Abstract
Learning new faces is crucial for effective social interaction. Although we are able to recognise familiar faces from a variety of viewpoints, viewpoint-invariance is not well understood. Responsiveness to mirror-symmetric views of faces has been proposed as a key computational step in achieving viewpoint-invariance. The aim of this study was to measure responses to symmetric views of real human faces in face-responsive regions of the human brain and ask whether these convey an advantage in the perception and recognition of faces. In Experiment 1 (n=20), we used fMRI to measure responses in face-responsive regions while participants viewed images of faces at different viewpoints (0°, 45°, 90°, 135°, 180°). Patterns of response to symmetrical views of the face (45° & 135° or 0° & 180°) were more similar than asymmetrical views (e.g. 0° & 45° or 135° & 180°) in the FFA and pSTS. In Experiment 2 (n=20) participants made perceptual similarity judgements to pairs of face images. Similar to the fMRI data, images with symmetrical viewpoints were reported as being more similar to each other than non-symmetric views. In Experiment 3 (n = 19), we asked whether this sensitivity to symmetric views would convey an advantage on an identity matching task for unfamiliar faces. We found performance on the unfamiliar face matching task was not significantly better for symmetric images. In Experiment 4 (n = 48), we asked whether symmetrical views convey an advantage when learning new faces. Using a recognition memory paradigm, participants learned new faces at different viewpoints. Performance was best when tested in directions symmetrical to the learning direction, compared to viewpoints that were asymmetric. Together these results demonstrate that the viewpoint symmetry found in face-responsive regions of the brain reflects the perceptual experience of participants and conveys an advantage when we learn new facial identities.
Meeting abstract presented at VSS 2018