… if we contemplate the classical problem of psychology, that of accounting for human knowledge, we cannot avoid being struck by the enormous disparity between knowledge and experience - in the case of language, between the generative grammar that expresses the linguistic competence of the native speaker and the meagre and degenerate data on the basis of which he has constructed this grammar for himself.
It is this problem that motivates Chomsky’s work in linguistics. When he began his linguistics work in the 50s, the main focus of psychological research was based on behaviorism: the notion that learning is based on stimulus/response patterns and on inductive processes that generalize from specific observations to more fully formed concepts. The idea that such simple processes could account for the rapid acquisition of language competence ought to have have been seen as absurd, but it is an idea that lingers to this day in somewhat revised form. Some of Chomsky’s early polemics against behaviorism, and B. F. Skinner in particular, can be found online, for example Review of B. F. Skinner’s Verbal Behavior and The Case Against B.F. Skinner.
Chomsky objected to behaviorism on two grounds. The first objection was, as just noted, the utter implausibility of extending stimulus/response models to the essentially unbounded and creative use of language. His second objection is that behaviorism makes the a priori choice to limit itself to descriptions of behavior, and explicitly rejects a program of explaining behavior. In particular, behaviorist attempts to explain language acquisition explicitly rejected the idea of explaining how language acquisition can be so rapid and accurate.
Though mid century stimulus/response theories of language acquisition have been discarded, there are new versions of tabula rasa (or nearly rasa anyway) theories based on the idea of probabilistic models. These suffer from the same problem: the limited and corrupt data available to the language learner cannot account, in a probabilistic learning model, for the rapid acquisition of linguistic competence. Anyone who has used Google Translate to view a web page in another language will see the problem at once. Translate uses probabilistic models, based on a huge and mostly sanitized corpus of translated work, to translate text. And it does a pretty good job, mostly, in making the translations intelligible. But the sorts of mistakes that it makes are ones that no native speaker would ever make, even in early childhood.
In Chomsky’s view, the only way to account for rapid and accurate language competence is by means of an innate human capability for language, one in which the range of possible grammars is extremely limited, and whose surface structure is obtained by selection from a small set of possible transformations from a deep structure. The rules governing deep structure, and the rules governing transformations between deep and surface structure, constitute Universal Grammar (UG).
Chomsky formulates language as a two-way mapping from semantics to phonetics, where semantics are, as it were, attached to the deep structure, and phonemic structure is attached to the surface structure of language. So there are four components of language: phonological, surface structure, deep structure, and semantics, and rules that govern transformations within and between each of the four components. And the immediate task of linguistics is to produce models of each of the four components and to discover transformation rules that account for the known properties of human language.
In this regard, Chomsky looks back approvingly on the attempts by Cartesian rationalists and their rationalist and romantic successors in the 17th and 18th centuries to explain language acquisition. He contrasts their explanatory goals with the limited descriptive goals of the modern “empiricist” school of linguistic and psychological work, and argues that the goal of science is explanation, not mere description (“meter reading”). His frequent reference to Cartesian rationalists has brought him unearned criticism of being “antiempirical” or even of being a Cartesian dualist. It is hard to take such criticisms seriously since even a superficial understanding of his writing on this subject shows that he sees linguistic work as depending crucially on emprirical evidence; he simply objects to the idea that the mere collection of data is a significant goal in itself. As for being a dualist: he is a methodological dualist, as described below, but is firmly grounded in the post-Newtonian world in which a mind-body dualism is a barely coherent concept.
When reading Chomsky’s work on UG it is important to keep in mind certain distinctions. His work, and his account of linguistics, deals almost entirely with the logical structure of language. So, for example, he describes phonological transformations as consisting of a short set of rules that are applied cyclically to an utterance; and application of a rule is constrained by, and modifies, a hierarchy of phrase markers in the (surface) grammar. But he explicitly says that it would be “absurd” to think that this is what actually happens when someone is uttering a sentence. The model is intended to provide insight into language competence, but must not be thought of as applying to language performance. So the first distinction is between competence and performance. Regarding language performance, Chomsky’s view is that at present we can do little more than spout platitudes.
A second distinction is between the logical model that represents UG and the strucures in the brain that give rise to UG. He points out that prior to investigating the intricate relationships between two entities it is necessary to have a clear understanding of those entities separately, to the extent possible. Chomsky’s work has been focused entirely on UG as a function of “mind”, and not as an output of “brain”.
So Language and Mind is a collection of talks and essays on these subjects, starting from a group of three talks given at Berkeley in 1967 (Linguistic Contributions to the Study of Mind: past, present, future). There are an additional three talks given in the 80s covering much the same ground, and a final more recent essay in which he talks about more recent advances in linguistics, including “Principles and Parameters” (P & P), which overcome certain technical difficulties with “classical” UG. The essays range from very informal introductory and overview material to fairly technical discussions of core issues in linguistics. There is a lot of overlap, as is common in Chomsky’s books. And there are lengthy refutations of a few of Chomsky’s critics; again, a customary feature.
This is as good a general introduction to Chomsky’s linguistics as any. For an early and more technical treatment, Aspects of the Theory of Syntax is not a bad place to start, but it deals solely with transformational / generative grammars.