Studying the neural representations of uncertainty
The study of the brain's representations of uncertainty is a central topic in neuroscience. Unlike other cases of representation, uncertainty is a property of an observer's representation of the world, posing specific methodological challenges. We analyze how the literature on uncertainty addresses those challenges and distinguish between "descriptive" and "process" approaches. Descriptive approaches treat uncertainty reported by subjects or inferred from stimuli as an independent variable used to test for a relationship to neural responses. By contrast, process approaches treat uncertainty derived from models of neural responses as a dependent variable used to test for a relationship to subjects' reports or stimuli. To compare those two approaches, we apply four criteria for neural representations: sensitivity, specificity, invariance, functionality. Experiments can be cataloged by their approach and whether they test for each criterion. Our analysis rigorously characterizes the study of neural representations of uncertainty, shaping research questions and guiding future experiments.