Considering the heterogeneous nature of textures in real-world environments, the ability to perceptually distill heterogeneous texture information into a single feature is an essential aspect of our visual experience. For example, despite wide variation in orientation among blades within a patch of grass, we can still perceptually extract the predominant orientation. The challenge of objectively measuring this ability was addressed by developing the Texture Coherence Paradigm. Observers were presented with an array of line segments in which a percentage of lines (signal) were oriented identically and the remaining lines oriented randomly (noise). The minimum percentage of signal required to correctly determine signal orientation was taken as an observer's coherence threshold. Naïve observers required only 16 to 20% signal for correct orientation discrimination. While it seemed that coherence thresholds varied as a function of signal orientation, results showed that Global Precedence of outer texture patch contours mediated the orientation anisotropies. This suggests that larger scales of analysis, i.e., at figural levels, have a significant effect on perception of inner local features and also that texture integration is isotropic with neutral patch
shapes. Observers required a relatively long exposure (200 ms) to reach maximal sensitivity. Taken together, this evidence suggests that a spatially integrative cooperative network of orientation analyzers mediates orientation perception in heterogeneous textures. Since integrative processes are thought to be higher level in nature, i.e., extrastriate, I presented varied textures to examine construction of featural representations. Features were inefficiently combined but focused selective attention assisted significantly. I also induced a texture orientation contrast effect, analogous to the motion aftereffect, in which sensitivity to textures was significantly changed depending on the texture previously viewed. This effect took time to develop and was relatively long lasting suggesting an attentional component. Further, the effect could be "switched off" by redirecting selective attention away from the inducer.