Having Linguistic Rules and Knowing Linguistic Facts

‘Knowledge’ doesn’t correctly describe our relation to linguistic rules. It is too thick a notion (for example, we don’t believe linguistic rules). On the other hand, ‘cognize’, without further elaboration, is too thin a notion, which is to say that it is too thin to play a role in a competence theory. One advantage of the term ‘knowledge’—and presumably Chomsky’s original motivation for using it—is that knowledge would play the right kind of role in a competence theory: Our competence would consist in a body of knowledge which we have and which we may or may not act upon—our performance need not conform to the linguistic rules that we know. Is there a way out of the dilemma? I’m going to make the case that the best way to talk about grammatical rules is simply to say that we have them. That doesn’t sound very deep, I know, but saying that we have individual rules leaves room for individual norm guidance in a way that ‘cognize’ does not. Saying we have a rule like subjacency is also thicker than merely saying we cognize it. Saying I have such a rule invites the interpretation that it is a rule for me—that I am normatively guided by it. The competence theory thus becomes a theory of the rules that we have. Whether we follow those rules is another matter entirely. Having Linguistic Rules and Knowing Linguistic Facts 2 The first time I saw Noam Chomsky give a talk, it was on the topic of knowledge of grammar. At the time (around 1980) I could not understand why he was using the term ‘knowledge’ in describing our relation to linguistic rules. Like most philosophers, I assumed that, to a first approximation, knowledge was justified true belief. The linguistic rules that Chomsky was talking about were not the sorts of things that people believed even upon reflection (e.g. the rule subjacency, which says that a move with an element like the word ‘what’ can’t jump over an S node and an NP node without an intervening landing site), nor was it clear that linguistic rules were the sorts of things that were true or correct, since there wasn’t a question of having the wrong linguistic rules. You just have the rules that you do. Finally, talk of justification didn’t make sense for grammatical rules; there is no issue of my getting the right rule in the wrong way (say from an unreliable teacher or peer). So linguistic rules weren’t believed, they weren’t true in any sense, and even if they were true and we did believe them there is no reason to think those beliefs would be justified. That is 0 for 3. Things are different for a prescriptive grammarian, for in that case there is a question of knowing the prescriptive rules of your language. Maybe some sort of official academy of language or someone in a power relation with respect to you determines what those rules are. In that case I might come to believe a rule like ‘never split an infinitive’, and (according to the prescriptivist) there could be a question of whether it is a “correct” rule for English, and there is even a question about whether I am justified in believing the rule (did I get it from a reliable source?). But for a generative linguist this picture is deeply confused. Generative linguists are engaged in an enterprise that is both descriptive and explanatory. It is descriptive in that they are interested in the linguistic rules that individual people actually have, and it is explanatory in that they are interested in why those people have the rules that they do. The explanation typically involves an innate language acquisition device that admits of parametric variation. For example, one common metaphor is that the language faculty is a largely prewired box with a finite number of discrete switches (parametric settings). When you are exposed to a language as a child the switches are set and you end up in a particular parametric state of the language faculty. Whatever we chose to call our relation to the resulting body of Vol. 5: Meaning, Understanding and Knowledge