Abstract
Our visual world feels stable, but with each eye movement, the retinotopic (eye-centered) locations of objects change dramatically. Increasing evidence suggests that spatial representations are coded retinotopically (Golomb et al, 2008 J.Neurosci.), even at higher stages of processing (Golomb & Kanwisher, in press Cerebral Cortex). What does this mean for the perception of object features? Here we report a systematic distortion of color perception occurring immediately after a saccade. Subjects viewed an array of colored stimuli and clicked on a colorwheel to report the color of the "cued" stimulus. The cue was always presented before an intervening saccade, and the task was to report the color at the spatiotopic (world-centered) location of the cue. However, if a differently colored distractor was present at the retinotopic location, the distribution of responses was shifted in color space toward that retinotopic color. This "feature mixing" was spatially specific, driven by the color of a distractor in the retinotopic – but not an equidistant control – location. Fitting the data with probabilistic models revealed that in addition to errors caused by random guessing and accidentally misreporting the wrong color, a significant shift in color space remained, as if the retinotopic distractor color were blending with the true color perceived at the spatiotopic location. These findings indicate that perceptual distortions around the time of a saccade are not limited to spatial or temporal judgments (Ross et al, 1997 Nature), but also include erroneous mixing of feature information from the wrong location. Importantly, however, we found no feature mixing in a retinotopic task: retinotopic colors were reported veridically, with no interference from spatiotopic distractors. Thus, our apparently spatiotopic behavior is based on underlying retinotopic representations that must be updated with each saccade, and this updating is imperfect, producing systematic color misperceptions even after spatial pointers have been updated.
Meeting abstract presented at VSS 2012