Abstract
Binocular disparity provides the human visual system with estimates of both the three-dimensional shape of surfaces, and of the depth between them. Additionally, binocular disparity cues provide a compelling sense of the volume occupied by objects in space. However, studies of stereoscopic vision have tended to examine the perceived depth of isolated points, or the perceived structure of surfaces in depth, without addressing the associated sense of volume. Comparatively little is known about how the visual system represents stereoscopic volumes. The experiments reported here address this issue by examining observers’ ability to judge changes in the range and distribution of disparity-defined volumes of dots. Observers were presented with Gaussian distributed random-dot volumes in a three interval, odd-one-out detection task. Each interval was presented for 200ms, and contained a stereoscopic volume within an area of 4.8 x 4.8 degrees. In two (standard) intervals, dot disparities were drawn from a Gaussian distribution of fixed standard deviation 1.1arcmin. In the third (target) interval a proportion of dot disparities were drawn from a uniform distribution with a range of between ±1.1arcmin and ± 7.7arcmin, with the remaining dots drawn from the same Gaussian distribution as the standard intervals. For some ranges, an entirely uniform distribution could not be distinguished from the Gaussian standards. Instead, the ability to detect the odd interval depended largely on the range of the distribution, not its shape. Changes in dot density, and in the standard deviation of the Gaussian distribution did not lead to a general sensitivity for distribution shape, but instead resulted in changes to the range of uniform distributions where the target interval could not be reliably detected. Our results suggest that the visual system makes use of an impoverished representation of the structure of stereoscopic volumes.
Meeting abstract presented at VSS 2012