Chapter 52 Reply to Putnam Jerry A. Fodor
Not many of Putnam's criticisms seem to me to require extended discussion; and many of those that do have been covered in Chomsky's reply. I shall therefore restrict myself almost exclusively to those of Putnam's remarks that concern things I said at this conference and in my book Language of Thought.1
Putnam on "Fodor's Tautology"
Professor Putnam thoughtfully reminds me (and Chomsky) that it is "not logically impossible that our heads should be as empty as the Tin Woodman's and we should still talk, love, and so on"; it was "a little careless" of me to suggest the contrary.
It would have been if I had. In fact, "tautology" is not a term that appears in my formal presentation, nor did I introduce it into the discussion, nor did I suggest (ever or anywhere) that "no one appreciated until Nelson Goodman" that there can be "no learning without some laws of learning." (What I did suggest is that the demonstration that there is no induction without an a priori delimitation of the field of projectible predicates is owing to Goodman and has profound implications for theories of learning. This remark, which has nothing to do with the existence or otherwise of laws of learning, is one I continue to endorse.) Finally, the sense of "tautology" at issue in the discussion was, of course, not "truth of logic," but rather "obvious truth, self-evident truth... etc." Putnam denies that his strictures on "tautology" are "just a quibble," but they’ll do until a real one comes along.Putnam on Fodor on the Innateness of All Concepts
Putnam (mis)construes an argument that I gave in Language of Thought. ΓU work with his version first and then come back to how he got it wrong and why that matters.
Fodorzs reasoning is as follows: Learning the meaning of a predicate is inferring what the semantic properties of that predicate are, that is, concluding (inductively) to some such generalization as
(A) For every x, P is true of x if and only if Q (x).
But if (A) is in brain language, so is Q. (P need not be; P is mentioned, not used in (A). But Q is used, not mentioned.) And if (A) is correct, Q is coextensive with P, and is so by virture of what P means (otherwise (A) is not a correct semantic characterization of the meaning of P). So Q is synonymous with P; P is not a new concept, because there is a predicate (namely, Q) in 'brain language" that is synonymous with it. But P is an arbitrary predicate the brain can Ieam to understand—so no new concepts can be learned!
Putnam has an analysis of what's wrong with this argument: "The assumption is as strong as what Fodor wishes to prove. So all we have to do is show how it could be false... and nothing is left but a simple case of begging the question." Now, I suppose that Putnam is being a little careless here, for it is not easy to see how an argument could be valid unless its "assumptions" were (at least) as strong as its conclusion. What Putnam must mean to say is that the assumptions are as tendentious as the conclusion. But that is surely just false, since the (operative) assumptions are that learning a predicate is learning its meaning, and that language learning is (inter alia) the projection and confirmation of hypotheses. And there is a tradition of making just such assumptions which goes back for literally hundreds of years in philosophy, and which is, to all intents and purposes, simply unquestioned in contemporary cognitive psychology. Indeed, what is puzzling about the argument (if anything is) is exactly that it requires only these fairly banal assumptions to arrive at the wildly paradoxical conclusion that all concepts are innate. (I assume that it's part of the philosopher's job to reveal the paradoxical lurking in the prima facie Untendentious—and then to make it go away.)
Now, what can Putnam offer to replace these trouble-making assumptions? Putnam does a very odd thing at this point. Instead of following the doctrine about language learning proposed in such of his papers as "The Meaning of 'Meaning'" (to which doctrine we shall presently return), he suggests that "even if we grant that the brain Ieams P by making an induction, it need not be an induction with the conclusion (A).
It would suffice that the brain instead conclude: (B) I will be doing OK with P if subroutine S is employed." Putnam does not tell us what sort of subroutine S is, and he is thunderously silent on where such subroutines come from, but he remarks (correctly) that S will have to be spedftable in machine language if the machine is to be able to execute it. We may add (as Putnam rather astonishingly does not) that if learning to execute the subroutine S is to be identifiable with learning P, then the machine (brain, etc.) must not only condude, but condude truly, that it is "doing OK" with P if it uses S. (There is, after all, a distinction between having a predicate and merely believing that you do.) Well, then, what might be the subroutine vis-a-vis P such that, when you have learned to execute that subroutine, your belief that you are doing OK with P is true! The dassic suggestion is, of course, that you must have a procedure for sorting things that do and don't satisfy P by reference to whether they exhibit some property Q. And now if we add that the fact that the possession of Q determines the satisfaction of P is supposed to be a consequence of the meaning of P, what we have is just (A) all over again. In effect, my (A) is a version of Putnam's (B). It is, moreover, the standard version of (B); and Putnam neither suggests how we are to avoid the paradoxes that arise from taking (B) on construal (A), nor proposes a version of (B) that provides an alternative to (A).Putnam seems to sense this sort of reply in the offing, because he says that I might daim "that the machine language description of how to use, say, 'tree' is (a form of) the predicate tree. But this is simply an extension of use designed to make his [my] thesis an uninteresting 'tautology.'" It doesn't, however, seem to be an extension of use at all. Putnam has, in effect, endorsed the view that what we Ieam when we Ieam "tree" is a set of procedures for using the word.
