Ulrich Koch (koch@mailhost.uni-koblenz.de)
Sun, 8 Oct 95 18:14:28 +0100
Steve, you were right, and I was wrong. I shouldn't have answered your question without looking up the ID schemata in P&S 87 and P&S 94. Your structure S / \ NP VP | | N V | | Kim left is correct for P&S 87 and P&S 94 (modulo obligatory null determiners), but the NP can't be licensed by the Head- Comp-Schema (Schema 2) as given in P&S 87 and P&S 94. (This is where I was wrong.) The reason is that, as you said, the mother's SUBCAT value would then be required to have length one. So the NP can only be licensed by the Head-Subj-Comp-Schema (Schema 3) from P&S 94. You said: > P&S 94, p.402 fn. 17 states that 'In > the parochial versions of Schemata 1 and 2, the INV value, _if any_, must > be minus; in the parochial versions of Schema 3, it must be plus.' The > crucial part here is the "if any". Nouns are not specified for INV. It's > unclear whether Carl and Ivan intended for the "if any" to apply to Schema > 3, but assuming this is the case then N projections can be licensed by it. That's what I assume. The NP is licensed by Schema 3. This is not possible in P&S 87, because the "if any" doesn't occur there. (I suspect this is why Carl and Ivan put it in P&S 94 in the first place.) > Then Schema 3 licenses the NP [Kim]-- along with bad NPs like [picture the > of Bill] (modulo LP constraints), where [picture], [the], and [of Bill] are > sisters. I _think_ so. (See? I'm getting more cautious! :-)) This analysis appears to be unwelcome -- it should really be: NP / \ DP N' | | \ D N PP | | |__\ the picture of Bill (Here the NP is licensed by Schema 1 (Head-Subj-Schema); the N', by Schema 2.) This analysis isn't ruled out by P&S 87 or P&S 94, but apparently, it isn't enforced either, although it should be. > In HPSG3 (i.e. Ch.9) this last problem could probably be solved by setting > up the new schemata properly. I think P&S 94, Ch. 9 solves this through lexical entries: [PHON <picture>, SPR <DP>, COMPS <(PP[of])>] That is, the DP is required to be a specifier, not a subject. So the Head-Specifier-Schema can license the NP, while the Head-Subj-Comp-Schema can't. > Recent variants by Ivan which dispense with > ID Schemata entirely and instead involve a richer hierarchy of phrase > sorts, also seem to work. As far as I can see, yes. > Hope you don't mind if I post this back to the hpsg list. See above. It's not your fault that I made a fool of myself. Best, Ulli
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:33:04 PST