Chapter 51 Discussion of Putnam's Comments Noam Chomsky

Putnam's discussion of what he calls "the innateness hypothesis" extends an earlier critical analysis of his to which he refers.¹ The earlier criticism, I believe, is based on a series of specific errors and a mistaken conception of the problem at hand.

According to Putnam, I advocate the "innateness hypothesis" as he formulates it, and I (and Fodor) attribute to "associationists"—the adversary—the mistake of "denying innate structure (laws of learning) altogether." The second of these claims is utterly false. I have repeatedly, consistently, and clearly insisted that all rational approaches to the problems of learning, including "assodationism" and many others that I discuss, attribute innate structure to the organism.³ I am sure that the same is true of Fodor.⁴ The question is not whether innate structure is a prerequisite for learning, but rather what it is. Furthermore, the literature is clear and explicit about this point.

For just this reason I have never used the phrase "the innateness hypothesis" in putting forth my views, nor am I committed to any particular version of whatever Putnam has in mind in using this phrase (which, to my knowledge, is his and his alone) as a point of doctrine. As a general principle, I am committed only to the "open- mindedness hypothesis" with regard to the genetically determined initial state for language learning (call it S⅛), and I am committed to particular explanatory hypotheses about So to the extent that they seem credible and empirically supported.

Putnam considers my two hypotheses H₁ and H₂, advanced to explain the formation of yes-or-no questions in English.

Putnam offers several arguments to the contrary, which I will consider in turn. The first has to do with the data available for language learning. I have argued that we can, under an appropriate idealization, think of the language learner as being supplied with a sample of well-formed sentences and (perhaps) a sample of ill-formed sentences— namely, corrections of the learner's mistakes. No doubt much more information is available, and may be necessary for language learning, although little is known about this matter. Nothing that Putnam says in this connection has the slightest bearing on my (rather innocuous) proposal, as it has actually been formulated. Thus his "false premise" that people object to all and only ungrammatical sentences is one that I have never proposed, and his discussion of deviance is compatible with my views on this subject, as expressed since the mid-1950s. Therefore, I will not comment further on these remarks, which have no relevance to the issue at hand or, as far as I can see, to my expressed views on language learning.

Putnam objects to my conclusion that "the whole burden of defining what a gram­mar is [falls] on the innate constraints," arguing rather that the grammar of a language is a property of the "language." I find it difficult to make much sense of this part of his discussion, which seems to me quite confused. Before considering his "different ap­proach," consider what he rejects. Is he proposing that only part of the burden of defining what a grammar is falls on the innate constraints? If so, which part? Which part of the burden falls elsewhere, why, and in what manner? No answer is suggested; therefore it is not clear that, and if so how, he is objecting to my conclusion. Note that he could hardly be claiming that none of the burden falls on the innate constraints, that is, that there are no innate constraints on what is a possible grammar, hence a possible human language. Thus even if language is constrained only by Putnam's ''general intelligence," ibfollows that the burden of defining what a language is falls on the innate constraints, and hence the burden of "defining what a grammar is" falls on the innate constraints, if grammar is, as he claims, a property of "language." Thus to begin with, it is quite unclear to what view Putnam believes he is objecting.

In fact, Putnam's counterproposal suggests that he has something different in mind, and that his objection is just misstated. His counterproposal is that "the grammar of a language is a property of the language, not a property of the brain of Homo sapiens" But this formulation refers to the grammar of a particular language, say English, not to the innate constraints on possible languages and grammars. Apparently, Putnam is con­fusing the grammars of particular languages (the topic of his counterproposal) with "universal grammar," his notion of "what a grammar is" (the topic of his objection). Let us turn now to his counterproposal, as he formulates it.

The two counterposed views, then, are these: (1) my view, that grammars are repre­sented in the brains of mature speakers, that languages are determined by these gram­mars, and that speakers of language can communicate to the extent that the languages characterized by the grammars in their brains are alike; (2) Putnam's view, that gram­mars are not represented in the brains of speakers but are properties of "languages."

