Abstract
The studies we report concern recognition of complex objects, such as faces, defined solely by motion cues. Dynamic object shape cues, such as structure-from-motion, are thought to be largely mediated by dorsal-stream areas, such as MT and MST. However, object recognition in general, and unfamiliar face recognition in particular, are strongly believed to be mediated by ventral stream areas. Thus, recognition tasks involving motion defined faces offer a unique opportunity to probe dorsal-ventral integration and its role in complex object recognition. Here, we report data from several psychophysical, neuropsychological, and functional imaging studies that we have conducted in exploring these questions. Our results show that (a) purely motion-defined unfamiliar faces can be recognized, (b) classic effects such as the Inversion Effect may also apply to the recognition of unfamiliar faces defined by motion, (c) intact cortical motion processing mechanisms are necessary for the perception of structure-from-motion objects, (d) intact cortical face processing mechanisms are necessary for the recognition and learning of motion defined faces, and finally, (e) motion-defined faces may not engage the Fusiform Face Area, but the Occipital Face Area. Taken together, our results make several important theoretical contributions. First, that dorsal-ventral integration is necessary for motion-defined object recognition. Second, putative face areas identified thus far with face photographs may not be responsive to dynamic 3D percepts. Finally, the presence of this integration suggests our simplistic hierarchical view of the ventral stream is incomplete.
This research was supported by an FRSQ Doctoral Fellowship to RF, CIHR and CFI grants to MP, SNSF grant to OB, and CIHR grants to AC