Integration of local features into a global shape

It is a well-established fact that the cortical filters selective for spatial frequency and orientation have a limited spatial extent. The present study investigated how information from local filters is integrated into global shapes. Specifically, we were interested in whether identification of a global pattern consisting of small oriented, spatially-bandpass features (Gabor patches) depends on the orientations of those features. The observer was presented with a C-like stimulus shape comprised of oriented Gabor elements on a blank background, and the performance measure was the threshold contrast for identifying the global orientation of the C-shape (four possible rotated orientations). Performance was significantly enhanced (i.e. contrast thresholds were lower) when the local features were aligned with the C-contour than when the features were orthogonal to it or they had mixed local orientations (randomly both aligned and orthogonal). Control experiments showed that the alignment of local features improves global pattern perception only when the features are close to each other. In contrast, a close spacing between the features interfered with global pattern perception when the local orientations were mixed.