It is difficult to compare these views, because Putnam's seems to me barely intelligi­ble and, insofar as it is clear, inconsistent with other positions that he maintains. Let us put aside the fact that such notions as "the English language" are not linguistically definable, but are rather sociopolitical in nature. Consider now Putnam's "different approach." Note first that Putnam agrees, of course, that language is neurally repre­sented (namely, in "the speech center in the left lobe," or the right lobe under early injury; see his "better argument"). It follows, then, that my language is a property of my brain. But Putnam claims that the grammar is a property of this language. Therefore, it is also a property of my brain, contrary to what Putnam asserts. If, as Putnam claims, grammars are not properties of brains but are properties of languages, then it follows that neither languages nor grammars are "properties of the brain of Homo sapiens," which is to say that my knowledge of English (and ability to use English) is not a property of my brain and is not represented in my brain, in the "speech center" or anywhere else.

One can, perhaps, choose to think of propositional calculus (Putnam's example) as a "mathematical object" with whatever kind of existence we attribute to such "objects," but that has nothing to do with the empirical problem of determining the properties of natural systems such as some human language, as represented (I assume) in the brains of individuals in their mature state, or the problem of determining the properties of Sθ, whatever these may be. Putnam gives no explanation of his alternative and allegedly "traditional" approach. I doubt that a coherent account is possible as a real alternative to the approach he wants to reject, which takes grammar to be a property of a brain and the "definition" of grammar to belong, in effect, to the theory of S}j. I see no need to comment further on Putnam's remarks about propositional calculus, except to note that even these are not free from error.⁶

Putnam proposes that the "declarative grammar of a language is the inductive defini­tion of a set of strings which is the set over which semantic, deductive-logical, induc­tive-logical (and so on) predicates are defined" and that it must facilitate these defini­tions, be computationally feasible, etc. Let us grant all of this, for the sake of discussion, putting aside an ample literature that is concerned with the alleged "parallel" between semantic and syntactic properties of natural language.⁷ From Putnam's suggestion, nothing FqUows about grammars being a property of "language" rather than "the speakers' brains," contrary to what Putnam asserts, without argument. The suggestion is entirely compatible with the view that grammars are represented in the brain, and represented in such a way that semantic (etc.) predicates have definitions whose clauses "paraUel... syntactic analysis" (though I think there are adequate grounds to suspect that the latter conclusion is incorrect—an empirical question, which I cannot consider here).

Putnam next turns to Washoe, arguing that she has developed structure-dependent rules. His discussion, however, is vitiated by an equivocation with respect to the notion "structure-dependent." Note that both of my hypotheses, H₁ and H₂, present rules that apply to a sentence, deforming its internal structure in some way (to be precise, the rules apply to the abstract structures underlying sentences, but we may put this refine­ment aside). Both the structure-independent rule H₁ and the structure-dependent rule H₂ make use of the concepts "sentence," "word," "first," and others; they differ in that H₂ requires in addition an analysis of the sentence into abstract phrases. A rule that does not modify the internal structure of a sentence is neither structure-dependent nor structure-independent. For example, a phrase structure rule, part of a phrase structure grammar in the technical sense of the term, is neither structure-dependent nor structure­independent.

The rule for conjunction that Putnam discusses in his Washoe comments takes two sentences p and q and combines them to form p&q ; in the framework of my discussion, it is a phrase structure rule rather than a transformational rule. It is neither structure­dependent nor structure-independent in my sense of these terms, since it does not require an internal analysis of the sentences to which it applies as a sequence of words or as a system of phrases. The rule does nothing to the internal structure of the sentences, and thus lies outside the bounds of the present discussion altogether.

Notice that in discussing question formation, I counterposed a structure-dependent and a structure-independent hypothesis, H₂ and H₁, respectively, and raised the ques­tion of why one is selected over the other on evidence compatible with both. In discussing conjunction, Putnam does not put forth competing hypotheses. The reason is that neither the notion "structure-dependent" nor the notion "structure-independent" applies in this case. There is no "structure-independent" counterpart to his rule, because it is neither structure-dependent nor structure-independent. Thus even if we were to grant that Washoe has learned her rule, and can form p & q (in principle) for arbitrary sentences p, q_l nothing at all follows with regard to structure dependence or the choice between H₁ and H₂. The other Washoe examples also fall outside the domain of our discussion. They have nothing to do with structure dependence or structure indepen­dence; they illustrate substitution of items in a fixed frame. There is, to my knowledge, no evidence that chimpanzees use structure-dependent (or structure-independent) rules, in the sense of my discussion. Clearly, Putnam's account involves no rules of either sort. Therefore, we can put aside the discussion of Washoe, which has no more relevance to the problem under consideration than the discussion of propositional calculus. Both concern a kind of syntax to which the concepts under discussion do not even apply (in the case of propositional calculus, context-free phrase structure grammar; in the case of Washoe, an extremely limited finite-state grammar, perhaps even without any cycles). The same is true of Putnam's Hebrew example, which involves a Rontransformational phrase structure rule like the rule introducing an abstract question marker in many treatments of English grammar.

