Strategies of anaphora resolution

First draft: November 1997 (OTS Working Papers) Revised: May 1999 Appeared in Hans Bennis, M. Everaert and E. Reuland (eds) Interface Strategies North Holland Amsterdam (2000).Several corrections were added at the proofs stage, not included here. The major points of this paper are: a. The definition of binding in terms of coindexation (or identity of variables) should be replaced by a definition based on the traditional logical-syntax concept of binding. Apart for conceptual reasons, the coindexation view faces empirical problems and disables stating binding generalizations. b. The traditional distinction between binding and coreference should be modified to distinguish between binding and covaluation. Covaluation is available regardless of the referential status of the antecedent, and shows up also in quantified contexts, as shown in Heim (1993). c. Covaluation is not governed by considerations of the computational system but by an interface strategy in the spirit proposed in Reinhart (1983). However, although this still requires reference-set computation, there is reason to believe, contrary to my previous view, this this strategy is not an instance of economy, as developed in recent syntactic theory.