so-called identification would turn out to be nothing more than a kind of ceremonial announcement or verbal baptism, something very different from an out-and-out claim to knowledge.

In view of this it appears that anyone wanting to regard all phenomenal identification as invariably true is forced to admit that memories, too, are sometimes infallible. If so, he must advance criteria to distinguish fallible from infallible memories, for he must be able to handle the puzzle raised by the example of the letter-square. It happens, however, that no such criteria have ever been advanced - and it is extremely difficult to imagine what such criteria would be like. Yet until we have such criteria, the conclusion to be drawn from the case of the letter-square seems inescapable, namely, that it is quite possible for even a sane man to make mistakes about the character of his immediate experience.

The temptation to regard sense experience as yielding infallible knowledge seems to arise from the historic confusion of knowledge with a kind of intellectual gazing: a sense impression has nothing hidden about it, nothing not presented to the eye of the soul, so it involves nothing about which one could be mistaken. But it is easy to see, again by considering examples familiar to psychologists, that knowledge which is a matter of having true, defensible opinions, is extremely different from intellectual gazing - and not just on the ground that there really is no eye of the soul. Consider, for instance, some of the experiments performed on congenitally blind adults whose vision has been restored by surgery. While these people learn to discriminate colors rather quickly, they often have an extremely difficult time with visual shapes. Senden found, for instance, that patients trained over a period of thirteen days to discriminate squares from triangles sometimes learned so little that they could not make these discriminations without methodically counting corners one after another.4 In fact some of these patients were quite unable visually to distinguish spheres from cubes! Though the shapes in point were physical ones, an empiricist could scarcely deny that the patients had the appropriate visual impressions. Yet if they were familiar with the idea of a sense impression, and were asked whether their current impressions were of spherical or cubical objects, they would no doubt have had to reply that they could not tell. If so, we could have had the spectacle of men intellectually wallowing in their immediate experience who nevertheless lacked the ability to appreciate the complex variety of what they were actually sensing.

The point here, though troublesome to traditional empiricists, is actually well known in the history of philosophy. And if, like Plato, Kant, and others, we make the indispensible distinction between having an impression or experience and thinking about it, attempting to classify it, and the like, we should find it very natural to admit the possibility that persons whose vision has been restored by medical treatment may very well be unable to make accurate discriminations among the visual impressions that are novel to them. Indeed, the spectacle just mentioned, of men having sensory experiences they cannot distinguish or identify, is nothing but memorable illustration of the Kantian point that percepts without concepts are blind.

In spite of all this I am fully prepared to admit that given sufficient training, it is extremely unlikely that a normal person could fail to distinguish surfaces so palpably different as circles and squares. But this concession is of little help to the traditional empiricist. The point seems to be securely established that judgments of phenomenal identification are not, in fact, infallible. We may come to have enormous confidence that, after a protracted period of training, a man's opinions about the character of his own experiences are never really wrong. But our confidence here is based on empirical considerations. There is no longer any reason to think that such opinions cannot be erroneous; rather, we have fairly good, though not fallible, reasons to think that they are normally reliable. But these reasons are neither wholly phenomenal, derived from introspection, nor purely logical, semantical, or a priori in some other subtle way.

There is, of course, a familiar defensive move that empiricists are generally anxious to make at this point. They often contend that error here implies misunderstanding of the language in which the identifications are made, so that if a man is not linguistically or perhaps conceptually befuddled in some way, what he says or thinks about his immediate experience is always true. Unfortunately, though there is often point to this contention, it is wholly useless in the present context. When introduced here, it simply redirects the challenge in question, allowing it to be focused on the matter of whether a man's confidence that he actually understands what he is saying is ever justified. Linguistic or conceptual befuddlement is, after all, just as serious an intellectual defect as out-and-out error, for confusion and error are both cases of ignorance, which is failure to appreciate the truth. It may well be a necessary truth that if one is not conceptually befuddled, one's judgments of phenomenal identification are always true. But because of the possibility of improper coordination between thought and sensation, illustrated by the above experiments, this alleged necessity does not support the traditional contention that immediate apprehension is an infallible source of knowledge.

The foregoing discussion illustrates the position I shall take in this book: subjective experiences, or introspective knowledge of them, are not sufficient to constitute the foundation of anything, let alone all our empirical knowledge. This is not to say, however, that introspective knowledge is an illusion; in fact, I shall take special pains to defend the legitimacy of such knowledge against its behavioristic critics. The defense I give will nevertheless be far out of line with the contentions of traditional empiricism. Not only shall I insist that subjective experience cannot provide the true foundation of our knowledge, but I shall insist that our knowledge has no foundations in the traditional sense. By this I do not mean that it is unfounded or baseless, in the sense that it is subject to every shift in the uncertain winds of custom or fancy. I simply mean that there is no such thing as an indubitable foundation on which knowledge of any sort can rest.


Although Wittgenstein's attack on a subjective basis for empirical knowledge is presumably consistent with the main lines of my argument against the infallibility of introspective claims, it involves further, more serious criticism; in fact some philosophers see it as constituting a critique of introspective knowledge generally.5 The fundamental point of the critique seems to be this. Anyone restricted merely to the domain of private experience has no possible way of checking up on, or even adding credibility to, his momentary apprehensions about his immediate experiences. Such a person could not, in fact, distinguish between knowing that an experience has, say, the property F and merely thinking of being or being under the impression that it has F. But if he could not make this distinction, and so justify the truth of his introspective claims, the very idea of introspective knowledge turns to dust and has, consequently, no significant place in serious epistemological deliberation.

In view of the far-reaching, indeed revolutionary, consequences of this contention, it is obviously important to scrutinize its credentials with exceptional care. As just presented, Wittgenstein's argument seems to rest heavily on the necessity of making a clear distinction between knowing that one has a certain experience and merely thinking that one has it.6 But this distinction is actually not of crucial importance. The contradictory of "He knows that p" is "He does not know that p"; and if one can make sense of the latter, one can give sense to the former, even if, for some reason, "He merely thinks that p" remains unintelligible. [The importance of this reservation to Wittgenstein's argument is brought out in Ch. IV; see esp. pp. 95-98.] Assuming that Wittgenstein was correct in insisting on the necessity of finding some kind of significant contrast for the claim that a man knows he has a certain experience, we may thus regard this contrast as adequately given by the claim that the man does not not possess this knowledge.

