Abstract
The importance of learning in controlling gaze is implicit in the task dependence of fixation patterns in natural vision (e.g. Land, 2004; Hayhoe & Ballard, 2005). However, this issue is relatively unexplored in unconstrained environments with no obvious task. Previously we studied whether experience influences gaze distribution in a real walking environment, in which individual pedestrians behaved in characteristic manners. Subjects walked around a central structure with pedestrians who veered towards the subject for approximately a 1 sec period. Three of the walkers (“Rogue”, “Unpredictable” and “Safe”) followed a script: Rogue went on a collision path every time he approached the subject, Unpredictable veered on half the occasions, and Safe never veered. Probability of a potential collision clearly modulates fixation probability. Subjects look at Rogues more than Unpredictable and Safe. Following a switch in roles of Rogue and Safe, subjects quickly learned to adjust gaze priorities according to collision probabilities. In our present study we compared performance in the real environment with a virtual environment where the potential collisions could be made in a more controlled way. Probability of fixating pedestrians was clearly increased with the increased probability of that pedestrian colliding. Further, subjects quickly adjusted to the manipulation of collision probabilities of pedestrians. Though there was ∼ 10 % overall reduction in probability of looking at pedestrians in VR compared to the real walking experiment, the pattern of results remained the same. Somewhat reduced fixation probabilities in VR might be the result of a less behaviorally relevant environment, or the need to devote more time controlling walking in it. However, the similar pattern of results as in the real walking experiment validates the virtual reality paradigm and confirms the pro-active nature of gaze deployment in the natural world.
This research was supported by National Institutes of Health Grants EY-05729 and RR-09283.