Abstract
Contour curvature maxima/minima and discontinuities, such as terminators, junctions, and corners, can serve as trackable features (TFs) that solve the aperture problem faced by motion-tuned cells (Pack et al, 2003, 2004). Another contour-dependant motion signal, component-motion, can account for many percepts of motion (e.g. Bowns 2001) and may work in conjunction with TFs to determine perceived motion. The goal of our experiments was to characterize the relationship of contour curvature as a TF in the context of rotational motion. A further goal was to locate the areas of the visual cortex that support the use of these TFs for motion processing. Behavioral experiments: Pairs of ellipses were rotated and 2AFC speed judgments were made. The perceived speed of rotation varied with ellipse aspect ratio. Modified ellipses and rectangles were also used to examine the interaction of curvature, component motion and perception. Data imply that component motion models are insufficient to account for the aspect ratio at which the veridical speed is observed. fMRI experiments: Modified ellipses called ‘bumps’, consisting of two half ellipses joined along their common major axis were used. Groups of rotating bumps defined by the degree of curvature discontinuity were presented. Stimulus groups were either rotated at the same speed or at subjectively equal speeds. 17 subjects were run in a standard fMRI block-design, GE 1.5T, one-shot EPI, FA 90 degrees, epochs 20s, TR = 2.5secs, 25 axial slices. Results: Areas of the brain where the BOLD signal varied parametrically with perceived speed included superior/anterior MT+. Areas of the brain that varied parametrically with curvature discontinuity (i.e. TF) included inferior/posterior MT+. We propose that contour curvature and other TFs are processed in more inferior/posterior regions of MT+ and V3A, and that perceived motion is computed by more superior/anterior regions of MT+/MST on the basis of TFs, component-motion and other cues.