Sellars-Firth Correspondence (1974)

Sellars-Firth Correspondence (1974)*

Wilfrid Sellars and Roderick Firth

Contents
January 13, 1974: Firth to Sellars
January 22, 1974: Sellars to Firth
February 2, 1974: Firth to Sellars
February 12, 1974: Sellars to Firth
February 22, 1974: Firth to Sellars
Department of Philosophy
Carleton College
Northfield, Minnesota
January 13, 1974
Dear Wilfrid,
1. I have been traveling almost continually since leaving Atlanta, and now, as you can infer, I am out here in Minnesota learning what real cold is like. I shall return home early in February to bask in the warmth of a New England winter. Am enjoying the students here. They have a kind of enthusiasm and lack of intellectual inhibition that I don't find at Harvard. More like the students I knew at Swarthmore many years ago.
2. I have been thinking about your APA paper ["Givenness and Explanatory Coherence," The Journal of Philosophy 70 (1973): 612-24], which seems even more profound, if that is possible, each time I reread it. It calls for a better response, and surely more of a response, than I made to it at the meeting. In fact it became clear to me there that I had misunderstood you at a critical point. I feel some despair when I ask myself whether it is really possible to get at the source of our differences -- if there are differences -- by written correspondence. An hour or so of conversation would be incomparably better. But perhaps if I try talking about the issues you will be able to see, at the very least, where I have misunderstood you.
3. I'll start from a point of agreement. In my comments on your paper I said that it is easier, in important respects, to tell whether something looks red to me now than whether it is really red. You replied that you agree with this. But you seemed to imply that my explanation of this fact -- or at least one common explanation -- is mistaken. The correct explanation, you said (I can't remember your exact words) is that in believing that something looks red we believe less than in believing that it is really red. This seemed to me at the time to imply the following thesis: Believing that some particular object, x, is really red is believing (1) that x looks red and (2) something else. (The thesis should be restricted to perceptual believing, Price's "accepting" and Chisholm's "taking.") This would explain the respect in which believing that something looks red is less than believing that something is really red. But on reflection I saw that you did not mean this. As you have often explicitly recognized in your writings, we may take x to be red when it looks not-red (e.g., gray) if we know the distorting effects of abnormal illumination. So my problem is to get perfectly clear just what one believes when one believes that x looks red if one is simultaneously taking x to be (say) white.
4. The traditional view, which you are criticizing, would not make it quite appropriate to say that we believe "less" when we believe that x looks red. What we believe is "different" but not precisely "less." We believe something about the "sensory content of our perceptual experience (in Chisholm's terminology, that we are "being appeared to redly") and this is just different from believing that x has the property red. In the traditional view we believe something less only in the epistemic respect on which we agree, namely, that it is easier to tell whether x looks red than whether x is really red.
5. On your view, I think, "less" is more appropriate than on the traditional view. I think you hold that (to put the point very loosely) to believe that x looks red is to believe something of the same kind that one believes when one believes that x is really red. Something of the same kind but not something incompatible with the fact that we can believe simultaneously that x looks red and that x is not red. I think you hold that the respect in which the two beliefs are of the same kind can be defined by reference to a similarity in the concepts involved in (required by) the two beliefs. In your paper you use the expression "mode of conceptualization" in contrast with "mode of sensing." If I understand you, to believe that x looks red is to believe something in the mode of conceptualization; and to believe simultaneously that x is not red is to believe something else in the mode of conceptualization. This is your way of pointing to a similarity between the two beliefs, the similarity denied by the traditional doctrine.
6. My problem boils down to this: I do not understand just what you think one believes when one believes that x looks red (while simultaneously believing that x is not red). On the traditional view, as I interpret it (the view you reject), it is not possible to state what one believes when one believes that x looks red without circularity. Looking red involves the nonanalyzable concept of having a red sense experience, being appeared to redly, etc. The doctrine implies that the rest of our concepts, including our concepts of color properties of physical objects, are not sufficient to enable us to formulate an analysis of "looks red." If you were to agree with this, I doubt that there would be much difference between us. In some contexts the fact that something looks red to me is a reason for my believing that it is red. If "looks red" involves the unanalyzable concept of sense experience, then at least we have irreducible propositions about sense experience supporting propositions about the external world. (This is pretty crude but there is no point in refining it unless you feel that something important is being obscured here.
7. You say, however, that what we mean by "looking red" is experiencing in the mode of conceptualization. I take this to imply that, contrary to the traditional view, the concept "looking red" does not involve a concept of sensing that is unanlyzable in the respect just discussed. I think you mean that the concept "looking red" is analyzable in terms of a property that we attribute to physical things. Yet you do not, so for as I know, offer an explicit analysis. (I do not have your books and papers here to check.) You do not say: "x looks red" means the same as "x is . . .", filling in the blank with an expression that identifies a property of physical things and does not include terms like "sense-experience," "being appeared to," and their synonyms.
