Prediction compensates for occlusion of a bounced ball

Abstract

Intercepting a rapidly moving target is complicated by the presence of sensory and motor delays that prevent the immediate use of new visual information. Although humans can compensate for these delays by predicting the future state of the visual environment, little is known about the underlying mechanisms. Previously, we investigated prediction in a virtual racquetball task in which the ball bounced once during its approach to the subject. Subjects were found to make predictive saccades to a location that the ball would arrive after the bounce. This location was fixated until the ball's arrival, and fixation was immediately followed with a pursuit movement. Although we manipulated post-bounce ball speed by changing ball elasticity and pre-bounce speed, subjects adjusted their predictive fixations to maintain an average duration of 170 ms between the bounce and the ball's arrival at the fixation, in all conditions. This constant-time strategy was accomplished by raising fixation height for balls that moved faster after the bounce. In this study, we test competing hypotheses for the constant-time strategy, for example, that this strategy allowed sufficient time to program the post-bounce pursuit. Pre-bounce speed varied on each trial (3 bounce-speeds x 30 repetitions = 90 trials per block). In a second block, the ball was always invisible for 100ms following the bounce. The data revealed that occlusion had only a minimal effect on pursuit performance. Fixation height was unaffected, and the predictive saccade was delayed by ~30 ms, allowing for a slightly extended viewing of the pre-bounce trajectory. Thus, one possibility is that subjects compensated for occlusion by placing greater emphasis on prediction using pre-bounce visual information. This interpretation is also consistent with the observation that, on some occlusion trials, pursuit was initiated in prediction of the ball’s reappearance. Meeting abstract presented at VSS 2013