Correspondence between Hector Castañeda and Wilfrid Sellars on Philosophy of Mind*
edited by Andrew Chrucky
- March 6, 1961: Castañeda to Sellars
- April 3, 1961: Sellars to Castañeda
- April 13, 1961: Castañeda to Sellars
- November 14, 1961: Sellars to Castañeda
- November 27, 1961: Castañeda to Sellars
- December 8, 1961: Sellars to Castañeda
- January 17, 1962: Castañeda to Sellars
- March 11, 1962: Sellars to Castañeda
- August 8, 1962: Castañeda to Sellars
March 6, 1961.
Professor Wilfrid Sellars
Department of Philosophy
New Haven, Connecticut
For the first time, I have been able to read your "Empiricism and the Philosophy of Mind,"1 which so much impressed me when I heard it exactly 5 years ago. I have been studying sections 56ff on private episodes. I liked your view and I still like it. However, I am perplexed by certain features of it.
What you say about theories and models seems to me substantially sound, and the proposal to consider inner episodes as theoretical entities is very insightful. But in the way you propose it, it seems to me to do injustice to self-knowledge. You regard the basic language, the language of the model, to be behavioral language, so that my own mental states as well as yours are for me theoretical entities. Since this clearly won't do, you correctly suggest that first-person statements about thoughts acquire another role: a reportive one. This you claim on pp.188 and 189 [Sections 58-59] explains the privacy of the mental. But it is not clear how this claim is justified. (Let me say right now that following your suggestion, I argued in my Buenos Aires lectures last November 1a that the theoretical relationship goes from a multi-personal psychological language as a "scientific" theory to a basic language in which I can talk, somewhat solipsistically and impersonally, about behavior of any body as well as about my own mental states.)
Your whole discussion of how the reportive role (and the privacy) comes about is:
. . . Dick, using the same behavioral evidence, can say in the language of the theory, "I am thinking 'p' " (or "I am thinking that p"). And it now turns out - need it have? - that Dick can be trained to give reasonably reliable self-descriptions, using the language of the theory, without having to observe his overt behavior. Jones brings this about, roughly, by applauding utterances by Dick of "I am thinking that p" when the behavioral evidence strongly supports the theoretical statement "Dick is thinking that p"; and by frowning on utterances of "I am thinking that p", when the evidence does not support this theoretical statement. Our ancestors begin to speak of the privileged access each of us has to our own thoughts. What began as a language with purely theoretical use has gained a reporting role. (EPM, p. 189)
What exactly is it that Jones reports in the new use of "I am thinking that p"? How is it that he can make correctly such a statement without observing his behavior? It is not easy to see how on your view these questions can be answered, and if they are not answered it is difficult to see what exactly your view is accomplishing.
As far as I can see there are the following avenues of explanation of how Dick can learn to say correctly "I am thinking that p" without observing his behavior. But notice that all this is an elaboration I am imposing on your remarks:
- Dick learns to say "I am thinking that p" on seeing that he is in the circumstances that normally accompany the behavior which on the theory are grounds for saying that he is thinking that p.
- In addition to the ability mentioned in (A), Dick is allowed to substitute his utterance of "I am thinking that p" for his original behavior, linguistic or otherwise, which were grounds for the theoretical statement "I am thinking that p".
- Dick acquires the ability to grasp the inner episodes which he reports in his first-person utterances.
(C) certainly will do. But I have the feeling that it is not the one you are driving at. For one thing, you have not argued in your paper that the ability to grasp inner episodes presupposes other people who correct one's statements about the episodes in view of their relations to public, outer episodes. It is true that Wittgenstein (and Malcolm and Rhees)2 thought that there was an argument for that sort of conclusion. But you know my discussion of that argument, and the petitio or irrelevancy of the argument.3 For another thing, (C) can be argued directly without the roundabout route of the theoretical language. (In this simplified way it is perhaps an empirical truth that the language of the inner is learned on the basis of the language of the public.)
(A) provides no account of privacy. Another person may learn to say, noninferentially, "Dick is thinking that p" just on looking at Dick's circumstances. No reason is given in (A) to suppose that Dick will be more accurate than another person in his assertion, which somehow is a prediction of Dick's behavior. On (A) a person's statements about his own mental statements will be refutable. For as you say, he has been applauded when the behavioral evidence supports the statement "I (Dick) am thinking that p", and he has been corrected when the evidence does not support it. Thus, when Dick has finally learned to say, without observation of his own behavior, that he is thinking, that p, if he is not grasping the inner episode, he has learned to say something which requires his behaving in some appropriate way. By the very principles of the teaching, if his behavior is not appropriate, either he has not learned to say, "I am thinking that p" in the reportive sense or he has made an occasional mistake.
In short, there is nothing that can be properly called Dick's private access to his own thoughts. Hence, I conclude, you have in mind something different from (A).
(B) is a very interesting, alternative. It makes use of Wittgenstein's suggestion to consider first-person 'pain'-statements as pain behavior. The view can be developed as follows. After Dick has learned to say noninferentially "I am thinking that p" on seeing his circumstances and his behavior, he may do something of the sort envisaged in (A). And then he is taught to substitute his utterance of "I am thinking that p" for his normal behavior, whenever he is in the appropriate ircumstances. Thus, "I am thinking that p" acquires a performative use, in addition to its theoretical one. Now, the ordinary sense of "I am thinking that p" is the conflation of both uses:
- it is a substitute thought-behavior,
- it reports that the substitution has taken place,
- it asserts the theoretical statement,
- it expresses that (i) and (ii) give (iii) an authority which the conclusions of sound arguments have, but
- it does not say or assert that (iii) is a conclusion.
I will call that the Sellars-Wittgenstein view. It seems to me a very strong one. It can account for the privacy of ones' own thoughts: Only the speaker can do (i). It also accounts for the incorrigibility of first-person statements about present thoughts. Because of i, my "I am thinking that p" is always evidence for the theoretical statement that I am thinking, provided the circumstances are the right ones. Nobody can correct me, once I have learned the language of thoughts. (Note that the two conditions -- private access and incorrigibility -- are distinct features of the mental.)
But there is another requirement of the mental which is not satisfied by the above view: the incorrigibility of first-person judgments about present thoughts. On this view, it is the utterance which constitutes the behavior, which is evidence for the inner episode, and I would know of the theoretical inner episode only through my overt statement "I am thinking that p". Clearly there are inner episodes or thoughts which are not accompanied by any behavior. Of course, Jones' theory does not have to deny that. But in your description of how "I am thinking" gains a reportive role, interpreted or developed as in (B), I could not know of them unless there are some overt signs: in either case the relevant behavior would consist in my utterance of "I am thinking that p". Yet the point is that there are inner episodes of whose occurrence I know prior to my making any statement (overtly), and which on many occasions I do not report. The point is that even if I reported all my inner episodes as described in (B), I would not know of their existence through my statements. Their occurrence entails my knowledge of them. This sort of incorrigibility which is one half of the important privacy of the mental is not accounted for by (B).
There are other points, but this has become too long. I am certain now that awareness has to be added to the basic language. So that the theoretical link you have rightfully (it seems to me) emphasized does not sew behavior to one's mental episodes or to the behavior and mental episodes of others.
. . .
Yours as ever,
[Table of Contents]
April 3, 1961.
Professor H.N. Castañeda
Department of Philosophy
College of Liberal Arts
Wayne State University
I have finally had time to go over your letter more thoroughly and to appreciate the problems and difficulties you raise. I quite agree that the passage in which I explain how the theoretical language of thoughts acquires a reporting role is terse and skeletal in the extreme. It obviously is not an effective way of making the point I wished to make, since you have not understood it. Let me make another try, taking my point of departure from view of certain features of the pre-Jonesean stage of my mythical community.
1. Everybody, as we would put it, always 'thinks out loud'. Life is built around talking. Although, as we post-Joneseans know, the overt verbal behavior involved in talking is the culmination of inner episodes which begin with thoughts, this fact is not realized by pre-Joneseans. There are, therefore, two key differences between pre- and post-Jonesean citizens. (a) Pre-Joneseans do not have the concept of thought as contrasted with talk, though, of course, they can distinguish between real talk and parroting talk. They no more conceive of talk as the expression of thought than, before chemical theory got under way, people conceived of chemical processes as the manifestation of molecular processes. (b) Pre-Joneseans have thoughts (though they are not aware of themselves as having them), but these thoughts never occur without culminating in overt verbal behavior. (The 'thinking out loud' makes sense only as contrasted with 'thinking to oneself'.)
2. From our vantage point, therefore, thoughts originally occurred as initial stages of the unitary process
Θ → V
where 'Θ' stands for thought and 'V' for overt verbal behavior. Doing, seeing, intending, recognizing, etc. were conceptual or rational as threaded on what we would call 'thinking out loud'.
3. But, of course, although 'Θ → V' is the natural unit with respect to acquiring a conceptual apparatus (after all, we still teach children to think by teaching them to talk), actually the process can be truncated. Behavior which was originally conceptual as built on 'thinking out loud' (Θ → V), can become conceptual as being built on Θ without V. Behavior which was originally a function of (Θ → V) becomes a function of Θ.
4. Thus, by the time Jones comes on the scene, thinking has acquired a measure of autonomy. He develops a theory according to which the overt verbal behavior involved in talk is the culmination of inner episodes (Θs) which are analogous to talk. According to the theory the difference between Dick hunting caribou talkingly and Dick hunting caribou silently is that in the former case the hunting behavior is built around (Θ → V) whereas in the latter it is built around Θ without V.
5. Thus the thought that there is a caribou ahead is analogous to a meaningful utterance of the sentence which corresponds in the language of the community to "There is a caribou ahead". But whereas once upon a time all hunting was built around overt utterances of hunting talk, now people hunt sometimes talkingly, sometimes silently.
6. So far only Jones has the theory of thoughts. Other people recognize that when Dick stands still, nostrils sniffing, he would say "There is a caribou ahead" if he were asked. But in one important sense they would have no explanation of this hypothetical fact. Compare knowing that x is soluble with knowing the theoretical explanation of why it is soluble.
7. Jones teaches his fellow citizens the theoretical framework of thoughts as inner episodes which naturally culminate in overt non-parroting verbal behavior.
8. The non-verbal behavior which, to one watching Tom who always hunts out loud but not in a position to hear him, would serve as evidence that at a certain time he was saying "There is a caribou ahead. I shall slow up and take cover" will, given the framework of the Jonesean theory, warrant the inference that Dick, who hunts silently, is thinking (but not saying) .There is a caribou ahead; I shall slow up and take cover..
9. Dick, understanding the theory, now has two inferential routes to the idea that someone (it may be himself) is thinking ·There is a caribou ahead; I shall slow down and take cover·: (1) The non-verbal behavior characteristic of that stage of a hunt. (2) The overt verbal behavior which is the natural culmination of the thought.
10. The stage is now set for the interpretation of the passages of EPM which you quote at length and puzzle over. The important thing to note is that the core of Dick's learning to report what he is thinking is a matter of his acquiring a tendency (ceteris paribus) to respond to his thought that-p by saying "I am thinking that p". Everything hinges on the force of the word 'respond' in this connection. It is being used as a technical term borrowed from learning theory. The following diagram will help clarify matters:
MΘi . → . MVi
where Θi is a thought that-p, MΘi is a meta-thought ·I am thinking that-p· and MVi is a meta-statement 'I am thinking that-p'.
