Ubiquity and Specificity of Reinforcement Signals throughout the Human Brain

SummaryReinforcements and punishments facilitate adaptive behavior in diverse domains ranging from perception to social interactions. A conventional approach to understanding the corresponding neural substrates focuses on the basal ganglia and its dopaminergic projections. Here, we show that reinforcement and punishment signals are surprisingly ubiquitous in the gray matter of nearly every subdivision of the human brain. Humans played either matching-pennies or rock-paper-scissors games against computerized opponents while being scanned using fMRI. Multivoxel pattern analysis was used to decode previous choices and their outcomes, and to predict upcoming choices. Whereas choices were decodable from a confined set of brain structures, their outcomes were decodable from nearly all cortical and subcortical structures. In addition, signals related to both reinforcements and punishments were recovered reliably in many areas and displayed patterns not consistent with salience-based explanations. Thus, reinforcement and punishment might play global modulatory roles in the entire brain.

► Reinforcement and punishment signals are broadly distributed in the entire brain ► Ubiquitous reward signals are revealed by multi-voxel pattern analysis on fMRI data ► Sensory and choice signals are more narrowly distributed in the brain ► Ubiquitous reinforcement signals cannot be explained by strategy or salience