The role of chromatic contrast and luminance polarity in stereoscopic segmentation.

We have investigated whether our ability to discriminate the stereoscopic depth of random-dot targets set amongst random-depth distractors is facilitated when target and distractors differ in particular combinations of colour and luminance polarity. For flat-plane targets, stereo-thresholds were found to be lower with a target-distractor colour/luminance difference, but only when that difference enabled the target elements to be identified in the monocular image, either by virtue of being less numerous than the distractors, or because the subject knew beforehand the target's colour/luminance. If neither of these conditions prevailed, stereoscopic thresholds were no different when target and distractors were segregated by colour/luminance than if they were not. For sine-wave disparity grating stimuli, in which subjects were required to discriminate the orientation of the depth corrugations, no advantage was found when target and distractors were segregated by colour/luminance under any condition. These results suggest that segregation by colour/luminance is only beneficial to the stereoscopic processing of random-element stimuli when the task can be performed by attending to a small number of target elements. A corollary to this conclusion is that stereopsis mechanisms do not automatically pre-filter the image into different colour/luminance maps.