Curvature population coding for complex shapes in human vision.

In the primate visual system relatively complex patterns such as curved shapes are first represented at intermediate levels of the ventral pathway. Furthermore, there is now evidence for the existence of curvature population coding in primate V4. We sought to determine whether similar encoding occurs in the human visual system by using a context-dependent lateral masking paradigm. In this paradigm a central closed contour comprising the test pattern is masked by surrounding larger or smaller patterns with various configurations. Results indicate that test thresholds are not affected by a circular control mask, and that elevations are greatest when curvature extrema of the mask are aligned with those of the target. These lateral interactions extend over greater than 1 degrees and are tuned for target shape. Masking increases with the number of local curvature extrema aligned with the target. Finally, masking persists when target and mask have orthogonal local orientations and increases with mask amplitude. These findings are incompatible with local orientation-selective interactions (V1-mediated) but are consistent with the existence of population codes based on curvature maxima at intermediate levels of processing (presumably V4) in human vision. The paradigm we introduce provides a new tool for evaluating the representation of complex percepts.