Visual discrimination of texture

Julesz (1975) proposed a theory of texture discrimination, based on an order statistics principle, which states that no two textures can be perceptually discriminated if they have identical second-order statistics. The experiments reported here demonstrate that this principle is not adequate to predict visual texture discriminability. Both letter and dot micropatterns were used to create texture pairs that either differed or were identical with respect to secondorder statistics. The subject’s task was to decide which quadrant of an array contained a disparate texture. In Experiments 1, 2, and 3, when controlling for spatial overlap, texture pairs having identical second-order statistics were discriminated more quickly than similar texture pairs having different second-order statistics, in contradiction to the principle. Although a significant effect in the direction predicted by the order statistics principle was found in Experiment 4 for texture pairs created from the dot micropatterns, other factors, such as spatial overlap, border differences, and goodness of pattern, must also be considered in predicting texture discriminability.