To the Editors:

John Searle’s response to Sylvain Bromberger’s letter [NYR, April 25] compounds the severe misinterpretations of work on language of the past fifty years presented in his earlier articles, corrected in the letter to which he responded.

Searle believes that this work (generative grammar within the biolinguistic framework, henceforth GG) was initially concerned with topics that interest him: following such rules as “drive on the right.” His claim that the “revolution” failed is based on the observation that no one interested in GG now pursues his project. That is true, for a simple reason. Searle’s project was never entertained, in fact never even mentioned in the past fifty years except to stress—explicitly, forcefully, and unambiguously—that GG adopted a conception of the nature, use, and acquisition of language in which his notions play no role at all.

True, the word “rules” was used in GG, adapted from traditional scientific grammar, but in a sense utterly unrelated to Searle’s usage. As Bromberger observed, rules are understood to be elements of the computational systems that determine the sound and meaning of the infinite array of expressions of a language; the information so derived is accessed by other systems in language use. That much is evident on the most cursory look at the earliest work, and everything since.

Consider Searle’s most “striking feature of the failure” of his “revolution”: that “even the apparently most well-substantiated rules …have been quietly given up,” specifically, the rule converting “John loves Mary” into “Mary is loved by John.” It is true, as he says, that “nobody thinks that anymore,” because no one ever did. The rule could not have been “given up” because nothing remotely like it was even formulable in the GG framework. An array of rules was indeed proposed to describe the properties of passive constructions in English, though with no resemblance to Searle’s proposal. As anticipated from the outset, there were serious flaws in the first attempts to deal with phenomena that had, in large measure, never even been noticed before the first efforts to construct explicit rule systems.

Since the 1950s, proposals have been revised and improved as more has been learned, though not “quietly”; rather as explicitly and loudly as possible. The reason is that these successive steps (still continuing) have been regarded as progress toward the original goal: to show that phenomena that appeared to require rules of great intricacy and diversity in fact followed from the interaction of far simpler rules that might be true invariants, holding for many constructions in typologically distinct languages. The long-term goal has been, and remains, to show that contrary to appearances, human languages are basically cast to the same mold, that they are instantiations of the same fixed biological endowment, and that they “grow in the mind” much like other biological systems, triggered and shaped by experience, but only in restricted ways.

For clarity, the invariant rules proposed in the past forty years have been called “principles.” But apart from invariance, simplicity, explanatory power, and empirical scope, the principles are “rules” in the earliest sense—though never in anything like Searle’s sense—and the general project remains effectively unchanged.

About twenty years ago, a great deal of work along these lines, with many participants here and abroad, crystallized in a conception of language that was sharply different from those that had guided linguistic research for thousands of years, including early GG. This “Principles and Parameters” (P&P) conception sought to eliminate language- and construction-particular rules entirely in favor of invariant principles associated with a finite range of options, yielding the particular human languages as these “parameters” are assigned values in the course of language acquisition. This was a welcome step forward, providing the first clear idea to how to resolve the tension that had driven research from the outset: between “descriptive adequacy” (which seemed to require elaborate and diverse rule systems) and “explanatory adequacy” (accounting for superficially varied phenomena on restricted and principled grounds). See Mark Baker’s recent Atoms of Language [Basic Books, 2001] for an expert and accessible account of the current state of these endeavors.

Searle bases his “abandonment” thesis on a statement of mine that he has cited several times but completely misunderstands: that the P&P approach “reject[s] the concept of rule and grammatical construction entirely: there are no rules for forming relative clauses in Hindi, verb phrases in Swahili, passives in Japanese, and so on.” This statement (which is twenty years old) is part of a summary of the goals of the P&P approach, pointing out that if it succeeds, language-particular constructions such as those mentioned will be shown to be taxonomic artifacts, rather like aquatic mammals. That would be an important step toward the original goals, which remain unchanged, along with the centrality of rules.


Searle persistently misconstrues a simple terminological point, and misses entirely the reasons or the change of terminology from “rules” to “principles,” discussed without ambiguity in the passage to which he refers. Fixated on his personal usage of the term “rule,” he fails to comprehend that the basic concepts and goals remain unchanged though there were far-reaching substantive changes in what the rules were taken to be (if invariant, called “principles”). If correct, these steps advance the earlier project considerably. Whatever their fate turns out to be, there is no serious doubt that they played a large part in the unprecedented flourishing of empirical and theoretical inquiry into language in the past twenty years, and in the important achievements in related areas that were reshaped and revitalized in these terms.

