Abstract
Longitudinal chromatic aberration (LCA) contributes to retinal-image degradation by introducing chromatic blur, which is strongest for short wavelengths of light. By blurring short wavelengths more than longer ones, LCA can cause an external stimulus designed to modulate only one cone type (e.g., (S)hort-wavelength cones) to produce a retinal stimulus that modulates multiple cones instead. We combine a model of LCA (Marimont & Wandell, 1994) with a cardinal model of chromatic detection to account for two results that were obtained with nominally S-cone isolating Gabor patches, constructed using the silent-substitution method (Estévez & Spekreijse, 1982): 1. Observers noted that Gabors appeared colorful (violet/greenish-yellow) at low spatial frequencies (SFs) and achromatic at higher SFs. 2. Forced-choice detection thresholds produced contrast sensitivity functions (CSFs) with two distinct branches: at low SFs the shape matched method-of-adjustment S-cone hue sensitivity curves, but at higher SFs the shape matched forced-choice luminance CSFs. For some observers, a dip in detection sensitivity occurs between these two branches. Calculations of the retinal-images produced by LCA show that our nominally S-cone isolating gratings contained (L)ong and (M)edium- wavelength cone contrasts that increased in magnitude with SF, relative to S-cone contrasts. These retinal-image cone contrasts were used as input to three cardinal, cone-opponent detection mechanisms: YB (S-cones opposed to LM), RG (L-M), and ID (“increment/decrement” or achromatic), using cone contrast weights compiled in Eskew, McLellan, & Giulianini (1999). The YB and ID response curves intersect around 2-3 cycles/degree, aligning with the sensitivity ‘dip’ in the S-cone detection CSFs. The dip results from a phase-reversal in the retinal-image cone contrasts produced by LCA (“spurious resolution”). We can account for the shapes of the S-cone detection CSFs, and the change in color appearance at threshold, using the calculated chromatic mechanism response curves.