An obvious first step in coming to terms with Wittgenstein's evident attack on the possibility of basic introspective knowledge is to ask how, in his opinion, one is ever able to distinguish knowledge from ignorance. His general answer to this question, if I understand him correctly, is that an assertion may be regarded as an expression of knowledge only if it is made in accordance with some appropriate rule, one by which the propriety of of applying the relevant words or concepts is to be appraised.7 If this is indeed his opinion, it appears that if the empiricist of the last chapter can show that his application of phenomenal term or concept is in accordance with some appropriate rule, then his contention that he actually does possess basic introspective knowledge stands a good chance of being justified even according to Wittgenstein's principles.


If we are to make sense of the kind of knowledge in question by reference to rules, we must be clear about the kind of rule involved. Wittgenstein, unfortunately, was not entirely explicit about this. Yet if we consider his general assumption that the use of a word is governed by "criteria," which are in some sense based on the word's definition, a natural interpretation of his doctrine is not difficult to formulate.8 Although his use of the term "criterion" is not without its puzzles,9 we can surely agree that the question whether a thing is correctly described as, say, a "lemon" would normally be answered by at least a tacit or indirect reference to the defining characteristics of lemons - and Wittgenstein would presumably call these characteristics "criteria." Thus, if I am assured that the term "lemon" properly applies to yellow, sour fruit of a certain characteristic size and shape, I can justify my use of this term to describe or identify something by showing that the thing in point possesses these defining characteristics. In doing so, I may be said to be relying on a rule, one to the effect, roughly, that fruit of such and such characteristics are correctly called "lemons." The general connection between knowledge and rules suggested by this example is that one can ultimately determine whether a thing actually is a K - whether the word "K" properly applies to it - only by reference to rules that specify the criteria for being a K. [I am only using the word "criterion" in an informal sense, which is slightly different from the technical sense I introduce in Ch. V, pp. 114f.]

It is crucially important to note that the rules involved here strictly authorize what might be called "intra-language" moves.10 This label highlights the special character of these rules, which is to authorize inferences. Thus, to continue with the same example, if one knows or has good reason to think that a thing is yellow, sour, of the appropriate size and shape, then the rule in point allows one to infer that the object is a lemon. And if, conversely, one knows that it is a lemon, then the rule probably allows one to infer that it is probably yellow, sour, and the like.11 (The utility of this latter rule is that it tells one what to expect when lemons are said to be in the vicinity.)

The idea that the rules with reference to which empirical claims are justified are all of this intra-language sort immediately leads to a very serious problem. In order to justify a claim by an intra-language rule one has to know that some other claim is justifiable, the claim, namely, that serves as the premise for the inference. Yet if this latter claim can be justified only by a rule of inference, one must know that still another claim is justifiable, and so on without end. This, however, seems to make it impossible to justify anything. To put it in another way, since intra-language rules merely authorize inferences, they cannot themselves justify any basic premises. Yet without such premises, we could never obtain justified conclusions; and without justified conclusions we could never have knowledge of the actual character of the world.

This kind of puzzle takes us immediately to the basis of traditional empiricism and partially accounts for the tenacity of the idea that empirical knowledge must be built on a phenomenal foundation. It is precisely phenomenal awareness, whose infallibility is so difficult to doubt that is supposed to justify the basic premises of empirical knowledge. One justifiably makes these basic claims as the result of one's awareness of a certain experience - an occurrence or datum which, unlike physical things, is entirely open to view. Once these primitive claims are made, rules of inference allow one to move on and construct a warranted picture of the world.

Although I have already advanced arguments against the idea that immediate experience provides an indubitable basis of knowledge, the assumption that knowledge requires some such basis is still very much alive. Since this assumption seems to rule out the plausibility of Wittgenstein's general approach by sustaining the problem mentioned above, a brief resolution of that problem must be attempted before proceeding any further with Wittgenstein's argument. I shall attempt to do this by showing that the assumption in point is untenable and that the problem it poses for Wittgenstein's approach is actually misconceived.

The line of reasoning behind the empiricist's assumption is, again, that while intra-language rules may validly take us from premise to conclusion, they cannot themselves establish empirical truth. If the premises you start with are false, you will have no guarantee that the conclusions you reach are not false either. Hence, to attain knowledge of the actual world, you must ultimately have premises whose truth is acceptable independently of any inference and whose status is accordingly indubitable. Only by having such premises can you gain a starting point that would make inference worthwhile. For convenience, these indispensable basic premises may be called "intrinsically acceptable." The possibility of empirical knowledge may then be said to depend on the availability of intrinsically acceptable premises.

If this line of thought is sound, it follows that utter skepticism can be ruled out only if one can locate basic empirical premises that are intrinsically acceptable. Although philosophers who attack skepticism in accordance with this appraoch generally think they are defending common sense, it is crucial to observe that they cannot actually be doing so. The reason for this is that, from the pont of view of common experience, there is no plausibility at all in the idea that intrinsically acceptable premises, as so defined, ever exist. Philosophers defending such premises fail to see this because they always ignore the complexity of the situation in which an empirical claim is evaluated.

I have already given arguments to show that introspective claims are not, in themselves, intrinsically infallible, they may be regarded as virtually certain if produced by a reliable (sane, clear-headed) observer, but their truth is not a consequence of the mere fact that they are confidently made. [Some philosophers argue that the truth of certain statements is a consequence of the fact that they are made with maximum understanding. I attacked this argument briefly on p. 37, and I shall attack it in detail in Ch. IV, esp. pp. 102-105.] To establish a similar conclusion regarding the observation claim of everday life only the sketchiest arguments are needed. Obviously, the mere fact that such a claim is made does not assure us of its truth. If we know that the observer is reliable, made his observation in good light, was reasonably close to the object, and so on, then we may immediately regard it as acceptable. But its acceptability is not intrinsic to the claim itself. Thus, philosophers who, like G. E. Moore,12 attempt to prove by direct inspection that they have hands do not proceed just by taking a quick look at their hands; they rather turn them over, look at both sides, pinch them, and the like. The certainty they arrive at is thus based on a whole group of observations, as well as on numerous tacit assumptions concerning the general reliability of their senses, the accuracy of their memories, the sort of things hands are supposed to be, and so on. I would venture to say that any spontaneous claim, observational or introspective, carries almost no presumption of truth when considered entirely by itself. If we accept such a claim as true, it is only because of our confidence that a complex body of background assumptions - concerning observers, standing conditions, the kind of object in question - and, often, a complex mass of further observations all point to the conclusion that it is true.

