Item

Abstract

Viewed pseudoscopically, an opaque square floating above a random-dot pattern appears as a rectangular cut-out. When the pattern moves upwards, an illusory gap is perceived at the upper edge of the square. In analogy with Julesz's "no-man's-land" we called this DSI-gap "trailing-edge no-man's-land". Its width, marked by subjects under well-defined conditions, indicates the 3-D computation time needed to determine spatial depth of the pattern, which virtually appears "from nowhere". Early data suggested that in 3-D vision there are two different processing pathways for fast and slow movements (as Gegenfurtner et al, 1996, Trends in Neurosciences 19 394-401 found in normal motion analysis). This could not be confirmed. 3-D computation times rather do not seem to depend on pattern velocity. Our data show interindividual differences and indicate computation times between 50 and 80ms. In any case, the minimum presentation time of 17ms, at which Julesz´ dynamic random-dot-stereograms are just recognizable, is too short to determine the position in depth in each single frame. The vision system rather seems to check that the depth situation has not changed, and maintains the percept of the floating square. Learning, and parameters like spatial frequency, disparity between the square and the moving pattern, and brightness within the range of photopic vision did not significantly influence the perceived width of the DSI-gap. In scotopic vision, however, it was increased in accordance with the Pulfrich effect. (Supported by the Deutsche Forschungsgemeinschaft.)