Abstract
PURPOSE: Exceptional human sensitivity to small deviations from circularity in closed contours has been well demonstrated psychophysically. Here we used fMRI to test whether circularity holds a special status in the neural coding of closed-contour shapes. METHODS: BOLD signals were recorded from 5 participants in 13 6-mm coronal slices with the most posterior slice anchored on the occipital pole. A region-of-interest analysis isolated the lateral-occipital complex (LOC) by contrasting BOLD signals from images of intact vs. scrambled tools. In key experiments, observers viewed closed contours that varied in basic shape (i.e. radial frequency) and deviation from circularity (i.e. radial amplitude). Experiments followed a block design where deviation from circularity was varied across blocks, and basic shape was either varied within block (multi-shape blocks) or held fixed (single-shape blocks). Observers performed size judgments to maintain attention. RESULTS: BOLD response in LOC for multi-shape blocks was lowest for pure circles and increased monotonically with deviation from circularity. Single-shape blocks showed similar results. Response in striate and extrastriate areas remained approximately constant across all conditions. CONCLUSIONS: Results are consistent with neural representations of closed-contour shapes that are centered on circular prototypes, and data suggest that prototype deviations constitute the basis of increased neural activity. Lack of circle-selective responses in striate and extrastriate areas suggests that LOC activity reflects active shape integration rather than passive inheritance from lower-level areas. Encoding prototype deviations is an efficient strategy and is likely a recurring theme throughout the visual hierarchy.
We wish to thank Canada's NSERC grant OP227224 for its support.