Andreas Kathol (kathol@ling.ohio-state.edu)
Wed, 19 Jul 1995 13:13:21 -0400 (EDT)
A few comments on Kai Lebeth's recent posting. > For determining the nominative case in constructions like: > > [Ein Held zu sein] macht Spass, I assume the following structural case > principle and a lexical entry for the copula: > > Structural Case Principle: > > The nominal C_DTR of a H_DTR, which head type is verb (infinite > and finite) receives nominative, iff the INDEX-value of the C_DTR > is coindexed with the PROM_ARG-value of the H_DTR and if the > CASE-value of the C_DTR is of type structural case. > > The nominal C_DTR of a H_DTR, which head type is verb (infinite or > finite) receives accusative, iff the INDEX-value of the C_DTR is not > (!!) coindexed, i.e. different, with the PROM_ARG-value of the H_DTR > and if the CASE-value of the C_DTR is of type structural case. > > The CASE-value of the nominal C_DTR of a H_DTR, which head type is > verb (infinite or finite), is licensed iff the CASE-value is of type > lexical case. > > > Lexical Entry of the copula (infinite form): > > SYN: LOC: (CAT: (HEAD: PROM_ARG: [1] & > SUBJ: <DP[1]structural-case>) & > SUBCAT: <DP[1]structural-case>) & > CONT: (INST: [2] state & > RELN: <[copula_rel & > INST: [2] > UND: [1]]>) 1. I think this proposal is interesting as it relates to similar proposals that have been made elsewhere. The idea of handling case assignment not lexically, i.e. strictly within the lexical entry for the head, but with an eye toward syntax, is rather reminiscent of Eun Jung Yoo's quite similar proposal for case assignment in Korean: @incollection{Yoo:93, author = "Eun-Jung Yoo", year = "1993", title = "Subcategorization and Case Marking in Korean", booktitle = "Papers in Syntax", editor = "Andreas Kathol and Carl Pollard", series = OSUWPL, publisher = "Department of Linguistics", address = "Ohio State University", number = "42", pages = "178--198"} (Stefan Mueller has suggested to handle subject-verb agreement also in syntactic, rather than strictly lexical terms: @misc{Mueller:94, author = "Stefan M{\"u}ller", year = "1994", title = "{H}ead-{D}riven {P}hrase {S}tructure {G}rammar f{\"u}r das {D}eutsche", note = "Unpubl. lecture notes", address = "Humboldt University, Berlin"} ) One place where this approach may indeed shed some light is the perennial problem of partial VP fronting with nominative arguments: (1) [Ein Fehler unterlaufen] ist ihr noch nie. a mistake-NOM happened-PSP is her-DAT yet never `A mistake has never happened to her.' (2) [Eine Concorde gelandet] ist hier noch nie. a Concorde-NOM landed-PSP is here yet never `A Concorde has never landed here.' There are a number of reasons to think that the participle forms a constituent with the nominative NP, as indicated--well, at least I'm not aware of any concrete proposal that does not make this assumption. It seems that Lebeth's account should cover such cases at least in principle, case assignment only "happens" in the context of syntactic realization. 2. Also, the idea of having a PROM_ARG feature is reminiscent of an idea I had a long time ago to group "ergative" arguments together, i.e. direct objects in transitive predicates and subjects in unaccusative ones in terms of a what I call "identifying" attribute "ERG". Along rather similar lines, I think one can simulate the effect of a split SUBCAT list by a similar identifying feature for the subject. Hence, instead of: +- VAL | SUBJ <NP> | COMPS <NP, NP> +- one would have: +- VAL | SUBCAT <[1]NP, NP, NP> +- SUBJ' <[1]> Whether or not a given element on the SUBCAT list is a subject is determined by whether or not structure sharing holds between that element and the value of SUBJ'. As far as I can tell, Lebeth's PROM_ARG attribute is pretty much equivalent to SUBJ' (modulo the sort of the value). Incidentally, I think it is fair to regard the idea of identifying features as a move toward the keyword approach to grammatical functions in LFG. 3. One thing that puzzles me, though, is whether/how Lebeth's proposal can deal with copula constructions in which the indices of the subject and the predicative nominal are distinct. Cf.: Diese Frau ist in Wirklichkeit ein Mann. this woman-FEM is in reality a man-MAS-NOM *einen Mann a man-MAS-ACC Because of the difference in gender, both indices cannot be the same, hence Lebeth's second condition should apply here, predicting the predicative nominal to receive accusative case, contrary to fact. > The nominal C_DTR of a H_DTR, which head type is verb (infinite or > finite) receives accusative, iff the INDEX-value of the C_DTR is not > (!!) coindexed, i.e. different, with the PROM_ARG-value of the H_DTR > and if the CASE-value of the C_DTR is of type structural case. --Andreas
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:33:03 PST