Both Principle B and competition are necessary to explain pronominal disjoint reference effects

In two experiments, we investigate the source of the constraint against coref-erence between a subject and a non-reflexive pronominal object. We contrast two broadclaims about what such pronominal disjoint reference effects reflect: A competition be-tween different means of linguistically encoding a locally coreferent relationship be-tween two nominals in a sentence, and a constraint against coreference that is invari-able across contexts. In our experiments we studied both the ways in which speakersprefer to express such locally coreferent relationships (Experiment 1), and the interpre-tations that are compatible with non-reflexive object pronouns (Experiment 2). We findthat across different discourse contexts, producers almost categorically avoid produc-ing locally coreferent pronouns (Experiment 1), and generally reject any context thatwould lead to coreference between a subject and an object in a context inference task(Experiment 2). However, we did see evidence that on a minority of trials, comprehen-ders allowed a locally coreferent meaning for the pronoun, and assigned it a mean-ing that is predicted by the competition view. Overall, our results are compatible withthe hypothesis that the pronominal disjoint reference effect reflects a semantic con-straint against local coreference and binding (a constraint such as Principle B), but thatspeakers can nonetheless draw inferences about likely meanings for these expressionsthrough general pragmatic reasoning processes.