Perceptual Collapse of Three-Dimensional Structure from Motion Parallax

Abstract

This study concerned how perceived depth collapses after asymptotic depth is reached and the information specifying depth is abruptly removed. The stimuli were random-dot, computer-generated three-dimensional objects and the depth information was motion parallax. Motion parallax was removed in two ways. In the first method, the depth of all object points was reassigned to zero, simulating a disk. In the second method, a rotation of the object was introduced in such a way that a degenerate case of motion parallax was produced. The results showed that judgments of depth slowly collapsed once motion parallax was removed. Over-all, judgments of depth required about the same duration to collapse as was required for the judgments to build up to asymptotic levels (about 750 msec.). Finally, depth collapsed more slowly when the motion parallax was removed by redefining the object as a disk than when removed using the other method.