Purchase this article with an account.
Lisa Betts, David Nichols, Hugh Wilson; Classification of fMRI activation patterns in face-sensitive cortex to the parts and location of faces. Journal of Vision 2009;9(8):553. doi: 10.1167/9.8.553.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The fusiform face area (FFA) and occipital face area (OFA) are known to respond more to faces than other objects, but the spatial structure of processing within these areas is not yet known. Previous physiological investigations in primates and neuroimaging studies in humans suggest that face-sensitive regions contain independent neural populations that are tuned to the internal features of the face, the shape of the head, or the full face, i.e., the conjunction of the features and the head outline. To test this hypothesis, we obtained fMRI data from eight participants while they viewed images of synthetic full faces, internal features, or head outlines. The FFA and OFA, defined as the regions with greater activation to photographs of faces compared to houses, were localized within individual subjects in a separate set of functional scans. We constructed linear pattern classifiers, based on all voxels in the regions of interest, using support vector machines, to test whether the FFA and OFA process the three different types of stimuli in a spatially distributed manner. Classification was significantly above chance for all types of stimuli according to a leave-one-out verification procedure. In an additional experiment, we found a dissociation in processing between the areas. Classification of the physical position of a face, in one of four visual quadrants, was better in OFA, than FFA, but classification of the type of stimulus, across position, was better in FFA than OFA. This suggests that the FFA and OFA are involved in different aspects of face processing, with OFA positioned earlier in the processing stream. The results are consistent with a columnar organization of faces and face parts, which in turn would support tasks such as viewpoint processing, gender classification, and identity discrimination.
This PDF is available to Subscribers Only