11. The connection between Θi and MΘi is in the first instance a conditioning and not an inference. As such it presupposes neither an awareness on Dick's part that he is thinking that-p nor any recognition on Dick's part that the circumstances are such as would usually involve his thinking that-p. It requires only that the reinforcer (applauder), in this case Jones, correctly infers that Dick is thinking that-p and, given that Dick happens to say "I am thinking that-p", applauds. Needless to say, Dick will be most unlikely simply to happen say "I am thinking that-p". He can be led to say it, for example by asking him "What are you thinking?" (remember that Dick knows the theory). But we need not suppose that he painstakingly argues to the conclusion "I am thinking that-p". All we need suppose is that circumstances can be arranged which bring about the joint occurrence of a thought that-p and a saying of "I am thinking that-p" with sufficient frequency for a conditioning by reward (applause) to occur. Thus the decisive feature is that the connection between Θi and (MΘi → MVi) is a direct non rational S-R connection. Certainly this S-R connection exists within a rich conceptual context, but unless it existed as an S-R connection, there could be no direct noninferential self-knowledge. I do not need to emphasize that the connection is only a necessary condition of direct, noninferential self-knowledge. It is not a sufficient condition, a point to which I will return in a moment.
12. Although to make the above points I have supposed that it is Jones who trains Dick to respond to the thought that-p with the statement "I am thinking that-p", actually Dick could have done it himself, another point to which I will return in a moment.
13. Now the important difference between a person who has merely been conditioned to respond to his thought that-p by saying "I have the thought that-p" and a person whose statement "I have the thought that-p" expresses direct self-knowledge is not that in the latter case the statement isn't occurring as a conditioned response. It is. The difference is that in the latter case the conditioning is itself caught up in a conceptual framework.4 Compare the case of a child who has merely been conditioned to respond to green objects (in standard conditions) by "This is green", as contrasted with a child whose utterance of "This is green" expresses direct noninferential knowledge. Indeed, the key to the account I have given of the direct-noninferential knowledge of inner episodes is the apparatus I developed in discussing the status of noninferential perceptual knowledge. (See the discussion of 'epistemic authority' in the section entitled "Does Empirical Knowledge have a Foundation?" [in EPM]) I simply assumed that anyone who worked through the latter would see how the relevant distinctions applied to direct self-knowledge.
14. An important difference between the two cases, of course, is that when the child is learning to respond to perceptible things with observation sentences, the latter do not have conceptual meaning for him until he has acquired the system of connections which are essential to the conceptual character of perceptual discourse. Thus the training must be done by one who already knows his way around in the language. Dick, on the other hand, is already using 'I have the thought that-p' as an expression in a theoretical language which in its turn rests on a rich conceptual structure. A case for you to ponder on is that in which a blind man who has learned the language of colored physical objects and the seeing of them has his visual apparatus put in order and shortly thereafter is able to see that we have a case in which language, already meaningfully used, acquired a 'reporting role' expressive of direct, noninferential knowledge which it did not, for that person, have before.
15. The above type of account explains the 'privileged access' a person has to his own inner episodes. For (although worlds are conceivable in which this is not the case) only the person who has a thought that-p can respond to it (in the manner discussed above) with the thought that he has the thought that-p.
16. As for incorrigibility, it does not extend, of course, to one's beliefs. These we can be mistaken about. As in the case of our direct knowledge of our sensations and feelings, the incorrigibility is in large part accounted for by the 'minimality' of the cognitive claim that is being made. There are no conceptual ways to go wrong, unlike the case of "There is an inkstand on the table".
17. "I (just) had the thought that-p" must be distinguished from "I believe that-p". It is also worth noting that the episode which is the thought that-p can occur in many contexts, and the incorrigibility must pertain to the episode characterized with a minimal commitment concerning this context. Notice that there is an important sense in which the thought **I (just) had the thought that-p** contains the thought that-p. This analytic incorrigibility must be distinguished carefully from the real incorrigibility which would be the counterpart of the incorrigibility of self-knowledge as it pertains to sensations and feelings.
I would be interested to learn more about what you have in mind by 'awareness'. If what you mean is direct noninferential knowledge of particular matter of fact, then my theory insists on awareness and offers an account of it. But perhaps my attempt to give an account which avoids 'the myth of the given' has not persuaded you.
I am afraid that the latter part of this letter is a bit disconnected and rambling. I hope you find it of some use.
[Table of Contents]
April 13, 1961.
Thank you very much for your extensive discussion of how knowledge of one's thoughts is noninferential within the framework of your views of inner episodes as (like) theoretical entities. You have clarified many points of doubt and questions I had. Some of the things you say are (it seems to me) worth adding to your famous essay.
There are, however, some things that are still unclear to me.
1. I am sure I understand your proposition that self-knowledge involves a meta-thought which is a conditioned response. And I would like to have an argument to show that that is so.
2. Using your own abbreviations, without subscripts, I can formulate three different propositions you mention and discuss, without separation:
- MΘ is a conditioned response to Θ (your section 11 top).
- MV is a conditioned response to Θ (your section 13 top).
- (MΘ → MV) is a conditioned response to Θ (your section 11, second half).
And we could add:
- MΘ, which is involved in (MΘ → MV), is a conditioned response to Θ.
3. You seem to be interested in establishing A as a necessary condition for noninferential self-knowledge. But in section 11 you at most establish C. It is clear that C does not entail A, even though perhaps A entails D, it is doubtful whether A entails C.
4. It might be thought that, given the way your Jones introduced thoughts as theoretical entities, when B holds we would be entitled to move to A. And it might be thought, besides, that A & B are logically equivalent to C. But neither thought seems to me to be justified. The latter would require that C entailed A, which is not the case.
Of course, you could define A & B as equivalent to C. But there is a good sense of 'conditioned response to' which would make that definition dangerously misleading. Inasmuch as the procedure you discuss in the second half of No. [section] 11 establishes C, you might, by definition, say that that is the same as A & B. But inasmuch as B holds (with that definition in mind), i.e., inasmuch as the utterance of "I have the thought that-p" is conditioned on the theoretical entity Θ, we have a good reason for saying that it does not issue from a thought. A conditioned response is normally one which involves no thought, one which is done without thinking. So, your Jones' theory of thoughts should require us not to suppose that MV issues from a thought, namely, MΘ.
Thus, inasmuch as you (may) define A & B in terms of C, C entails A & B, and vice versa, but then you are introducing a new sense of the term 'conditioned response to'. If you are using the latter in its normal sense, as you want to, as 'direct non-rational S-R connection', it would seem that conditioning to react by MV should not require MΘ.
5. But what perplexes me most is the same I mentioned in my first letter. In your new terminology my perplexity is "What does it mean to say that Dick has learned to react by MΘ to Θ?"
Compare the following case which is on all fours like your theoretical inner episodes:
(Z) Dick shows all the signs (criteria, symptoms, what you care to call them) of a person with a colony of certain filterable viruses lodged in his left kidney. He is taught the theory of viruses so that he can infer from his signs that he has a colony of the viruses in question. He is able, then, to make the theoretical statement "I have a colony of viruses . . .". Now, what does it mean to say that Dick can be conditioned to react to the colony of viruses, or to the presence of the colony (if you like)?
The central point is that theoretical statements are inferred. You try to minimize this in sec.11, but that is precisely what it seems to me you cannot take for granted.
In the situation described in (Z) I can understand (or so it seems to me) what you can mean by saying that Dick learns to use "I have a colony of viruses . . ." noninferentially. We conditioned him to say that on perceiving (or feeling) all or some of the signs in question. The important point here is that the S-R connection, if you want to say that, is not colony-utterance, but signs of the colony-utterance.
Dick no longer infers; you may say that he speaks as if "colony of viruses lodged in one's left kidney" were analyzable in terms of . . . .
Now, if that is what happens in the case of the theoretical entity Θ and relations, we would have to say that Dick is conditioned to utter "I am thinking that-p" or "I have just had) a thought that-p" on inspection of his behavior and circumstances. But this is just what we do not want.
6. In situation (Z) one can say that when Dick is conditioned to utter "I have a colony . . ." in that way, he is conditioned to react by uttering in the (presence of the) colony. But in this sense, we are merely devising an alternative way of talking, and the facts are as in my sec. 5 above: the R is an uttering and the S is the set of signs of the theoretical entity Θ.
7. In situation (Z) one can say that Dick has been taught to think that there is a colony upon thinking that such and such (the signs) are the case. We may feel tempted to say that his thought that there is a colony is a conditioned response to his perceiving (and/or feeling) the signs. (I wonder whether a line of talking of this sort is not related to your discussion in sec. 11. Probably not. For it is not helpful. We would have to say that in sec. 11 you describe how Dick could be conditioned to react by MΘ to his thought that his behavior and circumstances are such and such.)
8. In general, perhaps you are right in saying that self-knowledge is essentially (in part) being conditioned to respond by MΘ to Θ. But I do not see how that is to be done on the view that Θ (and therefore MΘ) is a theoretical entity.
Of course, it is consistent with the rest of your story to say that your Jones introduced the theory of thoughts with the postulate that (sometimes) Θ is followed (or reacted to naturally, I would suppose) by MΘ. But I would not know how to understand that postulate except as a way of saying that Θ's are not really theoretical entities. Remember that Θ's are one's thoughts.
In other words, as I said in my previous letter, it seems to me that your analogy with theoretical entities will do only for another person's thoughts, etc. Behavior and one's own inner episodes are the entities of the model, on whose analogy the others' inner episodes are conceived (or invented). But of course the language of my inner episodes is just a language about inner episodes (which are nobody's), and in a sense it is private.
9. I take issue with your last remark "But perhaps my attempt to give an account which avoids the 'myth of the given' has not persuaded you."
In the above suggested modification of your view there need not be any given in the sense I have understood to be objectionable: in the sense that each word gains its meaning completely by being associated with a given. The inner episodes (which are mine) which are on the side of the model, not on the side of the theory, will have to be thought through a conceptual framework. (In my discussion of the private language, your comments on which I still would like to receive, I tried to distinguish the issue about privacy from the issue about concept-empiricism.) 5
I hope you can spare some time to explain to me what exactly you mean by "MΘ is a conditioned response to Θ", and how it differs from the alternatives mentioned above, the only ones I can now think of.
Yours as ever,
[Table of Contents]
November 14, 1961.
If you still have a copy of your letter of April 13 at hand the following remarks may help clarify some of the difficulties you had with my account of direct self-knowledge.
1. I take it that you agree that direct self-knowledge involves occurrence of meta-thoughts and that your primary uneasiness concerns the concept of thoughts being conditioned responses to thoughts. I am not quite sure of this diagnosis, because I am not clear that the awarenesses you wish to introduce are thoughts, i.e. mental episodes having `propositional' form. (Have you had a chance to read the opening sections of my paper on "Being and Being Known"? 6
2. The discussion in sections 2ff of my letter of April 3 contains, I believe, the answer to the question you raise in your section 2 and 3. The first concept I introduce is that of a conditioning which relates the unit (MΘ → MV) with Θ. But given (1) that MΘ can occur without MV, and (2) that since the other term in the conditioning is a thought (Θ), the connection between (MΘ → MV) and Θ must catch hold of the former at the MΘ rather than the MV end, the direct connection between Θ and MΘ is the heart of the matter and is one which can survive the state in which MΘ occurs without its expression by MV. That is to say, it survives the training stage in which the utterance of MV in circumstances in which the trainers know that Θ occurs is the occasion for reward.
3. Your statement in section 4 that "a conditioned response is normally one which involves no thought, one which is done without thinking" is true only in the sense that the term `conditioned response' is most frequently used in connection with simple animal behavior. This is, however, an accidental feature of the concept. Again, it is most frequently used (by American psychologists in the Skinnerian tradition) in such a way that the overt is conditioned to the overt. This is bound up with the reluctance of many psychologists to speak in terms of establishing connections between postulated central processes; and in general, with an anti-theoretical bias. If the expression `conditioned response' (pace Pavlov) is to mean by definition connection between observables, then, of course, I can't use the term in making my points. I shall, therefore, substitute the more neutral term `associative connection'. If this is done, then your statement becomes: "an associative connection is normally one which involves no thought, one which is done without thinking." It now leaps to the eye that in one sense this is obviously true, and in another obviously false. Clearly to come to think of something by association is the opposite of coming to think of something in the course of rational inquiry (which is not to say that associative connections are not involved). But, and this is the essential point, there obviously is such a thing as the association of thoughts. For although there can be associative connections at the sub-conceptual level between images, and although Hume confused between images and ideas in the conceptual sense, what he and his successors called `association of ideas' was really 'association of thoughts' and in this sense "involves thought".
