Published in Action, Knowledge and Reality: Studies in Honor of Wilfrid Sellars, edited by Hector-Neri Castañeda (Indianapolis: Bobbs-Merrill, 1975): 295-347. Presented as The Matchette Foundation Lectures for 1971 at the University of Texas.

The Structure of Knowledge




1. There is, of course, a broad but technical sense in which even persons are 'things', though not mere things. There is, also, a metaphorical sense in which one can be said to treat a person as a mere thing. But this amounts to treating him, in Kant's phrase, as a means only, i.e. acting in ways which either disregard, or fail to value for their own sake, the traits by virtue of which we distinguish between a merely physical object and a conscious subject of purposes and intentions. In this lecture I shall be primarily concerned with our perceptual knowledge of merely material things, turning my attention in the following lecture to our knowlege of those 'things' which, however physical they may be, have, in addition, the features by virtue of which they are persons.

2. Before I zero in on my present topic, however, some remarks are in order on the broad, if technical, sense in which both merely material things and persons are 'things'. Epistemology cannot be severed from ontology as with a knife, and it is necessary to give some account of the basic categories which I shall be using, in order to provide a framework for the interpretation of what, after all, is but a fragment of a larger story.

3. The ideal aim of philosophizing is to become reflectively at home in the full complexity of the multi-dimensional conceptual system in terms of which we suffer, think, and act. I say 'reflectively', because there is a sense in which, by the sheer fact of leading an unexamined, but conventionally satisfying life, we are at home in this complexity. It is not until we have eaten the apple with which the serpent philosopher, that we begin to stumble on the familiar and to feel that haunting sense of alienation which is treasured by each new generation as its unique possession. This alienation, this gap between oneself and one's world, can only be resolved by eating the apple to the core; for after the first bite there is no return to innocence. There are many anodynes, but only one cure. We may philosophize well or ill, but we must philosophize.

4. The method is easy to characterize, but difficult in the extreme to follow. One begins by constructing simple models -- which are understood because we have constructed them -- of fragments of this multidimensional framework. These initial models are inevitably over-simple and largely, false. But the alternative to this path, with all its oversimplification and error, is to sketch the shifting surfaces of the functioning framework, and hope that insight comes by pasting the sketches together. This receptivity, however sensitive, and however important it may be as an element in philosophical method, must, by itself, fail to yield understanding. In much the same spirit, Plato warns that the poets, by concentrating on appearances, are precluded from understanding the actions and characters of men which they so contagiously depict. The real danger of oversimplified models is not that they are over-simple, but that we may be satisfied with them, and fail to compare them with regions of experience other than those which suggested them. And, indeed, the ultimate justification for system building in philosophy is the fact that no model for any region of discourse -- per-perceptual, discursive, practical -- can be ultimately satisfying unless its connection with each of the others is itself modeled. To press the metaphor to its limits, the completion of the philosophical enterprise would be a single model -- the working of which, again, we would understand because we had constructed it -- which would reproduce the full complexity of the framework in which we were once unreflectively at home.


5. The region within the encompassing framework on which I shall concentrate this evening is that of merely physical things and our knowledge of them. This knowledge is, in the first instance, perceptual or, it is better to say, at the perceptual level. For there is a widespread misconception, no longer as prevalentasit used to be, according to which perception -- in what is often called the 'strict' or 'basic' sense of this term -- gives a knowledge of singular truths which presupposes no knowledge of general truths. According to this misconception, all knowledge of general truths at the perceptual level is inductively grounded in the deliverances of perception.

6. Now I have no objection in principle to drawing a distinction between that which we perceive strictu sensu, and that which we perceive 'in a looser sense of the term'. For according to the methodology which I sketched above, one is entitled to 'regiment' discourse by constructing simple models. But any such distinction must, in Plato's words, carve reality at the joints. And, as I hope to show, no way of validly making this distinction supports the idea that there is a level of perceptual knowledge of singular truths which presupposes no knowledge of general truths about material things and our perception of them.

7. In short, knowledge 'at the perceptual level' essentially involves both knowledge of singular matters of fact and knowledge of general truths. Neither is possible without the other.


8. But enough by way of anticipation and methodology. The promised sketch of basic categories remains to be drawn. I shall be making use of them throughout this series, and, while they will not loom too large this evening, it will be useful to get them out into the open, so that the larger philosophical context of these remarks can be taken into account from the beginning.

9. What is a 'merely' material thing? It is, in the first instance, an individual, as is, of course, a person. But what is an individual? Questions of this ontological kind arouse a strong temptation to say that one is at the level of that which must be 'shown' rather than 'said'. And the temptation is not without insight. However, as is illustrated by Wittgenstein's own work, things can be said which aid the showing, and perhaps the most useful thing to say is that the linguistic correlates of individuals are singular terms.

10. Is everything an individual? The above remarks would suggest a negative answer, since not every linguistic expression is a singular term. It would therefore be wise to have a broader category in reserve for which we might use the word 'entity'. To do so is to countenance the idea that not all entities are individuals.1

1. If the question is now pressed: are there items which are not entities?, the ontological enterprise is really under way. But that is a story for another occasion.

11. Now some individuals are, in an important sense, 'reducible'. We feel comfortable about saying that they consist of 'simpler' individuals which are their 'parts'. One is tempted, therefore, to introduce the idea of a basic individual as one that has no individuals as its 'parts'.

12. Are there any basic individuals in this sense? Why might not individuals have parts, and these again, and so ad infinitum, as do the famous fleas which have other fleas to bite 'em. If one thinks of mathematical lines as individuals, do these not have parts, which are lines which have parts, etc.? But (1) a mathematical line is a set of points, and while sets have sub-sets which in their turn have sub-sets and so ad infinitum, it is doubtful whether sets are properly construed as individuals; and, in any case (2) there remain the points which, if they can be construed as individuals, and if they can be construed as 'parts', might be parts which have no parts, and, in this domain, basic individuals.

13. I mention, this, however, only to remind you of the dubious analogies which metaphysicians have often drawn between physical objects and mathematical entities. For my present purposes I shall simply stipulate that physical objects have ultimate parts. This dogma will shortly become more palatable, I hope, when I explain my use of the term 'physical object'.

14. To this I must add that it is of vital importance to distinguish between actual and potential parts. The distinction may be presented thus:

Consider the surface of this unmarked blackboard, taken at its face value as a continuous black expanse. Construe this surface as it would be construed by one who has no inkling of micro-physical theory nor of the puzzles which can lead philosophers to hold that such surfaces are 'subjective entities' which 'exist in the mind'. I draw a circle thus. Now I draw a diameter of this circle. Where before there was one undivided but divisible individual, the circle, there are now two individuals, the semi-circles.

15. Thus an object O which has no actual parts may be divisible, and, when divided, be superceded by, or, in a unique sense, become, two individuals O1 and O2. These new individuals may well be qualitatively quite different from the original object, and the composite which consists of them may be qualitatively quite different from the original undivided object. And this not only in the trivial sense in which a semi-circle is not a circle, but in the more puzzling sense in which a living thing, construed in Aristotelian style as non-composite, becomes, dismembered, a non-living whole of non-living parts.

16. I distinguished above between basic individuals and reducible individuals. I think that the concept of a basic individual serves as a reasonably good explication of the traditional concept of substance. But the above account of reducible individuals is much too restricted. For I have taken as my paradigm of a reducible individual an individual consisting of actual parts, presumably spatial. This restriction must be removed. To begin with, we must allow for temporal parts. I do not mean to imply that every physical object, for example, whether spatially composite or not, has temporal parts, for, at least as I am using the term 'physical object', this is false. Rabbits do not consist of rabbit-stages nor of rabbit-events, though the life history of a rabbit does consist of rabbit-events, and a revisionary metaphysics can doubtless be contrived in which doing things consists of temporally juxtaposing slabs of becoming (see [7]). I simply want to allow for such cases as that, for example, of a regiment which at different times has different soldiers as its parts.

17. In the second place, not every reducible individual is, in any ordinary sense, a whole of parts. Thus the average man is a reducible individual in the sense that statements about the average man can be paraphrased in a way which replaces reference to the average man by a general reference to particular men. Again the elephant is a reducible individual in that statements about the elephant can be paraphrased in a way which replaces reference to the elephant by a general reference to particular elephants. It is in this sense, also, that conjunctive individuals are, perhaps, reducible. Thus, although surface grammar obscures the fact, in:

Jack and Jill and Tommy are (constitute) a family,
the expression 'Jack and Jill and Tommy' functions as a singular term for the conjunctive individual Jack and Jill and Tommy, which the statement characterizes as a family. This example should be carefully distinguished from:
Jack and Jill and Tommy are human,
which is short for a conjunction of three sentences sharing the same predicate. A statement about conjunctive individuals may be paraphrasable into a conjunction of statements, but not of this simple form.

18. I said above that conjunctive individuals may be reducible, for it turns out that, unlikely though it may seem, they pose one of the central problems in the metaphysics of persons and sentient beings generally. For to say that conjunctive individuals are reducible is to say that statements about them can be paraphrased in ways which refer only to their 'constituents' (i.e., conjuncts). Thus, presumably,

Jack and Jill and Tommy are a family
can be paraphrased by a conjunction of statements which do not have conjunctive subjects, e.g.,
Jack is a man
Jack is married to Jill
Jill gave birth to Tommy

19. When it is said that some wholes have attributes which do not consist in their parts having such and such qualities and standing in such and such relations, it is, in effect, being denied that all conjunctive individuals are reducible.

20. It might be thought that by speaking of wholes and parts, rather than of conjunctive individuals and their conjuncts, I have changed the subject. But this is not the case, for reference to a 'whole' or a 'composite' is simply reference to a conjunctive individual the elements of which are presupposed to satisfy certain qualitative and relational conditions. Thus, a regiment is a conjunctive individual which consists of soldiers who stand in certain relations to one another which constitute a military peck order. But more of this later.


21. Well, given some such distinction between basic individuals (or substances) and reducible individuals, what shall we include in the former category? For the most part, I shall commit myself as I go along. But I shall begin by laying down that some physical objects are basic individuals, as are such quasi-physical objects as noises and flashes of lightning. More paradoxically, I shall also stipulate that persons are basic individuals.

22. What of scientific objects, the individuals postulated by micro-physical theory? Since I am usually classified as a Scientific Realist, it might be thought that in stipulating above that some physical objects are basic individuals, I was tacitly taking these basic individuals to be micro-physical particles. If so, the above claim that persons are basic individuals must have come as a surprise, for are not micro-physical particles actual parts of persons -- at least if persons are not to be equated with Cartesian minds? But though I am indeed a Scientific Realist, and think that the domain of basic individuals consists of those which scientific theory will 'in the long run' find it necessary to postulate, I also regard the conceptual framework in terms of which man experienced himself and the world long before the revolution in physics was even a twinkle in the eye of Democritus, as a coherent, delicately articulated whole which it is necessary to understand in its own terms before one can be in a position to determine the precise sense in which it, or part of it, is 'replaceable' by the world-picture presented by theoretical science.

23. Thus, for methodological reasons, I shall (to borrow Husserl's useful term) 'bracket' the theoretical picture of the world and concern myself with explicating what I have called the Manifest Image. (See [8]).

24. In the Manifest Image, physical objects have perceptible qualities -- roughly the proper sensibles and common sensibles of Aristotle. A dispositional property is, so to speak, an 'iffy' property; thus, water-solubility is the property of dissolving if placed in water. Note that for a property to be occurrent isn't just for it to be one that occurs to objects. The property of being magnetized occurs to soft iron when placed in a helix through which a current begins to pass. Nevertheless, the property of being magnetized, which occurs to the iron in these circumstances, is in our sense an 'iffy' property and is to be explicated, in first approximation, by means of the hypothetical: if an iron filing is placed in the vicinity, then, ceteris paribus, it moves to the core. Whereas an occurrent property in our technical sense is one that is not to be explicated by a hypothetical.

