Banks M S, Geisler W S, Bennett P J
School of Optometry, University of California, Berkeley 94720.
Vision Res. 1987;27(11):1915-24. doi: 10.1016/0042-6989(87)90057-5.
We examined the extent to which pre-neural factors constrain the detectability of sinusoidal gratings of different spatial frequencies and luminances. Contrast sensitivity functions were measured in two observers for foveally-presented grating patches. Spatial extent of the patches was inversely proportional to grating frequency. The observers' contrast sensitivity functions were then compared to the performance of an ideal discriminator (Geisler, 1984) which incorporated the effects of quantal fluctuations, optical transfer, ocular media transmittance, and the aperture, quantum efficiency, and spatial distribution of foveal photoreceptors. The sensitivity of the ideal discriminator was roughly 20-fold greater than that of the human observers, but the shapes of the ideal and human CSFs were quite similar from 5 to 40 c/deg and from 3.4 to 340 cd/m2. The similarity of shapes demonstrates that the high-frequency rolloff of the foveal CSF for gratings with a fixed number of cycles can be explained by the operation of pre-neural factors alone. Previous research has shown that grating summation area is inversely proportional to the square of spatial frequency. Thus, for gratings with fixed spatial extent the high-frequency rolloff can be explained by the pre-neural factors plus variations in grating summation area. These conclusions imply in turn that the neural transfer function is much flatter than previously thought and that private line connections from foveal photoreceptors to higher visual centers are common.