4. Your discussion of Dick and the virus infection is vitiated by the sentences. . . we condition him to say "I have a colony of viruses" on perceiving (or feeling) all or some of the signs in question. The important point here is that the S-R connection, if you want to say that, is not colony-utterance but signs of the colony-utterance.This is simply a petitio. You are assuming that we can't train Dick to say "I have a colony of viruses", when he has one, in a way which is not mediated by his noticing "signs of the colony". And you are assuming this because you are assuming a general principle (of which your example is intended to be an illustrative instance) to the effect that
(P) To train a subject, S, to say "I am in state φ" when he is in state φ, S must either observe that he is in state φ (where this is possible) or observe that he is in a state which is a sign of state φ.
This is false. Consider the implications of the following example:
4.1. We ask Dick to say "I am in state φ" every two minutes or so.
4.2. Sometimes Dick is in state φ, sometimes he isn't.
4.3. When he is, we know that he is by observing symptoms Ui . . . Uj and using a well confirmed theory.
4.4. We put him in a position where he cannot observe these symptoms.
4.5. When he says "I am in state φ" and actually is in state φ we reward him; in the contrary case we give him a mild shock.
4.6. It turns out that for some states φ (but by no means for all) we can bring about a connection between being in state φ and saying "I am in state φ".
4.7. Roughly the difference between the cases where it can be done and the cases where it can't is that in the favorable cases, being in state φ, causes neural impulses which feed into the central nervous system in such a way that they can be hooked up with the neural processes which culminate in the utterance of "I am in state φ", i.e. (on my view) with the thought **I am in state φ**.
4.8. The neural impulses from state f need not be accompanied by sensation or feeling.
5. If we modify your example to make it somewhat more episodic in character -- thus by supposing that virus colonies oscillate between growing rapidly and decreasing rapidly in number, and choose the growth of the colony, rather than its mere presence, as the state to be `reported' -- then my point is that Dick, by analogy with my myth, is trained to say "I have a rapidly growing colony of viruses" when such rapid growth occurs, and not when he notices observable signs of such growth. Contrary to your example, the `S-R connection' is not signs of the colony growth-utterance, but colony growth-utterance.
6. I believe that a careful rereading of my section 11 would highlight the point of the above discussion and blunt your charge that in talking about the establishing of a connection between MΘ and Θ I am merely devising an alternative way of talking about the establishing of a connection between MΘ, or, perhaps MV, and "the set of signs of the theoretical entity Θ".
7. I am puzzled by your remarks in section 8. Why should the idea that Θ and MΘ are theoretical entities be inconsistent with my account? You must be thinking that if an entity is a theoretical one, then it doesn't really exist, and can't, therefore, have genuine causal connection with other theoretical entities, let alone with overt verbal behavior. If so, you have missed an essential feature of EPM. Thus, whereas according to the 'positivistic' account of theoretical entities, 'Xs are theoretical entities' has roughly the sense of 'X is a non-logical symbol in a calculational framework for ordering our knowledge about (observation) entities', on my view `Xs are theoretical entities' has roughly the sense of `Xs are entities whose existence has been established by theoretical reasoning'. Thus whereas on positivistic views, 'Xs are theoretical entities' is inconsistent with 'Xs really exist', on my view the former entails the latter. Notice that the entities postulated by a bad theory (e.g. phlogiston) are not examples of `things having a purely theoretical existence'; they do not exist, period -- though at one time they may have been thought to exist.
8. I am taking the liberty of enclosing a paper of mine on the "Language of Theories".7 It contains a better formulation of my views on the above topic than is to be found in EPM.
9. I believe that I have put my finger on the fundamental sources of your puzzlement. If I am right about this, you should now see exactly what I had in mind in EPM when I concluded that the correct contrast between other people's mental states and our own is that between theoretical entities and theoretical entities plus, rather than, as you suggest, between theoretical entities and non-theoretical entities. It must be remembered that while we have direct noninferential knowledge of our mental states, we do not observe them. We have them, and we know that we have them. Needless to say, the conditioning discussed in this letter, while it is a necessary component in the ability to know what is going on in one's own mind, is not sufficient to account for it. For a discussion of the more that is involved, I refer you to section 13 of my letter, and to the sections of EPM to which it refers. I discuss the point in the paper on "Phenomenalism" in somewhat more general terms. 8
10. I have just been going over your critique of Wittgenstein and Malcolm on 'private languages'.9 It is a devastating analysis. I will send you some comments on it (or a later version, if you have one). My own view is that the most that a critique of private languages can hope to show is that in principle private language must be an offshoot from, and exist in the ambience of, a public language. It must be, that is to say, `derivative,' and `incomplete', -- a sub-language rather than a language.
[Table of Contents]
November 27, 1961.
Your letter of November 14th has certainly clarified a lot. I think that now I understand much better what you meant in EPM by saying that after teaching his theory to Dick, Jones trains Dick to use "I am thinking that p" in a reporting role. I can now press the queries of my first letter.
I.1. If I were to concede that the connection between my thought that p [Θ] and my thought that I am thinking that p [MΘ] is as contingent as just a non-rational association of thoughts, I would certainly concede that the following propositions are all self-consistent:
- Dick is made (trained) by Jones to say "I am thinking that p" [MV] when he shows all the signs of Θ, without observing those signs.
- After (1), Dick acquires the ability to have MΘ, corresponding to MV.
- After (2), Dick acquires the ability to utter MV and have MΘ even in the absence of the signs of Θ, on the occurrence of Θ.
- After (3), Dick acquired the ability to have MΘ, even without MV and even without the signs of Θ, just on the occurrence of Θ.
2. But it seems to me that you do not just claim that those four proposition express logical possibilities. You claim that they are true. Yet you have not, as far as I can see, tried to show that they are true. Notice, incidentally, your "the connection between (MΘ -> MV) and Θ must catch hold of the former at the MΘ rather than the MV end, . . ." (my italics). If your account is to be adequate, it certainly must be so. But your right to speak of `must' is precisely what is at stake.
3. From the statements you have made about Jones' "theory" (in EPM and in your letter of April 3) it does not seem to follow that (1)-(4) must be true.
II.1. Suppose Dick is having Θ without exhibiting any of the behavioral signs of Θ and that he does not utter V. According to Jones' theory of thoughts, nobody, including Dick, can know that he has Θ. This seems to be false.
2. Suppose further that Dick has MΘ. This in no way guarantees that he has (or just had) Θ. For even if he was successfully conditioned or made (e.g. by eating a specially spiced turnip pie) to develop the abilities mentioned in (1)-(4), all we can be sure of is that if Dick has Θ he will have MΘ, but not vice versa. Of course, you can add to Jones' theory the proposition that the occurrence of MΘ entails the occurrence of Θ. This entailment seems to me obvious, regardless of whether Θ is the mental act of judging, or the act of just considering (with or without judgment), provided MΘ is taken accordingly.
3. As far as I can see, just as (1)-(4) are self-consistent, so are their contraries and their contradictories. It seems to me that in your view, given that Θ and MΘ are only contingently associated (after all training is required for such association) the following is a live logical possibility:Dick thinks that he is thinking that it is raining, but he is mistaken, even though he has learned Jones' theory and has been subjected to Jones' conditioning.
4. On your view it seems to me that a careful person can sensibly answer the question "What are you thinking?" by the statement: "I do not know I have not observed my behavior well enough. I have an inclination to think that I am thinking that it is raining, but Jones conditioned me so long ago that I am not sure whether my meta-thought is the reaction to its corresponding thought (in content) or to something else, perhaps a headache. Why don't you observe my behavior and tell me what I am thinking, so that I can check my meta-thinking reactions?" Why shouldn't Dick's ability to have meta-thoughts need a booster training?
5. My claim is that "X has Θ at t and at t he has the ability to think that he has Θ(i.e., he has the concept of thought of thinking) and at t he attends to his mental goings on" entails "X knows occurrently at t that he has Θ" and since the latter entails "X has MΘ," so does the former. Also, "X has Θ at t and at t he has the ability to think that he has Θ" entails "X knows at t (in the dispositional sense) that he has Θ".
III.1. I take it that those entailments are versions of the propria persona principle. Now, I have been unable to find your argument against that principle. In "Being and Being Known" sec. 57 you say "I have argued that this principle is without foundation", the argument could only be in sec. 25-30, but it is not easy to locate. The only plausible candidate is:It is a serious mistake to suppose that merely by virtue of having sensations we experience sensations as sensations (do animals experience sensations as being sensations?) . . . (sec.25).
That this is the argument is suggested by your "I shall shortly be arguing that the same situation obtains of our concepts of intellectual acts" (sec. 25). But that argument seems to me inconclusive. All that it shows is that experience of sensations, like experience of physical objects, requires the employment of concepts. Either I have been utterly unable to understand you, or you should also conclude that chairs and brains are never experienced qua physical objects.
2. It seems to me that there are at least five different things which you have attacked, and I find it very difficult to separate the several arguments which are needed:
- The abstractive theory of concept formation;
- The propria persona principle;
- The thesis that sense is cognitive and belongs to the order of intentionality;
- Conceptual atomism (by this I mean, roughly, the thesis you oppose by your "One can't have one intellectual word in one's vocabulary without having many");
- "The idea that observation is constituted by certain self-authenticating non-verbal episodes, the authority of which is transmitted to verbal and quasi-verbal performances" (EPM, sec. 38; note that E is only one part of your statement of the myth of the given).
It is not clear to me that your argument against the myth of the given entails that B and E (interpreted in a certain way) are false.
This is all I can think of at the moment. Awaiting your comments,
[Table of Contents]
December 8, 1961.
Our long distance dialogue seems to be making headway, and I hasten to make my joinder to your letter of November 27.
Ad I.1. You do not seem to have taken seriously my caveat in sec. 9 of my letter of November 14 (which echoes sec. 11 and 13 of my letter of April 3) to the effect that the tendency to respond to Θ with MΘ is a necessary but not sufficient condition of the ability to know directly or non-inferentially that one has Θ. See particularly sec. 13 of April 3. I have been careful not to say that the connection between Θ and MΘ is "as contingent as just a non-rational association of thoughts". Please confirm this.
Also Ad I.1. In your formulation of propositions (1)-(4) you seem to overlook that Dick was taught the theory of thoughts before being trained to respond to Θ with MΘ. Thus, contrary to your proposition (2) he already has the ability to have MΘ before stage (1). In (3) and (4) I would replace 'ability' by 'tendency'.
Ad I. 2. I have, indeed, tried to show that something like propositions (1)-(4) is true. The general form of the argument is: The thesis is consistent with and, indeed, in the spirit of contemporary psychological theory, and, if true, would undercut classical puzzles about other and (to use Roger Buck's useful phrase) non-other minds. In short, I think that the thesis I advance "saves the appearances".
Ad II.1. It is indeed the case (and necessarily so) that, as long as Jones' theory remains a mere theory and has not acquired, for anybody, the additional role or dimension of use I am attempting to explicate, no one can know that someone (even himself) is thinking that-p without inferring this from behavioral evidence. But that all such knowledge must be inferential follows from the meta-theoretical premiseJones' theory is a mere theoryand not from the theory itself. Indeed, it is just because Jones' theory suggested to him the possibility of training Dick to use the language of the theory in the way described, and hence the possibility of teaching him to know what he is thinking otherwise than by using the theory as a theory, that Jones was led to his momentous experiment. Thus Jones' theory does not say that "nobody, including Dick, can know that he has Θ if he is having Θ without exhibiting any of the behavioral signs of Θ and . . . does not utter V".
