Interpretation of informational questions modulated by joint knowledge and intonational contours

Abstract We examine processes by which dialogue partners form and use representations of joint knowledge, or common ground, during on-line language processing. Eye-tracked participants interpreted wh-questions that inquired about task-relevant objects during interactive conversation. Some objects were known to both speaker and listener, and thus in common ground, whereas others were only known to the listener, and thus in privileged ground. Questions were produced with a typical, falling intonation (Experiment 1) or with either falling or rising intonation (Experiments 2–3). Unlike the falling contour, the rising contour can indicate a request for clarification about previously mentioned information. Participants interpreted falling-contour questions as asking about privileged-ground objects. By contrast, rising questions elicited more consideration of common-ground objects. Directly comparing questions that were produced during live conversation vs. questions that were pre-recorded revealed that this sensitivity to common vs. privileged ground emerged only during live conversation. Finally, individual difference analyses in all three experiments did not support the claim that individuals fail to take perspective when executive function is limited. Taken together, these findings provide evidence for the on-line integration of perspective and intonation during conversational language processing. The lack of perspective effects in non-interactive settings speaks to the inherently interactive nature of conversational processes.