8. If I am right in thinking that you do not propose such an analysis, is it because you think that "x looks red" has to be analyzed by reference to a property for which we have no name in ordinary speech? (This is suggested by your reaction to my talk about a primitive or primordial use of "red.") I suppose it might be argued that to believe that x looks red is to believe that x has a certain property (nameless,{1} and not analyzable in terms of sense-experience) which is compatible with x's being not red. But an adequate account of looking red must fit cases of non-delusory hallucination. If I believe that I am hallucinating a red rat, I believe that I am being appeared to redly, but not that there is something (x) which has such a nameless property.
9. It would be more plausible, because of this difficulty, to allow the "existential commitment" to be separated from the description of the "experience in the mode of conceptualization." Perhaps something like this would do. To believe that x looks red to me is to believe (i) that I see x, and (ii) that I have an experience as of an object with a certain property (the nameless property). Then in a case of non-delusory hallucination I would simply believe (ii). I would merely believe that I am having the specific kind of experience, not that the experience is related to x in whatever way is required for it to be true that x looks red to me. (ii) could be considered as a proposed analysis of "I am appeared to redly" when this idiom is being used in an acceptable way (as opposed to the way Chisholm uses it).
10. If you are suggesting something like this, my problem is to be clear what (ii) could mean, or, alternatively, what should be substituted for (ii). What could (ii) mean if not that I have a red sense experience? If the experience in question could be a species of believing, the problem would be solved. For (ii) we might try substituting (ii'): I take myself to be seeing an object with a certain property (the nameless one). But in the case of a non-delusory hallucination I do not take myself to be seeing (in the sense relevant to "x appears red") anything at all. Perhaps I am inclined to take myself to be seeing something red, but it seems implausible to hold that looks-expressions can be analyzed in terms of inclinations to believe.
11. I think I had better let the matter rest here for the present. I may have wandered so far from anything you would want to defend, that much or all of this may be irrelevant.
12. I shall only add that I have not thought of what I have called "the percept theory" is incompatible with the traditional analysis of perceptual experience into sense-experience and judgment (taking). I have thought of it as denying the traditional phenomenology of sense experience by affirming that sense-experience is much richer than philosophers and psychologists have traditionally recognized. In writing that is not yet published I make a good deal of use of the notion of a baptismal rule to clarify this point. When I look at Price's tomato, according to the percept theory, my sense-experience is not correctly described, as Price says it is, as a round red patch . . . A full description would have to take account of the fact that it looks juicy, etc. Which is not to say that I take it to be juicy, although I may also do that. We have to borrow words like "juicy" to describe the sense-experience; but it is still sense-experience that we are describing.
Thanks again for saying such stimulating things.
With best regards,
Rod
P.S. Although it is now very late, I'll add just a sentence or two about the possibility of a Firth-Bosanquet. I should think the sensory foundationalist would have to grant the possibility, indeed the actuality, of theories from which it follows inferentially that certain beliefs are warranted in virtue of their subject-matter. If such a theory is warranted, however, he would have to argue that it is warranted only because certain beliefs are self-warranted (non- inferentially warranted) in virtue of their subject matter. Thus ultimately, he would argue, all inferential warrant is derived from self-warrant. To settle this I should think you would have to produce a warranted theory (a theory of persons, as you call it) that does not ultimately derive its warrant from self-warranted propositions by some valid form of inference. (Valid forms of inference include the rules by which we distinguish good explanations from bad ones, the rules implicit in saying that of two "explanations" of the same phenomena, one is more warranted than the other.)
Table of Contents
Department of Philosophy
Faculty of Arts and Sciences
University of Pittsburgh
Pittsburgh, Pennsylvania 15260
January 22, 1974
Professor Roderick Firth
Department of Philosophy
Carleton College
Northfield, Minnesota 55057
Dear Rod:
I was very pleased to receive your letter. I believe that by reacting to some of the questions and alternatives you pose I can clear away some of the misunderstandings which have plagued our attempts to communicate. It is possible, as you suggest, that "an hour or so of conversation would be incomparably better" as a way of getting at the sources of our differences -- if any -- than a written exchange. On the other hand, the intuitive sense of understanding which arises in discussion is often, save on gross issues, delusive and, at best, ephemeral. Plato may be right that the subtleties which fill out the skeleton of a philosophical position belong to spoken dialogue, but even he was not above limning his ideas in writing. Perhaps this (I am afraid, hasty) response to your letter will help set the stage for a fruitful encounter.
By and large I shall take up the points you raise in the order in which you raised them. Assuming that you have a copy of your letter, I have numbered the paragraphs '1' to '12,' but will refer to the postscript as such. On the off chance that you do not have a copy, I am enclosing a xerox.
1. (Ad 3.) you begin by calling attention to a "point of agreement" at Atlanta; and I did indeed agree with your remarks that "it is easier, in important respects, to tell whether something looks red to me now than whether it is really red." You are also correct to ascribe to me the view that "[your] explanation of this fact -- or at least one common explanation -- is mistaken." Furthermore, I did, indeed, say, though perhaps not in exactly these words, that "in believing that something looks red we believe less than in believing that it is really red." You are also correct in your considered view that I did not mean by this that

. . . believing that some particular object, x is really red is believing (1) that x looks red and (2) something else.

