Categorial Minimalist Grammar

1. Convergence at first sight, and differences in a second time From the early days of the minimalist program of Chomsky (1993), a convergence with categorial grammars was noticed by Ep-stein and Berwick (1995). The striking similarity lies in the merge operation, which looks like the application rule of the AB grammars. In both cases, word order is a consequence of the consumption rules but in categorial grammars being the head of a compound expression also derives from the categories while in a minimalist setting it can be defined independently from the resource consumption. The most striking difference is the absence of movement in categorial grammar: how could this notion be captured? If it can, how could minimality conditions like shortest move be formulated? In the other direction the main difference is the atomicity of minimalist features: merge is not recursive, there cannot be a demand of a category which itself demand another category. Hence, after quite an optimistic wish of convergence, the task seems so tough that one may wonder why we are willing to do so? There are at least two reasons: • Of course, the main reason is that categorial grammar easily yield the logical form (or semantic representation) following some of Montague's idea. • A secondary reason is the formalisation of converging learning algorithms from positive structured examples, which can be defined for categorial grammars, but we shall not speak about it, although one of us did something in this direction Bonato and Retoré (2001). Because the categorial tradition is anchored in a formal and logical apparatus, the convergence between generative grammar and categorial grammars benefited from the development of resource