Students and colleagues sometimes seem puzzled about mentalistic assumptions in Chomskean linguistics. The following remarks are intended as a brief explanation. (Comments and other suggestions are welcome.)
Starting in the 1950s, a number of linguists following Noam Chomsky attempted to write explicit or generative grammars for parts of English (and other languages), finding certain principles that had never previously been mentioned in grammars, principles that had in some sense been taken for granted in the older grammars. The grammars of this time had phrase structure rules as well as certain movement rules, called transformations, which might for example transform phrase structures for active sentences into phrase structures for corresponding passive sentences. In the 1960s grammars were devised that assigned deep structures resulting from phrase structure rules and surface structures resulting from the operation of the transformation rules to the deep structures. More importantly, it was discovered that the transformation rules in such grammars were subject to certain general movement constraints and there began an intensive study of these constrains. But questions immediately arose as to how these constraints could be acquired by children learning language, given that linguists had had no inkling of these constraints before 1960. This was the key moment in giving rise to the new linguistics.
Chomsky and a number of other linguists argued as follows. Suppose that the constraints needed for English grammar are not learned. Suppose that they are somehow innate in a child’s language faculty. Then, since the child is capable of acquiring any language depending on its environment, the same constraints should be found in the grammar of any language and should be part of universal grammar. This hypothesis predicts that the study of English grammar can tell us things about French, or Japanese, or Turkish grammar, an empirical prediction.
Study of the grammars of these other languages found that the prediction was basically correct, with some modification in the original hypothesis.
Somewhat similar results were found with respect to phrase structure rules.
The upshot starting in the early 1970s was the so-called principles and parameters model. In this model, the principles of universal grammar consist in a certain model of phrase-structure (X-bar theory), an allowance for possible movements of items, and certain constraints. Parameters in the model allow for differences in whether the head of a phrase comes at the beginning or the end of the phrase, whether sentences must have subjects, etc. In this approach grammars of particular languages no longer have phrase structure rules or transformation rules. (So, there is no longer a passive transformation, for example.) The approach has been tremendously fruitful, leading to a great deal of information about languages that no one ever knew before.
This approach arose out of a mentalistic speculation about language learning. Without such speculation, it is unclear what could justify the prediction that certain principles discovered in the study of English should apply to other languages as well. (Those generative transformational linguists from the 1960s who failed to adopt that mentalistic approach have been completely unable to participate in the later development of the subject and have ended up quite marginalized. This includes several linguists who had made important contributions in the 1960s.)
By the way, it is easy to construct formal languages not subject to the principles of universal grammar (e.g., any language expressed in ordinary quantification theory). Theoretical language in science is not subject to such principles. Esperanto and other pidgin languages are not subject to such principles. On the other hand, the creoles that arise in children exposed to pidgin languages are subject to principles of universal grammar, as are the sign languages spontaneously developed by deaf children among themselves who have not had access to a standard sign language.
There has been a tremendous amount of work in the principles and parameters framework concerning the exact statement of relevant principles and the ways in which different languages differ.
Since about 1990 there has also been work attempting to see to what extent the basic principles of universal grammar are in some sense the optimal solution to a certain engineering problem of relating the semantic interface level and the phonetic interface level--the so-called “minimalist program.”
In the last section of his 1959 review of Skinner’s Verbal Behavior Chomsky says, “The fact that all normal children acquire essentially compable grammars of great complexity with remarkably rapidity suggests that human beings are somehow specially designed to do this. The study of linguistic structure may ultimately lead to some significant insights into the matter.” Chomsky’s 1962 lecture to the International Linguistics Conference begins with a relatively short summary of a mentalistic approach saying, among other things, that a speaker’s “competence can be represented, to an as yet undetermined extent, as a system of rules that we can call the grammar of his language”, etc. Furthermore, much of what was occurring in linguistics in the early 1960s was occurring at MIT where Chomsky regularly lectured on the material published as Cartesian Linguistics. But, although Chomsky in this way emphasized “mentalism” and considerations of language learning from the beginning, the actual practices of most early generative transformational linguists until, say 1970, seemed to be for the most part unaffected by such considerations.
Linguistic mentalism became more important to actual practice as more and more attention was paid to constraints on transformations (and other aspects of grammar) and how a child might acquire such constraints. Research in Chomskean linguistics from that point on became importantly dependent on such mentalistic considerations and as a matter of fact it was and is impossible to do interesting research in that framework e.g. about how French and Italian differ in certain respects without considering how children might acquire certain parameters, which parameter would have to be the default, for example. So, it is a simple fact (sociological and epistemological) that research in Chomskean linguistics did become tied to the mentalism, to such an extent that linguists in that tradition who avoided the mentalism were unable to do interesting research.)
This is even more true in the sort of thinking about universal grammar as it is affected by the minimalist program, which requires the specific assumption that the linguistic faculty is the interface between two other mental systems.1