This paper explores the consequences of unspecified syntactic features
for an analysis of Japanese light-verb constructions. In Japanese, th
eta-assigning nouns (verbal nouns or VNs) combine with the light verb
suru 'do' to form hundreds of clausal predicates. A VN can appear adja
cent to suru and caseless, or be marked with accusative case. In the s
econd instance, the VN may or may not form a constituent with one of i
ts complements. Prior research into the syntax of these constructions
has focused on (i) whether the relationship between the VN and the lig
ht verb is derived in the lexicon oi. is established in the syntax, an
d (ii) if rite latter, whether the several constructions involving VNs
are syntactically related The position adopted hel-e is that VN-plus-
suru combinations are syntactic, and further that the extraordinary sy
ntactic behavior of VNs, as well as their ability to generate three su
perficial structures from a single underlying form, is attributable to
their categorial features [N] or [V] remaining unspecified in the lex
icon.