Abstract
Relative stereo slant is better perceived than absolute slant, especially for vertical axis slant (Gillam,(P&P 1984,JEP 1988). This is attributed either to the unambiguity of relative slant (van Ee et al, Perception 1999) the presence of second derivatives of disparity (Howard and Rogers,1995) or the presence of a gradient of relative disparity (Gillam,1984,88).Slant contrast in the unslanted surface is attributed by Pierce and Howard (Perception,1997),van Ee and Erkelens (Vis Res,1996) and others to normalisation of the slanted surface while relative slant is maintained. To determine (a) the basis of relative slant effectiveness and (b) the adequacy of the normalisation explanation of contrast we measured stereo slant (using a monocular comparator) for each of two equal-sized textured surfaces separated or overlapped vertically or horizontally and with one surface subject to stereo slant. For V axis slant perceived relative slant was weak for horizontally separated surfaces but strong (with contrast) for horizontally overlapping and vertically separated surfaces. Thus only conditions providing gradients of relative disparity produced strong relative slant, ruling out other explanations. Contrast in the V axis (but not H axis) case appeared to involve repulsion. A second experiment confirmed this using fully overlapping surfaces with a fixed relative stereo slant of 45 deg. Absolute slant for each surface varied from 45 deg through 22.5 deg to zero. For H axis slant perceived relative slant was accurate and perceived absolute slant was always greater for the more slanted surface. For V axis slant relative slant was overestimated and perceived absolute slant was close to equal for the two surfaces regardless of the actual absolute slants present. Implications of repulsion and contrast anisotropy for stereo theory will be discussed.