Abstract
Anticorrelated random-dot stereograms (aRDS) are stimuli comprised of a binocular pair of images of randomly positioned dots in which the dots in one eye’s view are replaced by their photographic negatives. In some cases, the binocular energy model of neural responses predicts that the perception of depth should be reversed for aRDS. In contrast, the responses from second-order channels predict the perception of depth in the forward direction (Asher & Hibbard, 2008). Findings of reversed-depth perception are inconsistent, and highly dependent on the type of stimulus, its location and the duration of presentation. Zhaoping & Ackermann (2018) showed that the perception of reversed depth is more robust in the periphery compared to centrally presented targets, and argued that this reflects a reduction in top-down feedback in the periphery. We assessed the effects of stimulus eccentricity (0 to 15.5 degrees) and duration (80 or 700ms) on the perception of depth from correlated and anticorrelated RDS. Stimuli were either circular targets a horizontal depth edge. Observers indicated whether the target appeared near or far, and also whether they were confident in this judgement. For correlated stimuli, depth judgments were accurate at all eccentricities, but confidence fell for peripheral stimuli. For anticorrelated stimuli at the short presentation time, no depth was seen for the circular stimuli, and forward depth was seen for the horizontal edge stimuli. For long presentations, reversed depth was seen for circular stimuli, and forward depth for edge stimuli, at all eccentricities. Confidence fell to zero for stimuli presented at an eccentricity of 15.5 or beyond. Reverse depth in aRDS is possible with extended presentation times, but only for the circle condition for which second-order responses are unreliable. For peripherally presented random dot stereograms, reliable depth judgments were found in the absence of awareness of depth.