Abstract

We used random-dot cinematograms, composed of a distribution of direction vectors ranging over 90°, to measure direction discrimination for global flow. Observers can discriminate a change in mean direction of 34° for these displays. We created three displays with readily discriminable luminances (d′ = 2.0 between luminances). Increasing the luminance of all the dots in the display over this range had no effect on direction discrimination. Even randomly assigning these three luminances in each frame to dots within a single display had no effect on discrimination. However, if luminances were assigned to particular sets of directions, e.g., rightward vectors were bright, leftward vectors were dim, then the perceived direction of flow was profoundly affected. Was the motion system matching dots on the basis of luminance or on the basis of their spatial-temporal proximity? Using the randomized-luminance displays again, we set two of the three luminances to zero, plotting dots whenever they were assigned the remaining luminance value, increasing on average the spatial–temporal distances between dots compared to the displays containing all three luminances. This manipulation significantly degraded direction discrimination, suggesting that in the randomized-luminance displays, matches were made across luminances—a result expected if motion detectors respond to any detectable stimuli falling within their spatial–temporal window.

© 1992 Optical Society of America