Invariant versus context-specific representation of face shape and motion in the face network

Area of Science:

• Neuroscience
• Cognitive Science
• Visual Perception

Background:

• A prevailing theory posits distinct neural pathways for processing facial shape and motion.
• However, visual neurons exhibit non-classical receptive field effects, where one stimulus modulates responses to another, suggesting potential overlap in face processing.
• Investigating whether face-selective brain regions encode shape and motion invariantly or context-dependently is crucial for refining these models.

Purpose of the Study:

• To determine if face-selective brain areas encode facial shape and motion in an invariant or context-specific manner.
• To test the hypothesis that face processing regions may exhibit overlapping encoding of shape and motion.
• To challenge existing models by examining the interplay between shape and motion processing in the face network.

Main Methods:

• Employed a dual-strategy approach combining cross-decoding and context-sensitivity analyses.
• Utilized 3D synthetic face models with independently manipulated facial shape and motion.
• Analyzed data from twelve participants viewing dynamic facial stimuli.

Main Results:

• Shape and motion information were decodable from all tested face-selective regions, indicating overlapping neural representations.
• Distinct invariance patterns emerged: OFA (Occipital Face Area) showed motion-specific shape encoding, IFG (Inferior Frontal Gyrus) encoded shape invariant to motion, and FFA (Fusiform Face Area) processed both shape and motion invariantly.
• Controlling for pose-motion confounds was critical, as their conflation inflated invariance estimates across the network; substantial individual variability was also observed.

Conclusions:

• Findings challenge models proposing strictly separate pathways for facial shape and motion, supporting representational overlap.
• The context-specific nature of representations and significant individual variability necessitate revised computational models of face perception.
• Future models must integrate representational overlap, context sensitivity, and individual differences for a comprehensive understanding of the face network.