one of the main purposes of these lectures is to give grounds for the belief that the distinction between mind and matter is not so fundamental as is commonly supposed. in the preceding lecture i dealt in outline with the physical side of this problem. i attempted to show that what we call a material object is not itself a substance, but is a system of particulars analogous in their nature to sensations, and in fact often including actual sensations among their number. in this way the stuff of which physical objects are composed is brought into relation with the stuff of which part, at least, of our mental life is composed.

there is, however, a converse task which is equally necessary for our thesis, and that is, to show that the stuff of our mental life is devoid of many qualities which it is commonly supposed to have, and is not possessed of any attributes which make it incapable of forming part of the world of matter. in the present lecture i shall begin the arguments for this view.

corresponding to the supposed duality of matter and mind, there are, in orthodox psychology, two ways of knowing what exists. one of these, the way of sensation and external perception, is supposed to furnish data for our knowledge of matter, the other, called "introspection," is supposed to furnish data for knowledge of our mental processes. to common sense, this distinction seems clear and easy. when you see a friend coming along the street, you acquire knowledge of an external, physical fact; when you realize that you are glad to meet him, you acquire knowledge of a mental fact. your dreams and memories and thoughts, of which you are often conscious, are mental facts, and the process by which you become aware of them seems to be different from sensation. kant calls it the "inner sense"; sometimes it is spoken of as "consciousness of self"; but its commonest name in modern english psychology is "introspection." it is this supposed method of acquiring knowledge of our mental processes that i wish to analyse and examine in this lecture.

i will state at the outset the view which i shall aim at establishing. i believe that the stuff of our mental life, as opposed to its relations and structure, consists wholly of sensations and images. sensations are connected with matter in the way that i tried to explain in lecture v, i.e. each is a member of a system which is a certain physical object. images, though they usually have certain characteristics, especially lack of vividness, that distinguish them from sensations, are not invariably so distinguished, and cannot therefore be defined by these characteristics. images, as opposed to sensations, can only be defined by their different causation: they are caused by association with a sensation, not by a stimulus external to the nervous system—or perhaps one should say external to the brain, where the higher animals are concerned. the occurrence of a sensation or image does not in itself constitute knowledge but any sensation or image may come to be known if the conditions are suitable. when a sensation—like the hearing of a clap of thunder—is normally correlated with closely similar sensations in our neighbours, we regard it as giving knowledge of the external world, since we regard the whole set of similar sensations as due to a common external cause. but images and bodily sensations are not so correlated. bodily sensations can be brought into a correlation by physiology, and thus take their place ultimately among sources of knowledge of the physical world. but images cannot be made to fit in with the simultaneous sensations and images of others. apart from their hypothetical causes in the brain, they have a causal connection with physical objects, through the fact that they are copies of past sensations; but the physical objects with which they are thus connected are in the past, not in the present. these images remain private in a sense in which sensations are not. a sensation seems to give us knowledge of a present physical object, while an image does not, except when it amounts to a hallucination, and in this case the seeming is deceptive. thus the whole context of the two occurrences is different. but in themselves they do not differ profoundly, and there is no reason to invoke two different ways of knowing for the one and for the other. consequently introspection as a separate kind of knowledge disappears.

the criticism of introspection has been in the main the work of american psychologists. i will begin by summarizing an article which seems to me to afford a good specimen of their arguments, namely, "the case against introspection," by knight dunlap ("psychological review," vol xix, no. 5, pp. 404-413, september, 1912). after a few historical quotations, he comes to two modern defenders of introspection, stout and james. he quotes from stout such statements as the following: "psychical states as such become objects only when we attend to them in an introspective way. otherwise they are not themselves objects, but only constituents of the process by which objects are recognized" ("manual," 2nd edition, p. 134. the word "recognized" in dunlap's quotation should be "cognized.") "the object itself can never be identified with the present modification of the individual's consciousness by which it is cognized" (ib. p. 60). this is to be true even when we are thinking about modifications of our own consciousness; such modifications are to be always at least partially distinct from the conscious experience in which we think of them.