Given these prosaic considerations, it is not necessary to cite experimental evidence illustrating the delusions easily brought about by, for example, hypnosis to see that no spontaneous claim is acceptable wholly on its own merits.13 On the contrary, common experience is entirely adequate to show that clear-headed men never accept a claim merely because it is made, without regard to the peculiarities of the agent and of the conditions under which it is produced. For such men the acceptability of every claim is always determined by inference. [To say this is not to imply that one always does infer, or actually come to a reasoned conclusion, that a given observation claim is acceptable. It is rather to say that the acceptability of such a claim is to be justified by inference; that its acceptability is not intrinsic to it. This is entirely compatible with the obvious fact that human beings constantly accept claims as true without thinking about them at all.] If we are prepared to take these standards of acceptability seriously, we must accordingly admit that the traditional search for intrinsically acceptable empirical premises is completely misguided.

To rule out intrinsically acceptable claims on the grounds of common experience is to presuppose two things: first, that common experience can somehow provide an acceptable basis for knowledge and, second, that utter skepticism is untenable. If these presuppositions can be defended, it will therefore follow that the argument purporting to establish the need for intrinsically acceptable empirical claims must contain some crucial flaw. Although I shall attempt a justification of these presuppositions only in later chapters, [See esp. Ch. V, pp. 123-126, and Ch. VI, p. 137, footnote.] I can say now that the basic flaw in the empiricist's argument arises from a grossly oversimplified conception of the structure of empirical reasoning. In assuming that we must have unalterable and incontestable truth in order to infer something on which we can reasonably depend, it overlooks the important fact that successful empirical reasoning can proceed only against a background of general assumptions (many of them empirical [As indicated above, some of these assumptions will concern the reliability of the observer (whoever he is), the character of the standing conditions, and so on. For further discussion of this, see Ch. V, pp. 123-126.] ) in terms which we interpret our experience and assess the truth of what we say about it. The empiricist's basic error is thus his presumption that we could actually start out with a fistful of merely inferential rules and then, as innocent of the world as the youngest child, cast about for some self-justifying premise that will permit us to draw an inference. The error of this presumption is patent, because even though a premise may come to mind that is actually certain, we must have good reason to believe it is certain if we are to use it in an inference. Even if it were to bring with it the strongest feeling of confidence, or even wear a little lapel saying "I am true," we would plainly require some rational means of deciding whether the confidence or the label can be safely trusted.

It is granted that no empirical claim is stictly justifiable on its own merits but requires some kind of support from a body of other claims, we may accordingly infer that Wittgenstein's conception of knowledge as something requiring justification by reference to rules is not basically unacceptable. In fact, we may note that his insistence on the necessity for rules in this connection brings out the important fact that both confirmation and disconfirmation (or proof and disproof) involve relations between claims, between statements and propositions. To show that a statement is false, you have to establish some other claim with which the first one is inconsistent, and to show that a statement is true, or probably true, you have to show that it is rendered so by certain statements that formulate the relevant evidence. The basic point here is that confirmation and disconfirmation are logical relations; and such relations hold between terms belonging to the conceptual, rather than the natural, order.

Once it is seen that there are no intrinsically acceptable empirical claims and that confirmation as well as disconfirmation can be rendered only by other claims that also lack intrinsic acceptability, it becomes apparent that the process of firmly establishing an empirical claim can be, in principle, almost endless. Suppose, for instance, that I happen to be in some doubt as to whether the fruit I am holding is actually a lemon. In order to remove this doubt and confirm my tentative belief, I might appeal to my neighbors. If they agree that it is a lemon, and if I have no reason to doubt the honesty of their testimony, then I would ordinarily be considered justified in taking it to be a lemon. But I might, of course, be mistaken in trusting them. Should I later become aware of this mistake, I might consult still other persons, or perhaps do some research in a library. What I learn from these sources would normally settle my doubt, but it need not insure the truth of my belief. In fact, if I began to suspect that I had recently been hypnotized, and told to misinterpret any direct evidence bearing on the kind of fruit I am carrying, I might fall into an utter quandry. [In the paper referred to in note 13 of this chapter, an actual case of such a quandry was demonstrated experimentally. A man had been given a post-hypnotic suggestion that, after writing on some sheets of paper before him, he would forget having written on them and be unable to see that they were written on at all. During the experiment, he was subjected to close examination regarding what was written on the papers and, when his denial that anything was written on them met with constant objection by his interlocutors, he became both angry and confused, showing all the signs of a man whose perception of an obvious matter of fact meets with universal disbelief. A case of this sort should carefully be kept in mind by any philosopher who thinks that obvious matters of fact are always easy to settle.] To work my way out of this quandry I could appeal to other considerations and make further tests. This kind of appeal could, however, go on indefinitely, with a theoretical possibility of mistake at every turn. It is, of course, granted that in most cases I would not have to make such appeals in order to establish my claim beyond any reasonable doubt. But the important fact remains that the confidence I attain need never be logically immune to a rational challenge.


Assuming that the sort of rule to which Wittgenstein requires one to appeal for purposes of justifying an empirical claim is of the intra-language sort so far discussed, we may now consider whether a defender of phenomenal languages could possibly establish his claims to knowledge. In view of the foregoing discussion, we may pose our question as follows: "Could such a thinker justify a basic phenomenal claim, 'This is an A,' by relating it to some intra-language rule that he possesses?" The answer to this particular question seems to be "Yes."14 In order to justify his claim he might have recourse to a rule that relates the expression "A" to another expression "B." He could then argue that he knows that the item is an A because it is also a B and because it is a rule of his language that anything that is a B may properly be called "an A." Admittedly, the question whether it is indeed a B can also arise, and this question would have to be settled by further appeals. But the possibility of such constant queries and appeals is not peculiar to phenomenal language. As already indicated, it holds generally. To assume otherwise is simply to accept the immediately infallible in another form, and in so doing to reject the spirit of Wittgenstein's attack.

