Abstract
A human face is a multi-dimensional structure conveying several important social signals that are critical in day-to-day interactions. Recently, it has been proposed that two parallel pathways process form (ventral pathway) and motion (dorsal pathway) in faces. However, it is still unclear in what way the information carried by these two pathways may be independent. In this study, we investigated whether the ventral and dorsal pathways represent facial form and motion independently, using formal definitions of independent neural representation that can be tested through fMRI decoding analyses. We created highly controlled computer-generated stimuli, consisting of 3 identities showing 3 different facial motions (9 total dynamic stimuli). Two motions involved the exact same face expression, while the third motion involved a different expression. Twelve participants completed 3 hours of fMRI scanning during which they completed a one-back task. Employing MVPA, we assessed whether we could decode facial form and motion from face-selective areas in the two visual pathways. Our results suggest that most areas in the two pathways encode information about both face form and motion. In trials in which both motion and expression varied across stimuli, most areas showed evidence for invariant encoding, with the exception of aIT. In trials in which only motion varied across stimuli (i.e., no expression information was available), evidence of invariance was still observed in FFA (both motion and form) and IFG (only for identity), whereas results for other areas were inconclusive. In defiance of general opinion regarding aIT forming invariant representation of facial identity, representation of facial form untwined from motion was observed in IFG but not aIT. Individual-level analyses showed considerable differences across participants and in many cases qualified conclusions from group-level analyses, but they confirmed a key role of IFG in the invariant representation of face form.