Ad II. 2. I should, perhaps, have made more explicit the fact that the conditioning is a matter of 'MΘ only when Θ' as well as 'MΘ when Θ'. Indeed, it is more a matter of 'only when' than 'when'. Compare the case of the child learning the use of "This is green". He is not trained to respond always to all green objects (in standard conditions) with "This is green"; the response tendency is built into a system of response tendencies, Thus, only ceteris paribus does the trained child respond to the presentation of a green object with "This is green". On the other hand, he is trained to make this response only when presented with a green object (in standard conditions). It is this only when feature of the training which is the ultimate foundation of the trans-level inferenceThe thought that I am thinking that-p has occurred to me in such and such a manner and context. Therefore I am thinking that-p,the availability of which gives 'epistemic authority' to non-inferential thoughts of the form "I am thinking that-p". Compare the role of the availability of the trans-level inference (oversimplified)The thought that this is green has occurred to me in such and such a manner and context. Therefore (I am seeing that) this is green.which gives non-inferential thoughts of the form "(I am seeing that) this is green" the status of non-inferential knowledge.
Ad II.3. There are subtle and important issues in this neighborhood. The first point to be made is that we don't say that X is able to see that something is green unless not only has he been "subjected to conditioning" but the conditioning has been successful. Thus there is a logical absurdity inDick has the thought (which he has not arrived at by inference) that he is thinking that-p, but he is mistaken, even though he has been successfully trained to have thoughts of the form 'I am thinking that-p' only when he is thinking that-p,which parallels, at a more complicated level, the absurdity ofIt has thundered.The second point to be made is that even at the hypothetical stage when all thinking was thinking-out-loud, rationality involved not only the ability to say that-p, but also the ability to say such things as "I have just said that-p", "I shall say that-p", and, especially, "I am disposed to say that-p". And already at that stage there was the ability to have non-inferential knowledge (knowing-out-loud) of the form
Only when there is lightning does it thunder.
But there was no lightning.I have just said that-p.For non-inferential autobiographical knowledge of what one is thinking (whether one conceives of this thinking in pre-Jonesean or post-Jonesean terms) is essential to rationality. In short, the ability to think in the full sense of the term involves the ability to meta-think, and the ability to have noninferential knowledge of what one is thinking. Thus it is misleading to represent my position as one according to which non-inferential meta-thinking could be something which one was conditioned to do "so long ago that perhaps the conditioning has worn off, and one has random meta-thoughts." To lose the tendency to have appropriate meta-thoughts is to cease to be rational or lose one's mind.
Ad II.5. I am not clear why you think that I would disagree with this. Perhaps I can put my finger on a possible source of misunderstanding in the course of commenting on the concluding section of your letter.
Ad III.1. Unless I misunderstand you, you misinterpret the 'propria persona' principle. For the 'entailments' you mention (and which I accept) are not versions of the propria persona principle (which I reject). What I was criticizing under the latter head is the idea that if one has direct noninferential knowledge of an inner state or episode, the concept in terms of which that inner state or episode is thus non-inferentially known cannot be a mere analogical extension of a concept pertaining to public phenomena, but must directly and adequately express the true nature of the episode in question. Usually philosophers who have accepted this idea have done so under the influence of the abstractive theory -- but a believer in innate concepts could make a similar claim, arguing, perhaps, that if an inner episode is 'directly present to knowledge' its nature must 'lie open to scrutiny' and present no mysteries. He would grant, of course, that there might be mysteries, even insoluble mysteries, about its relationship to other episodes (e.g. brain states or disturbances in the animal spirits).
As I see it, the proposition 'Direct self-knowledge requires non-analogical concepts' is by no means self-evident, and owes its plausibility to some form of the idea that the necessary and sufficient condition of such direct knowledge is the 'direct presence' of the episode known. (Of course, if we mean by the direct presence of an item to the mind that some truths about it are being directly known (or in a position to be directly known) then the connection between 'direct knowledge' and 'direct presence' is analytic and unilluminating. Traditionally, however, more was packed into the notion of 'direct presence', the metaphor was not sufficiently sterilized. Often it was assumed that the only objects which were 'directly present' to the mind were its own states, e.g. impressions, feelings, thoughts. Thus it was assumed that the necessary and sufficient conditions of directly knowing that we are having such and such an impression or thought was the mere fact of having the impression or thought. At the back of these philosophers' minds was the idea that 'direct presence' leaves no room for factors which might distort or mystify.
These rough and ready remarks may help clarify the role of sec. 25 in the argument of "Being and Being Known". The form of the argument is (roughly):If one thinks that merely by having sensations one experiences them as sensations, one will conclude that our knowledge of sensations is non-analogical and adequate. But in previous sections I have surely shown that to have a sensation is not to experience it as a sensation (i.e, to know directly that it is a sensation), nor is it a sufficient condition of such knowledge. Therefore the suggestion that concepts pertaining to sensation are analogical merits consideration.That these concepts are analogical is argued by developing a theory to that effect which accounts for the facts which Thomists have attempted to explain by assimilating sensation to the order of intentionality.
Ad III.2. Whether or not what you take to be the 'propria persona principle' is false must await a further exchange. As for E, it is clear that the burden of my attack was on the idea of self-authentication. For I clearly think that there are non-verbal episodes (quasi-verbal in the sense that they are construed by analogy with verbal episodes) which are direct or non-inferential knowings. What I have attempted to do is show how their 'epistemic authority' involves 'trans-level' inference tickets of the form "The thought that-p occurred to me in such and such a manner and context, therefore, p.
[Table of Contents]
January 17, 1962.
Your last letter has clarified some points, but has confused me more concerning others. For this reason, I want to begin this letter with a somewhat fresh start on the part of your view that perplexes me most, later on I shall make definite comments on the paragraphs constituting your letter. I will refer to your letters by month and paragraph number.
1. I take it that your account of self-knowledge centers mainly on the thesis that MΘ is a conditioned (or acquired) response to Θ. This is, you have emphasized, only a necessary condition for MΘ to constitute direct noninferential self-knowledge of one's own thoughts. What else is needed is not clear to me, but this point will be discussed below.
Now, that conditioned-response thesis can take one of three forms:
(CR.1) One is conditioned to react by having MΘ when one has Θ.
(CR.2) One is conditioned to react by MΘ only when one has Θ;
(CR.3) One is conditioned to react by having MΘ when and only when one has Θ.
From December Ad II.2, I gather that you mean to hold (CR.3): you speak of "'MΘ only when Θ' as well as 'MΘ when Θ'" (my italics).
2. I have always understood that your view was not a conclusion that followed deductively from some given premises, but was a proposed way of looking at mental concepts (a theory, in the etymological sense), which was to be defended by showing how it "saves the appearances". My claim has been that there is an "appearance" that is not saved, so that your theory needs correction. (Remember that in my first letter of March 6, 1961, I expressed my acceptance of a great part of it.)
3. My claim is the one I formulated in my first letter: the ordinary language of the mind is such that in the case of thoughts (as well as in the case of experiences or states of consciousness) there is, inter alia, an incorrigibility and a privacy that is not saved by your theory. (By separate mail I am enclosing a new, much-revised version of my discussion of the private-language argument(s), on page 12B of which you will find a list of what I like to call "meaning postulates of pain", which I believe to be logically true also for thoughts and all feelings, mutatis mutandis.)10 Such "appearances" are the following necessary truths:
D. (Direct Access.) If X thinks at t that p, and at t X is capable of thinking that he thinks that p (i.e., has the concept of thought), and attends to his mental goings-on, then X knows occurrently at t that he thinks at t that p.
E. (Incorrigibility of judgment.) If X thinks at t that he thinks (or is thinking) that p, then at t X thinks (or is thinking) that p.
ES. (Incorrigibility of statement.) If X states at t that at t he is thinking that p, his statement is true.
4. My argument has assumed that the following propositions are logical truths:
(C1) 'X can be conditioned to react by R when S occurs' entails 'it is possible that S occurs without R'.
(C2) 'X can be conditioned to react by R only when S occurs' entails 'it is possible that R occurs without S'.
That (C1) and (C2) are the minimal assumptions I need became clear to me after reading your December letter.
Now, from E it follows by the contrapositive of (C2) that one cannot be conditioned to react by having MΘ only when one has Θ. That is, by the contrapositive of (C2) "appearance" E is incompatible with conditioned-response thesis (CR.2), and a fortiori, incompatible with (CR.3). Since I do not know how you handle attention in your view, I am not so sure that postulate D by (C1) entails the negation of (CR.1). But in spite of the caveat about attention, I feel confident of this incompatibility, for on your view (unless I have seriously misunderstood it) by simply attending to his mental goings-on Dick is not in a position to know noninferentially that he is thinking that p once he has merely learned the concept of 'thought' -- he has to be trained to react by MΘ to Θ; whereas in accordance with D such conditioning by Jones is wholly unnecessary. (There is a point in connection with this which appears for the first time, it seems to me, in December Ad II.3, which I will discuss below in sec. 12.)
In my first letter I discussed theory (B) in order to account for ES, and concentrated my attack on E. I still think that your conditioned-response account has to be supplemented in order to provide room for ES. (B) certainly furnishes an ingenious way.
5. Such is the argument which has been in my mind. That is why in my letter of November, part II, I insisted on those entailments embodied in D, E, and ES. I still fail to see how you can say that you accept them (December Ad III.1), but I am prepared to accept that I was mistaken in supposing that they were part and parcel of what you call the 'propria persona principle'. Of course, you can persuade me of the former as well by showing that (C1) and (C2) are false.
I proceed now to make orderly comments on your December letter.
6. In December Ad I.1, you reminded me that you have insisted on (CR.3) being only a necessary, but not a sufficient condition for the existence of direct noninferential knowledge. You can see from sec. 4 above why I have felt that the distinction is of no help.
Furthermore, as you say, the other conditions that on your account are also necessary and apparently constitute with (CR.3) a sufficient set, are discussed in April 13. But "the important difference" you discuss there is irrelevant to our present problem. And so is Section VIII of EPM to which you refer me in April 13. In both places you discuss, not how MΘ is a reaction to Θ, but how MV is a reaction to Θ. Clearly, whereas there could be a question about MV expressing the knowledge that Θ, since e.g. it may be a parrot-like utterance, there is no such question about MΘ, provided, that is, that postulate E is not rejected. In EPM you discuss the epistemic authority of a perceptual statement, and in April 13, you state that the discussion transfers to the epistemic authority of MV; but this does not help me. My problem is not MV, but the epistemic authority of MΘ, which may or may not culminate in MV. I am sure that you are not analyzing the epistemic authority of MΘ in terms of the epistemic authority of MV, unless I have misunderstood your discussion with Chisholm.11
I find no help in your remark in EPM, sec. 35: "overt or covert token of 'This is green'. . . ." Clearly, MΘ is not a covert tokening of MV.
To belabor the point, in EPM VIII you discuss how a token of "This is green" can express direct noninferential knowledge of the perceptual kind. You give two necessary conditions:
(a) The speaker has acquired the habit of uttering (or tendency to utter) "This is green" in perceptual situations: his eyes are open, in working order, he is awake, there are green objects before him, etc.
(b) The speaker is able at the time of utterance to mention facts which make his tokening of "This is green" a reliable symptom of green objects (in standard conditions). [my italics.]
But surely, (a) is not logically necessary for noninferential observational knowledge. The person in question need not speak any language to be able to have perceptual thoughts, which constitute perceptual knowledge; hence, he need not have acquired any habit of tokening any sentences, even covertly. By the same token, (b) fails, too, to be necessary.