In might be thought that I am simply begging the question here by assuming that the phenomenal thinker may indeed have rules. How, for instance, is he to distinguish between his actually having rules and his merely thinking that he has them? The question may, however, be attacked in all sorts of ways. For one thing, the thinker may employ a variation on Wittgenstein's own "This language game is played" theme, 15 arguing that he is fully confident that his language-game is under way, that he has absolutely no reason, concrete or otherwise, to think that he is not employing rules, and that he cannot, in point of fact, even conceive the possibility that he is operating with rules, since conceiving such a thing would require him to have them already. (He could not, therefore, "merely think" that he has rules; for thinking this, rather than that, is a rule governed activity, which may involve inconsistency and error.) Another approach to the question would be to declare that it is fundamentally misconceived. The thinker might, that is, reject the whole question, arguing that one's linguistic rules are not the sort of thing whose existence one must prove (such a proof would reequire rules anyway) but something one simply uses. Rules are not factual contentions, and to accept a rule is to adopt a procedure - a specific way of thinking, or organizing one's ideas. If the thinker actually has rules, his use of them will be shown by the inferences he draws concerning the character of his phenomenal data. If he draws inferences at all, he is thereby operating with rules, whether he is articulate in formulating them or not.


So far, I have taken the term "rule" in a fairly strict sense; I have regarded linguistic rules as the sort of thing that may justify an inference. There is, however, another sense of "rule" lurking in Wittgenstein's discussion. This sense concerns linguistic regularity, and it is brought into the picture as soon as it is asked whether a particular inference is regularly drawn from the same type of premise. This sense of the word "rule" is nothing like what has been called the "regulation" sense16 for it is essentially descriptive and cannot itself justify an inference or establish some claim. It is nevertheless of central importance to the issue of phenomenal languages, since the empiricist who wants to defend the legitimacy of such languages must be assured that he is at least consistent in the use of his phenomenal terms.

Such "rules" can exist, then, only when there is a certain linguistic practice. The question is, Could a solitary thinker have a consistent practice of applying his phenomenal terms? Although one would think that the answer to this is an obvious "Yes" - that the interesting question is not whether he can be consistent but whether he can defend his conviction that he is consistent - it is perhaps wise to consider two general lines of argument that might be urged against the very meaninglessness of the contention that a practice of this kind might exist, let alone be known to exist by the agent involved. The first objection concerns the meaning of the word "same" in the context of "applying a concept to the same phenomenal terms," and the second concerns the identity of the thinker who is supposed to use a phenomenal language.

Beginning with the first objection, we might ask what the word "same" is supposed to mean in a special phenomenal language. This term plainly belongs to ordinary English, and when it is applied to purely phenomenal objects, where familiar criteria are not involved, it appears to lose whatever sense it had. Yet if it does lose its familiar sense in this new context, it is obviously misused: for it is employed as though it had its usual sense - in fact, it is used in this context just because it has a very familiar but crucial job to perform. If it had utterly different meaning as used here, it would prevent the present issue from even being formulated.

Actually, this kind of objection concerning the meaning of "sameness in kind" - for that is what is presently involved - is not so troublesome to the phenomenal theorist in this context as it appears to be elsewhere. [Its importance in another context is discussed in Ch. III, sect. 2.] As used here, the expression has a purely formal significance, which is quite independent of any specific subject matter. Thus, the theorist can easily advance the following definitions:

"a and b are the same kind of thing" means "there is a kind of thing, A, of which both a and b are instances."
Since this definition is wholly general, mentioning no specific kind of thing, the peculiarities of a phenomenal subject-matter are logically immaterial to the meaning of "the same" as it appears here. Whether one is speaking of furniture, phenomenal data, or even sporting events, the expression "the same" as so defined retains precisely the same meaning.

Given, however, that "the same" has a purely formal significance, there is still the question of how the sameness of two or more things is to be determined. The answer to this immediately brings us back to the ancient puzzle about the relation between words and the world. Are a and b both lemons? Are they both instances of Lemonkind? Well, do they both satisfy the criteria for being lemons? Are they perhaps yellow, sour, of ovoid shape? As already indicated, to answer these questions is to establish the truth of other factual contentions, which may require justification as well. Questions of this kind can no doubt be answered to the satisfaction of a reasonable man, but the answers given will by no means be immune to critical examination.

Similar considerations naturally apply to the contentions of the silent thinker. If he is to justify his contention that two phenomenal items are the same in kind - both instances, that is, of some kind-concept - he need only appeal to certain other contentions, about which he is reasonably confident. There is simply no way of getting outside all conceptual schemes in order to see whether one's concepts are consistently and accurately applied to reality. Tests for semantic regularities must necessarily be carried out within the framework of some conceptual scheme, and the question whether a thoroughly intersubjective framework is consistently applied can be raised in just the embarrassing way that it is raised in connection with phenomenal languages - given, that is, that questions of this type are legitimate ones. The silent thinker cannot, it is true, assuage such doubts by appeal to the testimony of other persons. But a wide variety of appeals is still open to him, at least as regards any particular doubts he might have; and even if he could make an appeal to other persons, this would do nothing to settle any general doubt he might have concerning the consistent application of his language as a whole, for he would have to interpret their testimony within the framework of the scheme he happens to have.

Assuming, then, that a defender of phenomenal languages faces no greater obstacle in making sense of "same in kind" than anyone else, we have to consider the other line of objection mentioned above: Just who, or what, is this thinker that is supposed to have the solitary practice of applying phenomenal terms? The answer to this has already been given: the thinker in question is the X that has the phenomenal items in point, the person who, according to the empiricist reconstruction of the last chapter, has a certain body, a certain position in the world, and so on (see pp. 25-29). Of course, this answer cannot be given by the thinker himself at this stage of his deliberations. As already explained, his conception of his self is something he will proceed to construct on the basis of an ascertained "coherence" among a certain field of phenomenal items. It is not, therefore, binding on him to identify and describe himself at this stage of his argument. He will attempt to do this later on, when his phenomenal language is enriched by what are for him theoretical notions. At the present stage, he could conceive the general question regarding semantic regularities only along lines similar to this: "Might certain thoughts or conceptual episodes, such as 'This is a K' or 'The item that is a G is also a K,' consistently occur in connection with, and only with, K-kind phenomenal items?" If he can answer this question in the affirmative, and if, after forming his conception of his self, he can show that these regularities are simply manifestations of his practice, then the objection in point could evidently be met.

