COMPARISON OF FACE-TO-FACE AND VIDEO-MEDIATED INTERACTION

A series of experiments are reported in which pairs of subjects perfor med a collaborative task remotely and communicated either via video an d audio links or audio links only. Using the same task (the ''map task '), Boyle et al. (1994) found clear benefits of seeing the face compar ed with audio-only co-present interaction. Pairs who could see each ot her needed to say less to achieve the same level of performance as pai rs who could only hear each other. In contrast to these findings, in a ll three experiments reported here, users of video links produced long er and more interrupted dialogues than those who had audio links only, although there were no differences in performance. Performance was af fected when the video Links were of low bandwidth, resulting in transm ission delays. The drop in accuracy was correlated with a significant increase in levels of interrupted speech. We also compared the structu re of dialogues and the use of gaze in high-quality video-mediated com munication with those produced in face-to-face copresent interactions. Results show that both face-to-face and video-mediated speakers use v isual cues to check for mutual understanding. When they cannot see eac h other such checks need to be conducted verbally, accounting for the length effect in dialogues. However, despite using visual cues in the same way as face-to-face speakers, video does not provide the same adv antage of shorter and less interrupted dialogues. In addition, users o f video gaze far more overall than face-to-face speakers. We suggest t hat when speakers are not physically co-present they are less confiden t in general that they have mutual understanding, even though they can see their interlocutors, and therefore over-compensate by increasing the level of both verbal and nonverbal information.