Stereo and motion disparities interfere with positional averaging.

If required to judge the spatial separation between features that are sub-parts of larger objects, observers are biased in the direction of reporting the separation between the centroids of the objects. We predicted that this bias would be overcome by separating the target features from the larger objects, either by stereoscopic disparity or by relative motion. We measured the observer's perceptual biases for the "cluster illusion" with the target dots specified by: (1) contrast alone, (2) contrast and disparity, (3) disparity alone, (4) contrast and motion, and (5) motion alone. Disparity had the effect predicted, but the effects of motion were less clear.