Leveraging big data to understand the interaction of task and language during monologic spoken discourse in speakers with and without aphasia

Monologic spoken discourse allows us to evaluate every day speech while retaining some experimental constraint. It also has clinical relevance, providing cognitive-linguistic information not measured on typical standardised tests. Here, we leverage big behavioural data (AphasiaBank) to understand how discourse genres (narrative, procedural, expositional), and unique tasks within those genres, influence microstructural elements of discourse. We compare task × microstructure interaction across speakers with and without aphasia and evaluate the influence of aphasia type and overall aphasia severity on this interaction. Using multivariate statistical methods, we find that, for both speaker groups, discourse microstructure is most similar for tasks within the same discourse genre and that microstructure is largely dissociable across discourse genres. The aphasia group had more speaker variance per task, which was partially explained by aphasia type and overall aphasia severity. Our results provide necessary information for usage and interpretation of monologic discourse in research and clinical contexts.