Putnam later argues that my H₁ is itself structure-dependent, again equivocating on the term. I did χιot patent the terms and Putnam is free to use them as he likes, but in my usage, the rule is plainly not structure-dependent.

It is not clear why Putnam introduced propositional calculus and Washoe into the discussion of structure dependence. Perhaps his argument is that since the child (like Washoe, allegedly) can Ieam the rule for conjunction, and since this rule is “structure­dependent" (in Putnam's sense, though not mine), then the child will, by some kind of induction, choose the structure-dependent H₂ over the structure-independent H₁. I hesitate to suggest that this is Putnam's implicit argument (there is no explicit argu­ment), since it would be inconsistent with his assertion that both H₁ and H₂ are structure-dependent (in his sense); if this is so, then either could have been posited by "induction," so the original problem remains. Or perhaps Putnam means to suggest that the concept of "structure dependence" in his sense is a notion of "general intelligence" (since Washoe allegedly has it). But that is of no help to his argument, since Washoe also undoubtedly has the notion "before" in time and probably "first," so that these too, by the argument, form part of general intelligence. We are still faced with the problem of why the child selects H₂ over H₁, which "general intelligence" makes available (since it involves only the notions "before" or "first," applying to word sequences). Similarly, if both hypotheses are (as Putnam alleges) "structure-dependent" (in his sense), then we are still left with the original problem: Why is H₂ selected?

Whether or not Putnam has something like this in mind, in case anyone else might be misled into supposing that there is an argument here based on some kind of "induc­tion," let me add a few remarks. Imagine some new concept of "structure dependence" (call it SD*) under which the rule of conjunction and H₂ are structure-dependent (have the property SD*) but not H₁. Suppose further that the child Ieams the rule of conjunc­tion and others like it which have the property SD*. Can we then account for his choice of H₂, which has the property SD*, over H₁, which does not? Only if we suppose that the predicate SD* is "available" as a projectable predicate for induction. But that is to beg the very question at issue. That is, we can now ask why the child carries out an induction with the predicate SD* instead of another, equally good predicate SI*, which holds of the rule of conjunction and H₁, but not H₂ (for example, consider the property of being a rule that does not deform a sentence in accordance with its internal phrase structure). In short, this pseudo-argument requires that the predicate SD* but not SI* be available for "induction" (learning). The question then arises: why SD* but not SI*? But that is just a variant of our original problem—we have just another variant of the familiar Goodman paradox, except that in this case we cannot even tell which is "grue," SD* or SΓ, since neither seems a reasonable choice as a "projectable" predicate.

Putnam next turns to H₁ and H₂ directly, presenting his first real argument that the child "of course" uses structure-dependent rules. He argues that this follows from the fact that the child wants to Ieam "semantic rules" which cannot be stated without structure-dependent notions. Let us assume, for the sake of argument, that the semantic rules are structure-dependent. Does this explain why the child selects H₂ over H₁? Obviously not. Suppose that in fact English used the structure-independent rule H₁ to form yes-or-no questions. This would pose no problem at all for the formulation of the appropriate semantic rule. The rule for yes-or-no questions merely requires that these be distinguished from declaratives; they can be distinguished by H₁, by H₂, by painting them green, by standing on one's head while saying them, or in any other way, as far as the semantic rule is concerned. The rule asks: Is the corresponding declarative true or false? (Actually, the matter is more complex, but in no way that bears on this discussion.) We will turn in a moment to the matter of finding the "corresponding

declarative." But Putnam offers no argument at all to support his claim that H₂ facili­tates statement of the relevant semantic rule in a way that H₁ does not. Furthermore, there is no such argument, as the semantics makes clear. I should add that it is very common in discussions of language learning to appeal to "semantics" or "pragmatics" when problems arise. It is often not appreciated just what is at stake. Putnam's ar­gument, which is completely without force, is a clear example of this unfortunate tendency.

