On the detection of Gabor signals and discrimination of Gabor textures.

We have explored the detection and discrimination of Gabor signals in two different paradigms in order to determine the signals' relative importance for spatial vision. In the first experiment we determined the "detection efficiency" of five such signals presented in isolation, and our results support the conclusion that, for optimal detection, the Gaussian profile should only span (to 1/e decay values) approximately two cycles. In three subsequent experiments we have observed the discriminability of textures composed of such signals over a range of spatial frequencies, Gaussian windows and orientation differences. Our results are consistent with the detection data only insofar as the discrimination thresholds vary in a predictable way with the Gaussian window parameters. Finally, we have determined the degree to which texture discrimination, in terms of differences in both orientation and spatial frequency, can be predicted from the independent manipulations of both codes within a given Gaussian window size.