25. Thus, consider, to use a favorite example of mine, a pink ice cube. Many are tempted to identify its pinkness with a causal property, the property of causing normal observers in standard conditions to have sensations of pink, indeed, of a pink cube. Now there may be a place for some such move when the scientific revolution is taken into account. But it is a revisionary proposal, and it is, in my opinion, a sheer mistake to think of it as a correct analysis of the common-sense notions of color which function in 'basic' perceptual experiences. Different conceptual strata can, and indeed do, co-exist in our ordinary experience of the world. But this co-existence, peaceful though it is (at least until philosophical issues are pressed), must not be confused with compatability in any deeper sense.

26. Only a theory-intoxicated philosopher can look at a pink ice cube in daylight and suppose that to see it to be pink is to see it to have the power to cause normal observers to have sensations of pink when they look at it in daylight. And it is at least as absurd -- if not quite the same absurdity -- to suppose that to see it to be pink is to see it to look pink to normal observers in daylight, even though it is a conceptual truth that pink things look pink to normal observers in standard conditions which, until we become dwellers in modern caves, include daylight.

27. It should be noted that if physical objects are genuine individuals, they can scarcely have only iffy properties (powers, propensities, dispositional properties, and the like). They must have some non-dispositional or occurrent attributes. Nor, as Whitehead reminds us, will it do to limit their occurrent attributes to such 'primary' qualities as shape and size; for, to use an Aristotelian turn of phrase, geometrical qualities are 'formal' qualities and presuppose a 'content' or 'matter', thus color. Things which had 'primary' qualities without 'content' qualities would have "vacuous actuality." That Whitehead construed the occurrent content of physical objects in terms of feeling, rather than color, is a symptom of the revisionary character of his metaphysics.


28. Let me propose, then, as my paradigm of a physical object, a pink ice cube. It is cold, smooth, transparently pink, and cubical. In addition to these (and other) occurrent attributes, it has, of course, many properties of the iffy type illustrated above -- which it is also convenient to call 'causal properties'. Let us now bring into the picture a person who sees it.

29. In the Manifest Image, a person is a basic individual. (It should be clear that I regard Aristotle as the philosopher of the Manifest Image.) That which distinguishes man from merely material things and from brutes is his ability to think. Now the word 'think' is used in a number of distinguishable, but related, senses. Thus, it has a dispositional sense in which it is closely related to 'believe'. (What does he think about the war in Vietnam?) Thinking is often a deliberate action, as in thinking about (i.e., attempting to solve) a problem. Again, there is a sense of 'thought' in which thoughts just occur to one. We say "it suddenly occurred to me| that . . .," and, we can often add, "for no reason."

30. The importance of all this is that whereas we often contrast perception with thinking, there is, nevertheless, a proper sense in which perceiving essentially is or involves a thinking. Roughly, seeing this to be a pink ice cube involves a thinking this to be a pink ice cube.

31. I propose that we take very seriously the view that a thought, in the sense in which thoughts occur to one, is the occurrence in the mind of sentences in the language of "inner speech," or, as I shall call it, 'Mentalese'. These sentence events, as I have pointed out elsewhere, are not to be confused with verbal imagery. Although we are able to monitor, to some extent, our thoughts and to have direct non-inferential knowledge of them, this knowledge must not be construed on the perceptual model.


32. Before continuing, I must qualify the above remarks lest the animal lovers among us take them as libel and calumny. I count myself in their ranks and therefore hasten to add that of course there is a legitimate sense in which animals can be said to think and hence to be able, in something like the above sense, to see a pink ice cube and to see that it is pink. Furthermore, the point is important in its own right and not simply a rhetorical maneuver. For if one ties thinking too closely to language, the acquisition of linguistic skills by children becomes puzzling in ways which generate talk about 'innate grammatical theories'.

33. Not all 'organized behavior' is built on linguistic structures. The most that can be claimed is that what might be called 'conceptual thinking' is essentially tied to language, and that, for obvious reasons, the central or core concept of what thinking is pertains to conceptual thinking. Thus, our common-sense understanding of what sub-conceptual thinking -- e.g., that of babies and animals -- consists in, involves viewing them as engaged in 'rudimentary' forms of conceptual thinking. We interpret their behavior using conceptual thinking as a model but qualify this model in ad hoc and unsystematic ways which really amount to the introduction of a new notion which is nevertheless labeled 'thinking'. Such analogical extensions of concepts, when supported by experience, are by no means illegitimate. Indeed, it is essential to science. It is only when the negative analogies are overlooked that the danger of serious confusion and misunderstanding arises.

34. One must also take into account the sensibilities not only of animal lovers but of artists: painters, musicians, and the like. These are, of course, 'thinking beings.' But do they 'think' in the course of their distinctive creative activity?

35. One is tempted to say that the musician not only thinks about sound, but also 'in sound'. Thinking about sound, it might be admitted, can be construed on a linguistic model -- as 'inner speech' using the vocabulary of auditory qualities and relations. Indeed, much of the thinking that a composer does is conceptual thinking about the relationships of sound patterns, and since the notion of conceptual thinking as analogous to language leaves open the question of how precise the analogy is, it is surely not too far-fetched to take a linguistic approach to this aspect of the composer's activity.

36. But is there not also a more intimate relationship between composition (and other forms of musicianship) and sound: the aspect I referred to above as 'thinking in sound'? And is not this aspect also a mode of thought? Is not a performer a 'thinking being' even while he performs? The 'linguistic model' begins to look far too narrow and specialized to 'capture the nature of thinking, even at the strictly human level -- let alone in the sense in which animals think.

37. There is much food for reflection in these considerations. They raise questions which must be taken into account by any serious philosophy of mind. But the fundamental problems which they pose arise already at the perceptual level. For, as we shall see, visual perception itself is not just a conceptualizing of colored objects within visual range -- a 'thinking about' colored objects in a certain context -- but, in a sense most difficult to analyze, a thinking in color about colored objects. (The point stands out most clearly with respect to visual imagination; for imagination at its best is not simply a thinking about things which are not at hand and, more generally, thinking which is not concerned with what realty exists, i.e., thinking that takes the form of make-believe.) It is suffused with visual imagery and can also be said to be a 'thinking in color'.

38. What I am suggesting is that a correct theory of visual experience will throw light on the nature of visual imagination -- and this, in turn, on the unique relationship between thinking and imagery which characterizes creativity in music and the other arts.


39. Overt linguistic occurrences obviously involve a 'vehicle of meaning', the relevant matter-of-factual characteristics of speech which are idealized in the semanticist's concept of a 'sign design'. Once we abandon the naive idea that the vehicle of meaning in the case of inner speech is imagery, we are left with the question: What is it? Or is there, perhaps, no vehicle of meaning? In that case, the analogy between thought and speech would indeed limp.

40. Philosophers have often been tempted, in the case of overt speech, to reify the distinction between the words which are uttered and the 'meanings' which they convey. Clearly, something like this distinction must be drawn, though the crucial question would be just what sort of 'things' 'meanings' are. I have dealt with this issue in a number of places (most recently in [9], Ch. III) and will touch upon it lightly in a subsequent lecture, but for the present purposes it is sufficient to note that the analogy we are drawing between 'inner' and 'outer' speech requires that parallel questions be at least asked about inner speech. Must we not distinguish between sentence events in Mentalese and the meanings, they bear? We might be inclined to say that thoughts are not, so to speak, externally related to their meanings, as are sentence events in overt speech, but rather that they embody, in a unique way, the senses of the words in which they find expression. If so, a thinking that Nixon is President would be a mental event which embodies, in a unique way, the sense of 'Nixon is President'; or it is, perhaps, in a unique way, an event of the Nixon is President kind. Although this might seem to be a desperate attempt to preserve the analogical theory, the latter has sufficient merit to warrant strong measures. And although the formulations seem empty, it turns out, I shall argue, that they can be given empirical content.

41. I said a moment ago that seeing this to be a pink ice cube involves thinking this to be a pink ice cube. In the above terms, this means that seeing this to be a pink ice cube involves the occurrence of something like the Mentalese sentence, "This, over there, is a pink ice cube." Yet this can scarcely be all, for, we are inclined to expostulate, surely there is all the difference in the world between seeing something to be a pink ice cube and merely thinking something to be a pink ice cube.

42. Even if we take into account the fact that perception involves a causal dimension (in that, given our ability to think of something as a pink ice cube, and given that we are not blind, and given that the circumstances are propitious (daylight, unobstructed view, etc., etc.), the pink ice cube is the cause (in a relevant sense) of the thought that there is a pink ice cube over there), we feel that the distinctive character of seeing has not yet been captured.

43. Consider another example. I see that there is a red book in the corner. This time, since the book is not transparent, I do not see the other side of the book. Yet clearly I think of it, in the sense that in thinking of the object as a book, I think of it as something which necessarily has an opposite side, and that in thinking of the book as red, I am thinking not only of this facing side but also of the opposite side as red. Thus, we are tempted to say, most of the book is present to me as something merely thought of.2 Now I am, in the circumstances, caused to think of the book as red and, hence, not only of the facing side but of the opposite side as red. The facing surface, on the other hand, seems to be more intimately involved. Is it merely that the facing surface is the proper cause of the whole experience, so that insofar as I am thinking of the facing surface as red, my thinking corresponds to the proper cause? Is this what makes the seeing of the facing surface more than a mere thinking of the facing surface as red? Surely not.

2. Actually, this is the place to recognize an important role played by the imagination in perception. Must we not say that the other side of the book is imagined, rather than merely yhought of? Fair enough! Yet, as I have already indicated, the problems involved in distinguishing between seeing and merely thinking recur when one attempts to distinguish between imagining the opposite side to be red and merely thinking of the opposite side as red. If the former distinction requires the postulation of sense impressions. the latter requires the postulation of images. Thus, the heart of our problem concerns the facing side and is the problem of understanding the difference between merely thinking of the facing side as red and seeing it to be red.

44. Perhaps what we should do is recognize that the propositional act of thinking, is of a unique kind, a 'visual' thinking. This could be meant in two ways: (1) It could be claimed that the propositional act involves a unique concept which reflects in the very content of the act the element of passivity by virtue of which it is a response to visual (auditory, etc.) stimulation. But even if we grant that the thoughts involved in perception have a distinctive content, as I think they do, it is difficult to see how the addition of another conceptual item can account for the difference between seeing and merely thinking of.

45. A more promising line which might be taken is (2) that over and above its propositional character as the occurrence of a mental sentence, the thinking has an additional character by virtue of which it is a seeing as contrasted with a mere thinking. Although, as we shall see, this move is not incorrect, it essentially labels, rather than clarifies, the problem.

46. Notice, to begin with, that if we suppose this additional character to be that of being a seeing, it runs into the objection that the same difference between a perceptual experience and a mere thinking is present in cases where the experience is not a seeing, for example, that there is a red book over there, for the simple reason that there is no red book over there.

47. Superficially, at least, this objection can be overcome by referring to this additional character as that of being, not a seeing, but an ostensible seeing -- a matter of something's looking or (visually) appearing to be the case. An ostensible seeing is, roughly, an experience which would be a seeing if it were veridical, just as an ostensible memory is an experience which would be a remembering, if it were veridical. Thus, since our problem concerns that which distinguishes seeings and ostensible seeings alike from mere thinkings, it amounts to the problem 'What distinguishes ostensible seeings or lookings from mere thinkings?' And to answer 'the character of being an ostensible seeing' is scarcely illuminating.

