Abstract
Disparity-selective neurons have been reported in many areas of primate visual cortex; however, the circuitry of depth perception, in particular the areas that reflect stereopsis rather than more basic disparity mechanisms, remain elusive. Here we explored this proposed functional specialization in humans, by measuring fMRI activity related to depth judgments made at coarse and fine disparities.
Two horizontally aligned 8° × 8° dynamic RDSs were presented at 0°, 0.1°, 0.35° or 0.7° pedestal disparities and symmetrically arranged in depth around fixation. Additional depth was added to each plane with a sinusoidal profile (amplitude 0.2°, which was increased in one plane and decreased in the other by Δx/2). Subjects indicated which sinusoid had the greater amplitude, whilst maintaining central fixation. Consistent discrimination performance was achieved by adjusting the value of Δx. RDSs at zero disparity were used as the baseline. Functional and anatomical data were acquired in a 3T scanner at 2.2×2.2×2.5 mm and 1×1×1 mm resolutions, respectively. Visual areas were defined with standard retinotopic mapping.
The BOLD response increased linearly with pedestal disparity in V1, V2, hV4 and V7, but followed an inverted-U function in other areas. Further analysis of the pooled response from early (V1, V2 and V3), ventral (hV4, LO1 and LO2) and dorsal (V3A, V7 and MT) areas revealed a greater activation to nonzero than to zero pedestal disparity in the dorsal region. The response in the IPS also followed this pattern. By contrast, activation in early and ventral visual cortex did not change across pedestals. These results suggest that: (a) there are differences in sensitivity to fine and coarse disparity in the dorsal and ventral visual pathways; (b) dorsal visual areas are more strongly engaged when the task requires judgments between different values of coarse disparities, a result consistent with recent single-unit studies.
This research was supported by the MRC. HB is a Royal Society University Research Fellow.