Abstract
In a random dot stereogram (RDS), object surfaces in a three-dimensional scene are generated by images presented to left and right eyes that comprise interocularly corresponding random black and white dots. The spatial disparities between the corresponding dots determine the depth of the object surfaces. When a black dot in one monocular image corresponds to a (contrast-polarity reversed) white dot in the other, the RDS is called anti-correlated. Neurons in the primary visual cortex (V1) respond to such RDSs as if their preferred disparities become non-proferred and vice versa, thereby reversing the disparity signs reported to higher visual areas. Humans can perceive such reversed depths in peripheral but not in central vision (Zhaoping & Ackermann 2018). According to the central-peripheral dichotomy in recognition (Zhaoping 2017, 2019), V1 signals are fed forward to higher visual areas to suggest initial perceptual hypotheses about scenes; in central but not in peripheral vision, top-down feedback from higher to lower visual cortical areas aid recognition using analysis-by-synthsis, so that the would-be visual input for each hypothesis (e.g., a depth order) is fed back to V1 to verify whether it matches the actual visual input, and the hypothesis yielding a good or poor match is boosted or suppressed, respectively, this is termed the Feedforward-Feedback-Verify-reWeight (FFVW) process for recognition. Accordingly, V1's reversed depth signals are perceived peripherally, but not centrally since, by FFVW, they disagree with the would-be inputs expected by the brain from past experience. In this work, I show that the typically invisible reversed depth signals can influence depth perception in central vision in RDSs containing polarity-matched dots, polarity-reversed dots, and noise dots which are inter-ocularly independent. To report whether a central disk is in front or behind the surrounding ring in a noisy RDS, observers can see the defined depth order by the polarity-matched dots more (or less) clearly when the depth order in the reversed depth signals by the polarity-reversed dots is the same as, or opposite to, the defined depth order. Accordingly, the threshold signal-to-noise level needed in a noisy RDS to see depth can be lowered or raised respectively by adding congruent or incongruent reversed depth signals. The reversed depth signals are more perceptually more influential in dynamic RDSs with brief stereo-frames to make top-down feedback in FFVW less effective. Hence, the feedforward reversed depth signals from V1 can influence percept in central vision when the feedback veto can be reduced by brief inputs or mitigated by normal depth signals.
