Stephen M. Wechsler (wechsler@mail.utexas.edu)
Tue, 25 Jul 1995 18:02:12 -0500
Eun Jung Yoo notes the following, regarding Korean case: >In my paper, I proposed that most Korean verbs simply subcategorize for >structural NPs (i.e. NPs with structural case) for subjects and complements >and that their surface case is determined by the principle (i): > >(i) Case Principle (for Korean) > A unresolved structural NP which is a daughter of a phrase P is > [nom] if it is a SUBJ-DTR of P, and [acc] if it is a COMP-DTR of P. > ... >The main motivation for such a structural approach to Korean case >marking was to account for Korean raising verb constructions as in (ii): > >(ii) Mary-ka ku-lul o-n-ta-ko mit-nun-ta. > Mary-NOM he-ACC come-PRES-DECL-COMP believe-PRES-DECL > 'Mary believes him to come.' > I agree with the gist of Eun Jung Yoo's analysis: the 'ECM' construction has a flat structure (the ACC is the matrix object), and case is determined by something like grammatical relation. But there is evidence that the Korean construction is rather different from its apparent counterpart in English. Actually it appears to be a kind of 'null operator' (weak UDC) construction. Most importantly, the 'raised' ACC doesn't need to correspond to the subject of the lower predicate, as in this example with a locative (from Hong 1990): (i) John-un L.A.-lul hankwuksalam-i manhi san-ta-ko sayngkakhakoiss-ta. J.-Top LA-Acc Koreans-N many live-Dec-Comp think-Dec 'John thinks that many Korean live in L. A.' Not every speaker loves this example, but most find it OK. See Ki-Sun Hong 1990 (in Dziwirek et al _Grammatical Relations--a Cross-Theoretical Perspective) for more examples. Jae-Gyun Song's 1994 dissertation (UT Austin) gives further, rather surprising, evidence for the 'null operator': the ECM construction is a syntactic island. Also, in a paper at the last LSA, Yae-Sheik Lee and I looked at the semantic conditions on this construction. Roughly, the embedded predicate cannot denote a specific event; it must be generic or lexically stative (i-level rather than s-level). We tried to relate this to other weak UDC constructions like the tough-construction, which has a similar semantic condition. In HPSG terms this means that the apparent case problem for Korean should be handled similarly to other weak UDCs. Roughly, an 'ECM verb' like mit- 'believe' in Eun Jung Yoo's example above, would specify structure-sharing between indices of the NP[acc] and an item on the slash list of the embedded predicate: COMPS < NP[acc]/-[1], S[INHER|SLASH {NP/-[1]}] > Since only the index is shared, there is no case conflict. --Steve
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:33:03 PST