Perceived direction of motion vs. detection of global flow in random dot cinematograms

Purpose. A percept of global flow can be evoked by a random dot cinematogram whose spatially-intermingled elements move in diverse directions. The direction of this global flow has been assumed to be at or near the mean direction in which dots moved. A line-element model has successfully accounted for the precision with which one can discriminate the global flows evoked by different cinematograms. We tested the ability of such a model to account for the perceived direction itself. Methods. Subjects viewed cinematograms presented for varying numbers of frames (6-20). For signal cinematograms, direction distributions varied from 0 to 225 deg. Signal cinematograms were randomly intermixed with noise cinematograms, whose direction distributions were 360 deg (evoking no global flow). The subject used a rating scale to indicate whether the cinematogram was signal or noise. The subject also used a cursor to indicate direction of global flow. Results. Discrimination of signal from noise was generally, uniformly high [P(A)> 0.8]. Despite this evidence of a clear percept of global flow, perceived direction varied dramatically across conditions. Standard deviations of perceived direction ranged from 4-5 deg (at long presentations and narrow direction distributions) to as high as 40 deg. On individual trials, discrepancies between distribution mean and perceived direction occasionally exceeded 90 deg. Conclusions. The tendency to see global flow in random dot cinematograms seems to be dissociable from the accuracy with which the direction of that flow is registered. The line element model is only partially successful in accounting for such results. We are currently extending these results to non-uniform and multi-modal distributions of direction.