Society for Neuroscience 2004

K.Umeda1; S.Tanabe2; I.Fujita1,2*
1. Grad Sch Engin. Sci., Osaka Univ, Toyonaka, Japan
2. Grad Sch Frontier Biosci., Osaka Univ, Toyonaka, Japan

Relative disparity is defined as the difference in absolute binocular disparities of two visible features in a scene. Psychophysical depth judgment relies primarily on relative disparities. A small number of macaque V2 neurons encode relative disparity but majority do not, suggesting that further processing is required to account for depth perception. Here, we presented various combinations of two planes of depth to two alert macaque monkeys and recorded responses in area V4, an area that receives a major input from V2. The visual stimulus was a circular patch of dynamic random-dot stereogram, consisting of a center disk and a surrounding annulus. We studied the effect of shifting the surrounding annulus at different disparities on sensitivity to the absolute disparity of the center disk. Tuning curves for absolute disparity of the center disk were made individually for three different surround disparities. If a neuron codes the relative disparity between the center and the surround, the disparity-tuning curves should shift along the disparity axis by an equal amount and in the same direction as the shift in the surround disparity. If a neuron codes the absolute disparity of the center, no shift should occur. We quantified such shifts by calculating a shift ratio, defined as the amount of shift in the tuning curve divided by the amount of shift in the surround disparity. The distribution of shift ratios was unimodal and biased towards the direction expected for relative-disparity coding (median=0.37, n=152), and many neurons show a shift ratio of ~1. Comparison with V2 neurons indicates that V4 neurons as a population exhibit substantially higher sensitivity for relative disparity. The results suggest that 3D structural information in V4 is compressed from being absolute-disparity-based to being relative-disparity-based.