Abstract
People are extremely proficient at recognizing faces that are familiar to them, but are much worse at identifying unfamiliar faces. We used fMR-adaptation to ask whether this difference in recognition might be reflected in the relative view-dependence of face-selective regions in the brain. Differences in the response to faces and non-face objects were used to define face-selective regions in 20 subjects. We compared the response in each ROI to familiar and unfamiliar faces in 3 experimental conditions: 1) same identity, same image (same/same); 2) same identity, different image (same/different); 3) different identity, different image (different/different). Although the low-level image variation between the same/different and different/different conditions was comparable, these manipulations had no effect on the recognition of familiar faces. We predicted that, if the neural representation of faces is view-independent, adaptation to repeated images of the same face should be apparent even when they are shown from different views. Each experimental condition was repeated 8 times in a counterbalanced block design, with each block containing 10 images presented at a rate of 1/sec. We found a reduced response (adaptation) to the same/same condition compared to the different/different condition for both familiar and unfamiliar faces in the fusiform face area (FFA), but not in the superior temporal sulcus (STS). However, there was no significant difference in the response to the same/different and different/different conditions for familiar or unfamiliar faces. A whole-brain analysis showed a distributed pattern of view-dependent adaptation (same/same [[lt]]different/different) that extended beyond the face-selective areas, including other regions of the ventral visual stream and a region in the right inferior frontal lobe. However, this analysis failed to reveal any regions showing significant view-independent adaptation (same/different [[lt]]different/different). These results suggest that structural information about faces is represented in a distributed network using a view-dependent neural code.
Spyroula Spyrou for data collection Andre Gouws for technical assistance.