Fig. 5
Schematic figure depicting possible ways in which multiple value estimates, based on different periods of experience determined by different LRs, might be represented in the brain as indexed by fMRI. We consider how such representations might change as the environment’s volatility changes. Each row shows the representation of value estimates in nine example voxels in a stable and in a volatile environment. a According to the LR shift hypothesis, in a stable environment neurons in more voxels would compute value estimates based on lower LRs while they would shift toward higher LRs in a volatile environment. b The signal strength shift hypothesis predicts that the value estimates computed by the neurons of each voxel remain constant in all environments, but that those voxels with value estimates that are currently most relevant for the environment (high LR voxels in volatile environments and low LR voxels in stable environments) increase their signal strength. c The combination hypothesis suggests a combination of the two mechanisms in a, b
