[In the following review, Stabler provides favorable assessment and summary of Chomsky's Knowledge of Language and Language and Problems of Knowledge.]

Noam Chomsky has recently produced two more books about language for a general audience. (Earlier works of a similar character include Cartesian Linguistics (1966); Language and Mind (1968); Problems of Knowledge and Freedom (1971); Reflections on Language (1975); Language and Responsibility (1977); and Rules and Representations (1980).) They are both informal explorations of a wide range of issues relating to language and knowledge, refreshingly free of the academic parochialism that results from disciplinary inbreeding. Each covers new empirical ground, offering suggestions about how this material ought to be incorporated into the growing tradition of theoretical linguistics, and each offers some commentary on recent psychological and philosophical debates. Though the overarching perspectives of the two books are of course similar, they cover different material, and at different levels of sophistication—Knowledge of Language is the more demanding of the two. Either book could serve well as a language-oriented introduction to what is now called “cognitive science”. And although Chomsky's general orientation is familiar to those who know his earlier works, new empirical observations and stimulating new suggestions make each of these new works pleasant and valuable reading. Since this review is to be brief, I will just outline some of the main points discussed in each book, attempting to note especially the new and most significant material.

Knowledge of Language begins with an informal characterization of “Plato's problem”: “the problem of explaining how we can know so much given so little evidence”. The problem is illustrated with a variety of examples of what a competent speaker of English must know. The examples are motivated by current work in theoretical linguistics, but the discussion (with the exception of a technical section mentioned below) is readily intelligible to the general reader. The general conclusions about what a learner must bring to the learning situation, presumably on the basis of an innate endowment, are surprising enough to inspire in a novice an interest in getting a more sophisticated grasp of the theoretical background. This is exactly the effect that a good introduction should have!

In the second chapter of Knowledge of Language. Chomsky describes a shift in perspective on linguistic theory. Whereas it has sometimes been regarded as the abstract study of the sets of grammatical strings of a language, an “external” language or “E-language”, Chomsky urges persuasively that it is better regarded as the study of a speaker's “internal” knowledge of language, of “I-language”. It is interesting to note in this context that Chomsky's recent technical work, such as Barriers (Cambridge, Massachusetts: MIT Press, 1986), relies primarily on relative grammaticality judgments rather than on simple yes or no judgments about whether a sentence is well-formed. And he has explicitly defined principles, such as “n-subjacency”, which admit of more and less serious violations (in this case, depending on the value of n and certain other factors). So, for example, we have in the following three sentences one that is perfectly grammatical, one that is not good, and one that is very much worse:

What do you wonder about?

*What do you wonder how John fixed?

* Who do you wonder how fixed the car?

The severe ungrammaticality of the last example is explained by showing that it violates a certain principle of grammar, a principle that is not violated by the second example, which has only a weak I-subjacency violation. The first example violates no principles. So, from this perspective, the knowledge of language can be regarded as providing principles that indicate the level of grammaticality of a structure, and the idea that linguists study only a fixed and determinate set of perfectly grammatical strings is obviously inappropriate. In any case, on the view that linguists should be seen as studying I-language, linguistics is clearly a subdiscipline of psychology, and so it is appropriate for Chomsky to consider the psychological question of how a speaker's knowledge of language could be acquired, as well as such things as the relation of these matters to neurophysiology, as he does here.

A substantial shift in the approach of recent linguistic theory is described in some detail in the third chapter of Knowledge of Language: the shift from rule systems to principles. This important development has altered quite radically the character of linguistic theory. Chomsky here explains how the shift can be seen as a step towards a solution of the language acquisition problem, allowing us to identify specific parameters of variation in a much larger set of universal principles that apply to all possible human languages. Parts of this chapter will be difficult for the general reader, but these parts contain a number of points of interest for the linguist. Among other things, it is proposed that a constituent is “visible” to receive a theta-role only if it has case; that all theta-governed positions that receive theta-roles are filled at D-structure; that case is transferred to all elements of a chain and also by expletive-argument relations; and that structural case (nominative, objective) is assigned at S-structure, while inherent case (oblique, genitive) is realized at D-structure. Chomsky has not presented these views elsewhere, though some of them foreshadow the work in Barriers.

Chomsky devotes the fourth chapter of Knowledge of Language to critiques of alternative views. He considers at some length Saul Kripke's suggestion that the proposed conception of “knowledge of language” is undermined by Wittgensteinian considerations. Kripke argues that “If one person is considered in isolation, the notion of a rule as guiding a person can have no substantive content” (Saul Kripke, Wittgenstein on Rules and Private Language. Oxford: Basil Blackwell, 1983, p. 89). One of the ways Kripke defends this point is to claim that, apart from the intentions of the designer, there is no fact of the matter about what program a computing machine is following, because there could be no basis for distinguishing certain behaviour as a “malfunction”. Chomsky responds by arguing that this distinction can be drawn by “a more general account of the properties of the mind/brain, an account that defines ‘malfunction’ and ‘intrusion of extraneous factors,’ and is answerable to a wide range of empirical evidence” (p. 238). Unfortunately, he does not explain what sort of evidence could justify any particular definition of “malfunction”, and so some doubts may yet lurk in the philosophical reader.