But an argument can be quickly advanced against this procedure, too. Just what admittedly subjectless thoughts and phenomena are supposed to be involved here? Since the answer "Mine!" cannot possibly be given at this stage, must the class not include any phenomena whatsoever? Yet if it does so - if it includes even those which, if the thinker's language were fully developed, he would ascribe to other persons - then the existence of the appropriate regularities could never possibly be established without reference to nonphenomenal considerations, a reference that is explicitly excluded from the present context.

A defender of the traditional theory of awareness would naturally be tempted to reply to this question by contending that the phenomena in point are simply the objects of direct awareness. But this reply would obviously misfire in the present connection, and not just for the reason that no reference to the subject of awareness is presently allowable. The main difficulty is rather that an appeal to "awareness" solves nothing whatever. Any identification must be made within the framework of some conceptual system, not outside such a system by conceptually neutral act of direct awareness. [The limitations of direct awareness will be further explored in Ch. III, sect. 4, where the doctrine of "logically proper names" is discussed.]

It is possible, however, for the thinker to approach this question in a way that ought, by now, to be familiar. He need only ask: "How are identifications ever made? How is a domain of objects ever identified and circumscribed?" The answer to this, plainly, is that they are described in some way; they are picked out as the things falling under certain concepts. But if the private thinker has, as he claims, a phenomenal language including such terms as "K-kind," "A," and "B," then he would have no trouble specifying the objects in question: they are simply the things falling under these terms or their corresponding concepts. It is surely not binding on him to specify these objects in some language different from his own; and although we, who do not understand his language, will not know just what things he is talking about when he uses such terms as "K-kind," it does not follow from this that they are really meaningless, or really not terms at all.

All of this may sound very naive to a philosopher immersed in the intricacies of contemporary semantical theory. Such a philosopher might immediately wish to reply: "If we are serious in setting out an interpretation of a given language structure by specifying a domain of objects to which its various terms refer, we shall have to proceed in what is strictly a metalanguage.17 When we do this, we can then formulate semantical rules by which to determine the application of expressions in the object language. We will not therefore be limited to saying that a domain of objects is simply the class of things falling under certain concepts; we can rather identify the domain independently of these concepts by using the expression of the metalanguage. Thus, if we were using English as a metalanguage in order to outline the semantics of French, we might say that 'les chiens' denotes the class of dogs. It would then be by reference to this rule that we could determine whether a given Frenchman consistently applies 'chien' to the right things."

Although this argument is very familiar, it is clearly useless. Even if we actually had a hierarchy of metalanguages in which to specify the objects referred to by the terms of each sublanguage, we would still have to have one language the semantics of which is just "understood." It would be understood, moreover, without the aid of semantical rules. And if this understanding is not really understanding, if we could not strictly understand what we are talking about when we use this language, then, since each sublanguage can be ultimately understood only in terms of this one - the semantics of the first sublanguage being expressed in it, and so on down to the lowest level - we could not really understand any language at all. Hence, if the reference of words is ever to be understood, it must be possible to understand it without the apparatus of semantical rules.

Another difficulty with the semanticist's argument is that when we are concerned with natural languages, we are not free to lay down just any semantical rules; we must rather be able to justify every such rule that we advance. Yet when we consider how this justification is to be given, we see at once that knowledge of the reference of a term must be in hand before we are in a position to advance the appropriate semantical rule. Suppose, for instance, that we are concerned to justify the rule concerning "les chiens given in the last paragraph. How might we proceed? Surely by showing that the French expression applies to just those things, actual and possible, that we call "dogs." We have to consider possibilities as well as actualities because not all of the things to which an expression may legitimately apply need actually exist. To make a well-worn example, if there were no plucked chickens, the animals to which "featherless bipeds" might actually apply would be restricted to human beings, though the expression could also apply to a plucked bird should one exist. Because possible as well as actual applications of a given expression must be considered in establishing its distinctive reference, it becomes that in order to be fully justified in advancing the semantical rule concerning "les chiens" we must have some access to the criteria on the basis of which Frenchmen determine whether something describable in a certain way is or is not un chien. What we have to determine, in fact, is whether the things that are dogs, according to our criteria, are the same as the things that are chiens, according to the criteria of Frenchmen. And to determine this, we would have to be able to establish the extension of "les chiens without reference to the English language.

The basic point here, to which even the private thinker can appeal, is simply this: a kind of thing, such as Chien, Dog, or K, is definable only within some conceptual scheme or other, and it is only by reference to the criteria involved in the appropriate scheme that the decision whether a given thing belongs to a certain kind can possibly be reached. (Of course, once these criteria are ascertained, we may express them in other languages and so state semantical rules of the sort just considered.) The question, then, whether a given thing belongs to the domain of objects about which the silent thinker's phenomenal idiom allows him to speak, can properly be said to be answerable according to the criteria he happens to have, not to the criteria appropriate to some other conceptual scheme, whether public or not. [Since the silent thinker's domain of discourse is initially restricted to phenomenal items and does not allow him to speak of conscious agents, his linguistic rules can initially be formulated only as modal conditionals or equivalences: for example, "A's are necessarily B's."]


It might naturally be objected at this point that I have badly underestimated the importance of ostensive definition. If our language is actually to be applied to reality, it must surely be possible to indicate the reference of our basic descriptive terms by pointing out, grasping, or sometimes even manipulating, instances of the things to which they apply. This pointing, grasping, or manipulating is more primitive than language, does not require the apparatus of a conceptual scheme,18 and is therefore a logically fundamental means of indicating which objects are denoted by various basic terms. Since this pointing, grasping, and so on cannot be utilized by a private thinker, he labors under an irremedial handicap, and it is exactly this handicap that renders his theoretical labors futile.

