WILFRID SELLARS
REPLY
I'm going to follow yesterday's example and begin with Hilary's comments because they are freshest in my mind. My listing and illustrating of language-entry, inference and languagedeparture rules was intended to give only the bare bones of a theory of language. They clearly presuppose, for example, rules pertaining to wellformedness and the relation of deep structure to surface structure. More immediately relevant to Hilary's substantive remarks is the fact that I regard rules of inductive inference as rules for the reasoned change of a language. There could be linguistic communities which do not, as yet, have a rational method of changing their language; it changes, so to speak, in a Darwinian manner. If the language doesn't work in certain respects, it is modified - not because alternative rules are considered and adopted, but simply because those who didn't acquire the relevant new patterns of verbal behavior are less effective in competitive situations. Thus, as I see it, since thinking is, in a primary sense, languaging, induction must be construed, in the first instance, in terms of rules which pertain to languaging, yet do not fall in any simple way into my tidy trichotomy. If it had been my purpose to sketch a systematic account of the functioning of language in behavior, I would certainly have had a great deal more to say about the variety of rules involved. It must be remembered, then, that my aim (and this remark is directed partly at Dan) was to put forward verbal behaviorism not as a detailed model for a cognitive psychology, but as a tool for understanding those aspects of thought and language which have traditionally puzzled philosophers, and still generate more heat than light. In brief, i presented verbal behaviorism as a strategy for dealing with the basic concepts of semantical theory and the philosophy of mind. Thus, I will shortly indicate my agreement with many of Dan's comments, for they pertain to the fact, which I am happy to grant, that verbal behaviorism provides at best a coarsegrained explanatory framework. But, then, so is our common-sense world picture a very coarse-grained explanatory framework. The latter, for Synthese 27 (1974) 457-466. All Rights Reserved Copyright © 1974 by D. Reidel Publishing Company, Dordrecht-Holland
458
WILFRID SELLARS
example, doesn't contain the conceptual resources to enable us to explain why atom bombs work; so why should we expect a verbal behaviorist framework to accommodate an explanation of lightning calculators ? I was very impressed with Hilary's remarks about language as a social institution. I quite agree that there is the kind of division of linguistic labor to which he refers, and shall expand on the agreement in a moment. I also agree that the meaning of an utterance is not to be understood simply in terms of a momentary linguistic state of the utterer. Again, I was deliberately abstracting in my paper from the diachronic aspects of language. I am in complete agreement with Michael Dummett's remarks last evening about the importance of taking into account the ways in which languages change. As indicated above, I view induction as a rational way in which people change their minds; change them not only in the sense of generating new beliefs, but by reshaping their conceptual frameworks. I also find Hilary's discussion of the meaning of such stuff-sortals as 'gold' and 'water' both challenging and illuminating. But to comment on his remarks, I must first clear away some misunderstandings and call attention to what may be basic disagreements. Hilary sees as "the problem of semantics.., not what the meaning of a word is, but what the normal form description of the meaning of a word should be". With this I would agree to the following extent. Once one appreciates that and how the meaning statements with which classical logical theory has been primarily concerned I have the form '---'s (in L) are "...'s where " . . . " stands for a functional sortal, then the problem does indeed arise as to the 'normal forms' in which the criteria for such functional sortals are best to be formulated in the interests of clear and systematic linguistic theory. As I see it, however, despite~ his friendly gesture toward dot-quoting and illustrating sortals, Hilary trivializes the above strategy when he suggests that "as long as we have some theory as to what it is for a word to have a particular meaning.., we can introduce a sortal " A " for all words.., with that same normal form". For it obviously is not a matter of having some theory, but of having a correct theory, and the strategy I have offered is not something that can be applied after a correct theory of meaning has been established, it is (or purports to be) that correct theory
REPLY
459