As you point out, I have explicitly recognized that "we may take x to be red when it looks not-red (e.g. gray) if we know the distorting effect of abnormal illumination." You then pose the question, which remains at the center of your concern, as to "Just what one believes when one believes that x looks red, if one is simultaneously taking x to be (say) white."
2. Before continuing to follow your line of thought, I should perhaps, stress the fact that from "Empiricism and Philosophy of Mind" on I have defended the view that the analysis of 'looks red' involves the concept 'is (physically) red,' where '(physically) red,' like '(physically) blue,' is a 'simple' or 'primitive' or 'unanalyzable' concept. It should be remembered, however, though the point is not directly relevant to our problem, that this is a thesis about the 'common sense' or 'manifest' framework in terms of which we experienced ourselves and the world before we were confronted with the task of reconciling this framework with the claims of physical theory, which latter, though methodologically based upon the former, began to bite the hand that fed it. Though the common sense framework has been mixed, in useful but unperspicuous ways, with the framework of physical science, it can still be regarded, for our purposes, as the framework in which the perceptual knowledge which it is our task to analyze is embedded.
3. The analysis I originally proposed for

x looks red to S (at t)

was

If circumstances (including appropriate causal connections) were standard, S would be seeing x to be red at t.

I have since refined this account by distinguishing a sense of 'seeing x to be red' which entails 'seeing x's very redness' -- which the above definition requires -- from a looser sense which is compatible with 'not seeing x's very redness,' but the above will do for the moment.{2}
4. One essential feature of this analysis is that in both the definiendum and the definiens the term 'red' stands for physical redness, that is, the redness of physical objects -- which, once again, is 'primitive.' Another essential feature is that 'looks red' belongs at the same conceptual level as 'is seen to be red' and not at the same level as 'is red.' This feature may have been a source of misunderstanding, for since (in EPM) I defined 'looks red' in terms of 'is seen to be red,' and only subsequently, in my analysis of the latter, brought in my counterpart of the 'sense experience of red' which is the immediate crux of your account of 'looks red,' the essential role of the latter in my account of 'looks red,' may have been obscured. It should also be noted that since my basic pairing is of 'looks red' with 'is seen to be red,' I hold that the pairing between 'looks red' with 'is (really) red' rests on the fact that 'is seen to be red' entails 'is red.
5. (Ad 4.) You claim that "the traditional view would not make it quite appropriate to say that we believe 'less' when we believe that x looks red. What we believe is 'different' but not precisely (ital. W.S.) 'less."' Now if your model for "precisely less" is the one which tempted you to ascribe to me the view that

is red = looks red . . .

then I, of course, agree. What I actually had in mind can be expressed, in the first instance, by saying that 'looks red' pins down the world to a lesser extent than does 'is red.' But the way in which it does so comes closer than this would imply to the intended contrast between saying 'less' and saying 'more.' Thus, in, first approximation,

(A): x looks red to me now

entails

(1): x has some color

which, in turn, entails (for the moment I shall simply say, intuitively),

(2) If the circumstances (including appropriate causal connections) were standard, there is some color which I would be seeing x to have,

whereas

(B): x, which I see, is red

entails (trivially)

(1'): x is red

which, in turn, entails (intuitively)

(2'): If the circumstances were standard, I would be seeing x to be red.

In short (A) is more informative than (B) in the sense in which 'Tom is in New York' is more informative than 'Tom is in some city or other.' I suspect that even proponents of the traditional view would be uncomfortable with your claim that according to lt "we believe something less only in the epistemic respect on which we agree, namely, that it is easier to tell whether x looks red than whether x is really red.
6. (Ad 5.) You now attempt to explain why I think 'less' to be more appropriate than on the traditional view. The attempt is interesting because it contains a misplaced insight. You write, "I think you hold that (to put the point very loosely) to believe that x looks red is to believe something of the same kind that one believes when one believes that x is really red." This I do not hold; but, as pointed out above, I do hold that to believe that x looks red is to believe something of the same kind that one believes when one believes that he sees x to be red. (Once again the refinement sketched in paragraph 2 above must ultimately be taken into account.)
7. (Ad 5 continued.) In your attempted explanation of my view, you draw on my distinction between "experiencing in the mode of conceptualization" and "experiencing in the mode of sensing." You write:

If I understand you, to believe that x looks red is to believe something in the mode of conceptualization; and to believe simultaneously that x is not red is to believe something else in the mode of conceptualization. This is your way of pointing to a similarity between the two beliefs.