Familiar as this argument is, it is not difficult to see that it cannot possibly succeed. Collingwood pointed out years ago19 that a gesture of the kind in question is a linguistic act itself, whose reference is in no way more obvious than that of audible speech. When a man points, one must not just look at his finger, as dogs naturally do; one must look away from his finger, and then try to decide what he is attempting to draw one's attention to. Doing even this requires considerable sophistication, but it takes far more to appreciate the particular intent of the man's gesture, which must be understood if his so called ostensive definition is going to succeed. The foregoing discussion has proved, however, that this intent - which is to indicate that the object of the gesture is an example of something or other - could never be entirely given by a group of actual gestures, no matter how many might be made. The reason for this is that the meaning or denotation of a term can be determined only by reference to possible applications and these cannot be surveyed by a process of pointing. Whatever is merely pointed out or manipulated can be classified in an indefinite variety of ways, and a condition of understanding a certain classification is that one know just what is encompassed by it. To know the meaning, or conventional application, of an expression "K" we must accordingly have a grasp of the criteria for being a K, for it is only with reference to these criteria that we could possibly delimit the variety of things that, should they exist or come to our attention, would be correctly classifiable as K's. Since these criteria cannot possibly be given by a mere gesture of pointing, it is clear that a so-called ostensive definition is a misnomer. The most that ostension can do is direct our attention to a vaguely delimited region of space and time; if we are to understand what what is being pointed out, we must understand the relevant principles of classification.

Although these remarks make it plain that pointing is in no way sufficient to establish the meaning of any term, it might nevertheless be thought that it is necessary for the purpose of learning certain words. This view has at any rate been expressed by philosophers who attempt to draw philosphical conclusions about meaning from their conception of how a child might be taught certain words. The idea here is, however, false; there are no words whose meaning can be taught only with the aid of pointing gestures. As already indicated, the most that pointing can do is direct one's attention to a vaguely defined region of space, [I speak of a "vaguely defined region of space" because a mere gesture of pointing does not pick out a circumscribed thing or region. In pointing at a man I may be interpreted as pointing at his head, his eye, his pupil or eyelash, or even to the spatial area through which his head is moving.] and this can be accomplished by verbal signals just as well. The gesture of pointing is in no sense easier to understand than familiar verbal directions such as "Over there!"

It is important to observe in this connection that a child's first words are not really taught. As every parent knows, such words are learned "naturally" in a way that seems wholly mysterious. It may be granted that when a child is old enough to benefit from his parent's untutored and clumsy attempts to teach him words, he can best learn such nouns as "ball" when the appropriate things are in his view. But a gesture of pointing has no special utility in the learning situation. If a child hears the word "ball" when he is playing with one, his attention will already be directed to the appropriate object, and a gesture of pointing would only distract it. The actual process of learning language is really far more complicated than philosophers seem to think; and those who believe that the gesture of pointing has some special utility for language-learning should note that psychologists hardly ever mention pointing when discussing the subject of how people learn words. [If a gesture of pointing is understood as a linguistic act, comparable in significance to, say, "Look at that!," it would be considered a "mand" by the psychologist B. F. Skinner. Yet in Skinner's discussion of learning in his Verbal Behavior (New York: Appleton-Century-Crofts, 1957), Ch. 3, pointing is never even mentioned. In his view words are learned by progressive reinforcement of appropriate "operant" behavior.] Pointing is neither necessary nor sufficient for learning language, and the private thinker's inability to rely on gestures of pointing is thus of no significance at all.

This reference to language-learning prompts a further comment. It might be thought that the inability of an alleged private thinker to be be taught his language by some other agent shows that the idea of his actually having a language is senseless. But this view is clearly false, because it is certainly possible, logically, for a child to be born with the ability to speak his parent's language.20 Of course, if such a child does indeed possess a language at the hour of his birth, he must then be able to apply its terms correctly and consistently. In ordinary life this is something that we would determine by observing the child's verbal behavior. In the present case this is admittedly ruled out. Here we are concerned with the possibility of a phenomenal language, the correctness and consistency of whose use can be determined only by phenomenal considerations. They are the only legitimate considerations because, as already shown, the question whether a particular expression is correctly applied can be determined only by reference to criteria internal to the language in point. For the private thinker, these criteria are phenomenal, not public; and it would thus be logically inappropriate to demant that his phenomenal claims be assessed by criteria of any other kind.


The argument of the last section was built around one basic question: How, ultimately, can one ever justify the contention that the concepts of a given conceptual system are in fact consistently applied to the same kind of object? The topic of ostension was introduced because it is difficult to see how a justification of this sort could possibly be given by a mere appeal to intra-language rules. If, in order to assure oneself that "A" is correctly applied to X, one could only appeal to such rules as "Anything having an identifying feature B may be properly called 'an A,'" then one would have to know that X is in fact the feature B, which means that "B" properly applies to it. But if there is a general doubt about whether the conceptual system in point is correctly and consistently applied, then an appeal to this kind of rule is question begging, for the doubt about the application of "A" accrues also to the application of "B." Hence, if we are to take seriously a general doubt of this kind, it would seem that we must ultimately make an appeal to extra-conceptual considerations, perhaps to immediate awareness, pointing or something similar.

Having argued that any question whether a given concept is correctly and consistently applied to the same kind of object can be answered only by reference to the criteria distinctive of that kind of thing. I am obviously committed to reject any extra-conceptual meaning of ruling out such a doubt. But this does not mean that I am committed to take a general doubt of this sort seriously. On the contrary, if the preceding argument is sound, such general doubts are completely idle, baseless, and logically capable of casting any real suspicion on anything at all.

To see this, note that any serious doubt requires some kind of basis. In involving the claim that some aligned matter of fact is indeed doubtful, a justification is obviously in order. This justification can actually be given, however, only within the framework of some conceptual system. Consequently, if a man has only one conceptual system, one "language," he will not be able to formulate a serious doubt about the correct application of that system as a whole - for he would have to use the system in order both to formulate his doubt and to establish a justification for it. This does not mean that he will be unable to question the propriety of particular claims that are made in its terms, or even to revise large segments of the system while hanging by his bootstraps to the part that remains. But a general doubt about the correct and consistent application of his system as a whole will not be open to him - and this will be true of the private thinker too, if he has only one conceptual scheme with which to work.

In might be thought, however, that such general doubts could justifiably be made within the framework of some other conceptual system. Wittgenstein at any rate purported to do just this when he attacked the possibility of private languages. And on the face of it at least, an external attack of this sort seems entirely reasonable. Suppose, for example, that members of a native tribe were observed to apply a given expression to a very large, highly disconnected group of objects - to things that we could call "cows," "trees," "rocks," "manure," and so on. Would not this scattered sort of application show that their expression was not consistently applied? The answer, of course, is "No." To appreciate this answer, we have only to consider more closely the question at issue: Are certain expressions consistently applied to the same kind of objects? The crucial word here is "kind," for kinds are determined only with reference to some conceptual scheme. And while it may be true that the expression in point is not consistently applied to anything that we, or indeed any other civilized person, would regard as a single kind, this may not be true for the natives. In fact there is no theoretical limit to the variety of ways in which they may happen to carve reality.

