Carnegie-Mellon University

Department of History Schenley Park Pittsburgh, Pennsylvania 15213 (412) 621-2600 February 7, 1972

Professor Wilfrid Sellars
Philosophy Departwnt
University of Pittsburgh
Pittsburgh, Pa. 15213

Dear Wilfrid,

Your generously thorough reconstruction of and answers to my questions cannot but provoke a response from me, both to acknowledge confusions, and to commut on your exciting new suggestions, for the analysis of "aspectual" (aspectival?) attitudes. As some of the faults in my earlier letter resulted from sacrificing clarity to brevity, I feel this letter must be longer, and the first part of it rather tedious restatements or corrections of my original points. I leave to the end my remarks on your new suggestions for how to handle the "subtle grammar" of psychological attitude expressions, although some of those points will be anticipated in my first correction.

Well then, to acknowledge errors. You were right to be amazed that I could swallow "Smith as Smith." It proved indigestible; and shortly after writing to you I began to explore another way of expanding, (where necessary), sentences reporting a psychological attitude into sentences containing an as phrase, this tims avoiding the absurdity of "Smith as Smith." My current hope is that one can expand the simple

1. Ab, where A holds a place for a verb of psychological attitude with the name of some person as subject, and b in the object of this attitude, into either
2. Ab as something or other (in some capacity or other) or
3. A something or other as b.

(2) is very like your (7') on p. 5, except that, in my versions, what follows the as must be of the same ontological type as what precedes it, otherwise b could not occupy each place.

What I have been attempting, with indifferent success, is a classification of psychological verbs according as they favor expansion Into (2) or (3). The matter is complicated by the fact that if b is a proper noun or pronoun, (2) is automatically selected, whatever the verb. Even when b is a common noun, or description, the type of noun it is affects one's choice. Nouns which are obviously names of roles favor (3). Allowing for these effects, and contextual effect, it still seems to me that some attitude verbs favor (2), others (3). I have tentatively assigned likes, wants, believes in to (2), while admires, worships, and intends favor (3). (I am here assuming that the objects

-2-

of intending are not propositional, or, if they are, that propositions too can be combined by an as, as I think Chisholm's I connective combines them, in his analysis of intention.) But these results are very uncertain, and I am not sure what to do with them anyway. One might have hoped that the cut would coincide with some other distinction -- verbs which do and verbs which don't take a that construction, or a to construction. But it cuts across both these. One might say that to attempt such a classification was to attempt to sort psychological verbs into those with a preference for an opaque reading, (3), and those with a preference for a transparent reading, (2). But if there in any advantage in the expanded forms (2) and (3), over (1), it lies in the fact that both contain a place which admits of substitution of extensional identicals, and a place which does not, which amounts to a virtual rejection of the usual transparent-or-opaque dichotomy. The verbs now take double objects, one half referentially transparent, one half referentially opaque.

This whole botanising attempt depends upon acceptance of two assumptions which you seem to reject. First,, that at least some psychological O-verbs sometimes are substitution-intensional. You say my intuitions are wrong, if I think that

(4) Jones fears the examiner

does not admit of substitutions for 'examiner.' I think fearing is hard to classify along the lines sketched above -- in many contexts it would be naturally expanded into (2) not (3), so one could then substitute for 'examiner.' Extra information such as that given in

(5) Jones fears the examiner, who has threatened his life on several occasions,

would certainly block interpretation (3). But in a case like

(6) Jones admired the conductor but not the soloist,

it would surely be very odd or misleading, on the strength of the true identity statements

the conductor = the son who crossed to the concert hall on the pedestrian crossing, and
the soloist = the man who jaywalked to get to the concert hall,

to substitute in (6), giving us

(7) Jones admired the man who crossed to the hall on the pedestrian crossing, but not the man who jaywalked.

(I am aware, however, that in a case like this it my not be the verb admired, but the particular description given of the object, which blocks substitution.)