By the way, I do not understand what it is for "the conditioning itself to be caught up in a conceptual framework" (April 13).
7. I accept your correction in December 2.
8. See sec. 2 above for December Ad I.2.
9. I accept your reply in December Ad II.1.
10. Concerning December Ad II.2, see sec. 3, 4, and 6 above. The two inferences you compare are certainly valid, and perhaps the second one explains the epistemic authority or status as noninferential knowledge of thoughts of the form '(I am seeing that) this is green'. I am not too sure of this, but this is not the issue here. But the first inference certainly does not give epistemic authority to noninferential thoughts of the form 'I am thinking that-p'. For such inference is not an inference from symptoms to thoughts. It is a deductive inference, with an irrelevant premise conjunct (namely, "the thought occurred in such and such a manner and context"), justified by postulate E of the incorrigibility of judgment. I fail to see any translevel character in that inference.
Interestingly enough, your second inference presupposes direct noninferential knowledge of one's own thoughts of the perceptual kind, and your first inference presupposes direct noninferential knowledge of meta-thoughts. The latter presupposition is odd, but if you try to apply your "translevel inference" analysis to it you will get involved in an infinite vicious regress. But if you stop your inferential account with meta-thoughts of order n, and you hold postulate E as a claim in December Ad III.1, then there is no need of such trans- level inference, for having such meta-thoughts of order n entails that one has the meta-thought of order n-1, and so on; hence, one has the thought that p.
11. I should probably grant your initial point in December Ad II.3, that "we don't say that . . .". But I am not prepared to accept your suggestion that we do not say that because seeing green is logically involved with a conditioning to say "green". I may grant that in fact we teach our children to see that something is green by conditioning them to utter certain sounds which we say mean green. But this fact, which may very well be even a psychological law about human beings cannot be part of the analysis of seeing. After all, it is only a contingent fact that children have to be taught a language, and are not born with it, and it is a further contingent fact that human beings learn to think by learning a language.
The two absurdities that you compare are indeed logical self-contradictions: the first one is self-contradictory up to the word 'mistaken' by postulate E; hence, the conjunction "Dick has the thought . . . but he is mistaken, even though . . ." is also self-contradictory; the second absurdity is of the form "Thunder occurs; if thunder occurs, lightning occurs; and lightning does not occur". The causal character of the second conjunct is irrelevant. I have been unable to make out the point established by those absurdities.
It is not clear to me what the point is of your discussion of rationality in the second half of December Ad II.3. I think that I agree with you that if there was a hypothetical stage in which all was thinking-out-loud, rationality included the ability to have noninferential knowledge of one's own statements, and dispositions to make statements. But surely, such knowledge was not based on any inference, of the causal, translevel sort you mentioned in December Ad II.2. that would involve an infinite regress, since direct knowledge of one's own inclinations to make statements of a certain kind, i.e., direct knowledge of one's intentions and meanings, is direct knowledge of one's own thoughts.
12. I agree with you in that "the ability to think in the full sense of the term involves the ability to meta- think". This is an appearance that has to be saved by your account, but it is not saved if meta-thoughts are conditioned responses, as argued in sec. 4. I can only interpret your statements about the misleadingness of representing your position as if the conditioning could (logically) wear off, as indicative that you hold postulate B after all. But then, if (C2) is true, you cannot hold both (CR.2) and E.
13. Let us agree that we are not going to say that Dick has the full concept of thought unless he is successfully conditioned to react by MΘ only when he has Θ, and let us stipulate that he cannot really have MΘ unless he has the full concept of thought. This guarantees the existence in your account of a fact which saves "appearance" E. Now my counter-examples in part II of my November letter are refuted. But they can be reformulated. Instead of saving what I said about Dick in II.3, for instance, we should say:Dick has something which looks like the thought that he is thinking that it is raining, but perhaps it is not such a MΘ, for we do not know whether he is thinking that it is raining or whether his training to react with an MΘ only when Θ was successful or not.
Indeed, there is now a serious question as to how Jones can know that his training of Dick to meta-think, i.e., his teaching Dick the full concept of thought, has been successful or not. If Dick shows no behavioral signs of MΘ how can Jones be sure that Dick is coming along fine in the path to direct self-knowledge? How can Dick be sure that he has an MΘ, or that his MΘ is a reaction to a successful training to react by MΘ only when he has the corresponding Θ, and not to a feeling of inferiority or an itch?
14. Concerning December Ad II.5, see sec. 3 and 4 above.
15. December Ad III.1 is very perplexing. You characterize the 'propria persona principle' as a conjunction of two conditionals:
(PP.1) If one has direct noninferential knowledge of an inner state or episode, the concept in terms of which such inner state or episode is thus non-inferentially known cannot be a mere analogical extension of a concept pertaining to public phenomena . . . ;
(PP.2) If one has direct noninferential knowledge of an inner state or episode, the concept in terms of which such inner state or episode is thus noninferentially known must directly and adequately express the true nature of the state or episode.
16. I do not understand (PP.1); a fortiori, I do not understand the contrast between (PP.1) and (PP.2). I do not know what an analogical extension of a conception is, and why if a concept is formed by analogy it cannot express the true nature of the objects thought of through it.
17. I am not sure of (PP.2), but I gather by your remarks about 'true nature' and 'no room for factors that distort or mystify', that you have in mind, perhaps, one of the following propositions:
- If X knows directly and noninferentially that Y is P, then P is a property of Y exactly as it appears, independently of its being thought by X to be a property of Y.
- If X knows directly and noninferentially that Y is P, then P is a property of Y exactly as it appears to X to be a property of Y, independently of its appearing.
- If X knows directly and noninferentially that Y is P, then it is logically impossible for X to have thought falsely that Y is P.
I have tried to formulate other alternative interpretation of (PP.2), but I have not succeeded in finding any of them intelligible.
Proposition (ii) is ambiguous: in one sense it is the same as i. I will regard it to be different, by taking 'appearing' to mean 'being present to consciousness but not necessarily to thought'.
I accept (i), but reject (ii) for the case of feelings. When I know that I have a pain I know that the pain has a property which is independent of my thinking that it is pain, but not independent of my feeling it. But from this it does not follow that there is a distorting or mystifying factor. The rejection of (ii) is compatible with both (m) the view that unfelt feelings do not exist, and that there is no other reality underlying the felt properties of the feelings, and (n) the view that what appears as a felt property is a feeling-independent property of some entity, which may appear as some other property to other procedures of apprehension. But I see no reason to accept (n). I grant that what appears as an elephant to sight may be really an itch, or vice versa. It is an open question whether some procedure of apprehension, perhaps feeling (why not?), grasps the true nature of entities known directly. These are metaphysical questions which need a lot of argument, and it seems to me that the onus probandi lies rather on him who claims that a procedure of apprehension does not grasp the true nature of reality.
18. Alternative (iii) seems to me clearly mistaken in the case of perceptual facts, but true in the case of experiences, at least with the qualifications included in postulate D.
19. Toward the end of December Ad III.1 you say "Thus it was assumed that the necessary and sufficient condition of directly knowing that we are having such and such an impression or thought was the mere fact of having the impression or thought". If this is the 'propria persona' principle, I do not subscribe to it, witness my formulation of postulate D. But this view does not follow from either (PP.1) or (PP.2), or both, as far as I can see. The fact that we need concepts to think and to have propositional awareness does not in the least suggest that the concepts involved in direct noninferential knowledge can (or cannot) be analogical, or that such concepts, do (or do not) reveal the true nature of the objects thought of by means of them.
20. For that reason I do not get the point of the reformulation of your argument in sec. 25 of "Being and Being Known". See sec. 16 above.
In fact, your argument is a nonsequitur, and suggests a hidden fallacy of denying the antecedent:If one thinks that . . . one will conclude that our knowledge is nonanalogical and adequate.
" . . ." is false.
Therefore, the suggestion that the concepts pertaining to sensation are analogical merits consideration.
21. For December Ad III.2 see, especially, sec. 10 and 11.
I apologize for both the length and the delay of this letter. I only hope that such detailed comments will make it easier for you to spot both my misunderstandings and my mistakes once and for all. I want to end with a request for an explanation (with illustrations) of (PP.1) as well as of (PP.2).
Anxiously awaiting your clarifications,
[Table of Contents]
March 11, 1962.
I am very grateful for your long and careful letter of January 17. I think it was an excellent idea to review the correspondence with the aim of making a fresh start. Some misunderstandings of you by me and of me by you do seem to have been clarified, and by taking account of this progress the remaining issues posed in the earlier letters may stand out in clearer detail, and decisive new questions may occur to us. It seems to me that my role at this stage should be to comment point by point on your letter, leaving to a subsequent occasion a corresponding review on my part of the progress of the discussion.
Ad.1. It should be clear from the latter part of December Ad II.2 that I do not hold your CR.3. The ceteris paribus business and the emphasis on tendencies is essential. As a first approximation one may try
CR. 3' One is conditioned to react by having MΘ when (ceteris paribus) and only when one has Θ.
It is equally important to note that the "only when" must also be qualified, for one can obviously have a thought which is in a sense the same as the MΘ involved in direct knowledge of what one is thinking, though it is not occurring as a direct knowing of what one is thinking. To bring out this point, note that in the everybody-always-thinks-out-loud framework "That object is green" can formulate a hypothesis as well as express a seeing that the object in question is green. As the former it is, so to speak, simply a substitution instance of the function "x is (was, will be) green" which has as other instances such statements as "The other side of the moon is green", and "Cleopatra's nose was green". To express perceptual knowledge, "That object is green" must be more than the present tense of the function with a demonstrative subject; it must (among other things) be a response evoked by the very object perceived. (Objects, of course, can evoke such responses only from organisms which have been appropriately trained.) I have argued, "That object is green" doesn't express perceptual knowledge unless in some sense the perceiver knows that the object is thus responsible for the response. (I say, "in some sense", for I want to postpone until later in this letter a discussion of your regress argument in sec. 10.) This is bound up with the fact that to say that an utterance of "That is green" expresses a seeing that a certain object is green is to imply that the speaker could make the statement "I am under the impression that it is green" (or even, "I am under the impression that there is an object which is green on the facing side over there"). The phrase "under the impression that", in this context, gives expression to the passivity of the visually evoked propositional response, and suspends the endorsing implied by "I see that it is green". I could have used "It looks green" in place of "I am under the impression that . . ." but the logic of "looks" is more specialized and involves implications which are specific to sense perception and, in particular, visual sense perception. The generic "being under the impression that" is, however, equally appropriate to the case of introspective knowledge -- which is misconstrued if its analogy with perceptual knowledge is pushed too far.
Thus, "I have the thought that-p" can occur as an hypothesis, i.e. simply as a first person present tense instance of the function "x has (had, will have) the thought that-p" (I assume you will agree that thoughts can occur which we do not and (Freud) cannot (without therapy) directly know to occur, but which we can believe on evidence to occur.) To say of an MΘ episode of the form .I have the thought that-p. that it is a direct knowing that one has the thought that-p is to imply that the MΘ episode was evoked by the Θ episode of which it is the knowing, i.e, that the knower was "under the impression that" he had the thought that-p and in some sense (see above) knew that he was under this impression. Needless to say, from the point of view of this technical sense of "x is under the impression that", such statements as "Jones is under the impression that New York is bigger than Tokyo" are metaphorical extensions the point of which is the passivity (in a broad sense) of people with respect to most of their convictions.
Ad 2. Good. I'll see if I can save the appearance (if there is one to be saved) without modifying the theory.
Ad 3. Before getting down to the saving business, some comments on your D, E, and ES.
