Perceiving the stereo depth of simple stimuli isn't simple: The case of gratings.

Document Type

Conference Document



Embargo Period



Stereoscopic depth, Disparity, Binocular vision, 1-D stimuli




Cognition and Perception | Cognitive Neuroscience


Horizontal disparities are directly linked to perceived stereo depth of two-dimensional stimuli but, surprisingly, the same does not hold for 1-D stimuli. 1-D stimuli, such as lines and gratings, have ambiguous disparity signals, an analog of the aperture problem in motion. One consequence is that the depth seen between two stimuli, one 1-D and the other 2-D, can vary with the orientation of the 1-D stimulus even if horizontal disparities remain unchanged. How relative disparities and orientations jointly affect the perceived depth between two 1-D stimuli is unknown. To determine the computation humans use, we had observers discriminate the depth order of a test grating presented in the context of a reference grating. Stimulus onsets were staggered over time. A reference grating was presented parafoveally, together with an identical fixation stimulus. After 1 second, a target grating was added to the display for 180 ms at the same eccentricity as the reference grating. Importantly, no other stimulus was available to mediate relative disparity calculations. We measured the disparity of the target grating required for the target and reference gratings to be seen at the same depth. We found that the size of this depth-matching disparity did not depend on horizontal disparity but instead was proportional to 1/cosine of the orientation difference between the two gratings. The sign of the depth-matching disparity varied with the reference grating's clockwise versus counter-clockwise orientation relative to vertical. Cyclotorsion cannot account for the results; the largest possible ocular rotation would be much too small. Instead, the relative depth seen between the gratings is what would be expected from a normalization process that resolved the stereo aperture problem by notionally rotating the context stimuli to vertical, thus defining a standard functional disparity axis for computing relative disparities.