Value and prediction error estimation account for volatility effects in ACC: a model-based fMRI study.

In order to choose the best action for maximizing fitness, mammals can estimate the reward expectations (value) linked to available actions based on past environmental outcomes. Value updates are performed by comparing the current value with the actual environmental outcomes (prediction error). The anterior cingulate cortex (ACC) has been shown to be critically involved in the computation of value and its variability across time (volatility). Previously, we proposed a new neural model of the ACC based on single-unit ACC neurophysiology, the Reward Value and Prediction Model (RVPM). Here, using the RVPM in computer simulations and in a model-based fMRI study, we found that highly uncertain but non-volatile environments activate ACC more than volatile environments, demonstrating that value estimation by means of prediction error computation can account for the effect of volatility in ACC. These findings suggest that ACC response to volatility can be parsimoniously explained by basic ACC reward processing.