The Local Channel Structure of Human Stereopsis

((LL Kontsevich & CW Tyler))
Smith-Kettlewell Eye Research Institute, San Francisco

Rationale. At ARVO 96, we introduced the technique of peripheral Gaussian bars as a probe that is simultaneously local in retinal space and spatial frequency. For monocular vision, we validated the frequency selectivity implied for Gaussian bar targets by the log bandwidth structure of visual channels and low spatial frequency fall-off. The lack of side-lobes in the Gaussian bars makes them particularly suitable for the study of frequency-specific disparity tuning by a masking paradigm because there is no aliasing of the disparity signal by spurious peak concidences.

Methods. We measured local disparity tuning at a location 5º to the left of the fovea by a masking threshold paradigm, which gives a direct measure of the underlying channel tunings. Test bars were presented in a 2 sec cosine bell and mask bars were static. With the test bar of a particular width at 2X detection threshold, the contrast of the masking bar at a range of set disparities was varied to measure the mask contrast required to return the test to its detectability criterion. Masking thresholds were determined by a new minimum-entropy variant of the QUEST staircase that we have developed for optimal measurement of both the threshold and slope parameters of the psychometric function.

Results. The disparity tuning functions for zero-disparity test targets showed a remarkable ‘batwing’ form, with masking threshold increasing up to peaks at from 10 to 40 arc min (according to Gaussian bar width over a range from 4-60′), then plunging precipitously within about 10′ for larger disparities. Monocular masking by the sum of the dichoptic components showed a narrower spread matching the Gaussian bar width.

Conclusion. Local disparity masking extends over a much wider range of disparities than found in the dynamic cyclopean noise masking study of Stevenson et al (1992, VR). Disparity range is only weakly affected by test bar width, implying a disparity-specific selectivity of the mechanisms with little contamination by spread from their monocular spatial extent. The masking increase with disparity may imply transition from a fine to a coarse disparity system, a possibility that is being evaluated with fine- and coarse-disparity test targets.

Supported by NEI 7890 C5