Abstract
Purpose: Gurnsey and Browse (1987) showed that the textures comprising randomly rotated Ls, embedded in a background of randomly rotated Xs are more easily detected than Xs among Ls. Using a backward masking paradigm Rubenstein and Sagi (1990) showed that this effect is attenuated when the Ls and Xs are not randomly rotated. The initial purpose of this study was to assess whether discrimination asymmetries were indeed variability dependent.
Methods: Textures comprised micropatterns on a continuum from perfect Ls to perfect Xs allowing measurement of threshold differences in micropattern properties without a backward mask. Subjects were asked to detect L-type textures in X-type textures (and vice versa). In one condition the micropatterns were randomly oriented, and in another they were uniformly oriented. Texture scale was also manipulated.
Results. In contrast to the results of Rubenstein and Sagi (1990) discrimination asymmetries were found in all conditions tested. Surprisingly, the standard asymmetry was found when micropatterns were small but the asymmetry reversed when the micropatterns were large.
Conclusions: In contrast to the arguments of Rubenstein and Sagi (1990), orientation variability is not a necessary condition for texture discrimination asymmetries. Line terminators can also influence discrimination as noticed Rubenstein and Sagi (1996) and Julesz (1981). Furthermore, the scale of the textures appears to change the representation upon which discrimination depends.