View dependencies in the visual recognition of social interactions

Humans daily physically interact with other people (eg when shaking hands). Understanding the action of others is important for a successful social interaction. Little is known about the visual processes underlying the visual recognition of social interactions. Here we were examined the view dependency of social interaction recognition. We used motion capture to record predefined interactions (eg high five; handshake; hug) acted out by pairs of participants and created 3D models of these interactions. In the actual experiment participants were presented with these interactions one at a time and had to identify a predefined interaction (1IFC task). We manipulated the view point (front, side, top, 45 degree side view) from which participants saw the interaction and the presentation time of the social interactions. We recorded participants'accuracy (as measured by d prime) and reaction time to identify a predefined social interaction. We found that the d primes and reaction time significantly depended on the view point of the social interaction. The results suggest the existence of view dependencies in the visual recognition of social interactions