(a) I am uneasy about your equation of 'x is capable of thinking that he thinks that' with 'x has the concept of thought'. Obviously 'x has the concept of thought' doesn't entail 'x is capable of thinking that he thinks that E equals MC^2'. You would, of course, rejoin that it does entail 'x is capable of thinking that E equals MC^2'. This is indeed true, if properly qualified. But note that 'x is capable of thinking that his father should drop dead and x has the concept of thought' does not seem to entail 'x is (without further ado) capable of thinking that he thinks that his father should drop dead', though it does entail that no additional conceptual training would be required in addition to psychotherapy to bring about this ability.
(b) I regard the clause "and attends to his mental goings-on" as a dangerous one if perceptual overtones are allowed to take over. It is simply intended to imply a 'single-minded' occurrence of the question-thought *What am I thinking?* then -- with the reservations implied in (a) -- I would regard the principle as axiomatic. My suspicion is that the qualification 'single-minded' plays a fairly strong role. In any event I notice approvingly that you are implicitly rejecting the Cartesian notion that 'x thinks that-p' entails by itself either 'x knows occurrently that he thinks that-p' or even 'x has the thought that he thinks (i.e. has the thought) that-p'.
(c) I am uneasy about 'absolute incorrigibility'. If I am right in thinking that the concept of a thought is an analogical concept (in the 'theory' of thoughts) the fundamentum of which is the concept of a meaningful utterance, then it is worth noting that in pre-Jonesean eyes the state of affairs expressed by "x said '. . .' " (where x is presumed to know how to use the expression '. . .') differs from that expressed by "x parrotted '. . .' " not by virtue, obviously, of the acoustic character of the noises x is uttering, but by virtue of the different subjunctive conditionals they imply. Suppose that x can talk simultaneously out of different parts of his mouth, e.g. the right-hand side (RH) and the left-hand side (LH). At LH he says(A) I am uttering "It is raining".At RH he utters(B) It is raining.In a sense (A) is incorrigible, forx utters "I am uttering 'It is raining' ".entailsx utters "It is raining".But of course, the utterance with respect to which it hasthis'incorrigibility' is not the RH utterance (B), but a LH utterance which is a part of the whole LH utterance (A). This 'incorrigibility' of (A) abstracts from its reference beyond itself, so that (A) would have it even if (B) did not occur. This 'self-referential' incorrigibility must be distinguished from any incorrigibility (A) might have when construed as a claim about what is going on at RH. Notice that(A') I am asserting "It is raining".does not have self-referential incorrigibility, for
x asserts "I am asserting 'It is raining'".does not entailx asserts "It is raining".Again, notice that(A'') I am saying "It is raining".is self-referentially incorrigible -- where 'says' is weaker than 'asserts' but stronger that 'utters', for to assert "p or q" involves saying 'q' but, obviously, not asserting it. Notice, also, that when A'' is interpreted as a claim about RH, it would be false if (a) no utterance of 'it is raining' is taking place, (b) if such an utterance is taking place but only as a lapsus linguae.
Before I make something out of all this, let me remind you that to know (by seeing) that there is an object over there which is green and rectangular on the facing side involves background knowledge that the circumstances of perception are normal. I wish now to emphasize that the phrase 'circumstances of perception' must be construed to include not only the physical circumstances and the functioning of the sensory apparatus, but also the functioning of what might be called the 'conceptual apparatus' of the perceiver. Philosophers should pay more attention to the concept of madness, particularly its more extreme forms. But of this more later. The point I wish to make is that as one can know that other aspects of the circumstances of perception are normal, so one can know that one's conceptual apparatus is in good functioning order.
Now it is clear that pre-Joneseans would distinguish between utterances which are functioning meaningfully and utterances which are occurring as meaningless chatter even though the utterer knows the language. (I have in mind lapsus linguae rather than deliberate verbal play.) They would emphasize that the presumption is that utterances (on the part of one who has learned the language) are functioning meaningfully. There are various ways in which this presumption can be rebutted; it is, in Hart's sense, defeasible. Thus, given certain behavioral evidence pre-Joneseans might say of a particular person on a particular occasion, "He is unhinged -- his utterances are meaningless chatter". The point I wish to make -- and to which all the foregoing is a prelude -- is that the inner episodes which, as occurring in a properly functioning mind, are thoughts, can also occur as random or accidental "mental noise" (in the communications theory sense of 'noise'). The phrase is an apt one, for, as I see it, conceptual episodes are goings on in the neurological circuits formed by the educative process in all its stages. And the important thing to see is that the same episodes which, as goings-on in a properly functioning neural system, are thoughts proper can occur 'accidentally' or 'pathologically' as lapsus mentis. Let us distinguish, therefore, between 'T-episodes' and 'thoughts proper', the former being presumptively the latter as the utterances of one who knows how to use a language are presumptively sayings proper.
The point I am making is that to say (or think) that a particular inner episode is a thought is to make a claim which goes beyond its 'intrinsic' character as episode (cf. Ryle's distinction between flying south and migrating). One could be right in thinking that a T-episode was taking place, but mistaken in construing it as a thought. To thinkI have the thought that it is raining.is to presuppose that one's conceptual apparatus is working properly. And, since the proper working of one's conceptual apparatus involves CR3', i.e. reacting to having MΘ when (ceteris paribus) and only when one has Θ, one is, if the presupposition is correct, in a position to make the translevel inference.
I am under the impression that I have the thought that-p.
So, I have the thought that-p.
Before I return to a point by point discussion of your letter, let me use the above to elaborate a point first made in April 17. Suppose we distinguish between 'levels' of the conceptual apparatus, the lowest being the Θ-level, the next higher being the MΘ-level. (Compare the earlier example of the LH and the RH sides of the mouth.) In MΘ T-episodes occur which are presumptively MΘs; in the Θ-level T-episodes occur which are presumptively first level thoughts, i.e. Θs. Let us use asterisk quotes to form the names of T-episodes; Thus *It is raining* names the kind of T-episodes which is presumptively the thought that it is raining. Suppose, now, that a T-episode occurs which is an MΘ, e.g.I have an *It is raining* T-episode.This MΘ is self-referentially incorrigible in the sense thatx has an *I have an *It is raining* T-episode* T-episode.entailsx has an *It is raining* T-episode.This may not be obvious until one remembers that T-episodes are conceived on the analogy of utterances, so that an episode of the kind *I have an *It is raining* T-episode* contains an episode of the kind *It is raining* as an utterance of the kind "I am uttering 'it is raining'" contains an utterance of the kind "It is raining". Again, where MΘ isI have the thought ·It is raining·. (I have an ·It is raining· thought)(i.e., roughly,I have an *It is raining* T-episode as a well functioning rational being),it is also self-referentially incorrigible forx has the thought ·I have the thought·It is raining··.entailsx has the thought ·It is raining·.However, as before we distinguish between 'asserting' and 'saying' so here we must distinguish between having an ·It is raining· thought and 'mentally asserting' that it is raining. Roughly, an ·It is raining· thought is a mental assertion of ·It is raining· if it does not occur in a larger context (e.g. ·It is raining or it is snowing·) which suspends its assertive force.
To understand the high degree of incorrigibility MΘs have, we must take into account of the basic patterns of subjunctive conditionals involved in T-episodes being thoughts proper which parallel those involved in utterance-episodes being sayings proper. In short we must understand why, ceteris paribus, T-episodes are thoughts, and why, ceteris paribus, thoughts of the form ·I have the thought that-p·, where one is 'under the impression that' one has the thought that-p, are true. It must be granted that a human being could come into existence with all the developed intellectual abilities which normally take years to acquire. Nevertheless I would insist that all the distinctions and connections explored by theories of human concept formation through training are implicit in the concept of such a thinking being and can be pulled out by phenomenological reflection on what it is to think, see, have direct knowledge of one's mental states, etc. Furthermore, I would insist that although the ability to have T-episodes which satisfy all the relevant subjunctive conditionals, i.e. the ability to think, (a) could come into existence spontaneously, (b) could be a biological inheritance, the former, (a), is pointless as being a mere L-possibility, i.e. a supposition which has the minimal virtue of violating no principle of formal logic, however many other principles pertaining to the very conceivability of the phenomenon it violates, while the latter, (b) gets us in the same position by a mere circuitous route. After all, even the fact that it is logically possible for a complex protein molecule to come into existence by a sudden rearrangement of subatomic particles doesn't mean that the various ways in which such a molecule 'normally' gets built up by stages are related in a purely contingent manner to the concept of such a molecule.
Ad 4. (Finally!) I am now in a position to protest your assumptions (C1) and (C2). In the first place, one must distinguish between what is entailed by "x can be conditioned to react by R when S occurs" and what is entailed by "X has been conditioned to react by R when X occurs". And one must distinguish between logical and 'physical' or 'natural' possibility. It is indeed obvious that "x can be conditioned (but is not now conditioned) to react by R to S" entails that S without R is not only logically but also physically possible. It is also obvious that "x has been conditioned . . ." entails that S without R is logically possible. It is not so obvious to me that it entails that S without R is now (i.e. given that x is conditioned -- though he may later cease to be so) physically possible. In the second place, you are wrong in thinking that on my view "Dick is not in a position by simply attending to his mental goings-on to know noninferentially that he is thinking that-p once he has merely learned the concept of thought." I would, as indicated in Ad 1, introduce some qualifications, but these do not prevent my accepting the thesis that ceteris paribus Dick is in a position by simply attending to his mental processes to know that he is thinking that-p once he has acquired the concepts involved in the thought that-p. I suspect that you have too piecemeal a conception of conditioning. One does not need to be separately conditioned to respond to
Θ1 by MΘ1
Θ2 by MΘ2
Θ3 by MΘ3
. . . .
. . . .
Θi by MΘi
. . . .
Surely one can be trained to have a general disposition to respond to episodes of the Θ kind by corresponding episodes of the MΘ kind! And generalized disposition is part of what it is to "have the concept of thought" in the full, as opposed to 'theory of thoughts' sense. (Compare the seeing-man's concept of color with that of the blind man.) Thus, I see little force to an argument on which you lay great weight.
In the second paragraph of 4 you refer to theory B. In assessing its merits, you seem to overlook the fact that on the view I developed in EPM, 'I am thinking that-p' becomes, under Jonesean training, additional thought behavior (I don't like 'substitute' because it implies that the other forms of thought behavior are replaced) when it acquires the reporting role. It suggests that my theory would be improved by supplementing it with this consideration, you ask me to supplement the conditioned response account by itself!
Ad 6. In EPM, sec. 35, I had not yet introduced the concept of a thought, but I used 'covert tokening' in such a way that when the concept was introduced, 'covert tokening of "this is green"' would mean 'tokening of the proposition (thought) which is expressed verbally by a tokening of "this is green".
You say "I am sure that you are not analyzing the epistemic authority of MΘ in terms of the epistemic authority of MV, unless I have misunderstood your discussion with Chisholm". Agreed, but I am claiming that the concept of the epistemic authority of MΘ is a theoretical extension of the concept of the concept of the epistemic authority of MV. Roughly, the epistemic authority of MV is a matter of the manner and context of its occurrence. In the absence of facts about its occurrence which would make it non-authoritative (and these are of a limited and -- in principle -- ascertainable variety) it is authoritative. Having authority, as pointed out above, is a 'defeasible' or 'forensic' matter. One knows unless there are reasons to suppose one doesn't. It should be remembered that 'MV' already connotes 'non-parrotted'; thus MV may lack epistemic authority, but not because it is parrotted.
But above all I am startled at the unsympathetic construction you place on EPM VIII. Your comments are appropriate on a first reading, but less so when read in the light of what is to come later: the myth of the pre-Jonesean people who think out loud. In effect the concept of using language meaningfully which is being explicated is the pre-Jonesean concept, and the concept of perceptual knowledge which is being explicated is the pre-Jonesean concept of perception. These concepts differ from ours because they contain no reference, as ours do, to inner episodes (thoughts). Even pre-Joneseans can distinguish between 'merely uttering' and 'saying something'.
