Multiword Expressions and Lexicalism

Multiword expressions (MWEs) such as idioms exhibit a tension between phrase-like and word-like properties. Much recent work has treated idioms as exclusively phrase-like, by positing special idiom versions of the words they contain. In this paper, I argue that such approaches are unappealing, and suggest that by following the ideas of Abeille (1995), we can provide a more satisfying analysis that respects the special status of MWEs. This is implemented by replacing the context-free grammar standardly assumed for LFG c-structures with a Tree Adjoining Grammar (Joshi et al. 1975). This allows us to represent idioms in a single place, even when their parts can be individually modified and/or targetted by morphosyntactic operations.