Abstract
The ability to categorize patterns is crucial for making sense of the visual world as an orderly collection of discrete objects. Contextual information often facilitates perceptual categorization to disambiguate and organize visual input. In last year's VSS, we demonstrated neural correlates of categorization in the domain of faces. Our experimental strategy involved compiling 300 images that spanned a range of facial similarity from non-faces to genuine faces. Using fMRI, we found that the pattern of activity in the left fusiform changed in a graded fashion as the image stimuli became increasingly face-like, while the pattern of activity in the right fusiform showed a step-like response corresponding to a categorical difference between faces and non-faces. Here we investigated how contextual information can modulate brain activation patterns that are evoked by these stimuli. Behaviorally, we verified that subjects were more accurate at the face/non-face classification task when the 300 images were shown with surrounding context. Consistent with our previous finding, brain activation patterns in the right fusiform correlated with the perceived face categoricity. More interestingly, the pattern of activity in the left fusiform and right inferior occipital gyrus also became categorical in correspondence to the behavioral step-like difference between faces and non-faces. Particularly in the left fusiform, the surrounding context on one hand enhanced activation patterns induced by genuine faces, and on the other hand suppressed the activation patterns induced by the non-face stimuli that happened to look face-like. By contrast, the effect of contextual modulation was not significant in the left inferior occipital gyrus and the calcarine sulcus, ruling out activation change induced by bottom-up effects. These results suggest a categorically selective neural mechanism of contextual modulation and have important implications for models of real-world face-processing.