Critical spatial separation at the scale of V1 receptive fields determines motion segmentation

Integrating elements that belong to a single object while segregating overlapping objects is a fundamental challenge for the visual system, exemplified by the phenomenon of motion transparency. While the middle temporal (MT) cortex is central to motion processing, the role of V1 in motion transparency remains controversial. It is still unclear at what spatial scale segmentation for motion transparency occurs. To address these questions, we conducted human psychophysics experiments using locally paired-dot stimuli moving in two directions separated by 90°. Subjects performed a 3AFC task to report whether the visual stimulus had no motion, a single direction, or two distinct directions. We systematically manipulated the path length of the dots, and therefore, the spatial separation between the paired dots, and the retinal eccentricity of the visual stimulus. We found that as the spatial separation between the paired dots increased, subjects' perception shifted from a single direction to two distinct directions. Critically, we found that the spatial separation required for this perceptual transition increased with the eccentricity and closely matched the known receptive field sizes of V1 neurons at those eccentricities. Direction segmentation occurred only when the spatial separation between motion components exceeded the V1 receptive field size. Conversely, direction integration occurred when the spatial separation was smaller than the V1 receptive field size. Our results demonstrate that the receptive field size of V1 neurons sets the critical spatial scale for direction segmentation and suggest that V1 plays a key role in motion transparency and, more generally, in motion segmentation.