Seeing reversed depth in contrast-reversed random-dot stereograms in central vision

In contrast-reversed random-dot stereograms (CRRDSs), a black dot in one eye matches a white dot in the other eye. Disparity tuned neurons in primary visual cortex (V1) respond to CRRDS as if reporting reversed depth. However, humans generally perceive reversed depth only in peripheral vision (Zhaoping & Ackermann 2018, Perception), because feedback from higher visual areas to V1, to aid recognition and/or veto misleading V1 signals, is stronger in central vision. When contrast-reversed and contrast-matched random dots are mixed in a random-dot stereogram (RDS) that is presented so briefly that feedback is ineffective, the reversed depth signals are not vetoed in central vision. These signals then enhance or degrade depth perception depending on their agreement with normal depth signals from the contrast-matched dots (Zhaoping 2021, Vision Research). Here, we show that observers can perceive the reversed depth of a disk surface made entirely of contrast-reversed dots in central vision, likely because feedback is sufficiently compromised or downweighted during perceptual decision making. We achieved this by backward masking the (200 millisecond) RDS and, in some trials, making the RDS dynamic by replacing the random set of dots with an independent random set every 10 millisecond (without changing the disk's disparity). In each trial, the random dots for the disk were all contrast-reversed or all contrast-matched, the RDS was static or dynamic. Observers reported the reversed depth order between this disk and a surrounding ring more frequently than chance for both dynamic and static RDSs, but far more frequently for the former