Salience, Inference and Plural Anaphora

In this paper, I argue that the DRT Construction Rules for plural antecedents are redundant, because they are subsumed by an inference mechanism that must be made generally available for anaphora resolution. While Kamp and Reyle originally argued against a general inference mechanism for plural anaphora, I argue that the facts of compset anaphora require inference. Furthermore, I observe that compset anaphora is blocked by refset anaphora. I consider an alternative DRT account, in which inference coexists with Construction Rules. I argue that the Construction Rules are not necessary to capture the blocking generalization; rather, I argue that refset descriptions are semantically primed, thus deriving the blocking generalization in terms of general inference, together with an independently required mechanism of semantic priming. I argue that this alternative is to be preferred on grounds of theoretical parsimony. Furthermore, I present an argument that the general inference account correctly captures the fact that plural descriptions can be interpreted at the position of the plural pronoun, while the DRT account incorrectly requires that they be interpreted at the position of the antecedent.