48. Nevertheless, the view which we are considering is on the right track, insofar as it recognizes that the character of being an ostensible seeing or looking or (visual) appearing is a character which belongs to experiences which essentially have a prepositional component, involving, as they do, the occurrence in the mind of a Mentalese sentence. On the other hand, to ascribe, as is sometimes done, the character of being an 'appearing' or ostensible seeing to the propositional act alone, as though it were an 'intrinsic' character of the act, must be a mistake. Surely the propositional act itself is an ostensible seeing or looking only in the derivative sense that it is the propositional constituent of a total experience which is an ostensible seeing or looking.

49. On the other hand, it is equally mistaken to ascribe the character of being an ostensible seeing to the non-propositional element in the experience. For the expressions which the contexts '. . . ostensibly sees . . .', '. . .looks. . .', '. . .appears. . .', require, to complete them, are propositionnal in character:

Jones (ostensibly) sees there to be a red book in the corner.
There appears to Jones to be a red book in the corner.

50. Thus, if there is a non-propositional component in the total experience, it would be incorrect to refer to it by such terms as 'looks', 'appears', etc., unless these latter are given a new and technical usage. And one who does so would first have to make clear that there is such a non-propositional component and give some account of what it is.

51. Now I think it clear that, phenomenologically speaking, there is such a non-propositional component. But I also think that in the absence of what amounts to relatively sophisticated theory construction, it can only be characterized in a way which raises more problems than it solves.

52 Roderick Chisholm, in his various presentations of his views on the sensible appearances of things, seems to me to race over these distinctions. The phenomenological appeal is made, but since the language of 'looks', 'seems', 'appears', is used to characterize the discriminated items, the implication that they are propositional states is never explicitly discounted, though it is clear that he thinks of his looks and appearings as non-conceptual stages. But the failure to make an explicit distinction between the appearings which are propositional states and the appearings in the surreptitiously introduced new sense, in which they are not propositional states, bestows upon the latter an unearned non-problematic character with respect to the manner in which they involve the proper and common sensibles.

53. Chisholm correctly sees that the primary use of 'appears' is non-comparative. For in the comparative use we say, e.g., "This appears to be what white objects appear to be in such and such circumstances of perception." And while the whole sentence compares one appearing with others, not every sentence involving 'appears' can do so, for obvious reasons. There must be sentences to the effect that this appears, to be thus and so period, e.g., white. But one can grant that the primary use of 'appears' is non-comparative while arguing, as I have done, that the primary use of 'appears' applies to conceptual but not merely conceptual, episodes. In other words, one can grant the primacy of the non-comparative sense of 'appears', 'looks', etc. without applying these terms to the non-propositional component of the total perceptual experience.

54. I have argued in "Empiricism and the Philosophy of Mind" ([5]) that the non-propositional feature common to cases where

One sees that the object over there is red and triangular on the facing side;
The object over there looks to be red and triangular on the facing side;
There looks to one to be an object which and triangular on the facing side in front of one,
is primarily identified simply as this common non-propositional feature. I called it the descriptive (i.e.. non-propositional) core.

55. So far we would be little better off than if we simply said that ostensibly seeing that there is in front of one an object which is red and triangular on the facing side differs from merely thinking that there is an object in front of one which is red and triangular on the facing side, by virtue of being a thinking which is also an ostensible seeing. But we can say more. For, phenomenologically speaking, the descriptive core consists in the fact that something in some way red and triangular is in some way present to the perceiver other than as thought of.

56. The multiple indefiniteness in this description is disconcerting and makes it clear that the concept is a problematic one, in the sense of posing problems. But then I have argued in a number of places (for example, [4]) that the 'common-sense' picture of the world, in spite of its delicate coherence, is such as to pose problems which it lacks the resources to resolve. On the other hand, the above account of the non-propositional (descriptive) core is definite in its negative aspect. The mode of presence is not that of being thought of.

57. A scholastic might say that in perception (and ostensible perception) the relevant proper and common sensibles have being for sense as well as being for thought. Thus, when I see or ostensibly see something to be a pink ice cube, a pink cube has not only being for thought but also being for sense. The somehow presence of the pink cube could then be referred to as its being sensed. (The problematic nature of such 'sensing' should not be forgotten.)


58. What are the boundary conditions on such a theory of the 'descriptive core' of perceptual experience? The following strike me as relevant considerations:

(1) If we are to work within the framework of the Manifest Image, we must stipulate that the proper and common sensibles involved are to be construed as qualities of physical objects (noises, flashes, etc.). We are nevertheless free to introduce new theoretical uses of perceptual predicates in which the latter apply to items which are not, in the ordinary sense, physical objects -- though they may properly be said to be physical in an extended sense of the term.

(2) We are looking for characteristics which actually characterize the descriptive core. Thus, we must beware of using metaphors which carry with them the implications of being for thought, existence as thought-of, intentional in-existence. For if the pink and cubical item involved in hallucinating a pink ice cube had itself merely thought-of existence, it could not do the job the theory requires. This danger is present, as we saw, in the term 'being for sense'.

59. Classical sense-datum theory, construed as a postulational theory rather than as a phenomenological description, would satisfy these demands, provided:

(1) Sense data are not introduced as objects of perceptual knowledge. The objects of perceptual knowledge are the objects referred to in the propositional component of the perceptual experience, and these are physical objects, not private, subjective (let alone theoretical) items. For this reason, the term 'sensa' would be more suitable than 'sense data'.

(2) Sensa are not to be taken to have 'sensible qualities' in the primary sense of this phrase in which, for example, the pinkness of physical objects is primary. This, however, permits sensa to have theoretical counterparts of these qualities. Accordingly, the somehow presence of an item which is somehow physically pink and cubical would be explained in terms of the somehow presence (i.e., the being sensed) of an item which is pink and cubical in an analogical sense.3

3. This is at first sight disturbing, in that what we wanted was something physically pink and cubical. Surely, it might be said, to offer something which actually has a non-physical pinkness and cubicity, even though these are the theoretical counterparts of physical pinkness and cubicity, is to give us a stone instead of bread. There is a bite to this objection, but the full answer involves a general account of how the impact of the scientific image of the world is to be correctly interpreted. Briefly, it turns out that the color concepts which are posterior in the (historical) order of conception -- i.e.. color as color of sensa -- are prior in the order of being. Roughly, color is 'really' an attribute of sensory states of the perceiver, when these in their turn have been transposed into an adequate neuro-physiology of perception.

60. But our options are not restricted to such a revised form of sense-datum theory. Sensum theory construes sensing as a relation between a person and an item which is pink and cubical in an analogical sense. But once we realize that what is involved are theoretical counterparts of the proper and common sensible characteristics of physical objects, we see that the way is open to construe these items on a quite different model from the act-object model of sense-datum theory.

61. Thus, instead of saying that the non-propositional presence of a pink cube in an ostensible seeing is a matter of a relation of sensing between the perceiver and an item which is, in the indicated derivative sense, a pink cube, we can take the quite different tack of construing the object of sensing (a pink cube) as a manner of sensing. Sensing a pink cube would be interpreted as sensing a-pink-cube-ly.

62. This would be a cousin of what is known as the 'adverbial' theory of sensing. It differs, however, in two important respects.

(1) The usual adverbial theory would analyze our example in terms of sensing pinkly: pink, as a feature of the non-propositional content of the ostensible seeing of a pink ice cube, being interpreted as a manner of sensing.
But what was to be explained was the fact that the ostensible seeing presents us, in some way, not just with 'pinkness' but with a pink cube, i.e., something pink and cubical. Thus, to do its job, the adverbial theory would have to construe not 'pink' but 'a pink item' as the relevant adverb. But this is not all.
(2) Taking the preceding point fully into account, the adverb would have to be 'a pink cube' i.e., one would sense a-pink-cube-ly. Once we appreciate this we understand why the traditional adverbial theory limits itself to 'senses pinkly'; for it takes the sensing context to present the primary mode of being of colors, e.g., pink. The esse of blue, for example, is sensing bluely. Blue is a manner of sensing. But the philosophers who propound it would scarcely be happy to make the corresponding move in the case of shape. Does it make sense to say that the primary mode of being of triangularity, or, in our example, cubicity, is as a manner of sensing?

63. Thus, the virtue of the form of the adverbial theory which I am proposing is that it requires that the adverb be a phrase of the form 'a pink item' rather than 'pink', and, above all, it recognizes that the 'adverb' must take into account the primary qualities of perceptible things as well as the secondary. (The act-object or sensum theory has this advantage over the usual adverbial theory in that it can attribute sensa not only color but shape.) Once we realize that the predicates of the theory of sensing are theoretical predicates, which are introduced in terms of analogy with predicates standing for the proper and common sensible qualities of physical objects, we see that there is no reason why the introduced theoretical predicates should not have a quite different conceptual grammar.

64. According to this version of the adverbial theory of sensing, then, sensing a-pink-cube-ly is sensing in a way which is normally brought about by the physical presence to the senses of a pink and cubical physical object, but which can also be brought about in abnormal circumstances by objects that are neither pink nor cubical, and, finally, according to this form of the adverbial theory, the manners of sensing are analogous to the common and proper sensibles in that they have a common conceptual structure.  Thus, the color manners of sensing form a family of incompatibles, where the incompatibilities involved are to be understood in terms of the incompatibilities involved in the family of ordinary physical color attributes. And, correspondingly, the shape manners of sensing would exhibit, as do physical shapes, the abstract structure of a pure geometrical system.

65. The account which I have offered is obviously but a beginning, for it leaves almost untouched the intimate relation which exists between these two aspects of visual perception. Furthermore, it leaves completely untouched the relationship between each of these aspects and the neurophysiological processes which are at least as intimately related to them as they are to each other.

66. It is some progress, however, to have a sense of what remains to be done. As I once pointed out, the evolution of philosophy is as much the evolution of philosophical questions as it is of answers to pre-existing questions. In this sense, at least, I hope to have advanced the subject this afternoon.


1. One of the useful ways of emphasizing some of the points which I have been making in this lecture is to discuss an argument of Roderick Firth ([3]) to show that it makes good sense to suppose that physical redness can be defined in terms of 'looks red'.

2. It should be clear that according to the analysis which I have given, any such attempt is doomed to failure from the start -- if 'looks' is taken in its ordinary sense. For 'looks' applies to experiences which contain the thought 'such and such a physical object is red', and, if my analysis is correct, it becomes a truism that in this sense of 'looks' the concept of being red is logically prior to that of looking red.

3. Thus, if Firth's attempt is to get off the ground, he must be claiming that the non-propositional element in perceptual experience, for which he is borrowing the term 'looks', is itself red in a well-defined sense which is both other than physical redness and does not presuppose it. But, according to my analysis, the only well-defined sense of 'red' other than physical redness which is available is a theoretical construct which has been introduced on the analogy of physical redness and, therefore, does conceptually presuppose it.

4. Waiving this point, and granting Firth his well-defined, non-propositional sense of 'looks red and triangular', we can construe him as claiming that

O looks red and triangular to me
has something like the sense of
O is the cause of a red and triangular element in my perceptual experience.
There looks to me to be a red and triangular object in front of me
would have the sense of
There is a red and triangular element in my perceptual experience.

