Question under Discussion Uncertainty in Sentence Processing

Linguistic communication is a goal-driven process. Very often when humans try to understand a conversation or a piece of text, they ask “what is it about?”. Question under Discussion is a technical term that is used to capture this intuition, reflecting the idea that meaningful linguistic exchange involves structuring discourse context around relevant “issues” or questions. Communication between speakers is successful if conversational partners are cooperative and try to resolve the relevant issues together. Despite the important role QUD plays in linguistic theory, there exists a challenge in systematically evaluating the wide range of potential QUDs within naturalistic real-world discourse. Individual speakers may perceive different issues as relevant, even when presented with the same context, introducing uncertainty on what is the most salient issue. Such uncertainty can lead to miscommunication. The current project aims to develop methodologies to quantify QUD uncertainty in naturalistic discourse, and also investigate how QUD uncertainty impacts language comprehension in general. Four research questions will be addressed: (i) Can the concept of Question under Discussion be effectively quantified and analyzed in naturalistic discourse? (ii) How does QUD-(un)certainty influence people’s allocation of attention during language comprehension? (iii) Does QUD-(un)certainty account for the variation in how people make pragmatic inferences? (iv) Are there cross-linguistic and cross-cultural differences regarding how QUDs arise in contexts and how they impact language comprehension? To pursue these objectives, this project constructs open-access corpora for naturalistic discourse in two languages. Both corpora include rich annotation of QUD information. The second objective of the project is to develop behavioral methods to collect reading time data over the internet, which provides information to draw conclusions about people’s language comprehension behavior. Combining behavioral and computational methods, this project tests several linking hypotheses that connect QUD uncertainty to people’s reading time data, quantifying the effect of QUDs on human language comprehension. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.