Dr. R.D.Borsley (els003@bangor.ac.uk)
Mon, 4 Sep 1995 12:18:32 +0100 (BST)
A number of features of Bob Carpenter's remarks about phrase structure strike me as rather problematic. I would question the idea that the form-meaning relation is straightforward. Grammars relate sentence forms and sentence meanings, but what is understood when a sentence is used is the result of a complex interaction between the inherent meaning of the sentence and the context of use. That's why pragmatics exists. How a sentence is understood by somebody in a specific context is reasonably clear, but it's not obvious that the inherent meaning of a sentence is any clearer than its phrase structure. The notion of partial evaluation also strikes me as problematic. As far as I can see, it leads to a proliferation of rules and a loss of generalizations. If NP is replaced by Det + N whenever it appears on the right hand side of a rule then presumably it must be replaced by all the other expansions of NP. As a result each rule that introduces an NP will be replaced by as many rules as there are expansions for NP. (Actually, it's worse than this because a rule can introduce more than one NP.) You obviously loose the generalization that NP can consist of Det + N and the other generalizations about the structure of NP. You also loose the generalization that Det precedes N since you will have the rule VP ---> V Det N Det N (among others) for VP's containing a double object verb. If we don't mind a proliferation of rules and a loss of generalizations, then it's not clear why we should prefer HPSG to old-style PSG with thousands of rules. In fact if we don't mind a proliferation of rules and a loss of generalizations, syntax would be a completely different game. The suggestion that we should look to psycholinguists to solve our problems also strikes me as dubious. As I understand it from the Sag and Fodor paper, psycholinguistic work doesn't provide any evidence for or against empty categories. I suspect that is fairly typical. I think we should be looking for standard syntactic arguments, but of course if we're not concerned with capturing generalizations that will be difficult. Incidentally, it's not clear to me why speaker's acceptability judgements don't count as psycholinguistic data. One more point. The observation that there are extensionally equivalent grammars is not a new one. It was made by Quine in his 'Methodological reflections on current linguistic theory', and Chomsky argues in a number of places that it is of no great importance. Maybe he is right. Bob Borsley Department of Linguistics | URL: http://www.linguistics.bangor.ac.uk/ University of Wales, Bangor | FAX: +44 1248 38 29 28 Bangor LL57 2DG | Wales | Visit our web pages!
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:33:03 PST