Another point. I have, of course, granted that it is logically possible for a person to think without speaking a language, 12 but this formal-logical truth throws no light on the non-formal-logical conceptual connections between knowing a language and being able able to think, just as the formal-logical possibility of a perpetual motion machine throws no light on the non-formal-logical conceptual connections between drawing on a fixed quantity of energy, restoring this energy and continuing to work without end. In short, miracles are not self-contradictory, but it is not a 'purely contingent fact' that they do not occur.
By speaking of 'the conditioning itself' as 'caught up in a conceptual framework' I mean (a) that at least some of the items involved in the conditioning are conceptual (e.g. Θ and MΘ); (b) that the conditioned person comes to conceive of himself as so conditioned. (Of course, the word 'conditioned' has too rich a scientific meaning to be more than a metaphor for the 'can't help thinking' which is its common sense counterpart.)
Ad 10. I failed to see that in the case of the first inference the inference is not an inference from symptoms to thoughts . . . [but] is a deductive inference with an irrelevant premise conjunct . . . . You seem to me to be assuming thatThe thought that I am thinking that-p has occurred to me.entailsI am thinking that-p.That this is not so I have attempted to explain in Ad 3 (c) and the subsequent paragraphs.
Your remarks in "Interestingly enough . . ." puzzles me. I have never claimed that to have direct noninferential knowledge one must actually draw a translevel inference. I have simply said that one must be able to draw it. I know a vicious regress when I see it.
Another point to bear in mind is that thoughts do not occur in single file, they occur in structures of thoughts within an ambience of 'tip of the mind' dispositions. The classical notion of 'apperceptive mass' has much truth in it.
Ad 11. Again (cf. Ad 4) you are too quick to use the locution "It is a contingent fact that . . .". If you intend to limit conceptual connections to formal logical connections and 'analysis' to 'explicit definition' we shall have to thrash out some basic epistemological issues before following up some of the points at issue. The point dramatized by the 'absurdities' in question is that ceteris paribus 'x has the thought that he has the thought that-p' together with 'x has been trained to report what is going on in his mind' logically imply 'x has the thought that-p'.
Again, I have never claimed that a direct-knowing episode must be based on a translevel inference episode.
Above all, I not only do not deny, I want to insist that there is direct noninferential knowledge of what goes on in our minds. I am simply attempting to give a coherent and satisfactory account of what it is to have such knowledge.
Ad 15. Probably the phrase 'analogical extension of a concept' is misleading. What I had in mind is the formation of a 'theoretical' concept with particular reference to those cases in which there are strong analogies between the role of the theoretical concept in its framework and the role of a nontheoretical concept in the observation base of the theory. Two examples might help: (a) the discussion of 'molecule' in EPM; (b) the discussion of 'white and triangular sensation' in "Being and Being Known". 13
As for the term 'directly' and 'adequately', they can perhaps be explained by analogies. Although, as I emphasize in EPM, sec. 61, theoretical concepts are not to be construed as definite descriptions of properties which specify these properties in terms of their relations to other theoretical properties and to observation level constructs -- a construing which would entail that observational properties have an absolute priority and could never be replaced by theoretical predicates -- they can be construed as such in contexts where the problem of the reality of theoretical objects is not being explored. Consider, now, a red-green blind person. His statements about what objects are red and what objects are green can be true even though there is a sense in which his concepts of red and green are (a) indirect -- in that they have the form 'the quality which . . . .' -- ; (b) inadequate in the sense that these concepts could be replaced by concepts which provide a more adequate understanding of the situations to which they apply. The essential point is that as theories are improved and as theories are 'reduced' to theories about more basic entities (cf. the reduction of chemistry to physics) their concepts become more adequate, i.e. provide a richer and more detailed understanding of the world. The relevance is this. I hold that a concept of a thought is an inadequate concept in the sense in which the concepts of an 'unreduced' chemistry are inadequate. As you know, I hold that in the sense in which chemical compounds are really complex systems of sub-atomic particles, thoughts are really complex neurophysiological processes. This would have to be put more carefully if it were the point at issue, since -- see the Chisholm correspondence -- the concept of a thought has normative as well as descriptive or matter of fact components.
Thus I am committed to the view that direct self-knowledge can involve concepts which are in the sense explained inadequate. Many philosophers would find this view absurd. They would claim that direct knowledge can't involve concepts which are 'indirect' and 'inadequate' in the senses I have attempted to explicate. The propria persona principle is the thesis that direct knowledge involves concepts which are not 'indirect' and not 'inadequate'. Notice that although 'indirect' and 'inadequate' as applied to theoretical concepts are related adjectives, they do not coincide, for in an ideally satisfactory theory the concepts of the theory, though 'indirect' from the standpoint of the observation base, would not be inadequate.
You write "It is an open question whether some procedure of apprehension . . . grasps the . . . nature of entities known directly". (I omit the 'true' because it implies adequacy.) This is not an open question, but a tautology. I would certainly hold that in direct knowledge of states of affairs the nature of the state of affairs known is 'grasped' by 'some procedure of apprehension'. When, however, you interpolate 'perhaps feeling (why not?)'. I demur. I find it paradoxical and incorrect to say that I feel that I am thinking that-p. (It is even incorrect to say "I feel that I am feeling a pain".) There are cases, however, in which one "hears oneself thinking that-p", but on this point see EPM.
Ad 19. I agree that the assumption in question doesn't follow from PP-1 and PP-2. But you are wrong in assuming that by 'P if and only if Q' here I mean 'P is identical with Q'. That propositional thoughts are involved in direct self-knowledge is presupposed by the discussion, though the Thomists and Descartes were not as clear about this as they might have been. My point was that if one thinks that one couldn't be in a mental state without knowing directly that one was in that mental state, or at least being able to know this 'without further ado', then the idea that the ability to have direct knowledge of one's mental state involves concepts having an origin and structure which parallels that of theoretical concepts must seem absurd.
Ad 20. The argument is not a non-sequitur. Its form is
- If one thinks that . . ., one will conclude quite logically that the propria persona principle is true.
- But . . . is false.
- Hence one proposition which would be a good reason for accepting the propria persona principle is false.
- Hence, until another good reason for holding the propria persona principle is found, the suggestion that the concepts in terms of which we directly know our sensations are 'indirect' and 'analogical' is not estopped and, therefore merits consideration.
- Because relevant suggestions which are not ruled out by known truths and which might throw light on controversial issues merit consideration . . . .
Well, Hector, the above letter is an attempt to answer as carefully as you have raised them the questions posed in your last letter. I am afraid that by touching on so many topics I may have muddied the waters still further instead of providing the promised clarification. My suspicion is, however, that in order to clear up any one philosophical puzzle, one must locate it in the total landscape. Perhaps one or two of the things I have said in this wide ranging, free-wheeling "letter" may turn out to be just the right thing to advance the discussion. In short, like Hume, I have been beating about in the neighboring bushes.
[Table of Contents]
August 8, 1962.
Finally I am in a position to answer your long, careful letter of March 11. We have made tremendous progress. Your last letter, particularly, has helped to bring out into the open the fundamental issues that seem to divide us. So, instead of engaging in paragraph -by-paragraph comment on your letter, I will make scanty comments on those main issues, and add a few remarks on minor points.
1. It seems to me that the main issues are the following:
- A different conception of a priori connections.
- My postulate E and your translevel inference.
- The connection between language and thought.
- The impossibility of a private language.
- The 'propria persona' thesis.
2.You are absolutely right. We have to thrash out some basic epistemological issues. But to avoid unnecessary discussion I want to say at the outset that your suggested accusation that I "intend to limit conceptual connections to formal logical connections and 'analysis' to 'explicit definition'" is unjustified. Thus, fortunately, we can commence our thrashing out at a more advanced level. I have been using the term 'conceptually analytic' to cover:
You can see that already propositions of type (c) are not formally analytic, if the rules of inference involved are not substitutional. Note that analytic by explicit definition covers even a smaller class of propositions of type (c).
- truths of logic, i.e., propositions which are true and formulate principles of logic, like"(p)(q)(ph (p h q) hq)", where 'p' and 'q' are variables ranging over propositions and 'h' stands for 'if-(then)';
- instances of (a);
- propositions derivable from (a);
- propositions which express conceptual connections like "Everything red is colored", "Everything red is extended", "Everything red all over is not green all over (I mean here clear cut instances)";
- propositions entailed by any set of propositions of the previous type.
But it seems to me that you regard a much larger class as necessary propositions which express conceptual connections. I detect a strong inclination on your part to regard causal propositions as formulating conceptual connections.
The negations of those propositions of types (a) - (c) I have called "conceptually contradictory"; and those propositions which are neither conceptually analytic nor conceptually contradictory I have called contingent. This is exactly the sense in which I have employed the term in my previous letters, and this may help you understand my meaning more precisely.
3. Your discussion on my postulate E, scattered through your letter, has not convinced me that it is false. It seems to me conceptually analytic of type (d).
In E 'think' appears in the sense of 'judging' or 'considering', as I said in my previous letter. There are actually two different postulates:
E1. If X judges at t that he is judging at t that p, then X judges at t that p.And, of course, it is also conceptually analytic that:
E2. If X considers [entertains the proposition or possibility] that he is considering at t that p, then X is considering at t that p.J. If X judges at t that p, then X considers at t that p.
On page 3 of your last letter, you discuss X's utterances out of different parts of his mouth. I think that you are right in much of what you say about that but the connection with postulate E escapes me. To begin with, I am not at all concerned with utterances, but with thoughts. Second, even if I were to be concerned with utterances, I would be interested in X's utterance that he (X) is uttering that . . ., not with his utterance at his right-hand side that he is uttering at his left-hand side that . . . .
But I disagree with you when you claim that (a) does not entail (b).(a) X asserts "I am asserting 'It is raining'",That (a) entails (b) seems to me an instance of a conceptually analytic proposition which expresses a conceptual connection of the term 'assert' to itself. Your explanation ". . . to assert 'p or q' involves saying 'q' but, not asserting it" is true but it shows nothing against the entailment of (b) by (a). Clearly, to assert that one is asserting 'p or q' is not to assert p or to assert q, but it is to assert 'p or q' as required by the conceptual entailment from (a) to (b).
(b) X asserts "It is raining".
4. Your thesis that in order to think that one is thinking that p requires the ability to draw a translevel inference of the formI am under the impression that I have thought that p. So I have the thought that p,seems to me false. As a matter of fact, I lack the ability to draw such inference, for I do not have the concept of being under the impression that I have the thought that p; hence, I cannot think the premise, hence, I cannot think (or draw) the inference. But apparently I can think the conclusion without any difficulty. Hence, as far as I am concerned, "X has the ability to draw translevel inferences of the above form" is not conceptually, or even causally, indeed, not even materially implied, by "X thinks" or "X can think".
Honestly, I do not know what it would be like to be under the impression that I am thinking that say, 2 plus 2 equal 4 (or any other number, for that matter),
I was unable to make out your lapsus mentis. Your T-episodes seem to me chimerical. You introduce them by two remarks, one which makes of them as well as of thoughts neurophysiological episodes, and the other which makes them analogous to utterances-which-are-perhaps-mere-chatter-or-perhaps-expression-of-statement. But neither is really helpful. The former presupposes a metaphysical theory, namely the identification of thoughts with neurophysiological processes, and I see no reason at all in its favor. The latter fails to connect with my conception of language and thought. I can see how an utterance can be mere chatter and how it can be the expression of statement (or assertion), but I cannot see the parallel distinction in the case of thoughts. For the simple reason that the former contrast is that between utterance which expresses thought and utterance which fails to express thought; but when I consider just thought, without noises or marks there is no way of having thoughts which express thoughts and thoughts which do not. Could you offer a couple of complete definitions, or a list of the meaning postulates or conceptual truths which govern the concepts you are introducing?
