Quote Originally Posted by wm View Post
Transformational-generative grammar is far from parsimonious. In fact it seems to create more "band-aid" rules every time a grammatical exception pops up. Each of its rules has an extremely expansive ceteris paribus clause. As a result the rules can become rather hard to apply.
It might amount to a linguist’s aside not very germane to the discussion at hand, nevertheless…

If I understand correctly, and I might not as I have admittedly never taken a course in syntax (which is the bailiwick of the Chomskyan tradition), TGG takes for granted the existence of certain linguistic capabilities. It’s useful to consider how Chomskyans make use of the term ‘learning.’ They say that ceteris paribus, a child no more learns to speak than to walk. Rather, speech and walking manifest themselves. Grammatical and lexical exceptions are the elements of language that are learned.

There is some decent evidence for this perspective. The two best are probably the fact that children seem to invariably regularize and have to be taught not to in certain instances (like grammatical gender in Romance languages or umlaut plurals in English—“It’s ‘geese,’ not ‘gooses,’ Matthew.”) and creoles (which are by definition young languages and have very few irregularities). That evidence holds water to a certain extent, but there is counterevidence, too. When you have some knowledge of certain American Indian languages as I do and 1) see a lot more exceptions to rules than you typically do in Indo-European languages and 2) see pride of place in morphology rather than syntax, the foundations of the program appear a bit less stable.

There are a stock set of responses Chomskyans will make to objections. One is, “It is axiomatic that…”