The cost of making saccades
A considerable body of evidence indicates that, after every saccade, spatial representations are updated on the basis of a copy of the saccade command. Since this corollary discharge may not match perfectly the actual saccade, this operation presumably comes at a cost. Here, we report results obtained with a novel gaze-contingent display procedure, developed to measure the average error introduced by each saccade, overcoming several limitations of previous studies. Observers were instructed to search for two hidden targets in complete darkness. They were told that the two targets would appear as soon as they looked straight at them. In realty, the targets were briefly displayed at the current gaze position after a predetermined number of saccades. Two conditions were examined: in the first (visual localization), at the end of the trial subjects adjusted the position of two cursors on the screen to match the remembered locations of the two targets. In the second condition (saccadic localization), subjects saccaded back to the remembered position of the first target at the end of the trial. In the absence of visual references, this task can only be performed on the basis of extraretinal knowledge of eye position. We show that subjects were able to reliably localize the two targets even after multiple saccades. However, every saccade increased the radius of the localization region by approximately 20', suggesting that this is the intrinsic error associated with corollary discharge. This loss of accuracy did not occur when visual references were present: a single 5’ dot displayed throughout the trial at the center of the monitor maintained performance constant irrespective of the number of saccades. That very similar results were obtained in the two conditions of visual and saccadic localization suggests that common spatial representations are used for both tasks.