Abstract
The visual system could determine the 3D shape of an object in one of two ways: 1) by comparing local disparity signals or 2) by coding disparity gradients directly with higher-level 3D shape detectors. If 3D shape detectors are used, then for equivalent disparities, observers should have lower stereoacuity thresholds between shapes than within shapes. Two psychophysical tasks support this hypothesis. Stimuli were 3D surfaces whose top and bottom edges lay in the fixation plane and whose centers contained crossed disparity. Surface disparity was signaled by disparate luminance contours at the right and left edges or by disparate random dots on the surface (density = 5%). Surface disparity was generated from the equation y=axb, where a determines the total disparity range from back edge to peak and b determines the surface shape (b values of 1, 2 and 3 generated surfaces that were pointy, curved, and boxy, respectively). In the disparity interval task, all stimuli had the same surface shape (b value) and the total disparity range (a value) was varied. Observers judged whether a given stimulus was more or less curved (or pointy or boxy, in separate conditions) than the average surface. This task served as a baseline for measuring the stereoacuity of local detectors. In the shape task, all stimuli had the same disparity range (a value) and differed in shape (b value). Observers judged whether test shapes were more pointy or more boxy than the curved standard. If a difference between local disparity signals is used, then thresholds for the two tasks should be equivalent. This was not found. Thresholds were lower by a factor of 1.3 – 3.4 in the shape task, indicating better disparity sensitivity for 3D shapes than for local disparity differences. Thresholds for random dot stimuli were slightly higher than those for contour stimuli in all tasks. These outcomes offer direct psychophysical evidence for 3D shape detectors.
This work was supported by a grant from the UW Oshkosh Faculty Development Program