Abstract
In a random-dot stereogram (RDS), depth is by images presented to left and right eyes that comprise interocularly corresponding random black and white dots. The spatial disparities between the binocularly corresponding dots determine the object depths. If the dots are contrast-reversed, such that a black dot in one eye corresponds to a white dot in the other, disparity-tuned neurons in the primary visual cortex (V1) respond as if their preferred disparities become non-preferred and vice versa, reversing the disparity sign reported to higher visual areas. Humans cannot perceive the reversed depth in central vision. We demonstrate that, in central vision, reversed depth signals can augment or degrade depth perception in noisy RDSs when contrast-reversed and contrast-matched dots are mixed. When the reversed depth signals and the normal depth signals from the (contrast-matched dots) are congruent with each other, augmentation occurs; when they are incongruent with each other and when the RDS images are sufficiently brief, degradation occurs. These findings are consistent with the FFVW process, its feedback-Verify-reWeight component disambiguates noisy and ambiguous inputs from V1. Through an analysis-by-synthesis computation, this top-down feedback vetoes misleading V1 signals, including, in certain situations, the reversed depth signals. When the RDS is viewed too briefly to allow time for feedback, augmentation or degradation works by adding, respectively, the congruent or incongruent, reversed signals to the normal signals in the feedforward direction. With a sufficient viewing duration for effective feedback, the feedback veto removes the degradation on the percept of the normal signals when the reversed and normal signals disagree, or, when the normal depth signals are absent, makes the reversed depth imperceptible. Meanwhile, the analysis-by-synthesis in the feedback process also fills in or corrects imperfect signals, allowing the reversed signals, when congruent with the normal depth signals, to augment the percept of the latter.