The second assumption one would have to make to accept (2) and (3) as expansions of (1) in that as phrases form complex referring expressions out of two simple ones. This is to reject your adverbial account, in which as is treated as equivalent to a certain sort of because clause, adding an explanatory component (p. 5). If so, then to accept (2) and (3) as alternative expansions of (1) would amount to saying that every attribution of an attitude to something of form (1), implies either that there was some unspecified reason for that attitude, i.e. (2') "Ab, for some reason or other," or else it gave the reason for an attitude to now unspecified object, i.e., (3') "A something, because it is b." It does not seem correct to me to claim that every attitude is a reason-backred attitude, and (3') is particularly implausible as a reading of (1). I am making the dillferent claim that every attitude, reason-backed or not, is an attitude to something in some role it plays, or is taken to play (it may

-3-

of course be an attitude to it in any of its roles). At this stage I have taken account only of your preliminary analysis of as, p. 3-5, not of the modifications you introduce later in your letter, which I shall discuss later. For the moment all I want to note is that my acceptance of (2) and (3) as expansions of (1) depends upon my rejection of any because analysis.

The second fault I want to plead guilty to is carelessness in my statement of your view, and unimaginativeness in applying it. At the beginning and end of my letter I attributed to you, correctly I think, the view that intentional objects are propositions, either beliefs or intentions. But in the main part of the letter I spoke only of beliefs. I did not mean to suggest, what your writings clearly rule out, that you wished to reduce intentions to beliefs. (Incidentally, Casteneda seem to see your view that intentions result from a shall operation on beliefs as a sort of reductionism.) It was simply that it didn't occur to me to look for intentions in fear or admiration. I an still at a loss to find any in admiration or liking. Must I say "Would that I were like him!" if I admire anyone? However, the main issue of the propositional character of the objects of the attitudes is not affected by this oversight, and I do not find my original A_s any more satisfactory with the intention "Would that harm not befall me!" added to it. I note, in passing, that there may be a problem involved in attributing to the fearer both the belief that harm will befall him and the intention that it not befall him. Can one intend, even in the weak sense, what one expects not to happen? If no, then the weak sense in not a very close relative of the strong sense.

My third plain error lay in thinking there was any serious problem about crossreference in A_s. Whether one uses your method, involving the higher level predicate true of, or whether, like Kaplan, one adapts Church's Δ, that apparent problem can be solved. It does strike the unsophisticated eye as odd, however, to have sentences like yours on p. 8 and 9, in which part of what is asserted in claimed to be true, while the rest is just asserted. If someone actually said

"That Smith is tall is true, and Jones believes that Smith is the tallest man in Pittsburgh,"

I would ask, "Why the invidious comparison? I would have hoped that both conjuncts of your claim were true. If you think both are, why single out one for special truth mention?" But this, I realize, in a very unsophisticated reaction and depends heavily on the pragmatic implications of "is true."

It was wrong of me to say, top paragraph on p. 4, that if you chose to transform the behavioural relation between Jones and Smith into a relation between individual concepts, then fear would become intentional. What I meant was that the machinery would then allow you to make it opaque, so there would be no need to isolate opaqueness in the implied beliefs or intentions. The forms you give on p. 9 allow you to distinguish opaque from transparent fear thus:

(8) (Ei) that Jones shudders at i is true, and i ME Smith, and . . . .
(9) (Ei) that Jones shudders at i is true, and i-Smith and i ME Smith and . . . .

In a parallel way tallness, or at least smallness, might be said to be an intensional property, when it is not a property of individuals whose identity conditions are extensional, but a counterpart property of individual concepts. Smallness is truly attributed to Jumbo under the concept elephant, but not

-4-

under the concept animal. Whether or not something can be truly said to be large, to be feared, to be believed to be large, will depend on the description given of it.

Now I come to consider the obscurities of my attempt to show that the analysis of as phrases I attributed to you led to unacceptable results. To some extent this is to attack a straw man, especially since I now have your new analysis to replace it. But in view or your painstaking attempt to reconstruct my reasoning, I owe you an answer to the questions raised by you p. 9-11, 15-17.

The sentence to be analysed was

(10) Jones fears Smith as examiner.

The analysis attributed, A_s, was

(11) Jones shudders at Smith and/(when) he believes Smith is dangerous and/(when) he believes Smith is examiner.

In the counterfactual version this was strengthened by the addition of

(12) If Jones did not believe Smith examiner, he would not believe him dangerous, and would not shudder at him.