itself. For the crucial step in avoiding "meanings as objects" - an objective which Hilary shares - is that of interpreting both the 'means' o f " ' u n d ' (in German) means and" and the 'stands for' of "'dreieckig' (in German) stands for triangularity" as specialized forms of the copula. Obviously no progress would have been made if, for example, the functions ascribed to expressions by dot-quoted sortals were those of meaning or standing for specific entities, whether attributes or classes, abstract objects or Fregean concepts. To do so would not only be to misconstrue the senses of 'means' and 'stands for', but to stay at the surface level of giving translations, which, as Quine has so rightly stressed, presupposes a 'theory' of alien linguistic behavior and its similarity to and difference from domestic linguistic behavior. One gets beyond translation in the case of "'und' (in G) means and'" by formulating the syntactical uniformities with respect to 'and' which are endorsed by the English speaking community and which, it is claimed by the meaning statement, are paralleled by syntactical uniformities with respect to 'und' endorsed by the German speaking community.2 In general, one gets beneath the surface of translation by comparing specific patterns of rule-governed linguistic activity which, for my strategic purposes I classified under the headings: language-entry, inference, and language-departure. I certainly agree with Quine that the mere fact that a use is stipulated doesn't guarantee anything with respect to meaning. A scientist, for example, may stipulate a 'definition' which doesn't catch on, even in his own idiolect. And even if it does come off, the result is caught up in the Heraclitean flux of linguistic change. This does not mean that the use is no longer subject to rules, but simply that rules are not to be identified with stipulations. The primary symptom of the functioning of rules in our language is to be found in the use of appraisal words, like 'correct' and 'incorrect', which express our sense of grammaticality, of coherence, of what makes sense. Thus rules are living systems of appraisal which may or may not result from stipulation and, for the most part, don't. When 'Semantics' turns its attention to the search for normal forms for the "description of the meaning of a word" without having arrived at a correct account of meaning statements, what one often finds is an enriched and embellished version of Porphyry's Tree. This amounts to giving translations seasoned with a mixed bag of categories. The reference
460
WILFRID SELLARS
to the kind of rule-governed uniformities (positive and negative) which is the essence of the matter is usually implicit in these categories, though it may, at least in certain generic respects, be made explicit. On the whole, however, the emphasis is on translation and, as a result, those features of classical semantical theory which rightly exercise Quine are perpetuated. In the main portion of his remarks Hilary develops a numbers of interesting points about traditional theories of meaning. He claims (correctly, I think) that philosophers, by and large, have made two assumptions about meaning: (1) "... the meaning of a speaker's words does not extend beyond what he knows and believes", to which he adds "or possibly.., is disposed to do"; (2) "... meaning determines extension". He argues that when carefully scrutinized these two assumptions turn out to be "if not jointly inconsistent, at least inconsistent with the facts". He proceeds to sketch a theory which "preserves (2) and sacrifices (1)". Before I comment on this theory sketch, which I find highly suggestive, I must point out that Hilary is quite mistaken when he construes my theorizing about meaning as a theorizing about "individual competence". I have always stressed that language is a social institution, and that meaning is to be construed in social terms. Thus I certainly would not subscribe to the first of the above two assumptions. What a speaker's words mean (to use a less ambiguous expression which has the additional advantage of pointing away from Gricean labyrinths) is no more to be defined in terms of his beliefs and purposes than is, for example, the legal significance of his actions. Any adequate philosophy of mind must, indeed, be concerned with the relation of an individual's propensities for rule-governed behavior and the practices of his community. But this relation must be construed in such a way as to preserve, in a less metaphysical mode, something like Hegel's distinction between individual minds and 'objective spirit.' I am, therefore, in sympathy with Hilary's general enterprise. I also find his suggestion that "language... is not a tool like a saw.., that one person can operate", but rather "a tool like a battleship.., that takes a crew to operate", most illuminating. On the other hand, it seems to me that the points he is concerned to make have little to do with the contrast between individual and group, and concern rather the contrast between different groups, whether they belong, in a straightforward sense, to the ~ame larger community, or are as isolated as the populations of different possible worlds.
