Abstract
Purpose While the early stages of motion detection are selective for contrast polarity, this selectivity does not appear to be maintained at higher levels of processing (Edwards & Badcock, 1994; Vis Res 34, 2849–58.). This may reflect the fact that opposite polarities will often need to be integrated in the interpretation of optic flow. Indeed, the distributions of velocities carried by opposite polarities are broadly overlapping in global motion stimuli such as those used in the above study, suggesting that they relate to the same underlying motion. The current study investigated whether segmentation by contrast polarity might improve motion perception in other situations, using transparent motion stimuli in which the distributions of velocities carried by opposite polarities clearly differed. Method On each trial, two random dot kinematograms (RDKs) were presented sequentially. In one, the directions of motion of the dots were drawn from a single Gaussian distribution, while in the other the directions were drawn from two distributions with different means. The observer's task was to decide which interval contained two distributions of directions. The minimum difference between the means of the two distributions for which this could be accomplished was measured in two conditions: (i) all dots had the same polarity and (ii) the dots in the two distributions were presented in opposite polarities. Results Transparency thresholds were lower when the two distributions were presented in opposite polarities, suggesting that the independent processing of opposite polarities is maintained when the two signals carry distinct motion information. Conclusions The integration of opposite contrast polarities in the perception of global motion is not obligatory. Rather, both integration and segmentation of opposite polarities may both occur, according to task demands.