Abstract
Visual perception is qualitatively modified by experience, changing from part-based to holistic with familiarity or expertise. Yet the neural correlates of such effects in higher-level visual areas remain relatively unexplored. Here we examined the neural basis of flexible changes in neural tuning to parts and wholes using a manipulation of perceptual similarity, which has previously been shown to affect visual processing. Unfamiliar faces were arranged along arbitrarily-defined eye and mouth dimensions either equidistantly (“Diamond” stimulus space) or asymmetrically (“Kite” space) to differentially engage holistic or part-based processing, respectively. Although subjects were never informed of the stimulus space arrangements, they nonetheless showed a significantly smaller face inversion effect for the Kite space, suggesting that this stimulus set is processed in a more part-based manner. Functional magnetic resonance imaging (fMRI) of the Diamond versus Kite stimulus space during an unrelated target detection task revealed a corresponding switch from holistic to part-based tuning in face-selective right fusiform gyrus (the Fusiform Face Area, or “FFA”). In contrast, the left FFA consistently showed part-based tuning for both Diamond and Kite spaces. Thus, the flexible tuning of the right FFA for wholes and parts cannot be explained as a general response property of ventral visual areas, but likely reflects this region's special importance for holistic processing of faces. These data indicate that, at least in some high-level visual areas, recruitment of holistic or part-based representations can be modulated by perceptual similarity. Together with previous data on familiarity (Harris & Aguirre, 2008), these results underscore the flexibility of neural tuning within higher visual areas, suggesting that factors such as familiarity and perceptual similarity may interact for increased efficiency of sensory coding.