Abstract
To elucidate spatial frequency tuning of stereoscopic mechanisms, two research groups have measured disparity threshold functions, and used factor analytic techniques to estimate the number and tuning underlying stereo channels. One group measured separate thresholds for horizontal and vertical sinusoidal corrugations embedded in dots (0.1 to 1.6 cpd), for thirty observers (Read et al., 2016, JOV; Peterzell et al., 2017, Vision Research). The other obtained thresholds by having participants discriminate between 45° and 135° oblique sinusoidal corrugations embedded in carriers composed of 2-D fractal noise (0.24 to 2.39 cpd), for sixty-one observers (Reynaud & Hess, 2017, Frontiers. Comp. Neurosci.). Both found two significant factors, with rotated loadings tuned to narrow ranges of spatial frequency. However, the factors obtained by the two studies did not match when spatial frequency was classically defined in terms of retinal size, or cycles per degree of visual angle. The differences may reflect differences in methods, different mechanisms, or other procedural differences. However, when the two studies' factors were replotted in cycles per object (where "object" refers to the square aperture containing stimuli for both eyes, hence cycles relative to the object), the high frequency factor from the first study aligned with the low frequency factor from the second. When the results of the two studies are viewed together, using the relative measure of spatial frequency (cycles per object), the two studies may provide evidence for three factors spanning 1-50 cycles. Thus we tentatively hypothesize that cycles per object, rather than cycles per degree, could be the appropriate metric for describing some spatial-frequency-tuned disparity-sensitive processes. These findings lead to the possibility that the spatial frequency tuned mechanisms underlying variability are size-constant mechanisms, similar to those found for, e.g., faces. As these results are preliminary and tentative, additional studies are necessary to address the cycles per object hypothesis.
Meeting abstract presented at VSS 2018