REPLY
461
In making his points, Hilary concentrates on such mass or stuff words 3 as 'gold' and 'water', though he indicates that what he has to say could be extended to cover "almost all the nouns in our vocabulary, and a lot of verbs, adjectives, adverbs, and so on". I will concentrate on his examples. Thus, was the meaning of 'gold' in 1 B.C. the same for naive laymen and for jewelers? And did its meaning for laymen refer to its meaning for jewelers? I suspect that the answer to the first question is 'no' while the answer to the second question is 'yes'. (I postpone questions concerning the reference of 'gold'.) Clearly the layman would have admitted that the jeweler 'had a better idea of what gold is'. But what does this amount to ? Suppose 'gold' meant to the layman a substance which is ABC and rare. 'Rare' involves a reference to discriminators who can distinguish between sentences which are ABC and rare, and substances which are ABC but not rare. Let us follow Hilary in calling them jewelers, and suppose that they draw the distinction in terms of whether or not test T yields result R, the layman knowing nothing about such matters other than that the jeweler uses a test. Why not say, then, in first approximation, that 'gold' to the layman means substance which is ABC and rare as determined by the jeweler's test, whereas to the jeweler 'gold' means substance which is ABC and Rs when Td. The two meanings are obviously not different simpliciter. They are kindred and, which seems to be Hilary's point, such that a recognizable form of social division of labor is involved. Notice, however, that on my formulation the question of whether jewelers give deceitful answers to the question 'Is this gold or fool's gold?' does not arise. And, indeed, this possibility does seem irrelevant to the iss~es at stake. Now the symbiosis of laymen and experts with respect to the meanings of words can often be fruitfully viewed as the result of the persistence of earlier strata of meanings as science develops. Light is thrown on synchronic relations of meanings when we view language diachronically. Thus, did 'gold' mean for experts in the late 19th Century what it meant for experts in 1 B.C. ? Here I would stress what I have called the promissory note aspect of substance sortals. 4 Does the word 'gold' refer to what gold really is? (Notice that this question is not quite the same as, though for present purposes equivalent to Does the word 'gold' refer to (denote) items which are what gold really is?) The answer is, in a sense which requires careful explication: Yes. Words for nat-lral substances have as
462
W I L F R I D SELLARS
part of their function a classificatory role in the ongoing enterprise o f physico-chemical explanation, and it is in terms of this enterprise that 'what gold really is' is to be understood. Thus, what gold really is is not an extra-linguistic object, but rather the successor substance sortal which would inherit, in developed form, the role of the word 'gold' as it functions in science today. It would belong to a successor framework~ which, if it came to be realized, would satisfy the as yet ill-defined criteria which constitute the regulative idea of an ideally explanatory f r a m e w o r k one in which, to use a Bradleian turn of phrase, the intellect could rest. The ideal substance sortal, gold, can be referred to, but the concept of such reference must be dealt with gingerly. It is as tempting to reify what this ideal sortal would stand for, as it is to construe what 'triangular' stands for as an abstract extra-linguistic object. We have here one strand of the distinction, with which Locke wrestled, between a nominal and the real essence of gold. There is, indeed, a strong temptation to avoid the puzzling idea of successor sortals and simply posit the existence of real essences, entities which are construed as the true 'reference' of contemporary stuff-sortals. But while the distinction between real and nominal essences is a useful, indeed an important, one with which, in my Lockean phase, I was much taken, it is, if left at the intuitive levels, a place-holder which marks a problem rather than a solution. There is obviously no puzzle about saying that some of the stuffs which people once correctly called gold weren't really gold. Here we recognize both the kinship of and the difference between their use of 'gold' and ours. But does it make sense to say that some of the stuff we correctly call gold really isn't gold? Yet is not this possibility implied by the very idea of what gold really is? Is not the concept of the ideal denotations of 'gold' and 'water' correlative with the concept of the ideal meanings of these sortals? After a bow in this direction I must drop the topic, for thereis no time to develop it, even i f i could. I have, however, said enough to lay a foundation for some comments on Hilary's piece o f science fiction. Is water on Twin Earth the same thing as water on Earth? Are lakes on Twin Earth made of water? Suppose that at comparable stages in the evolution of Earth and Twin Earth both we and our twins used X, Y and Z as our criteria for water. If these criteria were all that 'water' meant, that would be the end of it. Their water would be the same as our water. But, ex hypothesi, the real
REPLY
463
