A stimulus artefact undermines the evidence for independent ON and OFF channels in stereopsis

Abstract

Early vision proceeds through distinct ON and OFF channels, which encode luminance increments and decrements respectively. It has been argued that these channels also contribute separately to stereoscopic vision. This is based on the fact that observers perform better on a noisy disparity discrimination task when the stimulus is a random-dot pattern consisting of equal numbers of black and white dots (a "mixed-polarity stimulus", argued to activate both ON and OFF stereo channels), than when it consists of all-white or all-black dots ("samepolarity", argued to activate only one). However, it is not clear how this theory can be reconciled with our current understanding of disparity encoding. Recently, a binocular convolutional neural network was able to replicate the mixed-polarity advantage shown by human observers, even though it was based on linear filters and contained no mechanisms which would respond separately to black or white dots. Here, we show that the stimuli used in all these experiments contain a subtle artefact. The interocular correlation between left and right images is actually lower for the same-polarity stimuli than for mixed-polarity stimuli with the same amount of disparity noise applied to the dots. Since our current theories suggest stereopsis is based on a correlation-like computation in primary visual cortex, it is then unsurprising that performance was better for the mixed-polarity stimuli. We conclude that there is currently no evidence supporting separate ON and OFF channels in stereopsis.

The mixed-polarity benefit of stereopsis arises in early visual cortex

“…Alternatively, the benefit could be explained by different image statistics of mixed- and single-contrast polarity RDSs. Read and Cumming (2018) showed that stimuli used in this study and all previous studies contain higher interocular image correlation at the target disparity if they have mixed contrast polarity compared to a single contrast polarity. A more correlated input drives binocular cells in primary visual cortex more strongly and produces a more reliable binocular disparity signal.…”

Mixed-polarity random-dot stereograms alter GABA and Glx concentration in the early visual cortex

“…Depth judgements are more accurate when binocular images are composed of both light and dark features, rather than just one or the other (Harris & Parker, 1995;Read, Vaz, & Serrano-Pedraza, 2011). This finding was initially interpreted as evidence of separate ON and OFF binocular channels; however, contradictory physiological (Schiller, 1992), behavioural (Read et al, 2011) and theoretical (Goncalves & Welchman, 2017;Read & Cumming, 2018) evidence has cast doubt on this hypothesis. Here we tested the potential for changes in the balance of excitation and inhibition that are produced by viewing these single-and mixed-polarity random-dot stereograms (RDS).…”

“…Critically, this neural network also reproduced the mixedpolarity benefit, suggesting that the difference in positive/negative activity may underlie this puzzling phenomenon. Read and Cumming (2018) have proposed that the mixed-polarity benefit arises from subtle changes in image correlation under some restricted circumstances. Given that the image correlation was the same between single-and mixed-polarity conditions, the difference in neurotransmitter concentration cannot be attributed to this hypothesis.…”

“…However, the convergence of ON and OFF channels in V1 to form simple cells (Schiller, 1992) seems to contradict this as a potential explanation for the mixed-polarity benefit.Recently, Goncalves and Welchman (2017) showed that it is possible to capture the mixed-polarity benefit using a simple linear -nonlinear processing architecture that did not depend on separate ON and OFF channels. Thereafter, Read and Cumming (2018) proposed that the 'mixed polarity' benefit could arise from subtle changes in image correlation that can occur in some circumstances. As current models of binocular processing are based on cross-correlation between the left and right eyes (Ohzawa, DeAngelis, & Freeman, 1990), they do not consider image features, but instead compute the inter-ocular cross-correlation between left and right images.…”

“…1b). Read and Cumming (2018) have proposed that the mixedpolarity benefit arises from differences in the interocular cross correlation between some single-and mixed-polarity RDS. The difference in correlation arises from an interaction between the range of luminance in the stereograms and the variability of the binocular disparity.…”