Abstract : Like regular one-word verbs, verb-particle combinations or phrasal verbs can be ambiguous in the lexicon. This study examines ways to disambiguate by adding specific semantic features or “classes d’objets” to the lexicon-grammar, along with characteristic syntactic selectional restrictions. We first enlarged the systematic description of the most productive particle used in English phrasal verb constructions: "up". The initial database, composed of 300 transitive and neutral occurrences of phrasal verbs with "up" was expanded, and in the final database of 721 "up"
expressions, 64% were shown to be ambiguous. Of these ambiguous expressions, over 60% involve just two or three homonyms, such as "crack up the audience" [= make laugh] vs. "crack up the car" [= damage], while another 20% comprise four or five homonyms. The rest involve six or more homonyms, with the expression "pick up" having fourteen distinct meanings. Following Gaston Gross (1994, 2004), Le Pesant & Mathieu-Colas (1998), and other researchers at LLI, we introduce hyperclasses, semantic classes, and domains to the object description of some of our data. These subcategorization refinements seem to help mitigate ambiguity and underscore the importance of a lexicon-grammar approach, which includes both syntactic and semantic information, to English phrasal verbs.