STUDYING INDEPENDENCE OF NEURAL PROCESSING OF FACE DIMENSIONS

A human face is a multi-dimensional structure conveying several important social signals that are critical in day-to-day interactions. Based on a review of the available neuroimaging literature, 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 dissertation, we investigated whether the ventral and dorsal pathways are tuned to process facial form and motion independently, using formal definitions of independent neural representation that can be tested through decoding analyses. We created highly controlled stimuli (computer-generated), consisting of 3 identities changing their facial expression, resulting in 3 different facial motions (9 total dynamic stimuli). Employing multi-voxel pattern analysis (MVPA) and training support vector machine classifiers, we assessed whether we could decode facial form and motion from different face-selective areas in the two visual pathways. We used a recommended two-test strategy by simultaneously testing against context-specificity and context-invariance of each process.

While our results support the idea that the ventral pathway processes form information, we found that further anterior areas show higher sensitivity to any change in facial form. This was reflected in an absence of invariant representations of face identities in aIT. Moreover, all areas in the dorsal pathway process both form and motion information from a face. Importantly, we found evidence for motion- invariant representations of facial form in IFG. IFG is one of the lesser studied face- selective areas in the face relevant literature. Hence, IFG’s role in face perception is an ongoing investigation and further research is required. Meanwhile, our results suggest that IFG is a viable candidate for computationally untangling processed information through ventral and dorsal pathways.

Major Professor: Dr. Fabian Soto