At Least at the Level of Inferior Temporal Cortex, the Stereo Correspondence Problem Is Solved

Stereoscopic vision requires the correspondence problem to be solved, i.e., discarding “false” matches between images of the two eyes, while keeping correct ones. To advance our understanding of the underlying neuronal mechanisms, we compared single neuron responses to correlated and anticorrelated random dot stereograms (RDSs). Inferior temporal neurons, which respond selectively to disparity-defined three-dimensional shapes, showed robust selectivity for correlated RDSs portraying concave or convex surfaces, but unlike neurons in areas V1, MT/V5, and MST, were not selective for anticorrelated RDSs. These results show that the correspondence problem is solved at least in far extrastriate cortex, as it is in the monkey's perception.