Greater sensitivity to convexities than concavities in human lateral occipital complex

Abstract

When subjects are asked to divide objects into natural parts, they tend to choose convex contour segments as parts. Furthermore, they are faster and more accurate in remembering and judging convex than concave contours (Driver & Baylis, 1995). One potential underlying neural mechanism is that cells in the human lateral occipital complex (LOC), an area thought to be critical for object recognition, are more sensitive to differences in convex than in concave contours. To test this hypothesis, we used event-related fMRI adaptation, in which neural sensitivity to the difference between two sequentially presented stimuli manifests itself in a higher BOLD response when the two stimuli are different than when they are the same or similar. We presented subjects with sequential pairs of convex or concave contours, viewed stereoscopically behind a circular aperture. Critically, the convex and concave contours were identical except for a stereo reversal that caused subjects to see either the convex or the concave contour as the figure. The contours were varied parametrically in either skew or aspect ratio, and the magnitude of the difference between the two stimuli in a given trial varied across trials. Subjects performed a same/different task on the pairs of contours. Reaction times were significantly shorter for convex than concave contours, confirming the behavioral advantage for convex stimuli. Concomitantly, the BOLD response in LOC was higher for different than for identical contour pairs when the contours were convex (p < 0.001), but not when they were concave (p = 0.89; interaction p = 0.03). This effect was more pronounced in anterior than in posterior LOC, and was absent in retinotopic cortex. These results indicate greater sensitivity to convexities than concavities in high-level human visual cortex, a difference that could underlie the psychophysical advantage for convex over concave contours.

Footnotes

 This work was supported by EY 13455 to Nancy Kanwisher. We thank Zoe Kourtzi for helpful comments and advice.