Now, of course, all believing is (trivially) 'in the mode of conceptualization' in the sense that to believe is to conceptualize. The question concerns what is conceptualized. What is conceptualized by a believing may or may not itself be a conceptualizing. Thus, consider the two statements, which we may suppose to be true,

Jones believes (occurrently) that Tom is tall
Jones believes (occurrently) that Tom believes (occurrently) that time is money

In the second, but not the first, what is conceptualized by Jones is a conceptualizing. Now I do, indeed, hold that what is conceptualized by

x looks red to Jones at t

is (in part) a conceptualizing (by Jones), for it attributes to Jones a state which would be a seeing x to be red if the circumstances were standard; and the state of seeing x to be red is a conceptualizing state (involving some counterpart of the proposition 'x is red') -- though (and this is crucial) not merely a conceptualizing state. To see x to be red is not merely to believe (take) x to be red; nor is it enough to add, for example, that the belief is true and warranted and that the believing is (appropriately) caused by the object, x. What more there is to 'seeing x to be red' than conceptualizing qualified along these lines is, of course, the key issue to which all this is leading. But before facing it, let me point out that, on my view, to believe that x is (or is not) red is, indeed, to conceptualize, but what is conceptualized is not itself a conceptualizing -- but rather a physical state of affairs, in which respect it differs from believing that x looks red.
8. (Ad 6.) You write: "My problem boils down to this: I do not understand just what you think one believes when one believes that x looks red (while simultaneously believing that x is not red.)" You then proceed to define the view which you take me to reject.

On the traditional view . . . it is not possible to state what one believes when one believes that x looks red without circularity. Looking red involves the unanalyzable concept of having a sense experience, being appeared to redly, etc. The doctrine implies that the rest of our concepts, including our concepts of color properties of physical objects, are not sufficient to enable us to formulate an analysis of looks red.

Here is the crunch. It is my view that the ur-concept{3} of 'seeing x to be red' in the sense which implies 'seeing its very redness' involves the idea that both x and its character of being (physically) red are somehow present in the experience in a mode other than that of being conceptualized, i.e. other than that of "intentional inexistence" or being a belief-content. It is thus my view that the ur- concept

seeing x to be red

is the concept of a unique togetherness of a conceptualizing of a (physically) red object and a somehow but non-intentional presence in experience of a (physically) red object. If x is seen to be red, x must, of course also exist in the physical world and be physically red and be, in an appropriate manner, causally related to the occurrence of the conceptualizing.
9. (Ad 6 continued.) Note that on my view this ur-concept of 'seeing x to be red' is an 'indefinite' concept, in that it does not spell out the 'somehow.' It does, however, emphasize that seeing is not merely true (and, it may be warranted) believing appropriately caused by the object. Thus philosophical theories about 'sensing red sense data' or 'sensing redly' are just that, i.e. theories. They introduce new expressions, either color predicates, which apply to new (postulated) objects (sense data), or predicate modifiers (adverbs) which apply to postulated states of the perceiver. The rules for the use of these new expressions draw upon postulated analogies between e.g. the 'colors' of sense data (or sensings) and the colors of physical objects. I suspect that it is this dimension of the situation which is reflected in your wrestling with "baptismal rules."
10. (Ad 6 continued.) Thus I agree with you that there is a unique concept of 'the presence of something red in experience' which is at the core of 'looks red,' and that without reference to it we cannot formulate an analysis of the latter. But (1) I take this concept to concern the somehow presence of a physically red object in experience (though a theory which seeks to explain this presence -- e.g. sense datum theory -- may flesh out the 'somehow' by the idea of the presence and experience of non-physical objects having counterpart qualities); (2) I think that the primary role of this ur-concept is to distinguish seeing an object from merely being caused to have true and reliable thoughts about its color; (3) I think that this ur-concept is neither an ur-concept of 'red' nor an ur-concept of 'looks red.' It is not the former because it includes the concept of physical redness; it is not the latter because 'looks red' belongs at the same conceptual level as 'is seen to be red.' Yet if I had to choose between the two alternatives I would prefer to say that it is an ur-concept of 'looks red' for the simple reason that once (like my man in the necktie shop) one discovers that seeing can "go wrong", this somehow presence of something (physically) red in veridical perception is used to explain (together with the idea of non-standard conditions) how we can have an experience which resembles veridical perception although no (physically) red object is there to be seen.
11. (Ad 6 continued.) You suggest that if I were to agree (as I do) that "the rest of our concepts, including our concepts of the color properties of physical objects, are not sufficient to enable us to formulate an analysis of 'looks red' . . . you doubt that there would be much difference between us." From my point of view, however, we may still be separated by a chasm. That this is so is suggested by the argument which immediately follows the above assessment. You advance an incontrovertible premise.

In some contexts the fact that something looks red to me is a reason for my believing that it is red.

You add what you take to be an implication of that on which we might, after all, agree,

. . . 'looks red' involves the unanalyzable concept of sense experience

From these premises you conclude

. . . then, at least, we have irreducible propositions about sense experience supporting propositions about the external world.

