Abstract
The visual quality of face and body information varies continually as we observe a person approach from a distance. We investigated whether and when neural responses to familiar and unfamiliar people could be discriminated in face- and body-selective brain regions during an approach. Participants were familiarized with identities using close-up and distant videos of the individuals. Next, in an fMRI scanner, they viewed 8s videos of familiar and unfamiliar people approaching from a distance (13.6m). Multi-voxel pattern analysis (MVPA) was applied to classify the neural activity patterns elicited in response to familiar and unfamiliar people in three ROIs: face-selective voxels, body-selective voxels, and a broader area of ventral temporal cortex defined by significant differences in responses to faces, bodies, and objects (“FBO voxels”). Using cross-validation, we computed classifier accuracy (d’) for discriminating neural responses to familiar and unfamiliar people at different times (i.e., viewing distances) in the video. Each 8s video scan time was divided into 4 TRs, each capturing 2s of video time, yielding four “distance windows.” Beginning by classifying neural activity collected across the entire video, we were able to discriminate responses to familiar and unfamiliar people in 10 of the 12 participants, for at least one of the ROIs. For these participants, the classifier was applied independently within the four distance windows. Classification accuracy depended jointly on the viewing distance window and the ROI, F(6,54)=2.73, p=.02, with no main effects of ROI or distance. In FBO and body-selective voxels, classification accuracy peaked in the third distance window (d’=0.88 and 0.64, respectively). In the face-selective ROI, accuracy peaked in the closest distance window (d’=0.56), when the face was most prominent in the video. The degree of neural response separation in response to familiar and unfamiliar people reflects the relative quality of face and body information over varying distances.
Meeting abstract presented at VSS 2015