We base the decisions we make on what we have learnt from our past actions, but also take into account other people's advice; if the advice turns out to have been wrong, we learn to take that person's advice with a grain of salt. These two forms of learning were assumed to have different neural bases but, as Behrens et al. show, the same associative mechanism underlies both reward-based and social learning.
To determine whether reward and social information are processed in similar ways, the authors measured the fMRI signals associated with reward prediction errors and social prediction errors, which reflect whether the actual outcome differed from the outcome that was expected on the basis of reward and social information, respectively. They also identified the fMRI signals associated with the volatility of both sources of information, which reflects the value of the information for learning. Reward prediction errors were associated with activation of the ventral striatum, the ventromedial prefrontal cortex (VMPFC) and the anterior cingulate sulcus (ACCs), whereas social prediction errors were encoded in the dorsomedial prefrontal cortex, the middle temporal gyrus and the superior temporal sulcus. However, the temporal patterns of activation that were associated with both types of prediction error were strikingly similar, indicating that social and reward prediction errors are processed similarly but in different brain regions.
This is a preview of subscription content, access via your institution