Abstract
In the natural world, the brain must handle inherent delays in visual processing. One compensatory strategy is to combine prior experience with current sensory data to predict a future visual state. Although prediction is known to occur in both saccadic and pursuit eye movements, the factors that contribute to prediction are poorly understood. Furthermore, because most experiments have investigated 2D planar movements with restricted head-movements, it is unclear how the findings generalize to more natural environments. In this study, subjects intercepted virtual balls in a simulated environment seen through a head-mounted display. On each trial, a launched ball bounced on the floor before its arrival at the subject. The subject used a racquet to hit the virtual ball at a target on the far wall. We varied the velocity of the balls from trial to trial, and the elasticity of the ball between blocks of trials. On 83% of the trials, subjects initiated a saccade prior to the bounce, to a location 8° to 13° above the bounce-point. This location predicted the ball’s post-bounce trajectory with high accuracy, so that the ball passed within 3.5° of the gaze point (+/- 0.53° SEM between subjects). The targeted location compensated almost exactly for the changes in trajectory resulting from ball velocity and elasticity, and the ball passed through the predicted location 141 (+/- 8) ms after the bounce in all conditions. This constant-time strategy meant that saccade height scaled linearly with predicted post-bounce ball velocity, suggesting that subjects used learnt knowledge of ball dynamics to predict where the ball would be after the bounce, and when it would get there. The accuracy of the prediction implicates a complex internal model of ball dynamics that accounts for changes in ball elasticity, 3D velocity, angle of incidence, and gravity.
Meeting abstract presented at VSS 2012