Computation Times of Binocular Depth Analysed by the "Delayed Stereopsis Illusion" (DSI)

Viewed through depth-reversing spectacles, non-transparent objects appear to cut gaps into a patterned background. In moving objects this gap is seen to extend beyond the occluded area (DSI): Its trailing edge appears to lag behind by a certain distance that can be measured to determine the processing time to accomplish stereopsis. The delay was quantified by our subjects by marking the trailing edge of the gap with a laser pointer. Dependent on the experimental conditions it amounts from 40 up to 125 ms. Why is this delay not perceived in normal stereopsis? If an object is moving before a background, the background usually maintains its position, it may be occluded, or not. Depth information thus might be extrapolated to the continuously uncovered regions of the patterned background. In depth-reversed vision, however, occlusion demands that the region of the background which is momentarily covered must jump behind the moving opaque object. As this object is perceived to retain its distance, the background, as it is getting uncovered, must jump back into the foreground where it can be perceived only after renewed calculation of binocular depth. In his random dot stereograms Julesz (Dialogues on Perception. MIT Press, Cambridge, 1994) coined the term “no-man`s-land” to designate those regions at both sides of the floating object where stereoptic depth information is missing, as they are visible with one eye only (“topologically caused no-man`s-land”). Our DSI, however, is caused by what we coin a “trailing edge no-man`s-land”: Though visible with both eyes, stereoptic depth is missing there because 3D computation is not yet finished. As its texture continuously emerges at the trailing edge, this no-man`s-land is concluded to be behind the moving object. It is therefore not ascribed to any depth position defined by present binocular depth cues, in contrast to Julesz` no-man`s-land. The dependence of DSI on eye movements, disparity, velocity, motion direction, surface texture, illuminance, spatial frequency, and fractal dimension of the objects involved is currently being investigated in model systems which allow to determine processing times of human stereopsis under well-defined conditions