Admittedly, if we knew something about the general structure of the natives' language, or even about their chief interests and beliefs, [The relevance of both interests and beliefs in this connection can be illustrated by reference to the word "carbon." A man with suitable interests might apply this word to precious diamonds, industrial diamonds (which may appear to be gravel), and ordinary chimney soot.] the number of these alternatives could perhaps be cut down to manageable size. But we would in any case have to determine the extension of the terms they use by reference to the ways they use them; and to the extent that their usage is puzzling, or seemingly inconsistent when measured by our way of viewing things, their conception of things is simply beyond our ken - which means that we are yet in no position to say whether they consistently apply their terms to what they consider the same kind of thing. In order to settle doubts about the semantic regularities involved in their verbal behavior, we must in fact approach the question from within, by learning their language and solving our problem by an appeal to the considerations they themselves invoke in appraising the success of their own verbal performances. Only in this way is it possible to discover whether they conceive it, and whether our language even has the resources to permit a definition of the kinds of thing to which many of their crucial terms apply.21

If a wholly general doubt about semantic regularity cannot legitimately be raised within the framework of the system to which it applies, and any other doubt raised within the framework of some other conceptual system is similarly self-stultifying - in that any specification of the kind of thing to which the terms in question are supposed to apply must be justified with reference to what are considered the semantical regularities governing their use - it then follows that a general doubt about the consistent application of an entire language cannot possibly be justified. Such a doubt must remain absolutely idle pointless, with no possibility whatever of gaining empirical support. Since a doubt of this kind is no less idle when directed to phenomenal languages, we must conclude that an external attack on such languages cannot succeed. This, however, seems to be the approach Wittgenstein took in attacking them.

Granting that general doubts about whether a phenomenal thinker does "as a rule" apply his terms consistently are, in this way, illegitimate, one might nevertheless insist that the foregoing argument renders particular doubts both unavoidable and irresoluble. If there is indeed no ultimate foundation of empirical knowledge, will not any attempt to answer a particular doubt inexorably lead to an infinite regress? As before, the answer to this question must be "No." While it is assuredly possible to raise doubt after doubt, any particular doubt that is worth heeding must be justifiable, which means that it, too, must be supported by contentions that are themselves subject to challenge and, potentially at least, in need of justification. I say "potentially at least in need of justification. I say "potentially in need of justification" because if we could never take something as not requiring justification for the moment, no doubts and no contentions could ever reasonably be advanced. Justification must therefore, as Wittgenstein said, come to an end somewhere -- if only for the time being. True, we can in principle challenge the credentials advanced in favor of any bona fide contention. But in doing so we must base our challenge on reasonable considerations and we must be willing to waive our doubts in the face of appropriate evidence. The idea that a justification may be self-justifying is nothing more than favorable results from the tests he is willing to make in support of some contention on which he bases his confidence. If he is sufficiently critical, sufficiently suspicious of dogma in all its guises, he can scarcely be accused of intellectual levity or gullibility.


The question that led to the foregoing discussion was whether a solitary thinker might in fact be consistent in applying basic phenomenal terms to elements of his immediate experience. [See p. 47.] So far, I have argued that any positive reason for doubting the consistency of such a practice cannot, by the very nature of the case, be formulated. The question arises, however, whether the idea of semantical regularities in a wholly phenomenal language actually makes sense. I think that it does. If, in line with the anti-behaviorist assumptions on which we have been operating since Chapter 1, one is prepared to grant that one's own feelings and sense impressions are covert occurrences or states, then the suggestion that one might attempt to identify them in thought by devising a special language for them does not appear to be incomprehensible, nor does it seem unlikely that the judgments of identification one might proceed to make in the new idiom would consistently correspond to the appropriate phenomenal occurrences. If one is prepared to admit, moreover, that other people also have covert feelings and sense impressions, and are also capable of silent thought, then it should not be excessively difficult to imagine another person employing such a language either.

Is is not necessary, after all, that these phenomenal occurrences be intrinsically private in the sense that no one but the private thinker could conceivably know that they occurred or existed. Although Wittgenstein apparently did attack the intelligibility of utterly private objects [It is important to recall that I am presently concerned with the question whether Wittgenstein's arguments refuses the view outlined at the end of Ch. 1 (see pp. 24-29). Whether his argument succeeds against the view he actually had in mind, which I take to be a form of logical atomism (see p. 67, footnote), is a very different question, not at issue here. Many philosophers apparently think that this argument rules out any kind of basic phenomenal language, and it is this opinion that I wish to refute here.] there is nothing in the contentions of the solitary thinker presently in question that would commit him to the existence of these peculiar things. Not only does his skimpy, basic conceptual apparatus lack the resources necessary to speak of other persons, but when it is enriched by nonphenomenal terms, so that discourse about selves and a public time and space can be carried out within it, he will endeavor to show that other persons can understand him when he speaks - in that language - of the feelings, thoughts, and sense impressions that he happens to have. Hence, since there is no initial presumption that the phenomenal items in question are intrinsically private in some profound epistemological sense, only a radical behaviorist should have special difficulty in understanding the idea that a practice of applying phenomenal terms might in fact be quite consistent.

It should be observed in this connection that even if, as is often claimed, we ordinary describe our experience in relation to patterns of behavior with which it is correlated, it does not follow that no other mode of describing our experience is possible. Thus, while itches are commonly distinguished from other feelings partly by reference to their tendency to bring about scratching, it is not inconceivable that a solitary thinker might classify those of his feelings that happen to be itches in a less indirect way. If the phenomenal theorist's strategy of enriching his basic language were accomplished in any significant degree, he might in fact reach the point of identifying the referent of one of his phenomenal terms "A" with what he later learns to call "an itch that I feel." The identity-statement he may later affirm "A's are itches that I feel," would of course be a contingent one, since "A's, unlike itches, are not definitionally tied to the tendency to scratch. But the merely contingent character of this statement will not prevent us from saying that the items he initially identifies as A's may be nothing more mysterious than what we might call "itches that he feels."