It was, after all, Wittgenstein (and not I) who suggested that the best candidate for meaning is rules-of-use. And Putnam is surely aware of a long philosophical tradition that identifies such rules with (one or another form of) operational definition of a term. (This tradition is, by the way, enthusiastically endorsed by the "procedural semanticists" whose work was at issue.)I think, however, that what Putnam must really have in mind is something quite different from what this argument suggests—something that has very little to do with all this business about "subroutines." I think what he really has in mind is that we should abandon (the classic form of) the proposal that to Ieam a word is to Ieam what it means; that is, he wants to distinguish between learning P and learning the meaning of P and to
Reply to Putnam 425 argue that the latter is not necessary for the former (not, at least, if the meaning of P determines logically necessary and sufficient conditions for P's applying). This move, of course, really is tendentious, but Putnam argues for it in "The Meaning of ,Meaning"' and, though I think his arguments there do not, in the long run, persuade, I won't try to deal with them here. For present purposes, my point is just that, on this account, not any old subroutine S will do vis-⅛-vis P. We require a very special kind of subroutine associated with "tree" such that something might satisfy the subroutine and fail, for all that, to be a tree. Of course, if one can show that then one has shown that (A) fails; hence that no argument that rests on (A) would prove that the meaning of "tree" is innate.
The trouble is, however, that on this view the meaning of "tree" isn't learned either. Indeed, on this view it is quite possible that nobody now knows, or ever will know, the meaning of "tree" (in the traditional sense of, roughly, the essential conditions for being a tree).
For, whether something is a tree depends (so the story goes) not on its having the properties we learned to associate with "tree" (in particular, it is not determined by the outcome of executing subroutine S) but rather on whether it has those properties that "the progress of science" will (or may) come to tell us that trees must have. And, of course, learning those properties (the ones which, as it were, give the real meaning of "tree") isn't part of learning "tree."This is where it becomes important that Putnam has misrepresented the argument I gave in Language of Thought. What I did there was not to endorse (A) (the principle which, on our present reading, Putnam has brought under attack) but simply to run it as an example of what you would be committed to if you were to hold that to Ieam a word is to Ieam its (for example, operational) definition. But, as I pointed out (ad nauseam and with explicit reference to Putnam's views),2 weaker assumptions than (A) might be made about what is learned when one Ieams P; and, given the structure of my argument, those weaker assumptions about what is learned will comport with correspondingly weaker conclusions about what is innate. What I endorsed was, in short, an argument scheme: you tell me what you think is learned (when P is learned), and I'll tell you what you must be assuming to be innately available to the learning device. You say: "meanings are learned," and I'll show you that you must assume that meanings are innate; you say: "subroutines are learned," and ΓU show you that you must assume that subroutines are innate; in effect, you tell me what sense of "concept" you have in mind when you speak of "concept learning," and I'll show you that you must take concepts in that sense to be innate. I don't think there is anything in Putnam's remarks that undermines this strategy; I can't find anything in Putnam's remarks that even bears on it.
Putnam on What Fodor Has Rendered Otiose
Putnam says that 'Todor's hypothesis of a language of thought'... is not... the same as Chomsky's hypothesis of a 'mental organ' for speaking; it even makes the latter hypothesis totally unnecessary." I don't understand what Putnam could have in mind here. Perhaps he has confused the question of whether cognitive processes (for example, language learning) presuppose a medium of representation with the question of whether such processes presuppose (unlearned) information couched in that medium. As far as I can see, Chomsky's thesis (which, for purposes of discussion, let's take to be that General Linguistic Theory is innate) entails mine on the principle: no (innate) information without (innate) representation. On the other hand, there could be an innate medium of representation without there being innate information about (for example) natural languages (or, I suppose, anything else). I remarked earlier that a tendency not to distinguish between issues about innateness and issues about internal representation is pretty general in Putnam's paper: it may be that this is just a case of it. Alternatively, it may be that Putnam has confused issues about the innateness of concepts with issues about the innateness of beliefs, the thesis that GLT (General Linguistic Theory) is innate being primarily a claim about the latter while the thesis that LOT (Language of Thought) is innate is primarily a claim about the former. Nor should I wish to suggest that these alternatives are exclusive, since it may be that Putnam is confused about all of them at once.
Putnam on God and Man
Putnam says: "I don't doubt that God is ultimately responsible for what we are... [but] this is such a messy miracle to attribute to Him! Why should He pack our heads with a billion different 'mental organs,' rather than just making us smart?" This is, however, a bad argument even on what appears to be the operative assumption: that God's aesthetic principles are indistinguishable from Putnam's. To see how bad it is, try applying it to any other species. Why didn't God make the spider smart instead of merely teaching it to eat flies and spin webs? Why endow the robin and the stickleback with a parochial talent for building nests instead of "general intelligence" and a bent for architecture? And what a messy miracle the bee's dance comes to. Clever gods would make clever bees, which could then invent navigation and the telephone in the fullness of time. Sloppy old God! Better consult a philosopher the next time 'round!
The point is, of course, that in all other spedes cognitive capacities are molded by selection pressures as Darwin taught us to expect. A truly general intelligence (a cognitive capacity fit to discover just any truths there are) would be a biological anomaly and an evolutionary enigma. Perhaps that is not what Putnam thinks we've got. Since he tells us nothing about what general intelligence is, we have no way of knowing.
The reasonable assumption, in any event, is that human beings have an ethology, just as other spedes do; that the morphology of our cognitive capacities reflects our specific (in both senses) modes of adaptation. Of course, we are in some respects uniquely badly situated to elucidate its structure (to carry through [what I take to be] the Kantian program). From in here it looks as though we're fit to think whatever thoughts there are to think (compare "a billion different mental organs"). It would, of course, precisely because we are in here. But there is surely good reason to suppose that this is hubris bred of an epistemological illusion. No doubt spiders think that webs exhaust the options.
We know more than spiders do; we can (and should) bear the biological precedents in mind. These precedents suggest that we must seem to angels the way that other species seem to us: organisms whose intelligence is shaped by their history and is therefore fragmentary, task-oriented, and domain-specific. I'll bet that's what the angels say when they are doing anthropology. Assuming that the angels bother.
Notes
1. J. Fodor, The Language of Thought (Cambridge, Mass.: Harvard University Press, 1979).
2. Ibid., ρ. 61.