at this point i wish to interrupt the account of knight dunlap's article in order to make some observations on my own account with reference to the above quotations from stout. in the first place, the conception of "psychical states" seems to me one which demands analysis of a somewhat destructive character. this analysis i shall give in later lectures as regards cognition; i have already given it as regards desire. in the second place, the conception of "objects" depends upon a certain view as to cognition which i believe to be wholly mistaken, namely, the view which i discussed in my first lecture in connection with brentano. in this view a single cognitive occurrence contains both content and object, the content being essentially mental, while the object is physical except in introspection and abstract thought. i have already criticized this view, and will not dwell upon it now, beyond saying that "the process by which objects are cognized" appears to be a very slippery phrase. when we "see a table," as common sense would say, the table as a physical object is not the "object" (in the psychological sense) of our perception. our perception is made up of sensations, images and beliefs, but the supposed "object" is something inferential, externally related, not logically bound up with what is occurring in us. this question of the nature of the object also affects the view we take of self-consciousness. obviously, a "conscious experience" is different from a physical object; therefore it is natural to assume that a thought or perception whose object is a conscious experience must be different from a thought or perception whose object is a physical object. but if the relation to the object is inferential and external, as i maintain, the difference between two thoughts may bear very little relation to the difference between their objects. and to speak of "the present modification of the individual's consciousness by which an object is cognized" is to suggest that the cognition of objects is a far more direct process, far more intimately bound up with the objects, than i believe it to be. all these points will be amplified when we come to the analysis of knowledge, but it is necessary briefly to state them now in order to suggest the atmosphere in which our analysis of "introspection" is to be carried on.

another point in which stout's remarks seem to me to suggest what i regard as mistakes is his use of "consciousness." there is a view which is prevalent among psychologists, to the effect that one can speak of "a conscious experience" in a curious dual sense, meaning, on the one hand, an experience which is conscious of something, and, on the other hand, an experience which has some intrinsic nature characteristic of what is called "consciousness." that is to say, a "conscious experience" is characterized on the one hand by relation to its object and on the other hand by being composed of a certain peculiar stuff, the stuff of "consciousness." and in many authors there is yet a third confusion: a "conscious experience," in this third sense, is an experience of which we are conscious. all these, it seems to me, need to be clearly separated. to say that one occurrence is "conscious" of another is, to my mind, to assert an external and rather remote relation between them. i might illustrate it by the relation of uncle and nephew a man becomes an uncle through no effort of his own, merely through an occurrence elsewhere. similarly, when you are said to be "conscious" of a table, the question whether this is really the case cannot be decided by examining only your state of mind: it is necessary also to ascertain whether your sensation is having those correlates which past experience causes you to assume, or whether the table happens, in this case, to be a mirage. and, as i explained in my first lecture, i do not believe that there is any "stuff" of consciousness, so that there is no intrinsic character by which a "conscious" experience could be distinguished from any other.

after these preliminaries, we can return to knight dunlap's article. his criticism of stout turns on the difficulty of giving any empirical meaning to such notions as the "mind" or the "subject"; he quotes from stout the sentence: "the most important drawback is that the mind, in watching its own workings, must necessarily have its attention divided between two objects," and he concludes: "without question, stout is bringing in here illicitly the concept of a single observer, and his introspection does not provide for the observation of this observer; for the process observed and the observer are distinct" (p. 407). the objections to any theory which brings in the single observer were considered in lecture i, and were acknowledged to be cogent. in so far, therefore, as stout's theory of introspection rests upon this assumption, we are compelled to reject it. but it is perfectly possible to believe in introspection without supposing that there is a single observer.

william james's theory of introspection, which dunlap next examines, does not assume a single observer. it changed after the publication of his "psychology," in consequence of his abandoning the dualism of thought and things. dunlap summarizes his theory as follows:

"the essential points in james's scheme of consciousness are subject, object, and a knowing of the object by the subject. the difference between james's scheme and other schemes involving the same terms is that james considers subject and object to be the same thing, but at different times in order to satisfy this requirement james supposes a realm of existence which he at first called 'states of consciousness' or 'thoughts,' and later, 'pure experience,' the latter term including both the 'thoughts' and the 'knowing.' this scheme, with all its magnificent artificiality, james held on to until the end, simply dropping the term consciousness and the dualism between the thought and an external reality"(p. 409).

he adds: "all that james's system really amounts to is the acknowledgment that a succession of things are known, and that they are known by something. this is all any one can claim, except for the fact that the things are known together, and that the knower for the different items is one and the same" (ib.).

in this statement, to my mind, dunlap concedes far more than james did in his later theory. i see no reason to suppose that "the knower for different items is one and the same," and i am convinced that this proposition could not possibly be ascertained except by introspection of the sort that dunlap rejects. the first of these points must wait until we come to the analysis of belief: the second must be considered now. dunlap's view is that there is a dualism of subject and object, but that the subject can never become object, and therefore there is no awareness of an awareness. he says in discussing the view that introspection reveals the occurrence of knowledge: "there can be no denial of the existence of the thing (knowing) which is alleged to be known or observed in this sort of 'introspection.' the allegation that the knowing is observed is that which may be denied. knowing there certainly is; known, the knowing certainly is not"(p. 410). and again: "i am never aware of an awareness" (ib.). and on the next page: "it may sound paradoxical to say that one cannot observe the process (or relation) of observation, and yet may be certain that there is such a process: but there is really no inconsistency in the saying. how do i know that there is awareness? by being aware of something. there is no meaning in the term 'awareness' which is not expressed in the statement 'i am aware of a colour (or what-not).'"

but the paradox cannot be so lightly disposed of. the statement "i am aware of a colour" is assumed by knight dunlap to be known to be true, but he does not explain how it comes to be known. the argument against him is not conclusive, since he may be able to show some valid way of inferring our awareness. but he does not suggest any such way. there is nothing odd in the hypothesis of beings which are aware of objects, but not of their own awareness; it is, indeed, highly probable that young children and the higher animals are such beings. but such beings cannot make the statement "i am aware of a colour," which we can make. we have, therefore, some knowledge which they lack. it is necessary to knight dunlap's position to maintain that this additional knowledge is purely inferential, but he makes no attempt to show how the inference is possible. it may, of course, be possible, but i cannot see how. to my mind the fact (which he admits) that we know there is awareness, is all but decisive against his theory, and in favour of the view that we can be aware of an awareness.

dunlap asserts (to return to james) that the real ground for james's original belief in introspection was his belief in two sorts of objects, namely, thoughts and things. he suggests that it was a mere inconsistency on james's part to adhere to introspection after abandoning the dualism of thoughts and things. i do not wholly agree with this view, but it is difficult to disentangle the difference as to introspection from the difference as to the nature of knowing. dunlap suggests (p. 411) that what is called introspection really consists of awareness of "images," visceral sensations, and so on. this view, in essence, seems to me sound. but then i hold that knowing itself consists of such constituents suitably related, and that in being aware of them we are sometimes being aware of instances of knowing. for this reason, much as i agree with his view as to what are the objects of which there is awareness, i cannot wholly agree with his conclusion as to the impossibility of introspection.

the behaviourists have challenged introspection even more vigorously than knight dunlap, and have gone so far as to deny the existence of images. but i think that they have confused various things which are very commonly confused, and that it is necessary to make several distinctions before we can arrive at what is true and what false in the criticism of introspection.

i wish to distinguish three distinct questions, any one of which may be meant when we ask whether introspection is a source of knowledge. the three questions are as follows:

(1) can we observe anything about ourselves which we cannot observe about other people, or is everything we can observe public, in the sense that another could also observe it if suitably placed?

(2) does everything that we can observe obey the laws of physics and form part of the physical world, or can we observe certain things that lie outside physics?

