Universal Grammar, parameters, and all that.

[Lionel Mandrake]

I'm starting to read a book about what Noam Chomsky has been up to this century. One of the ideas is that language acquisition is simplified by the existence of parameters, each of which must be set to one of only two values. One such parameter is the choice of whether heads of phrases come first or last. For example, in English a preposition (the head of a prepositional phrase) comes before its object. Other languages exist in which the equivalents of prepositions come after their objects, so one speaks of postpositions. The claim is made that every language must have all heads first or all heads last, never a mixture of the two.

I'm wondering what a linguist of this persuasion would make of the English word "ago", which seems to be a postposition, as in "Four Score and seven years ago, our fathers ... ." Similarly, Latin like English has lots of prepositions, but it also has a few postpositions, such as "causa" and "gratia". (MGM's motto "Ars gratia artis" is bad Latin; it should read "Ars artis gratia".)

[Igel]

1. How to handle that depends on what work you're reading.

Chomsky kept sliding around as the evidence (largely published by his students or others with rather less credit given) necessitated. This sounds like flip-flopping; it also sounds like revising your hypothesis in the light of additional data.

It's like "before" or "earlier." Descriptively it's an adverb, I'd say, with a required implicit argument being "the time referred to in the speech act." That kind of thing is also messy.

Note that principles and parameters is not an exceptionless set of rules. A word may be lexically specified and escape from the general parameter setting. There are a lot of exceptions. It works best with simple propositions.

P&P fails to account for the facts (i.e., requires that a lot of work be done by the lexicon) in cases of language change. Languages can flip from head initial to head final, and it's never a neat change. This makes P&P a rather unpleasant framework.

Later there'd have been all kinds of required movement to let the parameter setting work well--then you need to specify feature checking presumably in spec position, whatever the status of spec is these days. (It's been a while since I checked out of generative syntax; last I was around we were working with Move alpha in a Minimalist framework, and that was a bit more than a decade back.) A lot of that checking was, again, required by lexical specification, IIRC.

Give your framework in a bit more detail maybe I can update it or give a better answer. PM me, since I don't often get back to check on my posts.