Abstract
Visual information is primarily mapped in an eye-centered, retinotopic frame of reference. Because each saccade results in a shift of images on the retina, however, purely retinotopic representations cannot support the tracking of visual objects across saccades. The visual system therefore must compensate for the changing retinotopic locations of visual objects by updating their world-centered, spatiotopic information around the time each saccade is executed. In a series of experiments, we investigated how spatial attention is deployed to visual onsets during this updating period. If the focus of attention is updated rapidly in anticipation of an impending saccade, the spatiotopic coordinates of onsets may be transferred into “future-retinotopic” coordinates, thus facilitating target identification at two discrete locations. We tracked observers’ eyes as they executed saccades to a cued location and made a manual response to a visual probe. Within 200 ms of the saccade target appearing, a brief visual onset cue was displayed, followed by the target probe (a tilted bar). After completing the saccade observers were required to respond as quickly as possible to the orientation of the target. The cue was non-predictive of target location and, importantly, targets could be presented at one of three locations: the spatiotopic location of the cue, the future-retinotopic location of the cue, or directly between these locations. We found that the cue yielded faster responses to targets presented at the spatiotopic cue location, future-retinotopic cue location, and between these locations, when compared with eccentricity-matched control locations. These findings are contrary to the notion that the focus of attention shifts to the future location of attended objects just prior to a saccade. Our results also challenge the hypothesis that attention shifts predictively prior to saccades in an analogous way to neurons that predictively shift their receptive-fields.