Abstract
The exponents of functional features show a range of variation that cannot be captured even by dichotomies such as the clitic-affix distinction, not even when extended to include categories such as phrasal affix. A multidimensional parallel correspondence theory such as Lexical-Functional Grammar is eminently suited to dealing with the full range of variation. However, some of the patterns we find in languages involve extensions to the standard constructional formalism of LFG. In this paper we look at some challenging patterns of distribution exemplified by the definiteness feature and compare analyses expressed in terms of standard LFG formalism with those captured in a specification language approach.
Link to pdf of paper