If the idea of identifying what are in fact one's sensory experiences in a special language so meagerly constructed as to have no room for describing persons or public objects does make sense and if, further, no serious doubt about the consistent use of such a language can legitimately be raised, it seems clear that the phenomenal theorist's claim to a primitive phenomenal language cannot be dismissed on a priori grounds. This means that his initial steps toward constructing the position described in the last chapter eludes the kind of objection we have been considering. It may be that this initial step leads into a blind alley, but the step itself seems entirely coherent. It is not absurd in principle.


Although the Wittgensteinian arguments sketched in this chapter do not abolish the possibility of all phenomenal languages, my discussion of them has disclosed two points that should minimize the attractiveness of any appeal to subjective experience as the basis of all empirical knowledge. The first is that subjective experience cannot really provide the logically unshakable foundation of knowledge that empiricists have traditionally sought. In fact the very idea that there can be an ultimate foundation for knowledge has been thoroughly undermined. As I have argued at length, any empirical claim, phenomenal or not, is subject to possible correction by other claims. This means that although we may justifiably advance claims that concern nothing more than the momentary character of our immediate experience, we have not thereby reached the bedrock of unalterably certain knowledge that empiricists have tried to find in their subjective experience.

The second point disclosed by the foregoing discussion is perhaps more exciting: phenomenal claims are not intrinsically more reliable than claims of other kinds, such as those concerning persons and physical objects. Thus, as I have shown by describing a number of psychological experiments, it is quite possible for a person's introspective claims to be flatly erroneous. Not only might one be deficient in conceptual ability, but one might be drugged, dazed, insane, or hypnotized. In order to rule out such contingencies, one must have some way of defending the reliability of one's introspective claims. This is most easily done if an appeal to others is taken seriously: their corroboration of my claims adds significant credibility to them. Of course, the fact that their corroborative remarks are likely to be true is itself open to doubt. But to doubt it justifiably one must have some reason to think it perhaps false. To have this reason is obviously to have access to a far wider range of considerations than is available to one whose theorizing is carried out wholly in a phenomenal language. Yet this wider range of considerations in no way lessens one's chances of finding the truth. On the contrary, as I shall argue in subsequent chapters, the wider the range of considerations to which one can appeal, the sounder one's opinions are likely to be.

My view on this matter is almost the opposite to that of traditional empiricists. For them, the accuracy of a man's outlook varied inversely to its scope: the more it encompassed, the more it risked; and the more it risked, the less reliable it was. I prefer to argue that the possibility of error is not the same as the likelihood of error, and that in demonstrating error one is also demonstrating truth.

If you can show that a statement is false, or likely to be false, you have thereby shown that its negation is true or likely to be true. Hence for me any theory that increases the possibility of demonstrating error is far preferable to one that minimizes this possibility. As every gambler knows, when the stakes are low, so are the winnings. 22


1. See Ludwig Wittgenstein, Philosophical Investigations, trans. G. E. M. Anscombe (Oxford: Blackwell, 1953), Pt. I, secs. 243-270, and also Norman Malcolm, "Review of Wittgenstein's Philosophical Investigations," Phil. Rev., LXIII (1954), 530-539: repr. in Malcolm, Knowledge and Certainty (Englewood Cliffs, N.J.: Prentice-Hall, 1963), pp. 96-129.

2. A similar experiment is described in Donald O. Hebb, Organization of Behavior (New York: John Wiley, 1949), p. 36.

3. His likely behavior is verbal hesitance. See Hebb, ibid.

4. Cited in Hebb, ibid. p. 32.

5. E.g., by Norman Malcolm, "Knowledge of Other Minds," Journ. Phil., LV (1958), 977; repr. in Knowledge and Certainty, p. 99.

6. See Malcolm, "Review of Wittgenstein," Phil. Rev., p. 532; Knowledge and Certainty, p. 99.

7. See Wittgenstein, Pt. I, secs. 256-272, and also Malcolm, "Review of Wittgenstein," Phil. Rev., pp. 532ff; Knowledge and Certainty, pp. 98ff.

8. See Wittgenstein, Pt. I, sect. 354.

9. See Rogers Albritton, "On Wittgenstein Use of the Term 'Criterion'," Journ. Phil., LVI (1959), 845-857.

10. See Wilfrid Sellars, "Some Reflections on Language Games," Phil. Sci., XXI (1954), 204-228; repr. with changes in Sellars, Science, Perception, and Reality (London: Routledge & Kegan Paul, 1964), pp. 321-350.

11. See Michael Scriven, "The Logic of Criteria," Journ. Phil., LVI (1959), 857-868.

12. G. E. Moore, "Proof of an External World," Proc. of British Academy, XXV (1939); repr. in Moore, Philosophical Papers (London: Allen & Unwin, 1959), pp. 127-150.

13. See Milton H. Erickson, "Experimental Demonstrations of the Psychopathology of Everyday Life," in S. S. Tompkins and H. A. Murray, eds., Contemporary Psychopathology (Cambridge, Mass.: Harvard University Press, 1944). pp. 524-525.

14. Here I am indebted to Hector-Neri Castenada, "The Private Language Argument," in C. D. Rollins, ed., Knowledge and Experience (Pittsburgh: University of Pittsburgh Press, 1963), pp. 88-105.

15. See Wittgenstein, Pt. I, sec. 654.

16. See Max Black, "Notes on the Meaning of 'Rule'," Theoria, XXIV (1958), 107-136; repr. in Black, Models and Metaphors (Ithaca, N.Y.; Cornell University Press, (1962), pp. 95-139.

17. A clear discussion of modern semantics can be found in Richard M. Martin, Truth and Denotation (Chicago: University of Chicago Press, 1958).

18. This view seems to be defended by Stuart Hampshire in Thought and Action (London; Chatto & Windus, 1959), Ch. 1.

19. See R. G. Collingwood, The Principles of Art (Oxford: Clarendon Press, 1938), p. 227.

20. Defenders of Wittgenstein's argument generally admit this point; see Malcolm, "Review of Wittgenstein," Phil. Rev., p. 544; Knowledge and Certainty, p. 112.

21. See Willard Van Orman Quine, Word and Object (Cambridge, Mass.: John Wiley and M.I.T. Press, 1960), esp. pp. 53,77.

22. This mot occurs somewhere in John Wisdom, Other Minds (Oxford: Blackwell, 1952).