Abstract
Purpose. Symmetry is an important cue in face perception because faces have a high degree of bilateral and other symmetries. We ask to what extent the cortical areas responding to faces are responding on the basis of the symmetry cue alone, and conversely whether any cortical face areas show symmetry invariance?
Methods. We considered two types of symmetry: a) image symmetry, where one part of the image is the mirrored transform of the other part about an axis and b) object symmetry, where the correspondences are interpretable as parts of a symmetric 3D object. We used functional magnetic resonance imaging (fMRI) to compare cortical responses in human observers to front-view faces relative to a variety of symmetry manipulations of the same images.
Results. Both ventral and dorsal cortical areas were more activated by the face images than to Fourier-matched scrambled images: the classic fusiform (FFA) and occipital (OFA) face areas, the middle occipital gyri (MOG), and dorsal areas in the superior temporal and intra-occipital sulci. Contrasting faces and symmetrized scrambled versions showed a similar pattern of activation except in the right OFA, suggesting its involvement in facial symmetry processing. The response to front-view vs. 3/4-view faces (having the same 3D object symmetry but disrupted 2D image symmetry) showed strong responses in the MOG and IOS but little differential activation in the FFA or OFA, suggesting that face processing in the latter is holistic and viewpoint invariant. Contrasting upright vs. inverted faces showed robust differential activation at all sites.
Conclusions. Both dorsal and ventral face-responsive regions were robust to 2D symmetry manipulations, but only the FFA and OFA showed invariance under 3D manipulation, implying that they incorporate a 3D face representation. The strong face-inversion effects are attributable to the use of whole faces including neck and hair features.
Supported by NIH EY 13025