Abstract
Olzak & Thomas (1999) described linear-nonlinear-linear processing cascades specialized to process information about 2D spatial frequency information (texture). The nonlinear gain control pool and linear summing circuits are unlimited with respect to orientation, but limited in the extent to which they operate over a range of spatial frequencies. When plotted in polar coordinates, the summing circuits therefore appear as a series of concentric rings, or doughnuts. In the current study, we measured the spatial frequency bandwidths of the gain control pool and of the summing circuits. The task was a 2-alternative spatial frequency discrimination. The stimuli were simple or complex gratings with one component always appearing near 3 or near 15 cpd. The second component varied in its base frequency to measure bandwidths. Gain control pools were isolated by using the second component as fixed mask that did not differ between the two stimuli to be discriminated and performance was compared to single-component controls. Summing circuits were isolated using a configural effect test, which compared performance between two two-cue conditions that differed only in how the cues to discrimination were configured. The results indicated that the spatial frequency bandwidth of the gain control pool is quite broad with respect to spatial frequency, estimated to 2.3 octaves (half-height, full bandwidth) by fitting a Gaussian to the data. The summing circuits, however, were quire narrow, estimated to be only 0.35 octaves. Some asymmetries were observed.
OlzakL. A.ThomasJ. P. (1999). Neural recoding in human pattern vision: Model and mechanisms. Vision Research, 39, 231–256.
Supported by NEI grant EY13953.