There are serious questions that should be raised: Was the project that took shape fifty years ago properly conceived? Have the various efforts to pursue it been on the right track? Have they led to substantive progress in the biolinguistic program, as many (including me) believe? Searle cannot address these questions, because, despite the authoritative tone, he has never understood what the project was in the first place, or how it has been pursued. Some of the other issues that he mentions may be worth discussing, but not unless some ground rules are observed.

Noam Chomsky
Department of Linguistics and Philosophy
Massachusetts Institute of Technology
Cambridge, Massachusetts

John Searle replies:

I believe that Noam Chomsky’s letter seriously distorts the claims I made in my review of his book and evades the issues I tried to raise. I will briefly try to set the record straight.

  1. Rules. Chomsky thinks that I suppose the rules of grammar are regulative rules like the rule “drive on the right.” That is not true. In my article, I introduced a reference to driving on the right as an example to show how rules can function causally in behavior; but in the article I also distinguish between “regulative” and “constitutive” rules. Regulative rules regulate antecedently existing activities like driving. Constitutive rules create the very possibility of the activities they regulate, for without the rules there is no activity to regulate. Human languages—like chess, money, private property, and government—are matters of constitutive rules, because to speak English—or play chess—for example, one must follow (at least a large subset of) the rules.

The rules of generative grammar were intended to be constitutive in this sense. The chief difference between me and Chomsky was not over whether there really were constitutive rules of generative grammar, but their relation to consciousness. I thought that unconsciously functioning rules had to be the kind of rules that could be conscious; Chomsky disagreed. Indeed, four of the features of rules in Chomsky’s early work were that the rules functioned unconsciously, were not even accessible to consciousness, were constitutive, and functioned causally. The rules have to be causally real in order to explain linguistic behavior. The speaker’s competence consists in a mastery of the rules and his competence gives rise, though often imperfectly, to his performance. Competence is the competence to perform. And in order to function causally the rules have to be constitutive in my sense, because prior to the rules there is nothing to operate causally on. There is no set of physical properties sufficient to determine all and only English sentences (or games of chess or married couples or governments), because such things exist only within systems of rules.

Does Chomsky seriously deny that he held this conception? I went to MIT in the 1960s to do research with Chomsky precisely because I was writing a book on these and related issues (later published as Speech Acts: An Essay in the Philosophy of Language, Cambridge University Press, 1969). In discussions, lectures, writings, etc. there was never any doubt that the rules of grammar were causally real, constitutive, and unconscious. I later wrote an article in this journal, “Chomsky’s Revolution in Linguistics” [NYR, June 29, 1972], explaining these developments. Is he now really claiming that there never was a “revolution” of the sort I described, and there never were supposed to be real rules that constituted linguistic competence? If so why did he wait thirty years to say so? He has written about my views and reactions to his work on several occasions; so far as I know, he has never before denied that he had put forward the explanatory apparatus attributed to him.

The comparison between generative grammar and traditional grammar is illustrative. The objection to traditional grammar was not that it postulated real rules that people really followed, but rather that it was incomplete and unformalized. The competence of the native speaker was not explained but presupposed. We had to assume a contribution by the native speaker in order to understand the statement of the rules.

  1. Active and passive. Chomsky claims there never was a rule for statements in the passive voice of the sort I described. I left the technical details out of my account. For the sake of simplicity I described the rule as converting active sentences to passive ones. But of course that is not technically correct. Transformational rules do not operate on sentences but on underlying phrase markers, themselves generated by the phrase structure rules that are a component of the grammar. I thought I might spare the readers of this journal these details, because they did not seem relevant to the point I was trying to make. But there definitely was a passivization rule. I heard Chomsky explain it in detail in lectures at MIT and read it in his books.
  2. The paradigm change. In my exchange with Sylvain Bromberger we agreed that Chomsky’s conception of the subject matter of linguistics had not changed, but I argued both there and in my original review that the original paradigm had failed and that the current research program used a different explanatory apparatus. In his letter Chomsky denies this. But I think a close reading of his letter shows that he tacitly accepts it. He writes of his current conception of language that it is “sharply different from those that had guided linguistic research for thousands of years, including early GG” (generative grammar). And about the principles and parameters (P&P) approach he writes that it “sought to eliminate language- and construction-particular rules entirely in favor of invariant principles….” One wonders how the research program could eliminate language- and construction-particular rules entirely if it had not accepted them in the first place.