5. Now it should be clear that according to the view which I have been defending in this essay, one is not in a position to be perceptually aware of any fact, however minimal, unless one has a whole system of concepts which constitutes a Mentalese language of physical objects in Space and Time. But Firth, following Lewis, is willing to concede that "the vocabulary of 'looks' and 'seems' expressions that we use to describe sense experience is in some sense derivative from the vocabulary we use to describe the physical world" ([3]: 548). On the other hand, appealing to an apparently straightforward distinction between words and concepts, Firth insists that there is a domain of concepts pertaining to sensible qualities which are logically independent of concepts pertaining to physical objects. And, in the spirit of traditional empiricism, he finds the source of these concepts to be what I have called the descriptive or non-propositional core of the perceptual experience of physical objects. Thus he writes:

. . . if a philosopher maintains that 'The apple is red' can be analyzed as meaning "The apple would look red under such and such physical conditions," he is assuming that "looks red" is logically prior to "is red," i.e., that it is at least logically possible to have the concept "looks red" before we acquire the concept "is red." But if the coherence theory of concepts is correct and we cannot fully understand "looks red" unless we possess the contrasting concept "is red," then it would seem that it is not logically possible to have the concept "looks red" before we have the concept "is red." This paradox might even lead us to wonder, indeed, whether the conceptual interdependence of "looks" and "is" is enough to undermine Lewis' basic assumption that we can make "expressive judgments" (e.g., "I seem to see a door knob," "It looks as if I am seeing something red") without at the same time asserting something about the nature of "objective reality." It is these expressive judgments, according to Lewis, that enable us to escape the coherence theory of justification; and if it should turn out that these judgments all make some covert reference to physical objects, then -- depending, of course, on the kind of "covert reference" -- it might no longer be possible to make the epistemological distinction that Lewis requires. ([3]: 547.)

6. In the above passage, Firth is clearly confusing the proper sense of 'looks', in which it contrasts with 'is seen to be', with the contrived sense explored in the opening paragraphs of the Appendix, in which 'looks red' contrasts with 'is red' and means something like 'causes a red item in my visual experience'.

7. But, as already noted, this is not the point I wish to press. For even more significant is the fact that the next step in his argument is simply invalid and reveals an over-simplified conception of the relation between the meaning of word and the circumstances in which it is used. He writes,

It is a genetic fact, but a fact with philosophical implications, that when a child first begins to use the word 'red' with any consistency he applies it to things that look red to him whether these things are, as we should say, "really red," or whether they are merely made to appear red by abnormal conditions of observation. Thus the child calls white things "red" when he sees them through red glass. In fact at this stage the child says 'red' just in those circumstances in which we, as adults, could truthfully say ''looks red to me now," so that it would not be unreasonable to assert that the child is using 'red' to express a primitive form of the concept "looks red." ([3]: 547.)
He concedes that "to call this a 'primitive form' of the concept 'looks red' is to acknowledge that in some sense the child cannot fully understand adult usage until he is able to distinguish things that merely look red from things that really are red," but insists that "we must not suppose that the child somehow loses his primitive concept when he acquires a more sophisticated one." He then concludes his argument with the statement that "if we do not confuse baptismal rules with semantical rules (e.g., the semantical rule followed by the child who says 'red' when something looks red to him), the coherence theory of concepts does not seem to be incompatible with Lewis' theories of meaning and knowledge."

8. The absurdity of Firth's argument is best brought out by constructing a parallel argument which has as its conclusion,

. . . in fact, at this stage the child says 'red' just in those circumstances in which we, as adults, could truthfully say 'electromagnetic waves of wave length λ are striking his retina', so that it would not be unreasonable to assert that the child is using 'red' to express a primitive form of the concept 'electro-magnetic waves of wave length λ striking a retina'.



1. In my first lecture, I was exploring the nature of our perceptual knowledge of such elementary facts as that there is a pink ice cube in front of one or that there is a red book on the shelf. I emphasized that I was 'bracketing', that is, suspending commitment to, the stratum of concepts introduced by micro-physical theory and considering perception as it might have been considered by an epistemologist who lived in the days when atomic theory was but a gleam in the Democritean eye. In short, the model with which I was working was essentially an Aristotelian one, though I was not concerned with problems of historical exegesis.

2. I was emphasizing that in this model material things are colored in a sense which is not to be explicated in terms of a hypothetical reference to sensations of color. I asked you to contemplate a pink ice cube and urged the implausibility of the suggestion that to see it to be pink is to see it to have the power to cause normal observers in standard conditions to have sensations of pink or to sense pinkly. Indeed, I argued that the concept of the sensation of the pink cube is a theoretical concept, an element in a theoretical explanation, not only of how people can seem to see pink ice cubes, when no pink transparent material object is before their eyes, but how the veridical perception itself differs from a purely conceptual awareness of the pink ice cube.

3. I concluded by suggesting that the most satisfactory form of this theoretical account makes it a version of the 'adverbial' theory of sensing. According to this version, the theoretical concepts of manners of sensing are formed by analogy with concepts pertaining to the common and proper sensible attributes of physical objects as conceived in the Manifest Image of the world. Thus, the color manners of sensing form a family of incompatibles, where the incompatibilities involved are to be understood in terms of the incompatibilities within the family of ordinary physical color attributes. And, correspondingly, the shape manners of sensing would exhibit, as do physical shapes, the abstract structure of a pure geometrical system.

4. Thus, sensing a pink cube is a manner of sensing which is conceived by analogy (a transcategorial analogy!) with a pink physical cube and which, though normally caused by the presence of a pink and cubical transparent object in front of a normal perceiver's eyes, can also be brought about in abnormal circumstances by, say, a grey object illuminated by pink light or by a pink rhomboidal object viewed through a distorting medium, or in hallucination by, for example, a probing of a certain region of the brain with an electrode, or by the taking of an hallucinogenic drug after much talk of pink ice cubes.

5. I distinguished between the propositional and the non-propositional components of the visual experience. I characterized the former as a thinking that something is the case, where the thinking is construed as the occurrence, in the mind, of a sentence in Mentalese or, to use the traditional term, 'inner speech'. I said relatively little about Mentalese, save to emphasize the positive analogy between it and overt verbal behavior. I concentrated on the non-propositional aspect of visual experiences and was concerned to show that unless supplemented by theory construction, the phenomenology of perception takes us no further than the idea that somehow, something which is in some sense pink and cubical is present to the perceiver when, rather than merely thinking that there is a pink ice cube in a certain place, he sees, or seems to see, a pink ice cube in that place.

6. In this lecture I want to explore the topic of thinking as 'inner speech' (Mentalese) and lay the groundwork for discussion of the implications of the scientific revolution for this concept. Unless one takes a purely instrumental view of scientific objects, both sensing and thinking must be correctly located in a context of neurophysiological activity. The traditional mind-body problem has two dimensions which have often been run together -- or, at least, not carefully distinguished: (1) What is the relation of sensations to physical states of the body? (2) What is the relation of conceptual states (thinkings, inner speech) to physical states of the body? It should not be assumed that these two dimensions of the mind-body problem admit of the same type of solution.


7. How is the postulated analogy between Mentalese and overt linguistic behavior to be understood? To begin with, we must simplify our own model by abstracting from those features of language by virtue of which it is an instrument for influencing people. As Austin has emphasized, we can do things with words. We can inform or misinform, we can communicate our beliefs, we can make promises, etc., etc. Illocutionary and perlocutionary acts are actions. Like all actions, they are sometimes deliberate, sometimes unintended, sometimes thoughtless.

8. But though I am going to abstract from these features of linguistic behavior, I am not going to remove all reference to linguistic action from the positive analogy, for some thought processes are actions -- that is, are the sort of things that can be undertaken, that one can decide to do -- and, consequently, a place must be left in our model for linguistic actions proper. Roughly, I am excluding only those linguistic actions which are other-oriented and involve language as a means of communicating with, making commitments to, and influencing our fellowmen.

9. The simplified model which I propose to work with can be called 'Verbal Behaviorism'. It is not intended as an adequate account of thinking; it is, indeed, radically oversimplified. But I believe that it will prove a useful tool which will help us understand some of the features of thinking, and of our awareness of ourselves as thinking beings, which have been a source of puzzlement since the very dawn of philosophy.

10. According to this model, 'thinking that-p', where this means 'having the thought occur to one that-p', has as its primary sense saying 'p' and a derivative sense in which it stands for a short-term proximate propensity to say 'p'. Propensities tend to be actualized (a logical point about the term). When they are not, we speak of them as, for example, 'blocked'. For the purposes of the VB model, the relevant inhibiting factor which keeps the saying that-p from being actualized is that of not being in a thinking-out-loud frame of mind. One might use the model of a general 'on-off' switch which gets into the 'wiring diagram' when the child learns to keep his thoughts to himself.

11. Notice that I am treating that-clauses as quoted expressions; thus;

The thought that 2 plus 2 = 4 occurred to Jones
becomes, in the Verbal Behaviorist model,
Jones said (or had a short-term proximate disposition to say) '2 plus 2 = 4'.
I shall shortly take into account the fact that non-English-speaking people are not precluded from thinking that 2 plus 2 = 4. But our initial model will be provincial.

12. Picking up some themes from the above discussion of linguistic action, it is essential to note that just as thinking that-p, in the sense of having the thought occur to one that-p, is not a mental performance, something that one does or could do voluntarily, so, in the VB model, saying 'p' is not to be construed as an illocutionary act. It is to be construed, as I have elsewhere put it, as candidly thinking-out-loud-that-p and is not to be confused with asserting (to someone) that-p, telling someone that-p, or any of the verbal performances so lovingly collected by Austin.

13. Of course, in any ordinary sense, saying 'p' is a performance, because the phrase permits the utterance to which it applies to be either a spontaneous-thinking-out-loud-that-p or a deliberate use of words to achieve a purpose. I, on the other hand, am using the expression 'S says p' in a contrived sense in which these options are closed and the utterance specifically construed as a spontaneous or candid thinking-out-loud.

14. We can imagine a child to learn a rudimentary language in terms of which he can perceive, draw inferences, and act. In doing so, he begins by uttering noises which sound like words and sentences and ends by uttering noises which are words and sentences. We might use quoted words to describe what he is doing at both stages, but in the earlier stage we are classifying his utterances as sounds and only by courtesy and anticipation as words.

15. Only when the child has got the hang of how the sounds function in the language can he be properly characterized as saying 'this is a book', or 'it is not raining', or 'lightning, so shortly thunder', or 'you spanked me, so you don't love me'.

16. To say what a person says, or, more generally, to say what a kind of utterance says, is to give a functional classification of the utterance. This functional classification involves a special (illustrating) use of the expressions classified, or of synonyms -- where allowance must be made for degrees of synonymy or likeness of meaning -- of these expressions with which the addressee is familiar.

17. Some functional relationships are purely intra-linguistic (syntactical) and are correlated with the formation and transformation rules. Others concern language as a response to sensory stimulation by environmental objects -- thus, candidly saying, or having the short-term propensity to say, 'lo! this table is red'. Still others concern the connection of practical thinking with behavior. All these dimensions of functioning recur at the metalinguistic level in the language in which we respond to verbal behavior, draw inferences about verbal behavior, and engage in practical thinking about verbal behavior -- i.e.. practical thinking-out-loud (or propensities to think-out-loud) about thinking-out-loud (or propensities to think-out-loud).

18. Thus, when we characterize a person's utterances by using a quotation, we are implying that the utterance is an instance of certain specific ways of functioning. For example, it would be absurd to say

Tom said (as contrasted with 'uttered the noises') 'it is not raining' but has no propensity to avoid saying 'it is raining and not raining'.

19. In particular, to characterize a person's utterance by quoting sentences containing logical words is to imply that the corresponding sounds function properly in the Verbal Behavior in question, and is to imply that the uniformities characteristic of these ways of functioning are present in his sayings and proximate dispositions to say.