From my point of view, however, the conceptual tie of 'looks red' to 'is (physically) red' plays a decisive role in the interpretation of this argument. For if I am not completely off the track, and

x looks red to Jones

means something like

If the circumstances are standard, Jones is seeing x to be red

then the concept of 'looks red' not only draws upon the causal role which objects seen play in the seeing of them, but involves the contrast between those circumstances in which objects look what they are and those circumstances in which they look other than they are. (the discussion could easily be extended to include cases in which there looks to be an object in a certain place where no object is there.) Thus, from my point of view, the reasoning

x looks red to me now
So, in all probability x is red

contains a tacit premise to the effect that the circumstances are (in all probability) standard. Nor, on my view, is there any way to by-pass this premise in such a way as to replace it by premises all of which are 'looks' premises, qualitative or existential. I am by no means certain as to your present views on this matter, though I know that in the past you have taken a sympathetic stance towards phenomenalism.
12. However this may be, it is my present diagnosis that if any difference between us is fundamental and tends to carry other differences along with it, it is that I take '(physically) red' to be the basic concept pertaining to redness, for I take it to be (a) 'primitive' and (b) logically prior to both the ur-concept of

the somehow presence of a (physically) red object in experience

and the concept of

looks red,

whereas, according to you, "looking red involves the unanalyzable concept of having a red sense experience (ital. W.S.)" which concept, for you, is not only unanalyzable, but, as I understand it, the basic concept pertaining to redness. For, it should be noted, I would not object to saying that the concept of sensing redly is unanalyzable -- indeed I would insist that it is, if to analyze is to give an explicit definition. I would, however, deny that it is the basic concept pertaining to redness, for it is a concept embedded in the theory designed to explain how physically red, blue, etc., objects get perceived (and mis-perceived).
13. (Ad 7.) It should now be clear that I have attempted an explicit analysis of 'looking red,' and that I do think that this analysis (by virtue of its reference to seeing) involves a reference to a unique presence of a physically red object in experience. If the term 'sensing' is used for this 'somehow' presence, then I would agree that the analysis of 'looking red' involves a reference to sensing. Thus it is not precisely true to say, as you do in opening this paragraph, that on my view "what we mean by 'looking red' is experiencing in the mode of conceptualization." For while it does include experiencing in the mode of conceptualization, it also includes, given the above stipulation, experiencing in the mode of sensing.
14. (Ad 7 continued.) Again I do not "mean that the concept 'looking red' is analyzable in terms of a property that we attribute to physical objects" except in so far as on my view basic redness is a property of physical things, and a reference to physical redness is involved in

If the circumstances were standard, Jones would be seeing x to be red.

As I pointed out in paragraph 4 above, I believe that what has mislead you is the fact that I defined 'looks red' in terms of 'is seen to be red' and only then went on to probe for what distinguishes seeing from mere believing. It is my account of this latter distinction which led me to sensing as a unique mode of experience. The definition of 'looks' in terms of 'sees' is the heart of the matter, and is what kept me from such enterprises as seeking to define 'looks' in terms of named or unnamed properties of physical objects.
15. (Ad 8.) Since the hypothesis of this paragraph is false, there is little I can say about it. The same is true of 9. and 10. Yet they do provide an occasion for picking up a theme from the opening paragraphs. I return, then, to the problem you pose in your paragraph 6.

I do not understand just what you think one believes when one believes that x looks red (while simultaneously believing that x is not red [or, as you put it in paragraph 3, "if one is simultaneously taking x to be (say white).'])

My answer, in accordance with the views I have expressed in the course of this letter, can be broken down as follows.

(a) What one believes when one believes that x looks red is that one would be seeing x and seeing its very redness, if the circumstances (including causal connections) were standard.
(b) If one is simultaneously taking x to be white, one accepts something like the proposition 'this, x, is white,' where this acceptance involves a background belief that the circumstances are in a certain way non-standard, and that objects which look red in such circumstances are white. This acceptance, of course, need not be arrived at by a process of reasoning.

