Demanding an
Explana4on:

 Implicit
Causality
Biases
in
Discourse
Interpreta4on

To clarify the effects of IC biases on discourse interpretation by distinguishing (i) next mention biases and (ii) biases toward upcoming coherence relations. Problem: Previous passage-completion studies report strong biases regarding who will be mentioned next following implicit causality (IC) verbs with a ʻbecause ʼ prompt. However, these biases are reduced/eliminated with a full-stop prompt. (1) a. John scolded Mary because _________________ _. [strong bias to Mary] b. John scolded Mary. ______________________ _. [mixed biases] Proposal: In light of recent results showing two types of coherence-driven expectations in pronoun interpretation, we compare responses to contexts like (1a-b). We predict that IC biases depend both on expectations about upcoming continuation types (P(coherence)) and on biases for which event participant will be mentioned again conditioned on continuation type (P(referent | coherence)). Results: By categorizing responses by coherence relation, we localize the previously reported IC bias to Explanation relations. We find an additional IC bias concerning P(Explanation).This bias has gone unnoticed because previous work has not categorized responses by coherence