Abstract
In the real world, humans must allocate limited perceptual resources to ensure that gaze is in the right place at the right time. How is this achieved? We suggest that humans are very sensitive to contextual attentional demands and rapidly learn to deploy gaze proactively on the basis of the learnt statistical structure of the environment. Subjects walked along an oval path in a large room, for 12 laps, in the presence of other pedestrians who behaved in characteristic ways. One pedestrian (Rogue) walked on a collision course with the subject for a brief period (1sec) whenever they neared the subject. Another pedestrian (Risky) walked on a collision course with 50% probability, and another (Safe) walked normally. In another block of trials, subjects had the added task of staying a constant distance behind a lead pedestrian. The latency with which the pedestrians were fixated after appearing in the field of view changed rapidly for Rogue and Safe pedestrians. Latencies for Rogues shortened, and for Safe increased, by over 200msec. This was true whether on not the Leader was present. Subjects were nearly always fixating the Rogue pedestrian proactively, in advance of the path deviation. With no Leader, fixations durations changed with a similar time course, becoming longer for Rogues, and shorter for Safe. However, when subjects had the added task of following the Leader, fixations on all pedestrians were 100msec shorter and durations did not change across time. Thus Subjects handle the added load by reducing time spent fixating all pedestrians, but they maintain the same gaze priorities on first appearance and reduce latencies on Rogues even with the added task. Individual subjects behave similarly, despite the unconstrained context, suggesting that rapid gaze adaptation reflects a stable and lawful property of natural gaze behavior.
Supported by NIH EY05729.