Putnam argues that the child must use abstract phrase structure to understand the language, and that therefore H₂ is natural. He fails to add that the child also uses the notions "word" and "first" (presupposed by both H₁ and H₂) to understand the lan­guage; thus H₁ is no less "natural," in this regard. We then face again our original question: Why does the child use H₂, which employs analysis into phrases in addition to the notions presupposed in H₁? Putnam's argument is neutral with respect to this question, and therefore goes the way of the preceding ones.

Putnam next claims that (A) "the learning of grammar is dependent on the learning of semantics" He offers (A) as an apparent paraphrase of his earlier assertion that the grammar must provide for the definition of semantic predicates, but it is certainly no paraphrase of this assertion. Elsewhere, Putnam has been quite clear about the distinc­tion and has indeed advanced a very different and more plausible thesis.⁸ Indeed, it is not easy to reconcile (A) with Putnam's earlier observation that the inductive defini­tions of semantic notions "parallel or at least presuppose a syntactic analysis of the language." If the definitions of the semantic notions presuppose a syntactic analysis (that is, a formal grammar that assigns phrase structure, determines well-formedness, and so on), then how can the learning of this grammar be "dependent on" a (prior?) learning of semantics?⁹ But putting this question aside, suppose that (A) is true, in some sense that remains to be explained. Does anything follow concerning H₁ and H₂? Not as far as Putnam has argued or shown. The semantics of yes-or-no questions prefers neither H₁ nor H₂.

Putnam next argues that H₂ is preferable to H₁ because its "inverse" is simple, whereas the inverse of H₁ is "horribly complicated." He does not explain why he believes that this is so. As far as I can see, it is not; the inverses are very similar. In each case, the inverse operation requires that we find the position from which is (etc.) has been moved—a position immediately before the predicate. Given H₁, we seek the first such position (and if someone wanted to argue that the inverse of H₁ is in fact simpler, he might note that our search is facilitated by the presence of the word who [etc.] in this case). Given H₂, we will seek the "main" position, using the full phrase structure analysis. One can think of various algorithms, none of which, as far as I can see, differentiates between H₁ and H₂. Since Putnam offers no argument, I have to leave it at that.

Note, incidentally, that even if the inverse algorithm must be "structure-dependent," that has no bearing on the choice between H₁ and H₂, that is, on the question of whether it is the first occurrence of is (etc.) or the "main" occurrence that is proposed. We cannot argue that because (by assumption) the inverse is structure-dependent, then so is the rule. In fact, even if one were to put forth this illegitimate argument, it would not bear on the essential point. We could then rephrase our original query, asking why it is that the occurrence of is after the main noun phrase is moved, rather than the first occurrence after a noun phrase (that is, the leftmost occurrence in "The man who is here ista∏").

To allay any lingering confusion about this matter, consider the three relevant ques­tion forms:

(I) Is—the man here?

(II) Is—the man who is here tall?

(III) Is—the man who here is tall?

Both Putnam and I are assuming that the language learner is presented with many examples such as (I), and formulates either H₁ or H₂ to account for them. The facts of (II) and (HI) show that H₂ was correct. To apply the inverse algorithm in (I), (II), and (III), the child must be able to detect where is is missing in the form to the right of—in these expressions. The question has never been studied, but it seems likely that at the stage of language acquisition when children can freely form sentences such as (II) (using H₂), they would have no difficulty in determining where is is missing in any of the forms to the right of—in (I), (II), and (III). Indeed, I would not be surprised to Ieam that they can solve the problem more easily for (III) than for (II). But ability to solve this problem is all that is required for the inverse algorithm to operate. Therefore, Putnam's unargued assertion that the inverse operation for H₁ is "horribly complicated" as compared with the inverse for H₂ seems far from the mark. If in fact it is easier to solve the problem for (III) than for (II), we would have an additional puzzle for the Martian observer, who might have taken this as further support for the obvious hypothesis that H₁ is to be preferred.

