hide
Free keywords:
-
Abstract:
Human observers can discriminate depths of surfaces in 3-dimensional (3D) random-dot stereogram (RDS) scenes. In anti-correlated RDSs, a black dot in one eye corresponds to a white dot in the other eye, making V1 neurons respond as if the depth order between surfaces is reversed from that defined by binocular disparities (Cummings & Parker 1997). Zhaoping & Ackermann (2018) found that peripheral but not central vision can perceive this reversed depth according to V1 responses, when observers reported whether a disk was in front of a surrounding ring. They explained this finding in terms of a Central-peripheral Dichotomy (CPD) theory, proposing that feedback from higher to lower visual areas (e.g., V1) mainly targets central, rather than peripheral, visual fields and aids recognition by verifying potential perceptual outcomes using analysis-by-synthesis. Ac- cordingly, the reversed depth illusion from misleading V1 responses is vetoed in central vision by the verification process. Feedback verification is weaker in peripheral vision, and so the illusion is perceived. The verification process in central vision can be compromised by presenting difficult visual inputs briefly, followed by a mask — backward masking. This replaces the original details by the mask when feedback verification occurs, making verification difficult (DiLollo et al 2000). If the reversed depth illusion in central vision is indeed normally vetoed, then backward masking should make the illusion visible. We tested this prediction by making the RDS dynamic: the random set of dots in the RDS was replaced every 10 milliseconds by another random set while keeping the 3D scene unchanged. Each replacement random set acts as a backward mask for the previous random set. Indeed, in central vision, observers see the reversed depth illusion better in the dynamic than static RDSs. This finding supports the analysis-by-synthesis computation in the feedback process, and the CPD theory in particular.
