Abstract
Neuroimaging studies implicate the fusiform face area (FFA) in face individuation, but what computations allow it to represent distinct identities? Behavioral research indicates that face individuation relies on sensitivity to subtle spatial relations in faces, but no studies have shown that face-selective cortex represents this information. We used a sensitive fMR-adaptation paradigm to demonstrate that the FFA spontaneously codes spatial relations in faces. Participants (N = 16) viewed four kinds of adaptation blocks: relational blocks in which a single upright face was shown with various feature spacings, inverted blocks in which a single inverted face was shown with various feature spacings, same blocks in which a single upright face was presented repeatedly and different blocks in which a different upright face was shown on each trial. The FFA was identified using a standard localizer task. The FFA responded more strongly to a single face presented with various feature spacings (relational blocks) than to repeated presentations of an identical face (same blocks). Moreover the response to spacing variations was as strong as the response to a series of distinct identities (different). There were no interactions with hemisphere. These results suggest that the FFA can use variations in spatial relations to individuate faces. This sensitivity to spatial relations may also help the FFA integrate features into holistic face representations. We found little sensitivity of the FFA to spatial relations in inverted faces (inverted blocks). The orientation-selectivity is consistent with behavioral evidence that spatial relations are difficult to code in inverted faces, making the coding of upright and inverted faces qualitatively distinct, and with neuroimaging evidence that the FFA is the locus of behavioral face-inversion effects.
This work was supported by the Australian Research Council.