I rejected the strengthened versions for much the reason you give on p. 11, namely that if the counterfactual is added to the when version, then it amounts to the claim that the only basis Jones has for fearing Smith is his belief that he is examiner. This would make (10) incompatible with

(13) Jones fears Smith as secretary,

as long as being secretary and being examiner are not mutually entailing. But it in not incompatible.

You say on p. 10 that the counterfactual suggestions "comes close," but I do not see that any counterfactuals are involved in the because analysis you now offer. If there are implicit counterfactuals, I would be interested to know what they are. Perhaps

(14) If Jones didn't believe Smith examiner, and didn't believe him anything else dangerous, he would not fear him.??

As far as the unstrengthened version (11) was concerned, my first move was to raise the question of whether the beliefs spoken of in it were transparent or opaque. To settle this I looked at the situation reported in (10), not at (11) itself. An incomplete analysans can scarcely indicate its own completion. I asked, of the analysandum, whether the conditions you specify as necessary for transparent belief were met in that situation. I concluded that there was ground for calling the beliefs in (11) transparent. The fully spelt out reasoning was:

Jones fears Smith as examiner. (10)
Whatever a man fears he believes dangerous. (Assumed)
Therefore Jones believes Smith as examiner dangerous.

I assumed this was an ^obelief, that it was what, if anything, Smith says in his heart. But now we have a particular individual concept, Smith as examiner (assuming it qualifies as such), such that Jones ^obelieves it dangerous (understood in the appropriate way). By existential generalisation we then get

(Ei) ^oBfi

But it is also true that the concept Smith as examiner applies to the same individual as Smith, so we can get

-5-

(Ei) i ME Smith, and ^oBfi.

But thou, by your definition, we can say that Jones ^tbelieves Smith dangerous. Thus the belief in (11) is transparent. Since it is, we can substitute co-referential expressions for "Smith," and so conclude that Jones ^tbelieves that Smith as secretary in dangerous, indeed that Smith in all his guises in dangerous.

You say that if I get to this point I am home free, but I am not so sure of that now. All that I have shown is that

(15) Jones ^tbelieves Smith as secretary is dangerous

which does not imply a

(16) Jones ^obelieves Smith as secretary in dangerous.

That in what would lead us to the unwanted result that Jones, by virtue of fearing Smith as examiner, also fears Smith as secretary, despite the lack of connexion between the two roles.

You say, on p. 17, that

(17) Jones believes of Smith as examiner that he in dangerous (^tbelief)

entails

(18) Jones believes Smith is examiner (noncommittal between ^tbelief and ^obelief, but presumably ^obelief).

I do not see that (17) implies (18). Only if the belief in (17) were opaque would that be so.

I find myself now puzzled by the links between opaque and transparent belief, as you construe them. Can one belief be both? Presumably a term either allows of substitution or it does not, so a belief could not be both transparent and opaque. Yet in my previous argument the belief that Smith-as-examiner was dangerous was, as an ^obelief, the ground of the existential generalization which enabled us to get the ^tbelief that Smith was dangerous. Once we had that we could, by the same ME claim used to satisfy the definition, substitute Smith-as-examiner for Smith in the ^tbelief, thereby getting the ^tbelief that Smith-as-examiner was dangerous. So this belief seem opaque and transparent. This result does not depend upon the questionable acceptance of Smith-as--examiner as an individual concept, nor of the ME claims using it. It will hold whenever we know the particular belief which is the ground of the existential generalization needed to satisfy the definition for another ^tbelief. Should one be worried by this result, and is it a genuine result?

At last I come to what is of most interest, your new suggestions for the analysis of as phrases. I think it captures such of their content, and has exciting possibilities, but some things puzzle me. The comments I shall make are made as such in a spirit of inquiry as in criticism.

First I note that your strategy of transforming fears as into fears that will not work for many psychological O-verbs which do not admit of a that construction. There is no reason to expect one analysis to fit all these verbs with their differing logical grammars, but unless the as changes in import from verb to verb, one would hope for an analysis with enough internal structure, to do justice both to the differences between the verbs and to the fact that they can all take an as construction. I cannot see how to adapt your analysis of

-6-

fears..as to cover admires..as, likes..as, worship..as. Even if one allows "has an admiration of," one cannot get that into analogue of the suggested fears that construction. You could skip this step, and simply say that

(19) Jones admires Smith as an examiner

is to be analysed as

(20) Jones has an admiration for Smith because he believes he is an examiner.