+ + +
I shall have to postpone consideration of the huge topic which you raise in your postscript. It not only deserves, but requires, a lengthy discussion in its own right. I suspect that we shall have to come closer together on the nature of perceptual consciousness before we can fruitfully tackle lt. But this is not necessarily so, since there is no linear order in which philosophical problems arise.
In stopping at this point I am motivated not only by the realization that this letter is already much too long, but also by the hope that I can get it to you before I appear in Minneapolis next week. Perhaps we can find time to get together for the "hour or so of conversation" which might clear up our differences. If so, I hope that this letter will have been of some assistance in preparing the way.
With many thanks for your challenging letter,
Sincerely yours,
Wilfrid
Wilfrid Sellars
WS:rd
[Table of Contents]
Department of Philosophy
Carleton College
Northfield, Minnesota
February 2, 1974
Professor Wilfrid Sellars
Dept. of Philosophy
University of Pittsburgh
Pittsburgh, Pennsylvania 15260
Dear Wilfrid:
Many thanks for your splendid letter. This first response will be brief, partly because it is just a request for clarification, and partly because I am now immersed in term-end obligations. (I shall be back home on February 8, and during the spring term my best address will be 2 Patriots Drive, Lexington, Mass., 02173.)
I am beginning to wonder whether my own interpretation of perceptual experience, which I think is a (one) traditional interpretation among the philosophers I like to call Cartesians, really does differ from yours. But before I write to you about that I hope that you will clear up a difficulty I now think I find in your analysis of "x looks red to S (at t)." It seems to me that it is possible to deduce a contradiction from various things you say about the looks-is distinction. If I am right about this I think I can see various ways in which you might remove the contradiction; but how you do this will determine the extent of our disagreement.

* * *
You maintain that the following are logically compatible when x remains constant:

(i) x looks red to S at t
(ii) x, which S sees at t, is white

At the top of page 3 you offer this analysis of (i):

(iii) If circumstances (including appropriate causal connections) were standard, S would be seeing x to be red at t (S would be seeing x's very redness at t).

On page 4 you say that

(iv) x, which I see, is red (your (B))

entails

(v) If the circumstances were standard, I should be seeing x to be red. (Your (2!))

(iv) and (v) are intended, I assume, to provide an example of an entailment relationship that holds when other color terms are substituted in both cases for 'red'. Thus I infer that

(vi) x, which I see, is white
entails

(vii) If the circumstances were standard, I should be seeing x to be white

But if I am S at t, then, since (i) entails (iii)

(viii) x looks red to me now

entails a proposition incompatible with (vii), namely

(ix) If the circumstances were standard, I should be seeing x to be red.

Thus (vi) and (viii) entail logically incompatible propositions, viz., (vii) and (ix). And so, contrary to your original thesis, (i) and (ii) are not logically compatible.

* * *
Have I made a mistake here? I'll write again when I hear from you about this.
With best wishes,
Rod
P.S. Perhaps this is not a major point, but is your (2) on page 4 strong enough in view of the analysis proposed at the top of page 3?
[Table of Contents]
Department of Philosophy
Faculty of Arts and Sciences
University of Pittsburgh
Pittsburgh, Pennsylvania 15213
February 12, 1974
Professor Roderick Firth
2 Patriots Drive
Lexington, Massachusetts 02173
Dear Rod:
1. The logical points you raise should certainly be cleared up before we come to grips with our ostensible disagreement about the structure of the looks-is distinction. As I understand it, your argument hinges on the claim that subjunctive conditionals of the form

(1) If x were phi, x would be psi

contradict the corresponding subjunctive conditionals of the form

(2) If x were phi, x would not be psi

so that any proposition which entails one of the former would be inconsistent with a proposition which entails one of the latter.
2. The 'subjunctive' character of the conditionals is, of course, essential, for if they were construed as material conditionals, the resulting propositions (I use '-->' in place of the horseshoe)

(3) phi(x) --> psi(x)
(4) phi(x) --> ~psi(x)

would not be contradictory, but quite compatible, and, indeed, jointly equivalent to

(5) ~phi(x)

3. Thus, if I had not used subjunctive conditionals, and had merely claimed that

(6) x looks red to S

entails, by virtue of the analysis of 'looks' in terms of 'sees,' the material conditional

(7) The circumstances of S are standard --> S sees x to be red

(where 'S sees x to be red' carries with it the sense, to be explored, of 'sees its very redness'), and that

(8) x, which S sees, is white

entails the material conditional

(9) The circumstances of S are standard --> S is seeing x to be white

your argument would not have gone through, since (7) and (9) are quite compatible and, indeed, jointly equivalent to the denial of their common antecedent. Thus, (6) and (8) would not have been shown to be inconsistent. Furthermore, the idea that (6) and (8) jointly entail that the circumstances of S are not standard is itself not without a certain intuitive plausibility.
4. Well, then, did I mess things up by using subjunctive conditionals? As you are doubtless aware, this is a sticky topic. Contextual considerations play a key role, as does the fact that subjunctive conditionals license inferences. Could conditionals of the form (1) and (2), respectively, be simultaneously true? Could they simultaneously be good licenses? (Note, by way of anticipation, that the latter would not require that the licenses be simultaneously useable.)
5. Consider the following case, which I first put in formal dress:

(10) n is a square integer greater than 1 < (n is between 3 and 7 --> n = 4)
(11) n is an odd number < (n is between 3 and 7 --> n = 5)

6. In this example, (10) and (11) have, respectively, the forms

(12) p < (q --> r)
(13) s < (q --> r')

where r' is incompatible with r, while p is compatible with s.
7. Now it seems to me that although, in our example, q by itself entails neither r nor r', nevertheless two interestingly relevant subjunctive conditionals can both be true. Thus, given the truth of

(14) n is a square integer greater than 1

we can surely say that

(15) If n were between 3 and 7, n would be 4

And given the truth of

(16) n is an odd number

we can surely say that

(17) If n were between 3 and 7, n would be 5.

