Andreas Kathol (kathol@violet.berkeley.edu)
Mon, 25 Nov 1996 09:28:10 -0800 (PST)
Date: Mon, 25 Nov 1996 10:13:28 -0500 From: "Samuel L. Bayer" <sam@linus.mitre.org> Subject: Re: Bayer on coordination Hello Bob -- I don't subscribe to the HPSG list, but your note was forwarded to me by a friend. I wish to make a clarification on your comments about my Language article. I do *not* assume that all underspecification should be replaced by disjunction. There is a crucial difference in Categorial Grammar between argument positions and the categories that fill these positions, namely, the asymmetry introduced by the implicational paradigm that CG adopts. Argument positions need *not* be disjunctions. Your example, where the object position is underspecified with respect to definiteness and number, is as underspecified in my account as it is in any unification-based account, because the nature of functional categoires is implicational. That is, if I have a VP/NP[sg], I'm asserting that if you give me a singular NP, I'll give you a VP; in other words, the logic of categories in this case states that a singular NP *implies* a VP. This is the category of verbs which require singular arguments. If the NP is more specific than that (say, if it's singular and accusative and definite), that's fine, because more specific implies less specific for all the usual logical reasons, and the presence of such an NP implies the presence of (actually, *is*) the sort of NP which can fill the argument position in question. The situations where underspecification can't replace disjunction in my account are those in which the NP *itself* is ambiguous between definite and indefinite, for instance. So cases like the famous "sheep" case cannot be handled by underspecification in my account, and must be ambiguous. So in your terms, the SUBCAT list is special. As usual, I must credit Mark Johnson for his substantial contributions to this body of work, especially his work with my on features which appeared in the 1995 ACL proceedings. The coordination evidence presented there, I think, is especially compelling. Cheers, Samuel Bayer
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:33:30 PST