Differences in head orientation between speakers and listeners in multi-party conversations

We investigated whether people can identify who is speaking in multiparty conversations solely on the basis of the head orientations of the participants in the conversation. Two experiments were conducted using a virtual meeting room. Participants were shown the head orientations of four virtual humans and were asked to identify the speaker. In the first experiment participants were shown static frames, in the second an animation of an entire speaker turn. Head orientation data for both frame and animation experiment were taken from real-life meetings. Given that the animations provide more context, we assumed better scores for this experiment. However, participants did not score significantly better in the animation experiment (42.36%) compared to the static experiment (40.63%).