20. The functioning which gives the utterances of one who has learned a language their meaning can exist merely at the level of uniformities, as in the case of the fledgling speaker. Those who train him, thus, his parents, think about these functionings and attempt to ensure that his verbal behavior exemplifies them. In this respect, the trainer operates not only at the level of the trainee, thinking thoughts about things, but also at that higher level which is thinking thoughts about the functions by virtue of which first-level language has the meaning it does. In traditional terms, the trainer knows the rules which govern the correct functioning of language. The language learner begins by conforming to these rules without grasping them himself.

21. Only subsequently does the language learner become a full-fledged member of the linguistic community who thinks thoughts (theoretical and practical) not only about non-linguistic items but about linguistic items, i.e., from the point of view of our simple VB model, about first-level thoughts. He has then developed from being the object of training and criticism by others to the stage at which he can train and criticize other language users, and even himself. Indeed, he has now reached the level at which he can formulate new and sophisticated standards in terms of which to reshape his language and develop new modes of thought.

22. Notice that on the VB model of thinking, we can distinguish clearly between the functional role of utterances and the phonemic description of the linguistic materials which embody or are the 'vehicles' of these functions.

23. It is a most significant fact that the classical conception of thought as 'inner speech' (Mentalese) draws no such clear distinction between the conceptual functions of the Mentalese symbols and the materials which serve as the vehicle of these functions. Yet, if the analogy between thinking, classically construed, and overt linguistic behavior is to be a reasonably positive one, the idea that there must be inner-linguistic vehicles (materials) would seem to be a reasonable one. It is often thought that imagery is the vehicle of Mentalese -- but there doesn't seem to be enough imagery to go round. And, indeed, the idea of imageless thought is by no means incoherent. What might the vehicle be?


24. To our VB model there are, in addition to the objections, considered yesterday, to any attempts to construe thinking on the model of language, three familiar objections which must be given more attention:

(1) Surely, it will be said, thinking that-p isn't just saying that-p -- even candidly saying that-p as you have characterized it. For thinking-out-loud that-p involves knowing the meaning of what one says, and surely this is no matter of producing sound!
Answer: There is all the difference in the world between parroting words and thinking-out-loud in terms of words. But the difference is not that the latter involves a non-linguistic "knowing the meaning" of what one utters; rather it is that the utterances which one makes cohere with, each other and with the context in which they occur in a way which is absent in mere parroting. Furthermore, the relevant sense of 'knowing the meaning of words' (which is a form of what Ryle has called knowing how) must be carefully distinguished from knowing the meaning of words in the sense of being able to talk about them as a lexicographer might -- thus, defining them. Mastery of the language involves the latter as well as the former ability. Indeed they are both forms of know how, but at different levels -- the one at the 'object language' level, the other at the 'metalanguage' level.

25. I turn now to a second objection:

(2) Surely, it will be objected, we are often thinking when we are not saying anything. Our thoughts succeed one another with lightning rapidity. How can this be reconciled with the VB model?
Answer: It must be remembered that according to VB, thinking that-p is saying or having a short-term proximate propensity to say 'p'. To grasp how rapidly short-term propensities can shift, one need only think of the electromagnet in a doorbell and consider how rapidly it acquires and loses the propensity to attract the clapper.

26. A third objection runs as follows:

(3) Thinking, as was pointed out above, does not seem to occur in words. We are often conscious that we are thinking, for example, about a certain problem, without 'any words going through our mind'.
Answer: Only a very naive person would think of the flammability of gasoline as a hidden or inner flame, or of the propensity of an electron to jump from one orbit to another as hidden jumping. Thus, the Verbal Behaviorist would point out that the short-term propensity to say "I've just missed the bus" should not be construed as an 'inner' or 'hidden' saying "I've just missed the bus." Thus, the Verbal Behaviorist believes himself in a position to account for the classical conception of thoughts as analogous to linguistic activity but as, nevertheless, involving no actual equivalents of the words "in the mind." He sees the classical theory as an attempt to blend into one coherent picture items belonging to the radically different categories of act and propensity.

27. Above all, the VB model makes it clear how we know about thoughts. For in their primary mode of being, thoughts are publicly observable episodes: people saying things.

28. Thus, we can know what we think, in the primary sense, by literally hearing ourselves think. When we hear ourselves say (in a candid frame of mind) "I've just missed my bus," we are literally hearing the thought occur to us that we have just missed the bus.

29. And in hearing this, we would be thinking the higher-order thought:

The thought has just occurred to me that I have missed my bus.
And this higher-order thought would be an auditory perceptual response to one's actual thinking-out-loud "I've just missed the bus."

30. But what of those cases in which we know what we think, although we are not thinking-out-loud? The answer clearly does not consist in pointing out that thoughts in this sense are propensities to think-out-loud and that the propensities of things can be known by induction.

31. We know, for example, that acid turns litmus paper red by observing the relevant sequence in a number of cases and drawing a general conclusion from these observations.

32. And in more complicated cases we can, for example, know that our dog has the propensity to go and get his leash, because he, for example, sees us start to put on our rubber boots. And, in the most relevant type of case, people who know us very well are often able to predict what we are likely to think, given what is happening to us in the circumstances in which they see us.

33. Yet in reply to all this, the expostulation is surely justified that our own knowledge of what we are thinking (when we are not thinking-out-loud) is not inductive, is not an inference from our overt behavior, in the circumstances in which we find ourselves, to the existence of a certain propensity.

34. Yet the fact that we are able to learn the propensities of others (and ourselves) by inductive inference is the first step in the answer. But before developing it, we must consider how the account of perceptual experience sketched yesterday afternoon appears in the light of VB.

35. Thus, consider the case where

Jones sees there to be a red apple in front of him.
According to the account offered in the first lecture, this experience contains as its conceptual core the thought (which Jones is caused to have)
Here is a red apple.

36. In terms of our VB model, this becomes:

Jones thinks-out-loud: Lo! Here is a red apple.

37. Now to say that this visual thinking-out-loud that something is the case is epistemically justified or reasonable or has authority is clearly not to say that Jones has correctly inferred from certain premises, which he has good reason to believe, that there is a red apple in front of him. For we are dealing with a paradigm case of non-inferential belief. The authority of the thinking accrues to it in quite a different way. It can be traced to the fact that Jones has learned how to use the relevant words in perceptual situations.

38. It is for this reason that, when Jones candidly says, in response to visual stimulation,

Lo! Here is a red apple,
it is likely to be true that there is a red apple in front of him. I say likely to be true, because we all know of various ways in which things can go wrong. Jones is in front of a mirror; the supposed apple is a piece of wax; the illumination is abnormal and the object is really purple; or, there is nothing in front of him, but he has taken LSD, and people have been pounding his ears about red apples.

39. If we were to overhear him, and if we had reason to believe that none of these countervailing situations obtain, we would be justified in reasoning as follows,

Jones has thought-out-loud 'Lo! Here is a red apple'
(no countervailing conditions obtain);
So, there is good reason to believe that there is a red apple in front of him.
Note that although this is an inferential justification of our belief that there is a red apple in front of Jones, it is a special kind of inference. It has the form:
The thought that-p occurs to Jones in a certain context and manner;
So, it is reasonable to believe that-p.
The same proposition, that-p, is mentioned in both the premise and the conclusion. But the first mention concerns the fact of its occurrence as a propositional event in a context to which the basic features of language learning are relevant. From this premise, the inference is drawn that the proposition in question is one which it is reasonable to believe.

40. We looked at the above example from the standpoint of an external observer. Let us now look at it from the standpoint of Jones himself. As we saw in the preceding lecture, to be fully a master of his language, Jones must know these same facts about what is involved in learning to use perceptual sentences in perceptual contexts. Thus, Jones too must know that other specifiable4 things being equal, the fact that a person says 'Lo! Here is a red apple' is good reason to believe that he is indeed in the presence of a reel apple. Thus, Jones too, can reason:

I just thought-out-loud 'Lo! Here is a red apple'
(no countervailing conditions obtain);
So, there is good reason to believe that there is a red apple in front of me.

4. Which is not to say that there are no cases in which we would not know what to say -- e.g., electrode hallucinations.

41. Of course, the conclusion of this reasoning is not the thinking involved in his original perceptual experience. Like all justification arguments, it is a higher-order thinking. He did not originally infer that there is a red apple in front of him. Now, however, he is inferring from the character and context of his experience that it is veridical and that there is good reason to believe that there is indeed a red apple in front of him.

42. I wish now to argue that VB not only throws light on non-inferential perceptual knowledge but also gives us a strategy for coping with the problems posed by the existence of non-inferential knowledge of what we are thinking when we are not (candidly) thinking out loud.

43. The first step in the argument consists in pointing out that part of the process of learning to use a language is learning to make autobiographical statements. And not only autobiographical statements in general, but autobiographical statements about what one is thinking.

44. Now in the case of perception, non-inferential knowledge on the VB model, as we just saw, involves reliably responding to physical objects in standard conditions with the appropriate perceptual sentences; thus:

Lo! There is a red book over there,
Lo! Here is a pink ice cube,
in response to a red book or a pink ice cube, where the reliability of the response is a function of the way in which language is learned.

45. Now the more we know about a person, the better we are able to judge what (in the circumstances as we see them to be) he would be likely to say (think-out-loud) -- if he were in a thinking-out-loud frame of mind. It is obviously difficult to be accurate about this, particularly when we are dealing with sophisticated minds. But even here the difficulty is one in practice rather than of principle. And when it is children in the initial stage of learning a language who are the subjects, the difficulty in practice is substantially less than it becomes subsequently when they have learned to lie, deceive, and conceal their thoughts.

46. Can we not, as children, be trained by those who know us intimately (our parents), and who therefore know (ceteris paribus) what our short-term verbal propensities are (i.e., what we are thinking), to respond reliably to our own short-term propensities to say that-p, as well as to respond to our actual sayings of 'p'?

47. And can not this ability be generalized in such a way that we can reliably respond to new propensities, i. e., to thoughts other than those in terms of which we have been trained? And would not the fact that such responses are reliable constitute the core of the explanation of non-inferential knowledge of what one is thinking (in the proximate propensity sense), as the existence of reliable verbal responses to perceptible things is the core of the explication of non-inferential perceptual knowledge?

48. Many have thought that to explicate the concept of non-inferential knowledge of 'what is going on in one's mind at the present moment', one must return to the Cartesian framework. From the point of view sketched in this lecture, the essential feature of the latter framework is that it denies that 'thinking-out-loud' makes sense save as analyzable into 'thinking occurrent non-verbal thoughts and giving them expression in one's verbal behavior'. In short, the Cartesian argues that the concept of thinking-out-loud includes the concept of thoughts as Cartesian inner episodes. According to the VB position, on the other hand, the concept of thinking-out-loud stands on its own feet as the primary concept pertaining to thought, so that if a concept like the Cartesian concept of thought episodes which are not propensities to think-out-loud does turn out to be needed in giving a full account of "what thinking is," those who are inclined to accept something like the VB position would argue that it is a concept built on a VB foundation and is in some sense a derivative or secondary concept. As a useful parallel, consider the case of micro-physics. Macro-objects, we say, if we are scientific realists, are really systems of micro-physical particles. Yet our concepts of these particles are built -- not on direct observation -- but on a foundation of knowledge at the perceptual level. In short, though VB argues that even if there is a sense in which Cartesian thoughts are prior in the order of being to thinking-out-loud, the latter is prior in the order of knowing.


49. There are many delicate issues in this neighborhood, some of which can be left to slumber. Nevertheless, I introduced VB as a simple model, and while I have been polishing and defending it, it was introduced in order to be transcended. It correctly represents a basic stratum in our conception of what thinking is, but it is only a part of a larger picture, to which I now turn.

