Detecting shape deformation of moving patterns

This study measured thresholds for the discrimination of rigidly and nonridgidly rotating patterns in two dimensions. The stimuli employed were closed contours created by the sum of two 'radial frequency' components and sensitivity to their deformation was measured as a function of the difference in the angular velocities of the components. Results show that thresholds do not depend on the specific shape of the pattern. To quantify the influence of local computations versus global pooling, thresholds were measured with parts of the pattern covered by (invisible) pie-shaped apertures. One finds thresholds are not simply a function of the total amount of pattern visible but exhibit a dependence on the number of apertures. Moreover, sensitivity to deformation could neither be fully explained on the basis of local computations nor by linear global summation. A simultaneous masking paradigm was employed to elucidate potential mechanisms involved in the computation of deformation. While 1D masks (horizontal gratings and translating random dots) only marginally elevate thresholds, rotating and expanding motion significantly impairs sensitivity. This indicates that detectors tuned to radial and circular motion are involved in the computation of shape deformation.