essence of their water is different from that of our water. 6 If we were transported to Twin Earth (with the help of a little transpossible-world identity) we would say, on contemplating Twin Lake Michigan, "We have lots of the same stuff at home", and we would be wrong. And our mistake would be somehow connected with the fact that we acquired our dispositions and propensities with respect to 'water' in the neighborhood of our Chicago and not their Twin Chicago. But this 'indexicality' concerns the fact that the successor concept 'H20' which correctly applies to the stuff of which these objects (e.g. Lake Michigan - give or take a little pollution) consist, does not correctly apply to the superficially similar objects with which we would be confronted if we (per impossibile) were transported t o Twin Earth. As for Russell's puzzle about the universe coming into existence five minutes ago, it strikes me that if five minutes ago there was a clap of thunder, it is reasonable to suppose that it must have been preceded by a flash of lightning, and that this is why we feel conceptually uptight about the meaningfulness of the conjecture. I turn now to Dan's comments and will begin by emphasizing once again that I was putting forward verbal behaviorism as an account of language which would provide a model with respect to which categories pertaining to sense and reference, content and intentionality might be illuminated. I have always been deeply concerned with traditional puzzles about the intentionality of thought, but it wasn't until I began to construe the intention of a thought as analogous to the meaning of an expression, in the course of construing the framework of thoughts as a 'theoretical' framework related to the 'observation' framework of overt linguistic behavior (somewhat as a micro-physical theory is related to the macroframework of perceptible things) 7 that I began to see how they might be resolved. I had found a strategy, a program, and much of my work in the philosophy of mind since then has been a matter of carrying it out in detail. I pointed out in my paper that one of the motives for classical accounts of thoughts as 'inner episodes', was the general drive we have to find purely occurrent happenings which might account for changes in the propensities and dispositions of things. Thus, in our attempt to understand why soft iron in a helix through which bursts of current are passed becomes successively magnetic and non-magnetic, we succeeded in correlating these rapid changes in propensity with micro-physical episodes in the iron.
464
WILFRID SELLARS
Something rather like tiffs strategy is involved when verbal propensities are explained in terms of 'thoughts'. But if 'languaging' is the model for the 'thinking', which belongs to the desiderated fine-grained explanatory framework, this does not entail that thinking, in the latter sense, takes place in language. Thinking-out-loud does, trivially, take place in language, but thinking, in the desired sense, is at best analogous to languaging. I have stressed this from the beginning, and I am sorry that my failure to specify exactly which sections of the papers I sent to Dan bore directly on what I was going to say led him to overlook it. Nor have I spelled out how strong is the analogy between thinking-outloud and thinking. Theories typically are built around a model, and models, as I emphasized in E P M carry a commentary. Some features of the model are, of course, carried over into the new domain. Others are discarded. And it is often a rather intuitive matter as to what is carried over and what isn't, s Yet I do think that thoughts are analogous to languagings in those respects which are highlighted b y logical theory, transformational grammar and the concepts of language-entry and language-departure transitions. It is only in this sense that "[I commit myself] to the view that thoughts that are somehow verbal are, as psychological events and states, actual mediators in human behavioral reaction to experience". Thoughts as "actual mediators" are "somehow verbal", indeed, but the latter phrase must be compared to the opening phrase of "items which are somehow little particles constitute gases". I do not think, then, that Dan and I disagree as much as he suspects we do. Yet I would stress, perhaps more than he does, the extent to which the commonsense theory of thought episodes elaborated by classical philosophers of mind already contains the 'space' for at least some of the cash which a fine-grained neurophysiological account of the processes which mediate intelligent behavior will provide. Indeed, even the verbal behaviorist model has greater flexibility than Dan seems to allow. Consider, for example, a thinking-out-loud. Jones thinks-out-loud "the bus didn't stop!" Well, that takes about two seconds, Yet even on a verbal behaviorist account one has to construe verbal propensities as the sort of thing which can shift more quickly than that. There is room for the sophistication which has its counterpart in an advanced theory of reverberating neural circuits which, in different neural contexts, might initiate a
REPLY
465