50. In the case of the dispositions and propensities of material things, we distinguish between the propensities and dispositions themselves, which are iffy states definable in terms of test conditions and empirically ascertainable results, on the one hand, and the explanation of these dispositions and propensities, which theoretical physics has made available, on the other.

51. Consider the case of our repeatedly magnetized soft iron. The repeated onslaughts and flights of the iffy property of being such that if iron filings are present, then they cling is, from the theoretical point of view, accompanied by certain actual physical processes which are induced by the current and other physical processes which replace them when the current is discontinued.5

5. A careful formulation would distinguish between the iffy truths at the theoretical level which are correlated with the original iffy truths, and the non-iffy processes of a theoretical character which are nomologically tied to these theoretical iffy truths,

52. Can we not regard classical theories of mental acts construed as pure occurrents (as contrasted with short-term propensities to think-out-loud) as theories in a sense which is analogous to micro-physical theories? Indeed, cannot we regard our common-sense conception of thought processes as such a theory? Such a theory would be designed to explain propensities to think-out-loud6 as micro-physical theory is designed to explain the powers and propensities which we know things to have at the perceptual level, sophisticated by laboratory techniques.

6. And it is important to note that we all grant that there is such a thing as thinking-out-loud -- though Cartesians give an account of it which presupposes the concept of non-verbal conceptual episodes.

53. The theory of 'inner speech' or Mentalese would construe these postulated thought episodes or occurrences as items which have a strong positive analogy with the thinkings-out-loud to which the VB calls attention. And it is interesting to note that when we refer to the thoughts which are occurring in a person's mind, we find it quite natural to quote them; -- even though they are not overt sayings. Yet the negative analogy must not be neglected. They are not thought of as waggings of an inner tongue. Nor, as we have seen, is the vehicle of Mentalese to be construed as imagery.

54. I shall not elaborate the theory, for this is done by classical philosophies of mind. The points I am interested in making are points about the theory rather than in it.

55. Perhaps the most important point is that what the theory postulates in the way of new entities are processes and acts rather than individuals. In this sense, it remains within the manifest image. Persons remain the basic individuals of the system. We have simply enlarged our conception of what persons do, as compared with the VB model with which we began. In addition to sayings and short-term propensities to say, we now conceive persons to be characterized by purely occurrent episodes of thinking in this analogically introduced sense. We might be tempted to refer to them as 'inner' episodes. But the spatial metaphor is misleading. They are primarily 'in' the person as states of the person. To be sure, they are not perceptible, but neither is solubility, and solubility is not in any more interesting sense 'in' the salt. It is only when we come to think that some particular part of the body (e.g., the heart or the brain) is the locus of these activities that the term 'inner' acquires richer meaning. And this is what begins to happen when the scientific revolution makes its impact on our conception of the world. But before I turn to this and other topics, let me sum up the picture of man and the world which I have been developing.

56. I introduced the manifest image of man-in-the-world as, essentially, an image which has been purged of the scientific objects postulated by physical science. The basic individuals which it countenances are certain merely material things, living things other than persons (about which I have had little to say), and persons.

57. The attributes which it ascribes to material things include, in the first instance, the proper and common sensibles. But it also allows in its universal discourse attributes which are definable in terms of sensible attributes, and of which the most interesting are the powers and propensities of material things to change in certain perceptible ways when subject to influences which are themselves definable in terms of perceptible qualities and relations.

58. Then there are persons. These individuals have perceptible characteristics and behave in perceptible ways. The behavior on which we have concentrated is the use of language. We restricted ourselves to that use of language which we called thinking-out-loud, and we developed a VB model according to which the meaningfulness of verbal behavior is to be found in the coherence which it exhibits, not only within speech itself, but in its relation to the contexts in which it occurs and the actions to which it leads. Here again, dispositions and propensities pertaining to perceptible traits of individuals were taken into account -- in particular, the shifting short-term propensities to say things which, according to the VB model, are thinkings in a secondary sense of this term.

59. But notice that this austere conception of the person has been enriched in two important respects without introducing new individuals. Thus, sensings were introduced as theoretical states involved in the explanation, for example, of how it could seem to a person that there is a pink ice cube in front of him when in point of fact there is not. In both the veridical perception of a pink ice cube and a perceptual experience which would be veridical if there were such an object in front of one, the person senses a-pink-cubely, or, in more familiar terms, has a sensation of a pink cube (where 'of a pink cube' is to be construed depth-grammar-wise as an adjective) so that the expression might be parsed 'an of-a-pink-cube sensation'.

60. Again we began our account of thinking with a VB model but proceeded to sketch an account of mental acts as theoretical states or processes which provide some underpinning for the concept of Mentalese propositional episodes in terms of which the argument for sense impressions was couched.

61. Notice, however, that I have continued to abstract from the impact of the revolution in the physical sciences on the problem 'what is a person?' My appeal to micro-physics has been limited to suggesting that 'classical' thoughts as pure mental episodes can be interpreted as standing to propensities to say as micro-physical processes stand to the physical propensities of the familiar material things around us.

62. The enriched image of man-in-the-world, which includes sensings and Mentalese thinkings, but no individuals other than common-sense material things, living things other than persons, and persons, is what I have called the Manifest Image of man-in-the-world. Its coherence as a conceptual framework must be fully savored before one turns to the problem of assessing the impact on this Image of the scientific revolution.

63. An understandable consequence of the impact of the conceptual revolution in the physical sciences on the problem of the nature of the person was the resurgence of dualism -- a dualism, however, arising from neither religious nor ethical considerations, nor from abstruse philosophical puzzles concerning the nature of knowledge, but rather from the attempt to think through the implications of science. The passage of time has intensified the problem of understanding the unity of the person (though not, of course, for those who take an instrumentalist stance toward scientific theories). The more sophisticated we get, the more intricate our puzzlement. We now await the flowering of neurophysiology, the science which is inheriting the glamor of molecular genetics.

64. The one thing we can be sure of is that we will be confronted with new ways of looking at such 'familiar' facts as that neurons 'fire' and electro-chemical impulses are transmitted through networks and around circuits in the central nervous system. Science has already presented us with new and subtle categories. I suspect that newer and still more subtle categories will be needed to solve the problem of the place of persons in the scientific image of the world. The philosopher can attempt to see the future as in a glass darkly, but essentially his role must be the (by no means unfamiliar) one of a midwife.

65. Part of this midwifery will consist in submitting existing categories to critical examination, as Berkeley, Mach, and Poincare, to name but three, did in the case of Newtonian mechanics. Part will consist in conceptual experimentation on a cooperative basis which will parallel the free thinking in logic, mathematics, and, dare I say, ontology which provided the context in which relativity was born. Nevertheless, an essential part of the philosopher's role in the concerted attempt to understand how a person is really related to what his body really is, is to clarify what, in outline, we already know thinking and sensing to be: and it is to this task that I've addressed myself in these first two lectures.



1. The explication of knowledge as 'justified true belief, though it involves many pitfalls to which attention has been called in recent years, remains the orthodox or classical account and is, I believe, essentially sound. Thus, in the present lecture I shall assume that it can be formulated in such a way as to be immune from the type of counterexamples with which it has been bombarded since Gettier's pioneering paper in Analysis and turn my attention to another problem which has dogged its footsteps since the very beginning. This problem can be put in the form of two questions: If knowledge is justified true belief, how can there be such a thing as self-evident knowledge? And if there is no such thing as self-evident knowledge, how can any true belief be, in the relevant sense, justified?

2. But first let us beat about in the neighboring fields, perhaps to scare up some game, but, in any case, to refamiliarize ourselves with the terrain. Thus, are there not occasions on which a person can be said to be justified in believing something which he would not appropriately be said to know? Presumably, to be justified in believing something is to have good reasons for believing it, as contrasted with its contradictory. But how good? Adequate? Conclusive? If adequate, adequate for what? If conclusive, the conclusion of what is at stake?

3. We are all familiar with Austin's point concerning the performative character of 'I know'. We are also familiar with the fact that, whereas to say 'I promise to do A' is, other things being equal, to promise to do A, to say 'I know that-p' is not, other things being equal, to know that-p. Chisholm's distinction between the strict and the extended sense of "performative utterance" is helpful in this connection. According to Chisholm,

An utterance beginning with "I want" is not performative in [the] strict sense, for it cannot be said to be an "act" of wanting. But "I want" is often used to accomplish what one might accomplish by means of the strict performative "I request." Let us say, then, that "I want" may be a "performative utterance" in an extended sense of the latter expression. ([1]: 16-17.)
He asks in which, if either, of these senses an utterance of "I know" may be performative. After reminding us that 'I know' is not performative in the strict sense of the term, he allows that "[it] is often used to accomplish what one may accomplish by the strict performative 'I guarantee' or 'I give you my word.' " and "hence may be performative in an extended sense of the term" ([1]: 16-17).

4. He argues, however, that 'I know' is not always a substitute for 'I guarantee', pointing out that:

Just as an utterance of "I want" may serve both to say something about me and to get you to do something, an utterance of "I know" may serve both to say something about me and to provide you with guarantees. To suppose that the performance of the nondescriptive function is inconsistent with the simultaneous performance of the descriptive function might be called, therefore, an example of the performative fallacy. ([1]: 17.)
I think that Chisholm is quite right about this. On the other hand, it seems to me that he overlooks the possibility of a connection between 'I know' and 'I guarantee' other than the one he considers. 'I know that-p' might be related to 'I guarantee that-p' not just as an autobiographical description which on occasion performs the same role as the latter but as one which contains a reference to guaranteeing in its very meaning. Is it not possible to construe 'I know that-p' as essentially equivalent to 'p, and I have reasons good enough to support a guarantee' (i.e., to say 'I guarantee' or 'You can rely on my statement')? Such an account would enable us to recognize a performative element in the very meaning of the verb 'to know' without construing 'I know' as a performative in the strict sense. It would also preserve the symmetry between first person and other person uses of the verb 'to know' which seems to be a pre-analytic datum. Thus, 'He knows that-p' would entail 'He has reasons good enough to support a guarantee that-p'.7

7. Notice that the above account of the relation of 'I know' to a performative is not quite the same as Urmson's. According to the latter, as represented by Chisholm, to say that Mr. Jones knew some proposition to be true is to say that Mr. Jones was "in a position in which he was entitled to say 'I know'." This account, as Chisholm points out, brings us back to the original problem of how the first person use of the verb is to be construed.

5. Furthermore, this account would enable us to appreciate the context dependence of the adequacy involved. Reasons which might be adequately good to justify a guarantee on one occasion might not be adequate to justify a guarantee on another. Again, the presence of such a performative element in the very meaning of the verb 'to know' would account for the fact (if it is a fact) that we rarely think in terms of 'I know' in purely self-directed thinkings; that we rarely have thoughts of the form 'I know that-p' unless the question of a possible guarantee to someone other than ourselves has arisen. Of course, we can 'tell ourselves' that we know something, but, then, so can we be said to make promises to ourselves.


6. Yet even after justice has been done, perhaps along the above lines, to the performative element in the meaning of the verb 'to know', it seems to me that we must recognize a closely related use of this expression which, though it may have implications concerning action, is not in any of the above senses performative. For once the ethical issue of how good one's reasons for a belief must be in order to justify giving a guarantee is solved, there remains the problem of how good reasons must be to justify believing that-p, where to believe that-p is obviously not an action, let alone a performatory action in either the strict or the extended sense.

7. Confronted by this question, we are tempted to set apart a class of cases in which the reasons are not only good enough to justify believing that-p but good enough to make it absurd not to believe that-p (or, perhaps, to believe its contradictory). It is perhaps, some such concept as this which is (in addition to the truth condition) the non-performative core of the meaning of the verb 'to know'.