And since (14) and (16) can be simultaneously given as true (since they are compatible), surely (15) and (17) can be simultaneously available.
8. Notice that the availability, indeed truth, of (15) is contextually grounded in (14), and the availability, indeed truth, of (17) in (16). This dependence is made explicit if we write

(18) [Since n is a square integer greater than 1] if n were also between 3 and 7, n would be 4
(19) [Since n is an odd number] if n were also between 3 and 7, n would be 5.

When, however, the subjunctive conditional is detached, and the burden of its dependence is carried by the dialectical context, the 'also' loses its grammatical antecedent and disappears.
9. Now the two subjunctive conditionals (15) and (17) cannot be simultaneously put to use as inference licenses. The former, (15), authorizes

n is between 3 and 7
So, n is 4

The latter, (17), authorizes

n is between 3 and 7
So, n is 5

But the fact that (15) depends on (14) and (17) depends on (15) makes the simultaneous deployment of these arguments involve a simultaneous commitment to

(14) n is a square integer greater than 1
(15) n is an odd number
(20) n is between 3 and 7

and these, of course, constitute an inconsistent triad, though they are compatible taken by pairs. Thus, given the truth of (14) and (15), it follows that (20) is false.
10. Now if you find the above example persuasive, it should convince you that the following statements can all be true

(21) x looks red to S
(22) x, which S sees, is white
(23) If S were in standard conditions, S would be seeing x to be red
(24) If S were in standard conditions, S would be seeing x to be white.

That if they are all true,

(25) S is in standard conditions

must be false, would, of course, be no surprise. After all, (21) and (22) by themselves seem, intuitively, to have this consequence. But if I am right, the larger picture illuminates this intuition by grounding it in a certain construal of the relationship between 'looks' and 'is.'
11. Thus I may have made a direct case for our turning our attention to the specific character of the entailments to which I made appeal in my letter. Indeed, even if you do not find my answer to your difficulty persuasive, it might be worthwhile to by-pass the latter for the time being and consider some of the "various ways" in which you think I might "remove the contradiction." Of course, if I have made a simple logical error, it should be nailed before we go any further.
12. In conclusion let me briefly wave a red flag (the white one may come later). I think that my hope of convincing you that I have a viable theory of perception hinges on the fuller account to be given of 'seeing its very redness.' For it is this theme which serves as the bridge to my 'anti-Cartesian' account of sensing and its role in perceiving. Paragraphs 8 and 9 of my first letter and their reference to "The Structure of Knowledge: (1) Perception" are the immediately relevant passages, though the theme is also present in my attempt to cope in EPM with the "descriptive content" (sections 22 ff.) common to cases of seeing that x over there is red (which I was already interpreting -- with what I now recognize to be a cavalier disregard of ordinary usage -- as involving 'seeing its very redness') and cases in which something over there looks red or there looks to be something red over there. It was the concept of this "descriptive content," which somehow involves the quality red, which Jones developed into the theory of sense impressions which I describe in the concluding sections (60 ff.) of the essay.
13. Another red flag is the phrase "standard conditions." In EPM I rejected the claim that

(25) x is red

is to be analyzed as

(26) x looks red in standard conditions.

Yet I held it to be a conceptual truth that red things look red in standard conditions. Obviously, one way of insuring this is by defining standard conditions, in first approximation, as conditions in which phi objects look phi, where 'phi' ranges over the various colors. It was in connection with this move, which I made in section 18 of EPM (pp. 146-7 of Science, Perception and Reality) that I began to push for the 'coherence theory of concepts' which you so vigorously attacked in "Coherence, Certainty and Epistemic Priority," and which I hope will sooner or later move to the focus of our attention.
I hope that Minnesota has prepared you for the rigors of a New England winter. Certainly Pittsburgh seemed warmer on my return.
Cordially,
Wilfrid
Wilfrid Sellars
WS:rd
[Table of Contents]
Department of Philosophy
Enerson Hall
Harvard University
Canbridge, Massachusettes 02138
February 22, 1974
Professor Wilfrid Sellars
Department of Philosophy
University of Pittsburgh
Dear Wilfrid:
I have had to do, or at least have done, some more traveling since returning to this semi-tropical Boston climate. But I now look forward to a relatively uninterrupted term of philosophical dialogue with my typewriter.
It seems clear to me that you must use subjunctive conditionals to do what you want to do. You do not want to say that

(1) x looks red to S at t

can be analyzed into the truth-furctional conditional:

(2) If circumstances are standard then S sees x to be red at t (S sees x's very redness at t),

for then (l) would be true whenever circumstances are not standard. And so of course would

(3) x looks green to S at t,

as well as other statements produced by substituting other color terms for 'red'.
It also seems clear to me that your proposed subjunctive analysis of (1), viz.,

(4) If circumstances were standard, S would be seeing x to be red

is indeed logically incompatible with

(5) if circumstances were standard, S would be seeing x to be white.