Chomsky considers other critics of his approach more briefly. He agrees with Dennett that rule systems might be only “tacitly represented”, as the rules of addition are in a hand-calculator, but suggests that the best theories do not have this character: “Such possibilities cannot be ruled out a priori. The question is one of the best theory … objections of this nature … are beside the point” (p. 245). Chomsky rejects Quine's view that extensionally equivalent grammars cannot be empirically distinguished as representations of human linguistic competence unless they can be associated with distinct behavioural dispositions. He finds similar doubts about the legitimacy of attributing knowledge of language in Davidson, Dummett, and Searle. He says:

The argument at issue has two steps: The first step involves the tentative conclusion that the statements of the best theory of the language are true; the second, that the elements (rules, etc.) invoked to explain … behavior in the best theory we can construct in fact guide … behavior. (p. 249)

Chomsky argues that the first step by its very nature involves distinguishing “extensionally equivalent” rule systems. The second step should be a trivial matter: “I cannot see that anything is involved in attributing causal efficacy to rules beyond the claim that these rules are constituent elements of the states postulated in an explanatory theory of behavior and enter into our best account of this behavior” (p. 253). At this point, though, Chomsky does not consider the structure of the linguistic theory that explains our behavior in any detail. It is instructive to recall that earlier in the book he notes that the theory of how knowledge of language is put to use “breaks into two parts: a ‘perception problem’ and a ‘production problem’” (p. 25). He says, “The perception problem would be dealt with by construction of a parser that incorporates the rules of the I-language along with other elements: a certain organization of memory and access (perhaps a deterministic pushdown structure with a buffer of a certain size: see Marcus, 1980), certain heuristics, and so forth” (p. 25). If Marcus's work is to be taken as an example of the “incorporation” of the grammar that Chomsky envisions, then the relation between the grammar and the theories of linguistic performance will be quite remote—nearly as remote as the relation between rules for addition and the performance of the hand calculator. This is a familiar point, one that is surely familiar to Chomsky, so it is surprising that it is neglected here. The other aspect of language use, the production of linguistic behavior, is, Chomsky says, “considerably more obscure” (p. 25).

The fifth and final chapter of Knowledge of Language briefly notes that there seem to be domains unlike language in which knowledge acquisition seems to be a very difficult and precarious matter. Chomsky calls this “Orwell's problem,” and uses political and historical knowledge as his example.

The second book under review here, Language and Problems of Knowledge, is based on lectures given in Managua, Nicaragua in 1986. It provides a more basic and informal introduction to Chomsky's views on language than Knowledge of Language. Again, Plato's problem is introduced, but this time it is illustrated with examples from Spanish, and considerable attention is devoted to differences between Spanish and English and other languages. In this light, Chomsky is able to show rather clearly how the different aspects of the languages are sorted into those that must be learned and those that speakers do not learn but must bring to the learning environment. He uses examples from sound structure and semantics to illustrate that the presumption of an extensive unlearned competence is not peculiar to syntax, but seems to be essential in every case where we have very much evidence about uniformities in the acquisition of a body of knowledge. He uses the relation between declarative sentences and corresponding interrogatives to illustrate the surprising structure-dependence of linguistic principles, and uses this, in turn, to illustrate the importance of Plato's problem. He then considers “Descartes' problem”: “the problem of how language is used in the normal creative fashion” (p. 138). Here Chomsky takes the rather surprising line that the normal use of language is essentially non-deterministic. In other words, Descartes was right in thinking that we must be fundamentally different from, for example, computers that mimic human language performance, because a machine's performance is deterministic, compelled by its internal state and environment, whereas, in our language use, it is obvious that we are only “incited or inclined” by our state and environment. Chomsky says:

The human may often, or even always, do what it is inclined to do, but each of us knows from introspection that we have a choice in the matter over a large range. And we can determine by experiment that this is true of other humans as well. (p. 139)

Chomsky does not say what experiments would demonstrate the freedom of human action in a universe where the gross behavior of other large objects (like computers) is determined, and one might well wonder what he has in mind here. Chomsky goes on to consider the Cartesian proposal that the mind must be distinct from any physical object, but argues that we no longer have a coherent notion of physical object that allows us to formulate such a view, or any other interesting “mind/body” thesis. He suggests that Descartes' problem may simply be beyond the range of human intellectual capacities.

A good deal of Language and Problems of Knowledge is devoted to developing the basics of Spanish grammar in Chomsky's framework. Chomsky presents an argument for the view that simple Spanish sentences, like sentences of English and every other language, have (at some level of representation) an Aristotelian subject-predicate structure. The predicate contains the object in such a structure, so there is a subject-object asymmetry unlike, for example, the symmetry in binary atomic predications of standard first-order logics. A number of other universals are introduced informally and illustrated: binding principles, X-bar principles, the option of movement, the projection principle, and case theory. The treatment of the Spanish clitics in this framework is of particular interest. Examples of specific parameters of linguistic variation are also presented: the null subject parameter which distinguishes, for example, most Romance languages like Spanish and Italian from French; and the head-first paramater which distinguishes, for example, Spanish and Miskito.