But can this be further expanded into something analagous to your expansion of

(21) Jones fears Smith as examiner

into

(22) Jones has a fear of Smith because he believes he might examine, and by examining harm him?

But what could such an analogue be? Perhaps

(23) Jones has an admiration for Smith because he believes he might examine, and by examining do good (surely not "do him good").

It seems to me that what in closer to the sense of (19) is

(24) Jones has an admiration for Smith because he believes he "examines well," that is, that he performs well as an examiner, that is, that by Jones' standards for examiners, Smith does well.

This last form, however, contains the as you hope to eliminate. But I anticipate. My first point is simply that it in not obvious how to adapt your analysis to cover verbs which don't take a that-construction.

Secondly, I note that (20) is not equivalent in sense to (19), and nor, I think, does (22) capture (21). As I have suggested, it is not because Smith is an examiner that Jones admires him in some qualified way, it is because he examines conscientiously, or perhaps without undue anxiety, or even without waste of too much time. There is no knowing from (19) exactly what qualities Jones looks for in examiners. To know (19) is not to know why he admires him, but merely to know in what capacity Smith in admired. The question "why?" can still be raised. All the as phrase does in to put limits on what would then be an answer. "Because he grows fine roses" would not, but it would be a perfectly good reason for Jones having an admiration for Smith.

I am aware that the distinction between what Smith does, given in an as phrase, and the manner in which he does it, given in a because clause, is not a sharp one. But where one verb construction tells us both things at once, as in the case

(25) Jones has an admiration for Smith because he grows fine roses,

or

(26) Jones has an admiration for Smith because he always wins an argument,

such because sentences do not transform gracefully into the as sentence which your analysis would give us:

(27) Jones admires Smith as a grower of fine roses
(28) Jones admires Smith as a regular winner of arguments.

These have an odd ring to my admittedly biased ear. (27) suggests that there is some unspecified way in which Smith grows his fine roses, which, compared to the way others grow fine roses, excites Jones' admiration, while (28) suggests that Smith has a way of carrying off his regular wins (with modesty, perhaps) which Jones admires. I have now committed myself to the thesis I want to suggest, that the as phrase specifies the capacity or guise in which the object calls

-7-

forth the attitude, while a because clause, even when it has the form you specify, does something else, or something more than this, since it tells us how the object meets some standards or demands relevant to performance in that capacity. All this talk of performance fits cases like

(29) I admired Gielgud as Lear

better than it does cases where the intentional object is not a person. But even in such cases we can in an extended sense think of apples' "performance" as cookers or as eaters, and so on. If this thesis were correct, one would expect, what I think one finds, that what follows the as is rarely a long complicated description, but is a simple role label. Long details are relevant in because clauses, but out of place in as phrases. Thus my second point in that not all because clauses of the type you give can be transformed into sensible as phrases. This suggests that some restriction on as phrases is not captured by your because analysis. I have given you my view that as phrases specify capacity, and so imply relevant standards, but do not, like because clauses, tell us how such standards are met.

Thirdly I note that your suggestions for the analysis of as phrases following psychological verbs, p. 11-14, differs from that given earlier in your letter for as in causal contexts, which you at first suggested were parallel. In causal contexts you interpret the as as giving explanatory comment on the claim made by the main verb. But your later psychological analysis requires us to weaken the main verb whenever we transform an as into a because. Do you mean this to be treated as a genuine point of differrnce between the causal and the psychological cases? I might add that I find your causal examples unnatural -- I do not think one would use as to make the sort of claims you give on p. 4. Perhaps qua masks ambiguity. The smog example, p. 18, sounds more acceptable to me, and there I think you are right that it does transform into a because, although in other similar cases it does not, e.g.,

(30) As an allogen, smog is as unhealthy as pollen.

I am not offering any analysis of such cases, I merely inquire whether your view is that the because analysis which works for non-psychological cases has to be modified for psychological cases.

