Michael T. Wescoat (Osaka University)
Rcently, T. Mohanan posits the following four-class typology for noun-incorporation
A B C D
The incorporated noun is an argument NO YES YES YES
Modifier stranding is allowed NO NO YES YES
Doubling is allowed NO NO NO YES
Class A includes Samoan, C Southern Tiwa, and D Mohawk; Mohanan adds class B to accomodate observations she makes about Hindi. She claims that class B necessitates a grammatical function based approach. I maintain that on a broader empirical examination, Hindi actually falls in class C; i.e., in rejecting stranded modifiers, Mohanan overlooks semantically more plausible data that lead to the opposite conclusion. The felicity of stranded modifiers suggests that (a) there must be an object NP in phrase structure, (b) the need for grammatical functions is obviated, and (c) the typology above may be simplified by dropping B. Now NI of types A and D can be handled lexically. The remaining type C can be analyzed with either head movement or else a technique called LEXICAL SHARING. In the latter case, one may give a uniformly lexical treatment of all three type of NI - A, C, and D.