Abstract
Due to inherent processing delays, tracking a moving target with the eyes is a difficult task. To account for this, the oculomotor system uses not only the current sensory input, but also relies on recent experience. Such effects can even be found on a trial-by-trial level, where the prior trial is integrated in a reliability-weighted fashion. Here we show that even such a seemingly sophisticated integration of information for oculomotor control does not follow basic perceptual concepts as object consistency or size constancy. Participants saw two successive movements that they needed to track with their eyes. In the first screen, a car moved across a background which used perspective drawing to invoke a perception of depth. Importantly, the car could move at two different depth levels, which allowed us to dissociate retinal and perceptual size. We used a baseline psychophysical staircase procedure to create conditions where cars were perceptually identical in terms of size and velocity (perceptually-matched), but different in terms of their retinal size and velocity. In the second screen either another car or a Gaussian blob moved across a gray background (always at 10 deg/s), allowing us to directly measure the influence of the previous movement. Interestingly, the two perceptually-matched cars led to significant differences in subsequent oculomotor behavior: the retinally smaller & slower car in the back also led to slower pursuit in the subsequent trial. Interestingly, trial-by-trial effects were still present when the second stimulus was a Gaussian blob. This emphasizes the differences in the use of retinal information for oculomotor control and perception: While perception needs to integrate information to create the best possible percept, the trial-by-trial influence on oculomotor behavior seems to correct for the amount of retinal motion to reduce motion blur in subsequent trials to allow for better object recognition.