Fourthly, I note a possible difficulty with the procedure of transforming as in psychological contexts, into a because, with concommitant weakening of the main verb. What do we do when it is already weak?

(31) Jones has a fear of Smith as examiner, but a liking for him as a drinking companion.

Can we weaken "has a fear of" to "has a partial fear of"?!

A parallel difficulty holds for the reverse transformation from because clauses (of your sort) to as. We should strengthen the verb then, but it may be already of full strength. Would

(32) Jones fears Smith because he has threatened his life on several occasions

become

(33) Jones more than fears Smith as frequent threatener of his life?

Your analysis, unless supplemented with strengthening and weakening devices which can increase the range of psychological verbs, would seem to imply what is contrary to fact, that as never follows a weakened verb, nor because (of the right sort)

-8-

a strong verb.

Fifthly I want to explore the implications of your view for cases where we have both an as and a because in the sentence to be analysed, and in cases where we have a second as, and a second because. This is very tentative, since you can always add what rules you need to see that unwanted complexes are not produced. What I want to point out is that because builds on a previous because in a way different from an as or a previous as, so that some rules will be necessary if the transformations are to go smootbUy.

Let us begin with a mixed case

(34) Jones admires Smith as an examiner, because he always asks questions which enable the candidate to show his ability.

If we transform this to get rid of the as we would get

(35) Jones has an admiration of Smith because he is an examiner, because he always asks qaestions which . . . .

The second because cannot be construed as giving the reason why Smith is an examiner, which it would if it behaved like the second because in

(36) I like him because he likes me, because I once helped him out of a difficulty.

When we have a string of becauses, each may explain the fact given in the previous clause. The Kingdom was lost because the battle was lost because the rider was lost, because the horse was lost, because the horseshoe was lost, because a nail was lost. This cannot be what we want in (35) if it in to be a transformation of (34).

It might be construed as similar to the because in

(37) I like him because he likes me, because I am a reciprocating type.

Here the second because gives a reason, not for the fact asserted in the previous clause, but for the connection asserted in the first because. It is more difficult to lengthen the string of such becauses, and one is more likely to get mixed strings, as in

(38) I like him because he likes me, because I am a reciprocating type, since I am too timid to initiate any emotional tie.

Here the since clause gives a reason for the fact in the previous clause, while the second because gives a reason for the connection betveen the first two clauses. Where we do get a string of because clauses, each explaining the connection the previous because asserted, then each new clause would cite a fact of higher explanatory power, wider generality. However (35) is not of this sort. The second because does not provide us with a deeper explanation of why the connection given in the first because should hold. On the contrary, it introduces a more specific, not a more general fact, one with narrower rather than wider explanatory power. Smith's ability to ask good questions seems related to Smith's being examiner quite differently from my being a reciprocating type to my liking his. Given that I reciprocate, and that he likes me, it is to be expected that I like him, since liking will be a case of reciprocating. But given that Smith asks good questions, and that Smith is examiner, if it is to be expected that Jones has an admiration for Smith, it in not because Jones'

-9-

attitude is a case of any more general connection claimed in the 'explanatory' premise.

I am aware of the sketchy and incomplete nature of this argument, and its dependence on the exhaustiveness of my list (of 2) possible jobs for the second because. My point is that the because we get from an as does not seem to be related to other becauses, in the sentence, in either of the two ways in which becauses normally relate to one another. Is there another way in which a because can build on an earlier because, which is closer to the artificial as-derived case of (35)?

Let us now take (34) and perform the reverse transformation, from because to as. We would get, as a first attempt

(39) Jones admires Smith as examiner, as one who asks questions which enable the candidate to reveal his ability.

One trouble with this is that we have not strengthened the main verb in transforming the because into an as. To do this without altering the given as construction we would need a conjunction:

(40) Jones admires Smith as examiner, and more than admires him as one who asks questions which. . . .

Yet this seem plainly different in sense from (34) and is certainly not equivalent to (35).

It might be helpful to look at the relationship between as constructions, in cases of a multiple as occurring naturally, not by application of your theory. One might well may

(41) Jones admires Nixon as president, as conducter of our foreign policy, but not as leader of domestic policy.