8. I think the above discussion has served its primary purpose by highlighting the concept of having good reasons for believing that-p. For the solution of the problem which was posed in my opening remarks hinges ultimately on a distinction between two ways in which there can be, and one can have, good reasons for believing that-p.8

8. I have called attention elsewhere to the importance of distinguishing between questions concerning the reasonableness of believing that-p from questions concerning the reasonableness of 'acting on the proposition that-p', including guaranteeing that-p. The concept of acting on a proposition is clear only in simple cases, as when, for example, the proposition occurs as a premise in the agent's practical reasoning. When the agent takes probabilities into account, a far more complicated story is necessary to clarify the sense in which a person can be said to have acted on a given proposition. For a discussion of these problems, see my [6].

9. Now one pattern for justifying a belief in terms of good reasons can be called inferential. Consider the schema:

So, I have good reasons, all things considered, for believing q.
On reflection, this schema tends to expand into:
I have good reasons, all things considered, for believing p;
So, p;
So, I have good reasons, all things considered, for believing q.
Further reflection suggests that arguments conforming to this schema have a suppressed premise. What might it be? Consider the following expanded schema:
I have, all things considered, good reasons for believing p;
So, p;
p logically implies q;
So, I have, all things considered, good reasons for believing q.

10. The line of thought thus schematically represented would seem to involve the principle,

Logical implication transmits reasonableness.
In cases of this type, we are tempted to say, we have derivative good reasons, all things considered, for believing q. We say, in other words, that the reasonableness of believing q is 'inferential'.

11. Notice that the above line of thought is obviously an oversimplification, undoubtedly in several respects. In particular, it is important to note that if I have independent grounds for believing not-q, I may decide that I do not have good reasons, all things considered, for believing that-p. After all, if p implies q, not-q equally implies not-p. Yet in spite of its oversimplifications, the above train of thought takes us nearer to the distinctions necessary to solve our problem.

12. I have been considering the case where one proposition, p, logically implies another, q, and have claimed, with the above qualifications, that logical implication transmits reasonableness. Perhaps we can also take into account, with trepidation, 'probabilistic' implication, which would give us the following schema:

It is reasonable, all things considered, to believe p;
So, p;
p probabilistically implies q to a high degree;
So, all things considered, it is reasonable to believe q.
Probabilistic justification of beliefs in accordance with this pattern would, presumably, be illustrated by inductive arguments and theoretical explanations. In each case, we move from a premise of the form:
It is reasonable, all things considered, to believe E,
where 'E' formulates the evidence, to a conclusion of the form:
It is reasonable, all things considered, to believe H,
where 'H' formulates in the first case a law-like statement and in the second case a body of theoretical assumptions.


13. As has been pointed out since time immemorial, it is most implausible to suppose that all epistemic justification is inferential, at least in the sense of conforming to the patterns described above. Surely, it has been argued, there must be beliefs which we are justified in holding on grounds other than that they can be correctly inferred, inductively or deductively, from other beliefs which we are justified in holding. In traditional terms, if there is to be inferential knowledge, must there not be non-inferential knowledge -- beliefs, that is, the reasonableness of which does not rest on the reasonableness of beliefs which logically or probabilistically imply them?

14. We are clearly in the neighborhood of what has been called the 'self-evident', the 'self-certifying', in short, of 'intuitive knowledge'. It is in this neighborhood that we find what has come to be called the foundational picture of human knowledge. According to this picture, beliefs which have inferential reasonableness ultimately rely for their authority on a stratum of beliefs which are, in some sense, self-certifying. The reasonableness of moves from the level of the self-evident to higher levels would involve the principles of logic (deductive and inductive) and, perhaps, certain additional principles which are sui generis. They would have in common the character of transmitting authoritativeness from lower-level beliefs to higher-level beliefs.


15. Let us reflect on the concept of such a foundational level of knowledge. It involves the concept of beliefs which are reasonable, which have epistemic authority or correctness, but which are not reasonable or authoritative by virtue of the fact that they are beliefs in propositions which are implied by other propositions which it is reasonable to believe. Let us label them, for the moment, 'non-inferentially reasonable beliefs'.

16. How can there be such beliefs? For the concept of a reason seems so clearly tied to that of an inference or argument that the concept of non-inferential reasonableness seems to be a contradictio in adjecto. Surely, we are inclined to say, for a belief (or believing) to be reasonable, there must be a reason for the belief (or believing). And must not this reason be something other than the belief or believing for which it is the reason? And surely, we are inclined to say, to believe something because it is reasonable (to believe it) involves not only that there be a reason but that, in a relevant sense, one has or is in possession of the reason. Notice that I have deliberately formulated these expostulations in such a way as to highlight the ambiguities involved when one speaks of reasonable beliefs.

17. In attempting to cope with these challenges, I shall leave aside problems pertaining to inferential and non-inferential reasonableness in logic and mathematics and concentrate on the apparent need for 'self evidence' in the sphere of empirical matters of fact.

18. How might a self-justifying belief be construed? One suggestion, modified from Chisholm's Theory of Knowledge,9 is to the effect that the justification of such beliefs has the form,

What justifies me in claiming that my belief that a is F is reasonable is simply the fact that a is F.

9. [1]: 28. Chisholm's principle concerns 'what justifies us in counting it as evident that a is F'. But the 'evident' is defined on p. 22 as a special case of the 'reasonable'.

19. But this seems to point to the existence of inferences of the form,

It is a fact that a is F;
So, it is reasonable to believe that a is F,
and one might begin to wonder what principle authorizes this inference.

20. Something, clearly, has gone wrong. In order for any such argument to do the job, its premise would have to have authority; it would have to be something which it is reasonable to believe. But if we modify the schema to take this into account, it becomes:

It is reasonable to believe it to be a fact that a is F;
So, it is reasonable to believe that a is F,
which, in virtue of the equivalence of
believing a to be F
believing it to be a fact that a is F,
is obviously unilluminating.


21. Now many philosophers who have endorsed a concept of intuitive knowledge are clearly committed to the position that there is a level of cognition more basic than believing. This more basic level would consist of a sub-conceptual10 awareness of certain facts. In terms of the framework sketched in the preceding two lectures, there would be a level of cognition more basic than thinkings or tokenings of sentences in Mentalese -- more basic, in fact, than symbolic activity, literal or analogical. It would be a level of cognition unmediated by concepts; indeed it would be the very source of concepts in some such way as described by traditional theories of abstraction. It would be 'direct apprehension' of facts; their 'direct presence' to the mind.11

10. Where 'sub-conceptual' is far from being used as a pejorative term.

11. It is clearly some such position which is envisaged by many who explicitly reject the equation of knowledge with justified true belief. That it is implicit in Chisholm's position becomes clear not only when we reflect (as above, Sections 18-20) on what his principle concerning the directly evident might mean, but when we take into account his use of such phrases as "state of affairs" that " 'presents itself to him' " or that " 'is apprehended through itself' " ([1]: 28) and his general commitment to a fact ontology ([1], Chap. 7, passim), a 'fact', in the relevant sense, being a "state of affairs which exists" ([1]: 104). 'Exists' in this context should not be confused with the 'existential quantifier' but should be considered as a synonym for 'obtains'. It is obviously not self-contradictory to say that some states of affairs do not obtain.

22. Schematically we would have,

It is a fact (which I directly apprehend) that a is F;
So, it is reasonable to have the conceptual belief that a is F.
This multiplication of distinctions raises two serious problems: (1) What sort of entities are facts? Do they belong to the real (extra-conceptual) order? That 'fact' is roughly a synonym for 'truth', and 'true' is appropriately predicated of conceptual items (in overt speech or Mentalese) should give pause for thought.

23. Then there is also the question: (2) How is 'direct apprehension' to be understood? If the apprehending is distinguishable from the apprehended, is it not also 'separable'? Might not apprehending occur without any fact being apprehended? If so, an 'apprehending that-p' might not be an apprehending of the fact that-p. Hitting, in baseball, implies that something is hit. 'Swinging' does not. To hit is to swing successfully. Of course, 'apprehend', like 'see', is, in its ordinary sense, an achievement word. But does this not mean that, as in the case of 'see', there is a place for 'ostensibly apprehending', i.e., seeming to | apprehend, a concept which does not imply achievement?

24. Many who use the metaphor 'to see' in intellectual contexts overlook the fact that in its literal sense 'seeing' is a term for a successful conceptual activity which contrasts with 'seeming to see'. No piling on of additional metaphors (e.g., 'grasping', which implies an object grasped) can blunt this fact. Now the distinction between seeing and merely seeming to see implies a criterion. To rely on the metaphors of 'apprehending' or 'presence of the object' is to obscure the need of criteria for distinguishing between 'knowing' and 'seeming to know', which ultimately define what it means to speak of knowledge as a correct or well-founded thinking that something is the case.

25. If so, to know that we have apprehended a fact, we would have to know that the criteria which distinguish apprehending from seeming to apprehend were satisfied. In short, I suspect that the notion of a non- conceptual 'direct apprehension' of a 'fact' provides a merely verbal solution to our problem. The regress, is stopped by an ad hoc regress-stopper. Indeed, the very metaphors which promised the sought-for foundation contain within themselves a dialectical moment which takes us beyond them.


26. What is the alternative? I suggest that the key to our problem is provided by the Verbal Behaviorist model, developed in the preceding lecture. It is, we have seen, a simple, indeed radically over-simplified, model, but it will provide us, I believe, with the outline of a strategy for getting out of the classical labyrinth.

27. According to this model, it will be remembered, the primary sense of

The thought occurred to Jones that snow is white
Jones said 'snow is white',
where the verb 'to say' was stripped of some of its ordinary implications and roughly equated with 'to utter words candidly as one who knows the language'. In particular, it was purged of the illocutionary and perlocutionary forces which Austin and Grice find so central to their theory of meaning. 'To say', in this sense, was also equated with 'thinking-out-loud'.

28. According to the VB, as I described him, we must also introduce, in order to take account of those cases where one thinks silently, a secondary sense of

The thought occurred to Jones that snow is white,
in which it refers to a short-term proximate propensity to think-out-loud that snow is white. When this propensity is 'uninhibited', one thinks-out-loud, i.e., thinks in the primary sense of this term (as construed by VB). There can be many reasons why, on a particular occasion, this propensity is inhibited. But, for our purposes, the most important is the general inhibition acquired in childhood when, after being taught to think-out-loud, one is trained not to be a 'babbler'. One might use the model of an on-off switch which gets into the wiring diagram when the child learns to keep his thoughts to himself.

29. In the concluding sections of the preceding lecture (cf. Lecture II, section IV), I argued that yet another concept of 'having the thought occur to one that-p' can be introduced which stands to the second as the theoretical concept of electronic processes stands to the acquisition (and loss) of the power to attract iron filings (or a bell clapper) by a piece of soft iron in a coil of wire attached to an electric circuit. I argued that the classical concept of thought-episodes can be construed as part of a theoretical framework designed to explain the acquisition and loss of verbal propensities to think-out-loud. In approaching the problem of the status of non-inferential knowledge, however, I shall return to the VB model and concentrate, indeed, on the primary sense of having the thought occur to one that-p, i.e., think-out-loud that-p.

50. I argued in my first lecture that perceptual experience involves a sensory element which is in no way a form of thinking, however intimately it may be connected with thinking. This element consists of what I variously called 'sense impressions', 'sensations', or 'sensa'. I argued that these items, properly construed, belong in a theoretical framework designed to explain:

  1. the difference between merely thinking of (believing in the existence of) a perceptible state of affairs and seeing (or seeming to see) that such a state of affairs exists;
  2. how it can seem to a person that there is a pink ice cube in front of him when there isn't one -- either because there is something there which is either not pink or not cubical, or because there is nothing there and he is having a realistic hallucination.

