Abstract
How do we perceive the behavioral intentions of another pedestrian? In this study, a virtual avatar is programmed to pursue or evade the participant in an ambulatory virtual environment. The avatar is driven by our steering dynamics model (Fajen & Warren, 2007; Cohen, Cinelli, & Warren, VSS 2008, 2009). We investigate whether the perception of pursuit and evasion is based on the avatar's trajectory, which is contingent on the participant's movements, and on the direction of the avatar's head fixation. Participants wore a head-mounted display (63° H x 53° V) and walked toward an approaching avatar, while head position was recorded using an inertial/ultrasonic tracking system (70 ms latency). Avatars could (a) pursue or evade the participant, and (b) fixate or look straight ahead; pursuers fixated the participant and evaders fixated and walked to a point behind the participant. In Exp. 1, a single avatar appeared at 6, 7, 8, or 9 m, and participants reported whether it was pursuing or evading them. Mean d' was 2.5, and head fixation only contributed at 6 m. Mean RT was 2-3 s; head fixation provided a half-second advantage for pursuit at 7 and 8 m, but a half-second disadvantage for evasion at all distances. In Exp. 2, two, three, or four avatars appeared; one was a pursuer, and the others were evaders. The participant reported which avatar was pursuing them. RT increased with the number of distractors, and there were small improvements with fixation. The results indicate that pursuit/evasion is reliably perceived from the avatar's contingent trajectory alone, and that they are judged sequentially. Head fixation provides a modest additional advantage at close ranges.