One of the professional hazards of being a linguist is being asked what you think of Chomsky.* I never have a good answer because we didn’t actually study Chomsky in school. The linguistics department at Rice does what’s called ‘functional linguistics’, and Chomsky is ‘formal linguistics’. The names have to do with having different opinions about how language evolved and is produced, and why languages are similar to each other, but the explanation is a little esoteric.** At any rate, we didn’t learn much about him, and I didn’t get any clear idea of what he thought until later, when I read an interesting book called The Linguistics Wars.

Today, though, I realized on re-reading The Most Human Human that the author captures a distinction in computer science that I wasn’t aware of, which has a lot of similarity to the way the formal/functional distinction plays out in practice (as opposd to the esoteric theory): the distinction between computability theory and complexity theory. In computability theory, your concern is determining what is possible or impossible to compute, according to the principles of computation. Complexity theory deals with whether you could actually build a computer that could compute something, in, say, less than the total age of the universe. Chomskyan linguistics is similar to computability theory, and the type of linguistics I studied is more similar to complexity theory. Chomsky wasn’t interested in linguistic performance — he’s very clear about that. He studied what is and isn’t possible in language, under very idealized conditions. In practice, it means he mostly studied nicely-written sentences, and determined if they were allowed or not, and what that means about the language. (So, also, he mostly studies syntax.) Functional linguists study something more like complexity theory: what people actually do, when they’re speaking and writing, under the often sub-ideal conditions of the world. So they study speech errors and hesitations and the circumstances under which a seemingly-nonsensical sentence might make perfect sense and a lot of things that Chomskyans tend to just say are irrelevant.

I find this comparison very helpful, because (along with a better grounding in Chomskyan theory that I got from reading The Linguistics Wars) it helped me get beyond my knee-jerk belief that Chomsky is clearly irrelevant. Computability theory is very interesting. It’s where things like the Halting Problem and Godel’s Incompleteness Theorem live. But computability is sometimes not very relevant, if you could compute something but it would require more time than the age of the universe. Likewise, there are very large domains of linguistic endeavor that Chomsky simply ignored. Where I disagree with him is that he tended to think that these areas are not important at all — that performance is incidental — whereas I think it’s critical to understanding what language really is. So my answer is something like: Chomsky is good as far as he goes, but he doesn’t go very far in the real world — the way that computability theory is great as far as it goes, but it doesn’t go very far in the real world in which you’d like your computer program to finish before next week (let alone before the end of the universe).

* Not the most common one, though; that would be the question “How many languages do you speak?” My answer to this is decidedly unimpressive.

** In short: Functional linguists think it’s because language serves a common function, regardless of its specific implementation (that function being human communication), and that it evolved from and/or is produced by a collection of adapted general skills. Formal linguists believe that languages are similar because we all have a specific brain module called a language acquisition device — ie, that language is a specialized skill, not a general one, and that languages are similar because they’re all produced by this special module. I think it’s possible that both are true, although having been educated by functionalists, I tend to wonder whether the formal bit is actually necessary.