(3) can we observe anything which differs in its intrinsic nature from the constituents of the physical world, or is everything that we can observe composed of elements intrinsically similar to the constituents of what is called matter?

any one of these three questions may be used to define introspection. i should favour introspection in the sense of the first question, i.e. i think that some of the things we observe cannot, even theoretically, be observed by any one else. the second question, tentatively and for the present, i should answer in favour of introspection; i think that images, in the actual condition of science, cannot be brought under the causal laws of physics, though perhaps ultimately they may be. the third question i should answer adversely to introspection i think that observation shows us nothing that is not composed of sensations and images, and that images differ from sensations in their causal laws, not intrinsically. i shall deal with the three questions successively.

(1) publicity or privacy of what is observed. confining ourselves, for the moment, to sensations, we find that there are different degrees of publicity attaching to different sorts of sensations. if you feel a toothache when the other people in the room do not, you are in no way surprised; but if you hear a clap of thunder when they do not, you begin to be alarmed as to your mental condition. sight and hearing are the most public of the senses; smell only a trifle less so; touch, again, a trifle less, since two people can only touch the same spot successively, not simultaneously. taste has a sort of semi-publicity, since people seem to experience similar taste-sensations when they eat similar foods; but the publicity is incomplete, since two people cannot eat actually the same piece of food.

but when we pass on to bodily sensations—headache, toothache, hunger, thirst, the feeling of fatigue, and so on—we get quite away from publicity, into a region where other people can tell us what they feel, but we cannot directly observe their feeling. as a natural result of this state of affairs, it has come to be thought that the public senses give us knowledge of the outer world, while the private senses only give us knowledge as to our own bodies. as regards privacy, all images, of whatever sort, belong with the sensations which only give knowledge of our own bodies, i.e. each is only observable by one observer. this is the reason why images of sight and hearing are more obviously different from sensations of sight and hearing than images of bodily sensations are from bodily sensations; and that is why the argument in favour of images is more conclusive in such cases as sight and hearing than in such cases as inner speech.

the whole distinction of privacy and publicity, however, so long as we confine ourselves to sensations, is one of degree, not of kind. no two people, there is good empirical reason to think, ever have exactly similar sensations related to the same physical object at the same moment; on the other hand, even the most private sensation has correlations which would theoretically enable another observer to infer it.

that no sensation is ever completely public, results from differences of point of view. two people looking at the same table do not get the same sensation, because of perspective and the way the light falls. they get only correlated sensations. two people listening to the same sound do not hear exactly the same thing, because one is nearer to the source of the sound than the other, one has better hearing than the other, and so on. thus publicity in sensations consists, not in having precisely similar sensations, but in having more or less similar sensations correlated according to ascertainable laws. the sensations which strike us as public are those where the correlated sensations are very similar and the correlations are very easy to discover. but even the most private sensations have correlations with things that others can observe. the dentist does not observe your ache, but he can see the cavity which causes it, and could guess that you are suffering even if you did not tell him. this fact, however, cannot be used, as watson would apparently wish, to extrude from science observations which are private to one observer, since it is by means of many such observations that correlations are established, e.g. between toothaches and cavities. privacy, therefore does not by itself make a datum unamenable to scientific treatment. on this point, the argument against introspection must be rejected.

(2) does everything observable obey the laws of physics? we come now to the second ground of objection to introspection, namely, that its data do not obey the laws of physics. this, though less emphasized, is, i think, an objection which is really more strongly felt than the objection of privacy. and we obtain a definition of introspection more in harmony with usage if we define it as observation of data not subject to physical laws than if we define it by means of privacy. no one would regard a man as introspective because he was conscious of having a stomach ache. opponents of introspection do not mean to deny the obvious fact that we can observe bodily sensations which others cannot observe. for example, knight dunlap contends that images are really muscular contractions,* and evidently regards our awareness of muscular contractions as not coming under the head of introspection. i think it will be found that the essential characteristic of introspective data, in the sense which now concerns us, has to do with localization: either they are not localized at all, or they are localized, like visual images, in a place already physically occupied by something which would be inconsistent with them if they were regarded as part of the physical world. if you have a visual image of your friend sitting in a chair which in fact is empty, you cannot locate the image in your body, because it is visual, nor (as a physical phenomenon) in the chair, because the chair, as a physical object, is empty. thus it seems to follow that the physical world does not include all that we are aware of, and that images, which are introspective data, have to be regarded, for the present, as not obeying the laws of physics; this is, i think, one of the chief reasons why an attempt is made to reject them. i shall try to show in lecture viii that the purely empirical reasons for accepting images are overwhelming. but we cannot be nearly so certain that they will not ultimately be brought under the laws of physics. even if this should happen, however, they would still be distinguishable from sensations by their proximate causal laws, as gases remain distinguishable from solids.