given utterance at any moment during a certain time span. This sophistication is already adumbrated in commonsense notions concerning the curious combination of "speed' and 'duration' which is characteristic of thought. The commonsense theory doesn't add much to the explantory power of verbal behaviorism, but it does add something. In my presentation of the verbal behaviorist model, I stressed shortterm propensities to think-out-loud. But, of course, a more adequate account of verbal behaviorism as a proto-theory of persons would also stress less 'proximate' and, indeed, 'remote' verbal propensities. Furthermore, verbal behaviorism can certainly allow for the telescoping of abilities and propensities. I have no theory to offer about lightning calculators and such, but I would point out in connection with Mr. Tortoise and Mr. Hare that even the familiar process of learning to play the piano illustrates this telescoping. One starts out, as it were, by having to choose each motion of each finger, and ends up by having a propensity pertaining to a whole sequence of behavior which in no simple way consists of propensities pertaining to its behavioral components. The point is not merely that now one needs only choose to begin playing a waltz and the waltz unrolls. More relevant to Dan's concerns is the fact that, given an appropriate set, one can play phrases from any given part of the waltz, moving directly, for example, from the opening phrase to the closing phrase. The explanation of such connections of propensities may await neuro-physiological theory, but the categories for describing them are already at hand. If I had been seriously putting forward verbal behaviorism as a prototheory of persons, i.e. as a proto-psychology "of the higher processes" I would have developed it along these lines. However, since my immediate purpose was to use verbal behaviorism as a strategy for getting clearer about semantical categories, indeed, by construing meaning statements as functional classifications, I shoved the problems .with which Dan is concerned under the rug. I regret that I have so little to say "offthe cuff" on these matters. Above all I am sorry not to have heard the details of Dan's reaction to my treatment of quantification into belief contexts which he characterizes as "completely wrong". This topic was, indeed, the central theme of one of the manuscripts I sent him to prepare the way for the "real paper" which I knew would be late. But although it was the central theme of that manuscript, since it was a reply to Quine's critique of my earlier paper on
466
WILFRID SELLARS
the topic, 9 it m a d e essential use of the a p p r o a c h to language a n d t h o u g h t which I have presented today. T h e topic of quantification into belief ( a n d other modal) contexts was very m u c h o n m y m i n d w h e n I was reading m y paper, and, i n fact, I almost e m b a r k e d o n a n i m p r o m p t u discussion o f this topic to illustrate the usefulness of the distinctions I was drawing. B u t I could see t h a t I was r u n n i n g o u t o f m y time a n d y o u r patience a n d abstained. I can see n o w t h a t a n o p p o r t u n i t y for a h e a d - o n collision was lost, a n d t h a t the sparring which has actually t a k e n place has been a n a t t e m p t to c o m m u n i c a t e , rather t h a n a t a k i n g u p o f issues. F o r this I m u s t b e a r the full responsibility. NOTES 1 Which need to be separated from - though related to - the variety of other meaning statements which have been catalogued since Ogden and Richard's pioneering work. I am obviously idealizing the situation in a number of respects in order to highlight the difference between 'giving' the meaning of a word by the use of an illustrating functional sortal, and describing the complex mode of functioningwhich is the criterion for the applicability of the sortal. Thus I am not only abstracting from such facts as that 'and' in English carries with it temporal presuppositions, but also from consideration pertaining to degree of likeness of function. 8 I.e., in the ordinary sense, words, for 'substances'. 4 See, in particular, Sections 44-54 of 'Counterfactuals, Dispositions, and the Causal Modalities', Minnesota Studies in the Philosophy of Science, Vol. II (ed. by Herbert Feigl, Michael Striven, and Grover Maxwell), Minneapolis, 1958. 5 For an attempt to explicate the concept of a successor framework and related ideas, see Chapter 5 of Science and Metaphysics, London 1968, and 'Conceptual Change', in Pearce et aL (eds.), Conceptual Change, Dordrecht, 1973. It is worth noting in this connection that the idea of successor methods is as important as the idea of successor concepts. 6 To stipulate this does not, of course, commit us to the idea that either we (or our twins) now have an ideal knowledge of what water really is, but only to the idea that the conceptual route which began with our nominal essence substance which is X Y Z has diverged sufficiently from the conceptual route which began with their nominal essence substance which is X Y Z to ensure that these routes would not converge at a subsequent stage of inductive investigation. After all, one can know that a pain is not identical with the correlated physical state of the nervous system without having an adequate idea of the latter. 7 Cf. 'Empiricism and the Philosophy of Mind', in Minnesota Studies in the Philosophy o f Science, Vol. I (ed. by Herbert Feigl and Michael Scriven), Minneapolis, 1956. s For a more detailed account of how I construe the relationship between theories, models, and commentaries, see my 'Scientific Realism and Irenic Instrumentalism', in Boston Studies in the Philosophy of Science, Vol. 2 (ed. by Robert Cohen and Marx Wartofsky), New York 1965. 9 'Some Problems About Belief', in Words and Objections (ed. by Donald Davidson and 3aakko Hintikka), D. Reidel Publ. Co., Dordrecht, 1969. Quine's critique of this essay is contained in the same volume.