Towards a psycholinguistics of dialogue: defining reactionudtime and error rate in a dialogue corpus.

摘要

This study uses the multi-level coding of auddesigned corpus of unscripted task-orienteduddialogues to demonstrate that time to respondud(Inter-Move Interval, IMI) and rate of disfluencyudbehave like psycholinguistic measures, reactionudtime and error rate, in reflecting the speakers’udcognitive burdens. Multiple-regression analysesudshow that IMI is sensitive to social distanceudbetween interlocutors, to the difficulty of the taskudwhich the dialogue serves, and to comprehensionudof the prior utterance and production of theudcurrent one. Rate of simple overt disfluency, inudcontrast, shows social and task effects, with mostudof the uniquely explained variance associatedudwith planning and producing the currentudutterance. The results suggest that coded corporaudmay be useful in developing models of humanudinterlocutors.