Within the context of a role for this region in model-based computations, the findings by Nicolle et al. starkly demonstrate just how flexible the value computations in this region are: not only does vmPFC reflect valuation based on one’s own preferences when those are needed to guide choice, but the same region can also reflect the preferences of another person when those preferences

are relevant to the choice process. In addition to the valuation signals noted in vmPFC, Nicolle et al. also report a striking pattern of value-related BOLD activation in dmPFC. Specifically, on trials in which the subjects made choices on behalf of their partners, dmPFC responded to the difference in the self value for the two available prizes, while in trials in which subjects chose for themselves, dmPFC responded to the difference in their partner selleck products values. It is interesting to note that the self- versus other-oriented distinction was not reflected in the neural activations in either dmPFC or vmPFC. That is, although one value signal reflected the subjects’ own preferences for discounting and the other, arguably more social, value

Selleckchem ABT-263 signal reflected the preferences subjects attributed to their partners, each was encoded in vmPFC when relevant for choice and in dmPFC when it was not. The pattern of dmPFC activations is particularly surprising in this regard, given the role commonly attributed found to the region in supporting social cognition (Amodio and Frith, 2006). In particular, the ability to “mentalize,” or to attribute intentions, beliefs, and other mental states to other agents is consistently associated with activation of this region across fMRI and PET studies (Frith and Frith, 2003). However, the present results suggest that anterior dmPFC in the present task may not necessarily be “social” at all, but instead might facilitate the simulation of signals that are currently not relevant for choice, regardless of whether those signals correspond to representations about the

self or another person. Such an interpretation conforms to theories of dmPFC function that claim that its critical role lies in the creation of representations of the world that are decoupled from the sensory environment (Frith and Frith, 2003). Such a computational process could still underlie social inferences by allowing for the simulation of other agents, but importantly, its functional remit is not limited to social contexts, but rather to any situation in which simulation of events divorced from the sensory environment is required. The above-mentioned interpretation of the dmPFC findings raises an interesting question: Why are these value signals in dmPFC being computed in the first place? The presence of these activations is somewhat surprising in the task used by Nicolle et al., because the respective variables they represent are, at least superficially, irrelevant to the choice at hand.