Abstract
The response in the fusiform face area (FFA), as measured with fMRI, does not differ for viewing mammalian faces from different species. We investigated whether distributed patterns of response in ventral temporal cortex are distinct for different species of mammalian faces and whether these distinctive patterns are restricted to the FFA or extend into ventral temporal cortex that responds maximally to non-face objects. We measured neural activity using BOLD fMRI in 16 normal volunteers while they viewed pictures of male and female human faces, monkey faces, dog faces, houses, chairs, and shoes. Data were analyzed using a split-sample correlation method for multi-voxel pattern analysis (Haxby et al. Science, 2001). Patterns of response in ventral temporal cortex distinguished human from animal faces (76% accuracy, p<0.0001) and monkey from dog faces (76%, p<0.0001). Classification performance was equivalent when analysis was restricted to voxels in the FFA (70% and 69%, respectively, p<0.0001 in both cases) and to voxels that responded maximally to manmade objects (74% and 73%, respectively, p<0.0001). Patterns of response to male and female faces could not be distinguished in any of these regions (52%, 48%, and 50%, respectively). These results show that the average magnitude of response in and outside of the FFA is a weak indicator of the information carried by neural activity. The neural representation of information that distinguishes among the visual appearances of mammalian faces is not restricted to the FFA.