An incremental QUD approach to causal discourse expectations

During discourse processing, comprehenders are driven by the (implicit) questions raised in the discourse as it unfolds. By raising "why" questions, implicit causality (IC) verbs bias comprehenders to seek out causal explanations which drive discourse coherence. However, not every question can be actively pursued. We propose an incremental Question-under-Discussion (QUD) framework in which QUDs that drive discourse expectations cannot be raised by questions that are trapped in 'discourse islands' such as relative clauses. This framework is tested by manipulating the position of "why" questions raised by IC verbs in a comprehension task and a self-paced reading study. The results reveal that "why" questions raised by IC verbs can only generate QUDs that enhance offline causal inference rates and guide online causal discourse expectations to material that is at the same or lower structural level, supporting the framework's core predictions. This work contributes to discourse processing research by introducing an incremental QUD framework that is supported by our experimental results, namely that discourse embedding affects discourse-level inferences and expectations.