Abstract
Disparity-defined surfaces with the same physical slant often differ in perceived slant depending on their tilt: Surfaces with tilt 90 (rotated about horizontal axis) appear more slanted than surfaces with tilt 0 (rotated about vertical axis). This slant anisotropy is also observed with disparity-defined sinusoidal corrugations: disparity threshold is lower for low-spatial frequency corrugations when they are horizontal than when they are vertical (Bradshaw & Rogers, 1999). Slant anisotropy is not observed with real surfaces in which other depth cues are consistent with the disparity-defined surface (Bradshaw et al., 2002). This suggests that slant anisotropy is caused by the cue conflicts present in virtual disparity-defined surfaces and/or by differences in the slant prior for tilts 0 and 90. To examine these possibilities, we first made all other depth cues unreliable, thereby reducing their influence. We then varied the reliability of the disparity signal by blurring the dots anisotropically in a random-dot stereogram. Blurring the dots horizontally makes the calculation of horizontal disparity unreliable. Blurring vertically does not affect the calculation of horizontal disparity. In each two-interval trial, observers indicated whether a horizontal or vertical corrugation had more apparent depth. We varied the amplitude of the vertical corrugation to match the apparent amplitude in the horizontal corrugation. When the disparity signal was relatively unreliable (horizontally blurred dots), the vertical corrugation needed greater amplitude to have the same apparent depth as the horizontal corrugation: slant anisotropy was observed. However, when the disparity signal was relatively reliable (vertically blurred dots), the amplitude difference decreased substantially: slant anisotropy was diminished. Thus, the reliability of the horizontal-disparity signal strongly influences slant anisotropy. Our data are consistent with the hypothesis that the slant priors for tilts 0 and 90 differ, the former being more peaked at 0 deg.