A note on backward anaphora

In this paper I present four empirical observations about backward anaphora; building on Schlenker (2005), I suggest that the phenomenon should be though of in terms of a left–to–right interpretation process, in which discourse referents can be dynamically stored (by NPs) and retrieved (by pronouns) at two distinct and interacting levels of organization of the grammar: the «local» level of semantic computation of constituents, and the global discourse level