These comments exhaust Putnam's arguments concerning structure dependence. As far as I can see, none of them have any force. My conclusions, therefore, remain as stated.

Next, Putnam turns to the question of "general intelligence," beginning with the following assertion:

(IV) "So far I have assumed that there is such a thing as general intelligence," including "multipurpose learning strategies, heuristics, and so forth."

Actually, (TV) is a rather misleading assertion. All that Putnam has so far assumed is that So, whatever it may be, contains only the general mechanisms for learning. Recall that he gives no hint as to what these are. To invoke an unspecified "general intelligence" or unspecified "multipurpose learning strategies" is no more illuminating than his refer­ence, at one point, to divine intervention. We have no way of knowing what, if anything, Putnam has assumed. The point is worth stressing, since it illustrates a com­mon fallacy in discussions of this sort. The use of words such as "general intelligence" does not constitute an empirical assumption unless these notions are somehow clarified. As matters now stand, very little is asserted by (IV).

Putnam claims that his "multipurpose learning strategies" enable us to Ieam and create physics. He seems to feel that I should also grant something of the sort, since I insist, naturally, that these achievements are possible. But I am not committed to an empty claim. If Putnam tells us what these "multipurpose learning strategies" are, even in the most vague and informal way, I will be glad to join him in inquiring as to their efficacy in accounting for our learning of physics, etc. In the absence of any proposal, I have nothing to say about the problem. Nor does Putnam, it is crucial to emphasize.

There are, in fact, striking and obvious differences between language learning and the learning (or discovery) of physics. In the first case, a rich and complex system of rules and principles is attained in a uniform way, rapidly, effortlessly, on the basis of limited and rather degenerate evidence. In the second case, we are forced to proceed on the basis of consciously articulated principles subjected to careful verification with the intervention of individual insight and often genius. It is clear enough that the cognitive domains in question are quite different. Humans are designed to Ieam language, which is nothing other than what their minds construct when placed in appropriate conditions; they are not designed in anything like the same way to Ieam physics. Gross observa­tions suffice, to suggest that very different principles of "learning" are involved.

As for the proper delimitation of cognitive domains and their nature, I have nothing to add here to earlier discussion, at the Royaumont conference and elsewhere.¹⁰ Where a rich and intricate system of belief and knowledge is rapidly attained in a uniform way on the basis of limited and degenerate evidence, it makes sense to suppose that some "mental organ" with special design is involved, and to try to determine the nature and properties of this "organ" and the cognitive domain related to it, as well as its relations to other systems that form part of the general structure of mind. Progress in delimiting these domains and determining their nature may come through studies analogous to those I have discussed in the case of language, or perhaps in other ways. Putnam asserts that the number of domains is "virtually unlimited" and that the strategies we use "must be extraordinarily multipurpose," although he adds that "we have presently no idea what they are." I know no more about these strategies than Putnam does, or about the delimitation of domains, or about their number or specific character. As far as I can see, we differ here only in that I am disinclined to put forth what appear superficially to be empirical hypotheses where, as we both admit, we have "no idea" as to what the facts may be. I would urge that Putnam too should adopt the "open-mindedness hypothesis" and refrain from putting forth assertions such as (IV) and others that appear in that section of his "Comments."

Putnam argues that if there are such cognitive domains as "learning a grammar," "recognizing faces," and others that are "so small" and have such "highly specific­purpose" learning strategies, then "it becomes really a miracle that evolution endowed us with all these skills," since most of them (for example, mathematics) weren't used until after the evolution of the race was complete. I see no miracle here. Consider the human ability to handle fairly deep properties of the number system. I suppose that this ability is genetically determined for humans, though it is hard to imagine that it contrib­uted to differential reproduction. But we need not suppose that this is a miracle, if true. These skills may well have arisen as a concomitant of structural properties of the brain that developed for other reasons. Suppose that there was selection for bigger brains, more cortical surface, hemispheric specialization for analytic processing, or many other structural properties that can be imagined. The brain that evolved might well have all sorts of special properties that are not individually selected; there would be no miracle in this, but only the normal workings of evolution. We have no idea, at present, how physical laws apply when IO¹⁰ neurons are placed in an object the size of a basketball, under the special conditions that arose during human evolution. It might be that they apply in such a way to afford the brains that evolved (under selection for size, particular kinds of complexity, etc.) the ability to deal with properties of the number system, continuity, abstract geometrical space, certain parts of natural science, and so on. There are innumerable problems here, but I see no need to appeal to miracles. Nor do the problems that arise seem qualitatively different from familiar problems in accounting for the evolution of physical structures in organisms.

