Conference on Cognitive Computational Neuroscience

Most events in our lives are probabilistic. Estimating their probability is crucial for adaptive behavior. Humans can estimate such probabilities, and even explicitly report their estimate when asked. The neural basis of such probability estimates remains elusive. Here, we found a neural code of probability estimate in humans during a changing probability estimation task. This neural code explains the measured fMRI activity in prefrontal and parietal cortex. It contains a majority of non-linear responses tuned to certain preferred values of probability, which contrast with the mostly linear responses observed for another quantity—the confidence associated with each estimate. This neural code may serve as a basis for probability estimation, which underlies many behaviors.