Abstract
The most dominant neural model of face processing posits that the face network is composed of two pathways that process different types of facial information (Haxby et al. 2000): A ventral pathway processes invariant aspects such as identity and gender and a dorsal pathway processes changeable aspects such as expression and gaze. This model is primarily based on studies that presented static images of faces. Recent studies that presented dynamic faces show that the dorsal stream is highly responsive to dynamic faces, whereas the ventral stream shows similar response to static and dynamic faces (Pitcher et al. 2011). These recent findings raise the question of what is the primary division of labor between the two pathways: is it to motion and form? to changeable and invariant facial aspects? or the interaction of both. To answer this question, we presented dynamic and static faces while subjects performed either an expression (positive/negative) or a gender (male/female) discrimination task. Univariate analysis revealed higher response to the dynamic than static faces in the dorsal pathway, whereas the ventral pathway responded similarly to the dynamic and static conditions. Both face streams showed no effect of task. Multivariate analysis further revealed that the dorsal but not the ventral regions carry information about motion, indicated by higher correlation within than between moving and static stimuli. Neither the ventral nor the dorsal streams carried information on whether subjects performed an expression or a gender task. Finally, the pattern of response in the motion area, MT, was correlated with the response of the dorsal face stream, further indicating its primary response to motion. These findings suggest that the primary division of labor between the two face streams is to motion and form rather than to changeable and invariant aspects, as current models posit.
Meeting abstract presented at VSS 2016