Here it seems the second as specifies the sub-role, included within the role of president, which has been performed in a way, unspecified in (41), which leads to Jones' admiration. Were we to add a because clause, it would tell us why Nixon's foreign policy pleases Jones. I suggest, then, that to pile on as phrases is to specify roles within roles, and this is quite different from giving reasons for reasons. If we tried transforming (41) along your lines, we would get

(42) Jones has an admiration of Nixon because he is president, because he is conducter of foreign policy and not because he is leader of domestic policy.

Such a sequence of becauses make sense neither by taking the second to give a reason why Nixon is president, nor by taking it as explaining why that fact should excite Jones' admiration. I have suggested that a second because clause gives a reason for an earlier reason in one of two senses:

(a) It gives a reason for the earlier reason as a fact (in 36)
(b) It gives a reason for the earlier reason as a reason (in 37)

But (42) makes no sense along either lines.

Another way to bring out the difference between a pile-up of as phrases and a pile-up of because clauses is to point out that (41) implies

(43) Jones admires Nixon as conducter of foreign policy.

That is, the first as can be dropped without changing the truth value. But

-10-

to drop the first because from (36) would give

(44) I like him because I helped him

which in surely not implied. To drop the first because from (37) would give us

(45) I like him because I am the reciprocating type.

This in all right, but incomplete as a reason, whereas (43) in not incomplete as a role specification.

All of this makes me think that one cannot successfully absorb as phrases into because clauses. If one does, one has to admit that the because clauses in question behave in different ways from familiar reason-giving because clauses. What would be the point of reducing a non-propositional structure to a propositional one, if the latter is artificial and only to be understood by appeal back to the non-propositional as?

I have now completed my comments on your because analysis of as. There remains just one last remark on the deep issue between us, the relation between language as thought and language as communication. (Note uneliminable as constructions!) You say that I was wrong to assume that the fundamental intentronal context, the semantic, involves taking "red" as a ·red.· I can see that "taking" might suggest more choice on the language-user's part, more real options, than do normally obtain, but my question could as easily have been raised in other terms. If the fundamental intentional context involves recognizing or using "red" as a "red," then the fundamental case will be just an non-propositional.

Here I suppose you might say that

(44) Jones recognises "red" as a ·red·

is equivalent to

(45) Jones recognises that "red" in a ·red·.

But one can only recognise that p, if p in true, and is it true that "red" is a ·red·? Or is it only as a word, and as an English word, that "red" is a ·red·? Perhaps only words can occur between quotation mark, in which case the as phrase in

(46) Jones recognises that "red," as a word, is a ·red·

would be redundant. But then the quotes themselves would be tantamount to an as phrase. The noise made by someone saying "red," if made by something else which is not a language-user, (nor adjunct to one, like a record) would not be a "red." Part of what learning to speak involves is learning what noises to treat as words, known or unknown. I do not see a plausible way of changing the "as a word," or "as an English word," needed in the dot-quate belief, into a because construction or other propositional construction. The closest one could get would be

(47) "Red," in as far as it is an English word, is a ·red·.

But this introduces a counterpart as, as a propositional connective, like Chisholm's I, and now we need to explain what it is for a noise to be an English word. What else is this than its being used as a role player for one of the roles that language (usually along with other languages) makes possible? The conventions of use and of use-preserving criticism create the roles the words fill. To have words is to treat certain behaviors as subject to linguistic criticism, and to similarly criticisable responses. I do not want to claim that such

-11-

"treating" is deliberate, in the sense that we pause and deliberate whether or not to respond to noises as words. But that does not show it not to be action. The things Austin called illocutionary acts include many which are typically not deliberate (find, rate, grant, doubt, accept, agree, understand, mean, refer), and there need be nothing non-candid about a speech act. They are performances in the sense that they are convention constituted and convention-criticised, not in the sense that they are rehearsed, or selected after deliberation.

These last remarks are very sketchy and dogmatic, but it would make a monster letter more monstrous still to offer arguments in their support on this occasion. I have said enough to indicate that taking the duck-rabbit as duck (or as rabbit), although not done at will, can still be a stubborn performance.

Cordially,

Annette C. Baier