* "psychological review," 1916, "thought-content and

feeling," p. 59. see also ib., 1912, "the nature of

perceived relations," where he says: "'introspection,'

divested of its mythological suggestion of the observing of

consciousness, is really the observation of bodily

sensations (sensibles) and feelings (feelables)"(p. 427 n.).

(3) can we observe anything intrinsically different from sensations? we come now to our third question concerning introspection. it is commonly thought that by looking within we can observe all sorts of things that are radically different from the constituents of the physical world, e.g. thoughts, beliefs, desires, pleasures, pains and emotions. the difference between mind and matter is increased partly by emphasizing these supposed introspective data, partly by the supposition that matter is composed of atoms or electrons or whatever units physics may at the moment prefer. as against this latter supposition, i contend that the ultimate constituents of matter are not atoms or electrons, but sensations, and other things similar to sensations as regards extent and duration. as against the view that introspection reveals a mental world radically different from sensations, i propose to argue that thoughts, beliefs, desires, pleasures, pains and emotions are all built up out of sensations and images alone, and that there is reason to think that images do not differ from sensations in their intrinsic character. we thus effect a mutual rapprochement of mind and matter, and reduce the ultimate data of introspection (in our second sense) to images alone. on this third view of the meaning of introspection, therefore, our decision is wholly against it.

there remain two points to be considered concerning introspection. the first is as to how far it is trustworthy; the second is as to whether, even granting that it reveals no radically different stuff from that revealed by what might be called external perception, it may not reveal different relations, and thus acquire almost as much importance as is traditionally assigned to it.

to begin with the trustworthiness of introspection. it is common among certain schools to regard the knowledge of our own mental processes as incomparably more certain than our knowledge of the "external" world; this view is to be found in the british philosophy which descends from hume, and is present, somewhat veiled, in kant and his followers. there seems no reason whatever to accept this view. our spontaneous, unsophisticated beliefs, whether as to ourselves or as to the outer world, are always extremely rash and very liable to error. the acquisition of caution is equally necessary and equally difficult in both directions. not only are we often un aware of entertaining a belief or desire which exists in us; we are often actually mistaken. the fallibility of introspection as regards what we desire is made evident by psycho-analysis; its fallibility as to what we know is easily demonstrated. an autobiography, when confronted by a careful editor with documentary evidence, is usually found to be full of obviously inadvertent errors. any of us confronted by a forgotten letter written some years ago will be astonished to find how much more foolish our opinions were than we had remembered them as being. and as to the analysis of our mental operations—believing, desiring, willing, or what not—introspection unaided gives very little help: it is necessary to construct hypotheses and test them by their consequences, just as we do in physical science. introspection, therefore, though it is one among our sources of knowledge, is not, in isolation, in any degree more trustworthy than "external" perception.

i come now to our second question: does introspection give us materials for the knowledge of relations other than those arrived at by reflecting upon external perception? it might be contended that the essence of what is "mental" consists of relations, such as knowing for example, and that our knowledge concerning these essentially mental relations is entirely derived from introspection. if "knowing" were an unanalysable relation, this view would be incontrovertible, since clearly no such relation forms part of the subject matter of physics. but it would seem that "knowing" is really various relations, all of them complex. therefore, until they have been analysed, our present question must remain unanswered i shall return to it at the end of the present course of lectures.