Dr. R.D.Borsley (els003@bangor.ac.uk)
Mon, 11 Sep 1995 10:22:42 +0100 (BST)
I'm glad Geoff provided the references for me. It saved me checking them out. I have a number of comments on your points. I think most syntacticians would see syntactic generalizations as essentially something theory-independent in terms of which different theories (or different versions of the same theory) can be evaluated. Other things being equal, the better of two theories is the one that captures the most generalizations. Of course, it may not be immediately apparent what the generalizations are - GB claims that a single generalization (the ECP) accounts for the ungrammaticality of both 'Who do you think that did it?' and 'Have John may done it?' while everyone else denies this - but I don't think this affects the main point. You ask how do we decide between an analysis with and an analysis without empty categories? I would say in the same way as we have always decided things - by seeing which works best. But of course working best largely means capturing generalizations best. If we're not concerned to capture generalizations, it will certainly be hard to find arguments in this area, but it will be hard to find any arguments for anything. As far as I can see, there will be no arguments for separate ID and LP rules rather than PS rules, no arguments for general principles like the HFP and NLFP rather than stipulations built into individual rules, no arguments for a single head-complement schema (as in HPSG) rather than a lot of different head-complement rules (as in GPSG), etc. To repeat, syntax would be completely different if we weren't concerned with generalizations. Incidentally Bob Levine and Tom Hukari believe they have an argument for empty categories and I believe I have one based on Welsh mutation. I guess, though, that they both involve the idea of capturing generalizations - mine certainly does. You appear to be saying in connection with partial evaluation that a large number of specific rules are preferable to a smaller number of more general rules from a processing point of view. Is this true? I seem to recall Berwick arguing just the opposite. Come to think of it, that could mean that it is true! 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 - 20:33:04 PST