Modifiers and specifiers in categorial unification grammar

This paper deals with the relationship between phrase structure and feature distribution in categorial grammar extended with unification. It focuses in particular on issues relating to the notions 'modifier' and 'specifier'. The reformulation of categorial grammar as a unification-based grammar formalism (Shieber 1986) opens the possibility of providing an elegant and well-understood theory of feature distribution for these grammars. An important ingredient of this theory is that specifiers and modifiers are defined in terms of their properties with respect to feature distribution. This leads to predictions about distributional properties of these elements which differ radically from the predictions made in phrase-structure theories such as X grammar (Jackendoff 1977). We argue that the introduction of partially specified categories, a natural thing to do from a unification point of view, leads to an account of some otherwise puzzling facts about modifier and specifier placement. Furthermore, the use of partially specified categories makes functor composition and head operations (Pollard 1984) to some extent superfluous.