In theAspectsframework, strict subcategorization is regarded as a feature on a verb, say [+_NP] forexpress; various
othertreatmentshaveappeared over theyears. The presentapproach suggests that wecananalyzeexpressas havingthe
lexical structure shown in (2.7).
This lexical VP has a phonological reflex only in the V position (the only syntactic position that has a pre-subscript to
phonological structure). So on the sur-face it looks like it is just a verb. But it is not: it is actually a lexical VP with an
open object position.^85
A similar approach recommends itself for governed prepositions. For instance,depend (on)is like the idioms in (18)
except that it lacks the NP argument.
To su mup, we have discovered a cline of lexical VPs, going fro mverbs that obligatorily subcategorize co mple ments,
throughverbswith governedprepositions, throughidioms, throughconstructional idioms thatspecify part oftheVP's
complement,totheresultative.Each step from oneto thenext is perfectly natural, but at theends of theclinewehave
entities that appear entirely different—and have regularly been so analyzed in the literature. The present approach
permits us to see subcategorization by a verb, by an idiom, and by a construction as falling under a common
formalization: leaving positions open in a lexical VP. (Williams 1994) and Langacker 1998) point out the existence of
such clines in other areas of the lexicon.)
LEXICAL STORAGE VS. ONLINE CONSTRUCTION 177
(^85) I had better say why (26b) is grammaticaldespitethe absenceof a syntacticobject. Under standard assumptions, the object positionis in fact occupied by an NP: a trace or
its equivalent, bound to the wordwhat. The semantics of the trace, however, comes via its being equated with the clausethat he was disappointed.Incidentally, the second
semantic argumentin (27) is simplyco-subscripted withthe syntacticargument, instead of being marked optional or obligatory as in previous examples. There are probably
better realizations of this detail.