Abstract
The visual system combines slant cues to form a single percept in a fashion that approaches statistical optimality. We asked whether in combining cues the system fuses the cues such that access to the single-cue estimates is lost. Previous work on this issue observed fusion for disparity and texture using a 3-interval oddity task. We wondered whether the 3-interval task made it too difficult for the observer to retain single-cue estimates. To test this, we used a simpler 2-interval task that reduced the memory load. Observers were presented two slanted planar stimuli in succession; one with a conflict between the disparity and texture, and one without. The conflict interval was a two-cue stimulus (disparity inconsistent with texture). The no-conflict interval was one of three types of stimuli: a single-cue stimulus (disparity only or texture only) or a two-cue, no-conflict stimulus (disparity consistent with texture). Disparity-only stimuli were sparse random-dot stereograms. Texture-only stimuli were monocular Voronoi patterns. Two-cue stimuli were binocular Voronoi patterns. On each trial, observers indicated the interval with greater slant. A staircase procedure adjusted the slant of the no-conflict stimulus until the two stimuli had equal apparent slants. If disparity and texture were fused (such that the individual estimates were lost), the results with the three types of no-conflict stimuli would be the same. If the two cues were not fused, the results with the three types would differ because the observer could use the cue(s) available in the no-conflict interval to do the match with the conflict interval. We observed both partial and full fusion.