Abstract
Howard and Duke (Vision Research, 2003) generated binocular displays that contained a grey transparent square offset from a vertical bar in one eye, and a vertical bar with a gap in the other eye. Observers perceived a transparent square in depth that gave rise to quantitative percepts of depth. The authors argued these displays constituted a new form of stereopsis, since they were “without conventional disparity” and that the depth experienced was due to “transparency rather than occlusion”. Although there were no vertical contours to support conventional disparity computations, one possibility they did not examine experimentally is that the depth experienced in these displays was due to the disparity of the ends of the horizontal contours rather than a new transparency-based mechanism of stereopsis. To test this possibility, we compared the results of their experiment to three new displays that contained similar horizontal contour terminations, but were inconsistent with transparency. For one of our stimuli perceived depth was indistinguishable from Howard & Dukes' results. In the second and third cases depth was seen (although somewhat less precisely) even though in one case the luminance polarity of matching contours of the disparate square was opposite in the two eyes. We conclude that transparency is not necessary to generate depth in these displays, and that the metrical depth observed in these patterns arose from disparate horizontal contour terminations (cf. Gillam, Nature, 1994). These and previous results also demonstrate that disparate contours of opposite contrast polarity can generate depth even when these contours are not consistent with occlusion.
Supported by ARC DP0211698 to B. Gillam, UNSW Vice-Chancellor's and NSERC PDF to P.M Grove