31. I've explored problems pertaining to the nature and status of this sensory element on many occasions (most recently in [9], Chapter I, and in [11], which is a reply to [2]), but further exploration of this theme would leave no time for the problem at hand.

32. What is important for our purposes is that perceptual experience also involves a conceptual or propositional component -- a 'thinking' in a suitably broad sense of this accordion term. In perception, the thought is caused to occur to one that, for example, there is a pink ice cube in front of one. It is misleading to call such a thought a 'perceptual judgment' -- for this implies question-answering activity of estimating, for example, the size of an object. (I judge that the room is ten feet tall.) Perhaps the best term is 'taking something to be the case'. Thus, on the occasion of sensing a certain color configuration, one takes there to be an object or situation of a certain description in one's physical environment.

33. Consider once again the case, discussed in Lecture II, sections 35-41, where

Jones sees there to be a red apple in front of him.
As we explained there, given that Jones has learned how to use the relevant words in perceptual situations, he is justified in reasoning as follows:
I just thought-out-loud 'Lo! Here is a red apple'
(no countervailing conditions obtain);
So, there is good reason to believe that there is a red apple in front of me.

34. Of course, the conclusion of this reasoning is not the thinking involved in his original perceptual experience. Like all justification arguments, it is a higher-order thinking. He did not originally infer that there is a red apple in front of him. Now, however, he is inferring from the character and context of his experience that it is veridical and that there is good reason to believe that there is indeed a red apple in front of him.

35. Notice that although the justification of the belief that there is a red apple in front of (Jones) is an inferential justification, it has the peculiar character that its essential premise asserts the occurrence of the very same belief in a specific context.12 It is this fact which gives the appearance that such beliefs are self-justifying and hence gives the justification the appearance of being non-inferential.

12. I called attention to this feature of the justification involved in 'non-inferential' knowledge in [10], Chapter 3. Thus, I wrote ". . . one only knows what one has a right to think to be the case. Thus, to say that one directly knows that-p is to say that his right to the conviction that-p essentially involves the fact that the idea that-p occurred to the knower in a specific way" ([10]: 88). I suggested that this "kind of credibility" be called 'trans-level credibility', and the pattern of inference involved in the reasoning which 'mobilizes this credibility, 'trans-level inference'. A similar point was less clearly made in Sections 32-39 of [5].

36. It is, as I see it, precisely this feature of the unique pattern of justification in question which, misinterpreted, leads Chisholm to formulate as his principle for the 'directly evident',

What justifies me in counting it as evident that a is F is simply the fact that a is F. ([1]: 28.)
To be sure, Chisholm's examples of the 'directly evident' are not taken from the domain of perceptual beliefs, but rather, in true Cartesian spirit, from one's knowledge about what is going on in one's mind at the present moment. Indeed, he rejects the idea that particular perceptual beliefs of the kind which I illustrated by my example of the red apple are ever directly evident.

37. On the other hand, though he does think that particular perceptual beliefs of this type can at best be indirectly evident, he does think that they can be reasonable. Should we say 'directly reasonable'? I, of course, would answer in the affirmative. Yet it is not clear to me that Chisholm would be happy with this suggestion. If (as he should) he has at the back of his mind the reasoning;

There (visually) appears to me to be a red apple here;
So, it is reasonable for me (to believe) that there is a red apple here,
then he should not object to speaking of the reasonableness in question as 'direct', for the premise does not contain a predicate of epistemic evaluation. If, on the other hand (as he should not), he has at the back of his mind the following reasoning,
It is evident to me that there (visually) appears to me to be a red apple here;
So, it is reasonable for me (to believe) that there is a red apple here,
we could expect him to object to speaking of this reasonableness as 'direct'.

38. This tension sets the stage for a corresponding comment on Chisholm's third epistemic principle, which concerns the case where what we visually take to be the case is the presence of something having a "sensible characteristic F" (where 'F' ranges over the familiar Aristotelian list of proper and common sensibles). The principle reads as follows:

(C) If there is a certain sensible characteristic F such that S believes that he perceives something to be F, then it is evident to S that he is perceiving something to have that characteristic F, and also evident that there is something that is F.

39. I shall not pause to quibble over such matters as whether, in the light of Chisholm's definition of 'evident', it can ever be evident to me that I am perceiving something to be pink or that something in front of me is pink -- even if the claim is limited to the facing side. A high degree of reasonableness will, do. The point which I wish to stress is that once again the question arises, does Chisholm think of the evidence involved in the principle as 'direct' or 'indirect'? This time it is clear that he thinks of it as indirect. As I see it, then, he has at the back of his mind the following reasoning:

It is evident to me that there appears to me to be a pink object here;
So, it is evident to me that I perceive a pink object to be here and evident to me that there is a pink object here.
The contrasting reasoning would be:
There appears to me to be a pink object here;
So, it is evident to me that I perceive a pink object to be here and evident to me that there is a pink object here.

40. Now I suspect that what has misled Chisholm is the fact that if I were to argue,

There appears to me to be a pink cube here;
So, it is highly reasonable for me (to believe) that there is a pink object here.
a skeptic could be expected to challenge me by asking 'What right have you to accept your conclusion, unless you have a right to accept the premise? Are you not implying that you know that there appears to you to be a pink object here; and must not this claim be a tacit premise in your argument?' But, surely, the skeptic would just be mistaken -- not, indeed in asserting that in some sense I imply that I know that there appears to me to be a pink object here, but in asserting that this implication must be taken to be a premise in my reasoning, if it is to be valid, and, hence, if the corresponding epistemic principle is to be true. But in that case, the latter principle would be not Chisholm's (C), but rather:
(C') If it is evident to S that there is a certain sensible characteristic F. . .

41. The larger import of the above reply to the skeptic will be sketched in my concluding remarks. For the moment, let me say that from my point of view something very like Chisholm's principle (C) is sound but concerns the direct evidence (or, better, direct high degree of reasonableness) of certain perceptual beliefs. Let me formulate it as follows:

(S) If there is a certain sensible characteristic F such that S believes that he perceives something to be F, then it is evident to S that there is something that is F and, hence, that he is perceiving something to b
e F.

42. Notice that I have reversed the relative position of the two clauses in the consequent as they appear in Chisholm's principle. This is because, on my interpretation, the core of the principle is

(S1) If I ostensibly see there to be an F object here then it is highly reasonable for me (to believe) that there is an F object here.
And the move to
(S2) If I ostensibly see there to be an F object here, then it is highly reasonable for me (to believe) that I see there to be an F object here
is justified by the conceptual tie between 'ostensibly see', 'see', and truth.


43. Chisholm's principle (C) and his other epistemic principles pertaining to perception and memory are themselves justified, as he sees it, by the fact that unless they, or something like them, are true, then there could be no such thing as perceptual knowledge to the effect, to use his example, that there is a cat on the roof. We have here a justification of the 'this or nothing' kind familiar to the Kantian tradition. The principles also seem, on occasion, to be treated as candidates for the status of synthetic a priori (and even, one suspects, self-evident) truth.

44. As I see it, on the other hand, these epistemic principles can be placed in a naturalistic setting and their authority construed in terms of the nature of concept formation and of the acquisition of relevant linguistic skills. The model which I have been using is, indeed, a very simple one, and I have largely limited my use of it to the epistemic authority of perceptual beliefs. But if the strategy which I have suggested is successful, it is a relatively simple matter to extend it to memory beliefs. I have discussed the case of non-inferential knowledge of our own mental states in some detail, using this same general strategy, on a number of occasions (most recently in [9], esp. pp. 71 ff, 151 ff).

45. But, surely, it will be urged, facts about learning languages and acquiring linguistic skills are themselves empirical facts; and to know these facts involves perception, memory, indeed, all the epistemic activities the justification of which is at stake. Must we not conclude that any such account as I give of the principle that perceptual beliefs occurring in perceptual contexts are likely to be true is circular? It must, indeed, be granted that principles pertaining to the epistemic authority of perceptual and memory beliefs are not the sort of thing which could be arrived at by inductive reasoning from perceptual belief. But the best way to make this point is positive. We have to be in this framework to be thinking and perceiving beings at all. I suspect that it is this plain truth which is the real underpinning of the idea that the authority of epistemic principles rests on the fact that unless they were true we could not see that a cat is on the roof.

46. I pointed out a moment ago that we have to be in the framework of these (and other) principles to be thinking, perceiving, and, I now add, acting beings at all. But surely this makes it clear that the exploration of these principles is but part and parcel of the task of explicating the concept of a rational animal or, in VB terms, of a language-using organism whose language is about the world in which it is used. It is only in the light of this larger task that the problem of the status of epistemic principles reveals its true meaning.

47. From the perspective of this larger task, the metaphor of 'foundation and superstructure' is seen to be a false extrapolation, to use a Deweyan turn of phrase, from specific 'problematic situations' with respect to which it is appropriate. And when we concern ourselves, as Philosophy ultimately demands, with how it is with man and his world, as contrasted with the catch-as-catch-can procedures which generate man's awareness of himself and his world, surely we can say, as I wrote some fifteen years ago in an earlier essay on this topic,

There is clearly some point to the picture of human knowledge as resting on a level of propositions -- observation reports -- which do not rest on other propositions in the same way as other propositions rest on them. On the other hand, I do wish to insist that the metaphor of 'foundation' is misleading in that it keeps us from seeing that if there is a logical dimension in which other empirical propositions rest on observation reports, there is another logical dimension in which the latter rest on the former.

Above all, the picture is misleading because of its static character. One seems forced to choose between the picture of an elephant which rests on a tortoise (What supports the tortoise?) and the picture of a great Hegelian serpent of knowledge with its tail in its mouth (Where did it begin?). Neither will do. For empirical knowledge, like its sophisticated extension, science, is rational, not because it has a foundation but because it is a self-correcting enterprise which can put any claim in jeopardy, though not all at once. ([5], section 38; quoted from [10]: 170.)


[1] R. M. Chisholm, Theory of Knowledge (Englewood Cliffs, N. J.: Prentice-Hall, 1966).

[2] James W. Cornman, "Sellars, Scientific Realism, and Sensa," Review of Metaphysics 24(1970).

[3] Roderick Firth, "Coherence, Certainty, and Epistemic Priority," Journal of Philosophy 61(1964): 545-57.

[4] Wilfrid Sellars, "Counterfactuals, Dispositions, and the Causal Modalities," in Herbert Feigl, Michael Scriven, and Grover Maxwell, eds., Minnesota Studies in the Philosophy of Science, Vol. II (Minneapolis: University of Minnesota Press, 1957).

[5] ______, "Empiricism and the Philosophy of Mind," in Herbert Feigl and Michael Scriven, eds., Minnesota Studies in the Philosophy of Science, Vol. I (Minneapolis: University of Minnesota Press, 1956). Reprinted as Chapter V of [10].

[6] ______, "Induction as Vindication," Philosophy of Science 31(1964): 197-232.

[7] ______, "Metaphysics and the Concept of a Person," in Karel Lambert, ed., The Logical Way of Doing Things (New Haven: Yale University Press, 1969).

[8] ______, "Philosophy and the Scientific Image of Man," in Robert Colodny, ed., Frontiers of Science and Philosophy (Pittsburgh: University of Pittsburgh Press, 1962). Reprinted as Chapter I of [10].

[9] ______, Science and Metaphysics (London: Routledge and Kegan Paul, 1967).

[10] ______, Science, Perception and Reality (London: Routledge and Kegan Paul, and New York: Humanities Press, 1963).

[11] ______, "Science, Sense Impressions, and Sensa: A Reply to Cornman," Review of Metaphysics 25(1971).