Putnam's further remarks about evolution seem to me mystifying. He feels that I have "dismissed" Piaget's concerns about evolution, but that is quite false. Rather, I remarked that the structures I have been led to postulate for S⅛, though 'biologically unexplained," are not, as Piaget asserts, 'biologically inexplicable." Furthermore, I see no specific problem that arises in this connection beyond those that are familiar (if often mysterious) in the case of physical organs. Putnam's further discussion seems to indicate that he agrees. Therefore, I assume that he has somehow misunderstood what I said about this matter.

In my earlier discussion of Putnam's criticisms of the "innateness hypothesis" (see note 2), I noted that his views about evolution seemed to me very curious. Thus in the paper to which he refers,¹¹ Putnam asserts that "invoking 'innateness’ only postpones the problem of learning; it does not solve it" This is a very odd principle, one that would never be put forth in connection with the development of physical organs. If, in fact, the general properties of binocular vision or the fact that we grow arms instead of wings is genetically determined, then it would be senseless to say that "invoking 'innateness' only postpones the problem of the learning of binocular vision or the learning of arms rather than wings." There is no such problem to be "solved." True, a problem remains, but it is not the problem of learning; it is the problem of explaining the origin and development of structures that are innate. I see no reason to take a different approach when we study higher mental faculties. If, indeed, certain properties of language are genetically determined, then "invoking 'innateness'" does not "postpone the problem of learning" with regard to these properties, but rather is the proper move, since there is no "problem of learning" in these respects. Putnam seems to believe otherwise, but I have no idea why.

I will not comment on Putnam's "better argument," except to observe that it does not bear even in a remote way on the questions that I discussed and that seem to me to be the interesting ones, namely, what is the nature of S⅛, how does it relate to other faculties of mind or to "general intelligence" (whatever it may be), and so on.

Putnam summarizes the view that he has been putting forward as follows: "Every­thing Chomsky ascribes to an innate template of language, a 'mental organ' specifically designed to enable us to talk, can, for all we know, be explained by general intelli­gence." And he suggests that this conclusion agrees "in broad outline" with Piaget's views. At the level of vagueness at which he discusses the problem, I would not dis­agree, once his specific arguments are dismissed as fallacious. Thus I agree that "for all we know" some notion of "general intelligence" about which we have "no idea" might explain everything I have ascribed to S⅛. Similarly, there would be little point in debating the claim that "for all we know" some mysterious force, as to the character of which we have "no idea," might explain everything that physicists try to explain in terms of their complex constructions. Thus, contrary to what Putnam believes, I would not deny his contention. We differ only in that I dismiss it, whereas in contrast he seems to think the contention is important—why, I do not know.

There is much to say about Putnam's discussion of Piaget, but I will not go into the matter here. One point deserves mention, however. Putnam feels that Piaget's approach converges with mine in that the notion of "reflective abstraction" relates to the use of language in inference. At the risk of seeming ungracious, I must demur. My uneasiness with "reflective abstraction" is not that it is placed "apart from language" (as Putnam asserts), but rather that I do not know what the phrase means, to what processes it refers, or what are its principles, any more than I know what Putnam has in mind when he speaks of "general intelligence," "multipurpose learning strategies," and the like. Hence it is impossible for me to take a position on the potential convergence that Putnam perceives.

Putnam argues that Fodor and I misused the term "tautology." He fails to note that the term did not appear in any presented paper, but was introduced in the informal discussion (by whom, I do not recall) and was then used by all participants not in the technical sense of "logical truth" but in the informal sense of "obvious truth." Since one cannot speak with warning-quotes, this may not be explicit in the transcript, but it surely is obvious enough from the context. Since Putnam agrees that the contention at issue is an obvious truth, there is no disagreement here.

