Date:
Location:
Nonlinear Mixed Selectivity Coding in the Human Brain
Recent work in neurophysiology has identified neurons that are tuned to heterogeneous combinations of task and stimulus variables, which computational modeling suggests may endow these neurons with vast representational flexibility. However, the prevalence of this neural coding principle remains unstudied in the human brain. In two studies, fMRI was used to probe the existence of this nonlinear mixed selectivity coding in the human brain in the context of task-directed visual processing and visual feature binding. Results suggest that frontoparietal multiple-demand regions may nonlinearly multiplex task and stimulus information to a higher degree than purely visual regions in occipitotemporal cortex, and that early visual cortex may contain neural populations that are nonlinearly tuned to color and shape features, suggesting a potential ingredient in how the brain solves the visual binding problem.