Or, to be more accurate, (4) and (5) are logically incompatible if the term "standard circumstances" is going to play the role that you (and I) want it to play in connection with the looks-is distinction. We could of course define standard circumstances so broadly that standard circumstances would not uniquely determine how a thing of some particular color looks, or vice versa. But then (4) could not be an analysis of (1).
One possible solution to the difficulty, as you suggest, would be to treat (4) as short for a more complex form of statement, viz.,

(6) Since - - - -, if circumstances were standard, S would be seeing x to be red.

If (5) is similarly treated as short for

(7) Since - - - -, if circumstances were standard, S would be seeing x to be white,

and the blanks in (6) and (7) are filled in by different expressions, then (6) and (7) may not be logically incompatible. (I say "may not" because there are of course some pairs of different expressions that would make (6) and (7) logically incompatible.) But you do not complete your argument from the mathematical analogy by suggesting what these two expressions might be if (6), when fully spelled out, is to be a plausible analysis of (1). My own experiments have not yielded anything that looks hopeful to me; but I'd be glad to discuss the matter farther with you if you wish.

* * * *
When I last wrote to you I thought, though without much reflection, that the most likely way out of the difficulty would be to alter (4) -- i.e., your proposed analysis of (l) -- to read something like this:

(8) S is in a state such that, if S were in that state in standard circumstances, and if S were seeing x, S would be seeing x to be red (seeing x's very redness).

Then perhaps your analysis of "x, which S sees, is white," could be altered to read something like this:

(9) x, which S sees, is in such a state that if S were seeing x in that state in standard circumstances, S would be seeing x to be white (S would be seeing x's very whiteness).

When formulated in these ways I think it is not very difficult to construe (8) and (9) as logically compatible. It is a matter of specifying just what is supposed to be held constant (so to speak) if the actual world is altered to satisfy the if-clauses. In (8) we hold constant the state such that, and we also hold constant whatever (minimal) relation between S and x must hold if S is to see x. But we allow x's color to change (if causally necessary). Thus if x looks red to S, S is in a state such that if circumstances were standard, and S continues to see x, x would be red. (This, i.e., x's being red, is a logically necessary condition, if I understand you, of S's seeing x to be red, i.e., of S's seeing x's very redness.) But the fact that x would be red in those circumstances is of course logically compatible with the fact that S is actually white. (8) is thus logically compatible with (9) if (9) is construed so that the color of x is supposed to be held constant. Perhaps, indeed, this should be made explicit by substituting for (9):

(10) x is so colored that if S were seeing x while x is so colored in standard circumstances, S would be seeing x to be white.

Or perhaps this would be sufficient:

(11) x has a color such that if S were seeing x in standard circumstances S would be seeing x to be white.

Perhaps something could be worked out along these lines that would allow (l) to be logically compatible with "x, which S sees, is white." I'd be glad to exchange thoughts with you about this if the general line seems to you to be promising and not too far from what you have in mind. If we do work on this, I suspect that I'll be a better critic than originator of ideas; for my intuition (as a Csrtesian) is that counterfactuals like (8) are just too "speculative" to represent what we mean by "x looks red." But I grant that I could easily be wrong about this.
With best wishes,
Rod
Roderick Firth
[Table of Contents] [Return to Main Menu]
Notes
{*} Edited in Hypertext by Andrew Chrucky. This is one of several correspondences contributed to the Sellars Archives by Willem deVries. [Back]
{1} To say that it is nameless is not of course to say that nothing can be done to identify it. I assume that your story about the man who works in the clothing store can be construed as a way of identifying it. It is the property that he attributes to suits and ties before he learns about the effects of changes in illumination. [Back]
{2} It is a familiar fact that 'Jones sees that x is f' does not entail 'Jones sees x.' 'Jones sees x to be f' does have this entailment, but, to turn to the case in which we are interested, 'Jones sees x to be red' does not entail that 'x looks red to Jones.' I use the modifying clause 'and sees its very redness' to pick out the sense of 'seeing something to be red' which is the basic concept pertaining to visual perception. Some philosophers speak of 'seeing x as red,' but this seems to have a different job, besides failing to imply that x is red. [Back]
{3} The term 'ur-concept' is used here not (as in the Atlanta paper) to indicate priority to the contrast between the concepts of 'seeing' and 'seeming to see,' but rather to indicate the schematic nature of the concept. It is this schematic concept which is 'subsequently' (from the standpoint of a rational reconstruction) fleshed out by theories of 'sensing.' For a discussion of the schematic nature of this ur- concept and of attempts to flesh it out by theory, see sections 41 ff., of "The Structure of Knowledge: (1) Perception."
[Table of Contents] [Return to Main Menu]