Sublexical modality and the structure of lexical semantic representations

This paper argues for a largely unnoted distinction between relational and modal components in the lexical semantics of verbs. We hypothesize that many verbs encode two kinds of semantic information: a relationship among participants in a situation and a subset of circumstances or time indices at which this relationship is evaluated. The latter we term sublexical modality. We show that linking regularities between semantic arguments and syntactic functions provide corroborating evidence in favor of this semantic distinction, noting cases in which the semantic grounding of linking through participant-role properties apparently fails. This semantic grounding can be preserved, however, once we abstract away from sublexical modality in lexical semantic representations. Semantically-based linking constraints are insensitive to the sublexical modality component of lexical entries and depend only on information in a predicator's "situational core".