Search for a command to run...
Prior research has shown that during tasks with very salient saccadic targets, like popout visual-search and double-step saccade tasks in both humans and monkeys, two-saccade sequences occur, where an initial saccade to a distractor is followed soon after by an ultra-short-latency (USL) second saccade to the popout target. This indicates that the USL second saccade uses visual information obtained before the first saccade and corrects the erroneous decision to make the first saccade to the distractor. Here, we demonstrate that such USL second saccades also often occur when monkeys perform active visual search without a popout visual target. Unlike in prior studies where the search target was extremely salient and therefore all second saccades showed a high accuracy toward the target, in our task, we show that USL saccades are preferentially directed towards the search-target and the transfer of target selection and location information across the first saccade is very accurate. USL saccades foveate the target more often than regular-latency saccades especially when the saccade starts far away from the target. We demonstrate various properties of USL saccades that reveal how the error-correction process is very accurate and how concurrent processing of saccade goals may proceed on the basis of computations in visual/oculomotor priority maps. Our results expand our understanding of saccadic targeting and information transfer across saccades during goal-directed active vision, lay the ground for future neurophysiological studies, and suggest that error-processing in different sensory and response modalities show similar patterns and may be accounted for by similar models.