Neural signatures of variable beliefs increase with task learning in v1

Previous work has suggested that “top-down,” especially task-dependent, sources of variability are some of the dominant modes of trial-to-trial variability in sensory neurons’ responses. It has also been hypothesized that such top-down modulators can be understood as representing “variable beliefs” about task-relevant features of the stimulus in a probabilistic inference framework of perception, where task-based expectations act as a prior [1].