It is generally believed that during the process of choosing between alternative actions, neurons in a brain area called the striatum, represent the values of these possible actions. This hypothesis is based on a large number of studies, in which the activities of neurons in the striatum were correlated with the values of the actions in a learning task. Here we show that the conclusions of all these studies are subject to at least one of two critical statistical errors. First, we show that even weak correlations in time in the neuronal activity may result in an erroneous identification of action-value representations. Second, we show that experiments and analyses designed to dissociate action-value representation from the representation of other decision variables cannot do so. We suggest solutions to identifying action-value representations that are not subject to these confounds. Applying one solution to previously identified action-value neurons we fail to detect any action-value representations. We conclude that the claim that striatal neurons encode action-values should be reconsidered.
Paper of the month
Loewenstein's Lab: Striatal action-value neurons reconsidered
eLIFE on May (2018)