Abstract
The visual system solves the aperture problem by means of integrating local 1-dimensional motion signals into a global 2-dimensional motion. Although it is well known that motion pooling occurs within first-order (luminance-based) motion or within second-order (non-luminance-based) motion of the same type, whether it generally occurs across different motion types in a feature-invariant manner remains a matter of controversy in the past plaid literature. Furthermore, this issue has neither tested with objective performance measures, nor under the condition where local 1-dimensional motion signals are cleanly separated in space with no possibility of local artifacts. Here we challenged this problem by measuring direction-discrimination performance of a four-aperture motion stimulus. The stimulus consisted of four oscillating bars, simulating a situation where the contour of a 12.8 x 12.8 deg diamond was translating along a circular path, and seen through four Gaussian apertures (SD: 1.07 deg), each located at the center of an edge. The attribute that defined each oscillating bar was either luminance, temporal frequency of dynamic random dots, or binocular disparity of dynamic random-dot stereogram. The results indicate that observers could judge the direction of global circular translation (clockwise or anti-clockwise) not only when all the edges were defined by the same attribute, but also when adjacent bars were defined by different attributes, although the attribute had some influence on the performance of direction discrimination and the quality of perceived global motion. In addition, motion pooling between first-order and second-order motions was possible even when first-order motion that did not contain obvious positional shift of features that might be detected by second-order mechanism. These results indicate that second-order motion signals do contribute the solution of aperture problem either solely or in corporate with first-order signals.