Abstract
Neurons that modulate their firing rate in response to binocular disparity have been observed in many visual cortical areas and are thought to form the neural substrate of stereopsis. V1 neurons which show disparity selectivity for random-dot stereograms (RDS), also modulate their firing rate in response to disparity changes in anti-correlated RDS. Disparity-tuning is often inverted as predicted by binocular energy models. However, the way that image correlation affects our perception of stereo depth is unclear, because anti-correlated stimuli give rise to either weak or no perception of depth. To gain further insight into the relation between activation of binocular cells in V1 and depth perception, we employed the Panum's limiting case (PLC) surface invented by Braddick. Here a horizontal dot pair in one eye matches a single dot in the other. One sees a pair of transparent surfaces with a depth difference proportional to the separation in the monocular pair. In the variant employed here, one element of the monocular pair has reversed constrast (e.g. black). We find that this gives rise to a reversal of depth sign of the reverse-contrast dots. Observers reported whether the reverse-contrast (black) dots were near, far, or unclear (the matching white dot surface appeared in the fixation plane). Tests were conducted with dynamic dots in 2 s trials. The results show a clear reversal of depth for the opposite-contrast dots. However, there are two important observations: 1. The depth of the opposite contrast dots did not vary with dot offset; and 2. The dots appear to form an amorphous cloud rather than a surface. Both observations contrast with same-contrast PLC-RDS results. We postulate that under these conditions the post-V1 mechanisms that integrate and reconcile the outputs of V1 disparity mechanisms fail. Hence our inability to see a coherent surface or quantitative depth. Our perception corresponds qualitatively (up to depth sign) to the incoherent output of V1 disparity cells.