I have had the recurrent thought that your view of thought is too analogous to your view of perceptual knowledge. I find no objection to the claim that the ability to frame a perceptual judgment implies the ability to frame appearance or "being-under-the-impression" judgments. But I do not see why psychological self-knowledge should require a similar ability.
Another thing puzzles me. Let us suppose that you are right in claiming that my ability to know that I am thinking that 2 times 3 equals 6 involves my ability to draw a translevel inference from my knowing that I am under the impression that I am thinking that 2 times 3 equals 6. But is there also another translevel inference involved in the actual knowledge of the premise for that translevel inference? Surely, my having the ability to draw the inference implies my having the ability to know the premise. But then either I am able to know that I am under the impression that I am thinking that . . . without being able to draw another translevel inference, or I can know that only if I can draw a second-order translevel inference. In the first case there is self-knowledge without translevel inference, in the second case you must provide a discussion or analysis or definition (implicit, recursive or explicit) of the concept of 'being under the impression that one is thinking that p'. I am sure that I lack the ability to draw that second-order translevel inference, because I lack the requisite concepts to formulate the relevant premises.
5. At the end of Ad 3, you claim that the possibility of the spontaneous coming into existence of the ability to think, "has the minimal virtue of violating no principles of formal logic, however many other principles pertaining to the very conceivability of the phenomenon it violates". The first part of your assertion seems obvious to me; the second part, however, is not obvious. Could you formulate those principles pertaining to the conceivability of the phenomenon which are violated? I will have to satisfy myself that the principles you have in mind are conceptually analytic of either type (c) or (d) or (e) mentioned above. I am prepared to concede right now that it is, perhaps, a psycho-sociological law that thinking requires the possession of a language. I am prepared to concede that a language as complex as, e.g., German or English, cannot (causally) be possessed unless one has been carefully trained to master it. But these causal truths are factual discoveries over and above the possession of the conceptual connections constitutive of the concepts involved.
Here I can have no concrete objection before you present me with the argument establishing your claim.
6. In Ad 6 you state with admirable clarity one of the main issues that divide us. I am grateful for this statement. You assume that it is impossible for a person to learn to think without learning to understand a (public) language by a training which connects sentences in the language with the environment . . . . And you claim that this impossibility is a non-formal-logical conceptual connection between knowing a language and being able to think. Such conceptual connection as you claim exists is not obvious to most people. Indeed, according to some, your assumption is very much the same thesis that Wittgenstein established for the first time. Again, I do not say that you are making a false assumption. I only request that you produce the proof of your assumption. I have sent you a long discussion of argument which purport to show that the impossibility of a private language is conceptually self-contradictory. Please note that in my discussion of Rhees' paper I have examined Rhees' claim that learning a language requires training by members of one's community. 14
7. Concerning the 'propria persona' thesis I have moved from the umbra to the penumbra. I see enough of what you mean by indirect and inadequate concepts to think that I hold some version of the 'propria persona' thesis. Which version I hold is, of course, difficult to say at this stage.
You characterize the inadequacy of a concept in two way which seem to me different. In the example of the red-green blind person I can understand what you mean by saying that his concepts of red and green can be replaced by concepts which "provide a richer and more detailed understanding of the world" (Ad 15). But I do not see that the same situation holds for the concepts of chemistry which have been 'reduced' to the concepts of physics. Could you give me a complete statement of the reduction of some chemical concept of physics? I have no idea of what you mean by reduction. Has the concept water been reduced to H2O? Has the concept temperature been reduced to the concepts of moleculaes, motion, average, kinetic energy? Has the concept of Hydrogen (H) been reduced to the concepts of electron, nucleus, electronic orbit? In none of these cases, as far as I know, do we have a replacement of concepts, but only the introduction of new concepts. Note that while the red-green blind man need no longer use gadgets to determine that an object is red or green, or need not ask other people but instead he can determine what color the object is by seeing it, the physicist has to measure temperature with thermometers, and from those measurements he goes on to determine further facts about molecules. While the red-green blind man can now check his gadgets and reports given him by other people, the physicist is in no position to do anything analogous.
But there is no point in my discussing these examples. As soon as you give me the complete statement of one such reduction, I will be in a position to express opinion.
As far as I can see, my concepts of thought and pain, for instance, are not inadequate in either sense. Indeed, to substitute my concept of pain by some other concept would be to lose important detail of my understanding of the world, and my understanding of the world would be definitely poorer. The same is true of my concept of thought. The replacement of those concepts is precisely the reverse of the red-green blind man: it would be analogous to losing the concept of red which a non-red-or-green blind man has in order to become a blind man with an indirect concept.
Your other characterization has so far been unhelpful. I do not understand how a thought can be a brain process, but even if it were (somehow), it would be a brain process at most in the way in which temperature is the average kinetic energy of a multitude of molecules, and that 'reduction' is not a conceptual connection, but at most a causal association of two characteristics. Certainly, such 'reduction' does not involve a replacement of concepts for more adequate ones, in the sense of concepts which "provide a richer and more detailed understanding of the world ". That 'reduction' is nothing more than the introduction of concepts which have to sit beside the older ones in order to yield a more detailed (i.e., adequate) knowledge of the facts of the universe.
My concepts of pain and thought are direct inasmuch as my concepts of my thought and of my pain are direct. I know of only one argument to prove that one's concept of pain cannot be direct. It is what Malcolm called "Wittgenstein's external argument" in his discussion of the Philosophical Investigations. 15 But Malcolm's argument is inconclusive. I have tried to improve it in several ways, but the enterprise has turned out helpless. I will be grateful to you if you can send me any argument you may have in this connection.
Of course, I agree with you that, in a sense, if there ever was a stage of development in which everybody was just able to think out loud, then if the concept of thought was not innate or got by some accident, then it may very well have been framed on the analogy with the concept of statement. But this does not make it indirect, inasmuch as those thinkers had the ability to have noninferential knowledge (knowing-out-loud, of course) of the form 'I have just said that-p', since the knowledge of one is saying, even if out-loud, is direct knowledge of one's intentions or thoughts, not based on inference from circumstances and utterances to utterance-cum-thought or statement.
8. A minor point perplexes me yet. You meet the charge of vicious regress against your translevel inference by sying that you do not hold that direct knowledge requires the actual drawing of a translevel inference, but only the ability to draw one given the circumstances one is in. Yet (on page 4) for instance, you insist that perceptual knowledge involves a lot of background knowledge, about one's circumstances as well as about the functioning of one's own sensory mechanism and conceptual apparatus. This knowledge, true, may not be in consciousness, in episodes of knowing or thoughts; it may be dispositional knowledge. But since the issue is the ability to draw a translevel inference, the knowledge in question is knowledge-that. And I am prepared to argue that there is no knowledge-that in the dispositional sense without there being at the time of acquisition episodes of knowing-that. Furthermore, my knowledge of the conditions of my sensory mechanism and my knowledge of the condition of my conceptual apparatus at time t is not helpful when I am perceiving an object at time t', later than t. So that when I directly know at t by seeing that the object over there is a small tree the background knowledge that is involved in my perceptual knowledge is the knowledge at t that my sensory mechanism is at t in such and such conditions and my knowledge at t that my conceptual apparatus is at t in such and such conditions and my circumstances are at t, etc. So, you can see why your view of the background knowledge seems to me too strong.
Besides, part of this background knowledge must be that I am not dreaming at time t. Now, how can I distinguish whether my situation is one of dreaming or not?
9. I am still unclear as to how you evade the difficulty mentioned in section 13 of my previous letter.
This also became longer than I planned. But the important disagreements (or apparent disagreements) are at least filed. I will be awaiting your reply with great interest.
Editor's note: The copy of the correspondence, which I received from Hector Castañeda, contained some omissions of page references and sources.These I have added as needed. All underlining has been changed to italicizing; and all names or abbreviations of books have been italicized. [Revised version: Aug. 4, 2006] [Back]
1 Referred to, hereafter as EPM in H. Feigl and M. Scriven, eds., Minnesota Studies in the Philosophy of Science, Vol. 1 (Minneapolis: University of Minnesota Press, 1956). Reprinted in W. Sellars, Science, Perception and Reality (London: Routledge and Kegan Paul, 1964). [Back]
1a Hector-Neri Castañeda, "The Private Language Argument," in C.D. Rollins, ed., Knowledge and Experience (Pittsburgh: University Press, 1963). Reprinted in Essays on Wittgenstein, ed. E.D. Klemke (University of Illinois Press, 1971) [editor's note] [Back]
2 See L. Wittgenstein,Philosophical Investigations (Oxford: Blackwell, 1953), Pt. I, secs. 243-270; N. Malcolm, "Discussion of Wittegenstein's Philosophical Investigations", The Philosophical Review, 63 (1954), 530-559; id., Dreaming (London: Routledge and Kegan Paul, 1959). R. Rhees, "Can there be a Private Language?", Aristotelian Society Supplementary Volume, 28 (1954), 77-94. [Back]
3 Hector-Neri Castañeda, "The Private Language Argument," in C.D. Rollins, ed., Knowledge and Experience (Pittsburgh: University Press, 1963). [Back]
4 The meaning of this sentence was the initial subject of the correspondence between David Rosenthal and Wilfrid Sellars, published as "The Rosenthal-Sellars Correspondence on Intentionality," in Intentionality, Mind and Language, ed. Ausonio Marras (Chicago: University of Illinois Press, 1972.)[editor's note] [Back]
5 Hector Castañeda, "The Private Language Argument", in C.D. Rollins, ed.,Knowledge and Experience (Pittsburgh: University Press, 1963). [Back]
6 Wilfrid Sellars, "Being and Being Known," Proceedings of the American Catholic Philosophical Association 34 (1960), 28-49. Reprinted in Wilfrid Sellars, Science, Perception and Reality (London: Routledge and Kegan Paul, 1964). [Back]
7 Wilfrid Sellars, "Language of Theories," in H. Feigl and G. Maxwell, eds., Current Issues in the Philosophy of Science (New York: Holt, Rinehart & Winston, 1961). Reprinted in Wilfrid Sellars, Science, Perception and Reality (London: Routledge and Kegan Paul, 1964). [Back]
8 Wilfrid Sellars, "Phenomenalism, in Science, Perception and Reality (London: Routledge and Kegan Paul, 1964).[Back]
9 "The Private Language Argument", in C.D. Rollins, ed., Knowledge and Experience (Pittsburgh: University Press, 1963). [Back]
10 Hector Castañeda, "The Private Language Argument", in C.D. Rollins, ed., Knowledge and Experience (Pittsburgh: University Press, 1963). [Back]
11 See "The Chisholm-Sellars Correspondence on Intentionality," in H. Feigl, M. Scriven, and G. Maxwell, eds., Minnesota Studies in the Philosophy of Science, Vol. II (Minneapolis: University of Minnesota Press, 1958): 521-539. [Back]
12 I will go further and admit that it is physically (naturally) possible for a person to learn to think without learning to use a (public) language in the sense of using symbols to communicate -- though not without learning to understand a (public) language by a training which connects sentences in the language with the environment by rewarding appropriate behavior in that environment which occurs when the subject hears (or otherwise receives) the sentences, i.e. the connections may be established in a more roundabout way than stressed in the simple model assumed in EPM. [Back]
13 See EPM, sec. 51 "Being and Being Known", secs. 19ff.[Back]
14 Hector Castañeda, "The Private Language Argument," in C.D. Rollins, ed., Knowledge and Experience (Pittsburgh: University Press, 1963). R. Rhees, "Can there be a Private Language?" Aristotelian Society Supplementary Volume 28 (1954), 77-94. [Back]
15 N. Malcolm, "Discussion of Wittgenstein's Philosophical Investigations," The Philosophical Review 63 (1954), 530-559.[Back]
[Table of Contents] [Return to Main Page]