Classical generative grammar partitions linguistic competence into three basic components: lexical knowledge, phrase structure rules and transformational rules (Chomsky 1965, 1981). One of the fundamental debates over the years, and one which is still alive today, concerns the division of labour between information and processes that reside in the lexicon and those rules and processes that are part of syntax.

In this book, I explore a view of the architecture of grammar whereby the lexicon is eliminated as a module with its own special primitives and modes of combination. By this, I do not intend to deny that there are items within the language that need to be listed/memorized, or that they are associated with grammatical information. Rather, I will seek to claim that to the extent that lexical behaviour is systematic and generalizable, this is due to syntactic modes of combination and not to distinct lexicon-internal processes (Hale and Keyser 1993, etc.). The general ideology is not novel; I am attempting to implement an old idea in the light of current, accumulated knowledge concerning the nature of ‘lexical’ generalizations and patterns. In pursuing, as I will, a radically unstructured view of the lexicon, I engage with recent ideas of constructionalism (Goldberg 1995, Marantz 1997b, Borer 2005) and make my own proposal based on what I take to be the core empirical issues of ‘thematic’ roles, event structure (aktionsart) and selection.

One of the things I will take for granted in this work is that human beings' linguistic competence includes, minimally and crucially, a (linguistically specific) combinatorial system.