Putnam concludes his paper with the claim that Fodofs hypothesis of a "language of thought" ∏φkes my hypothesis concerning a '“mental organ' for speaking... totally unnecessary." He offers no hint of an argument in support of this contention. He would be right if the "language of thought" had, in general, the properties of So- But it is exactly this question that Putnam has failed to address, once errors in argument and incorrect statements of fact are eliminated. I do not state categorically that the thesis is false; only that no argument to support it has been offered by Putnam or anyone else, to my knowledge, whereas there are empirical (though, obviously, nondemonstrative) arguments to the effect that S⅛ has certain properties for which there are no known significant analogues elsewhere. Furthermore, there are real and generally unappre­ciated difficulties in the thesis that intelligence is "undifferentiated."¹² Perhaps the reason why we can offer no specific analogues elsewhere to the properties postulated for So is that we just do not know enough about other aspects of cognition, or perhaps postulation of these properties is incorrect. Or perhaps the reason is, as I suspect, that the "mental organ" of language really has special properties, hardly a surprising conclu­sion, though of course far from a necessary truth.

Perhaps I may conclude with a personal remark My old friend Hilary Putnam and I have been debating these issues for quite a few years. He begins his discussion here with some kind remarks, for which I am grateful. He even goes so far as to say that if I am unable to provide arguments for what he calls "the innateness hypothesis," then probably no case can be made for it. As noted, I do not feel that he has established any of his points; rather, it seems to me that my arguments stand, as given, with just the qualifications and strictures given. But let me return the compliment. Putnam has re­markable intellectual gifts and an awesome command of many fields of knowledge. Furthermore, more than any other philosopher to my knowledge, he has concerned himself with the problem to which his present comments are addressed, seeking to establish that "general intelligence" or "multipurpose learning strategies" suffice to account for the specific workings of language. I feel that to date, he has not made a case for his contentions, and indeed, has not even succeeded in making clear what these contentions are. Perhaps, then, we may conclude...

Notes

1. See H. Putnam, Mind, Language, and Reality (Cambridge: Cambridge University Press, 1975), chapter 5.

2. See chapters 3 and 6 of my Language and Mind, extended edition New York: Harcourt Brace Jovanovich, 1972).

3. For a quite typical example, see chapter 1, section 8 of my Aspects of the Theory of Syntax (Cambridge, Mass.: MIT Press, 1965).

4. See J. A. Fodor, T. G. Bever, and M. F. Garrett, The Psychology of Language (New Yoric McGraw-HilL 1974), pp. 436 ff.

5. We might even argue that his proposal though nearly vacuous, can indeed be assessed, and rejected. The sole content of his proposal as it stands, is that the properties of Sθ are simply "general learning mechanisms," which apply freely in all cognitive domains. But there is evidence that S⅛ contains mechanisms and structures for which it is difficult to find even a vague analogue outside of language, for example, the specific principles of transformational grammar postulated to explain the structure dependence of rules (and much else). Consequently, the belief that the properties of Sθ are in fact "general learning mechanisms" is quite implausible with regard to its minimal empirical content.

6. For example, when he asserts that "p & ~ p" is "deviant for semantic reasons," and "would not be 'uttered' for obvious semantic reasons." It is not deviant at all, and might well be "uttered" as a line in any proof by reductio ad absurdum.

7. See, for example, Otto Jespersea The Philosophy of Grammar (London: Allen & Unwia 1924); and for some recent discussioα my "Questions of Form and Interpretation," Unguistic Analysis, vol. 1, no. 1,1975.

8. See Putnam's Mind, Languages, and Reality, volume 2, chapter 4.

9. Perhaps Putnam means that the learning of grammar is "dependent on" the learning of semantics in the sense that semantics provides the goal or motive for the learning of syntax. But if this is what he has in mind, the argument again fails, since as already noted, there is no problem in stating the semantics with the structure-independent rule.

10. See chapter 1 of my Reflections on Language (New York Pantheon, 1975).

11. Putnam, Mind, Language, and Reality_lAapter 5.

12. See chapters 1 and 4 of my Reflections on Language for some discussion.