Abstract
Stereomotion perception relies on two primary sources of information: interocular velocity differences (IOVD) and changes in disparity over time (CDOT). IOVD results from comparison of two monocular flow fields (MFF). Binocular disparity results from comparison of two monocular image structures and CDOT characterizes how this disparity evolves. CDOT-based stereomotion perception has been theorized to be a slower process than IOVD-based perception. If so, CDOT should be an inferior source of information for fast-moving objects. If CDOT is inferior for fast objects, is there a spatiotemporal domain in which CDOT is superior? Ten participants were each presented with three types of displays that depicted two squares approaching from different distances in depth at different constant velocities yielding different times-to-contact (TTC). Squares disappeared during approach and participants specified which square would have contacted them first. One display type was dynamic random-dot stereograms that isolated CDOT by rerandomizing points each frame (CDOT-only). Another was evolving (i.e., without rerandomization) random-dot configurations (MFF+IOVD). Corresponding points were not used, as they are in stereograms, so disparity did not specify approach. The third was evolving random-dot stereograms, so motion was defined by MFF, IOVD, and CDOT (COMBINED). For all three displays, in half of the trials, the two squares moved at speeds ranging 26–32 cm/s. In the other half, speeds ranged 73–127 cm/s. For fast stimuli, performance as measured by proportion correct was comparable for MFF+IOVD and COMBINED trials, but CDOT-only trials were significantly worse. Performance with slow stimuli was comparable for CDOT-only and COMBINED trials, but MFF+IOVD trials were significantly worse. Optimal performance levels were similar for both speed conditions. To yield invariant performance levels across speeds, the visual system appears to use primarily CDOT to perceive motion of slower objects, but MFF and IOVD for faster objects.
Meeting abstract presented at VSS 2015