Again, when he writes that the P&P approach “reject[s] the concept of rule and grammatical construction entirely: there are no rules for forming relative clauses in Hindi, verb phrases in Swahili, passives in Japanese, and so on,” the clear implication is that the previous explanatory model accepted rules for forming relative clauses, etc., and this acceptance is now found wanting, and has been abandoned. Chomsky cannot have it both ways. He cannot say there never was a paradigm shift, and besides the new paradigm is better.

  1. Normative criteria. Let us suppose for the sake of argument that Chomsky is right about the history of the subject and that he never thought there were any rules of grammar in the literal sense of “rule.” Then it seems to me the situation becomes worse and not better. Here is why. Actual sentences of actual human languages are human artifacts; they are not natural biological phenomena like photosynthesis or synaptic transmission. The fact that a string of words is a sentence of English is not a brute fact like the fact that I have a pain; it is a special sort of fact, an institutional fact, in that it presupposes the existence of a human institution, the English language with its constitutive rules. (It is not an objection to this point to claim that the institutions are realized in individual human brains as internal or “I-languages.”) As human artifacts created within human institutions, sentences of French, or English, or any other language have a special normative status: their utterances will fit or fail to fit certain normative criteria such as being ill-formed or well-formed. For example, in my dialect of English one criterion would be whether a singular subject will have a singular verb. To say “he are here” would not fit the criterion. What is the source of these normative criteria? There have to be constitutive rules (or some such apparatus) of natural languages in order to account for the existence of such normative criteria.

Chomsky would like to treat linguistics as a natural science, but this kind of normative standard has no analogue in natural science. Under certain stimuli my nervous system will produce a sensation of pain, under certain other stimuli it will also produce an utterance of an English sentence, but there is a huge difference. There is no analogue to grammaticality and ungrammaticality where pains are concerned. There has to be something like “rules for using singular or plural verbs,” etc. to account for the fact that speakers can do it rightly or wrongly.


I could not figure out from the book I reviewed what Chomsky thinks the relation of competence to performance is. How do the “invariant principles” relate to the actual performance of speech acts? Perhaps he discusses it in another work, but there has to be some account of this central question in linguistics: How does the speaker’s mastery (internalization, knowledge) of the rules of the language relate to his speech behavior? And any such account has to deal with the normativity of the phenomena.

  1. Computation and natural science. What about Chomsky’s appeal to “computational procedures”? Is that sufficient to account for the data we need to explain? As I pointed out in my article the notion of computation is not defined in Chomsky’s book, and is ambiguous in popular speech. There is a pre-Turing sense in which “compute” just means figure out. And there is a contemporary technical sense in which computation is defined in terms of the implementation of algorithms, using symbols, normally in a binary system.

Which is Chomsky’s use? It can hardly be the former because the computational rules would then be of the sort that he is denying. But if the latter, then computation exists only relative to an observer. Chomsky would like linguistics to become a natural science, but “computation” does not name a phenomenon of natural science like photosynthesis or neuron firing, rather it names an abstract mathematical process that we have found ways to implement in commercial hardware and which we can often interpret as occurring in nature. Indeed, just about any law-like process that can be described precisely can be described computationally. Thus some scientists describe the stomach in computational terms (“the gut brain”) but we all know there is no mental life in the stomach and no zeroes and ones. Computation is not discovered in nature, rather it is assigned to it.

Is it reasonable to suppose on the basis of what we know that the attempt to make linguistics a natural science is a plausible or even well-defined research program? The problem is not merely that Chomsky is involved in some very implausible assumptions—for example, that all concepts are innate, including concepts like “bureaucrat” and “carburetor”—but that several of the fundamental notions he discusses, such as sentence, grammaticality, and, above all, computation are observer-relative, or normative, or both, in ways that have no echo in natural science. It is often tempting in the human sciences to aspire to being a natural science; and there is indeed a natural science, about which we know very little, of the foundations of language in the neurobiology of the human brain. But the idea that linguistics itself might be a natural science rests on doubtful assumptions. I have tried to make clear some of those doubts, and nothing Chomsky says in either his book or his letter has answered my doubts.

This Issue

July 18, 2002