ABSTRACT Current theories of grammar generally assume a constraint on lexical selection, i.e. the determination of syntactic properties of one lexical item by another, which was originally proposed by Chomsky in 1965 as "local subcategorization" -- that the two lexical items must be sister constituents, immediately adjacent in underlying syntactic structure. A number of instances have emerged in recent research, however, which suggest that this constraint may be too restrictive, that categorial selection can under some circumstances be non-local. The aim of this project is to investigate the limits of non-local lexical selection, and then to develop a general system of non- local selection which correctly characterizes the constraints. The investigators will examine an existing data-base for English, the Brandeis Verb List, which specifies the syntactic forms of the complements of roughly 600 verbs and which appears to contain numerous instances of non-local selection, and they will formulate the lexical-conceptual structures and the argument structures for these verbs. They will then construct functions that map these semantic structures into the syntactic structures specified in the Brandeis Verb List. After entering the lexical-conceptual structures and argument structures for the verbs into a computer, the investigators will program the functions from semantic categories to syntactic categories in an attempt to reconstruct the original database from the modified database. The study will bear on the representation of case, prepositions, and complementizers in lexical entries.
