1, z = 2.74, p < 0.01; Figure 2C). These data identify

a dorsal-ventral functional organization of the medial PFC that does not follow a frame of reference of self versus other, but instead is tied to a frame of reference of executed versus modeled choices. To explore the data Wnt inhibitors clinical trials that underlies this functional gradient, we looked at activity that correlated with subjective preference-related activity separately under each choice condition (choice for self or for other). In blocks where subjects chose on behalf of themselves, activity in vmPFC correlated with the difference between the subject’s valuation (discounted by their own discount rate) for the chosen and unchosen options (Figure 3A), thus reflecting their personal choice preferences (Montreal Neurological Institute [MNI] atlas 12, 53, −11, t = 3.31, z = 2.76). Simultaneously, dmPFC activity also exhibited a value difference correlate but here values were discounted according to the partner’s discount rate, and the relevant value difference was between the partner’s preferred and nonpreferred choices (MNI 3, 41, 25, t = 5.00, z = 3.75). Hence in KU-57788 clinical trial self-choice trials,

despite the fact that the partner’s valuation bore no relevance to the task, dmPFC activity nevertheless reflected the experimental subject’s estimate of their partner’s preferences. As predicted by the gradient analysis, we observed a dramatically different pattern of activity during the delegated choice condition (Figure 3B). When subjects now made choices on behalf of their partner, these regions precisely swapped agents, such that the vmPFC now maintained an estimate of the partner’s values, expressed in the frame of reference of the choices made on behalf of the partner (MNI −6, 23, −11, t = 7.36, z = 4.94). Conversely, the dmPFC now reflected the subject’s own values, signed according to the

choices that the subject themselves would have preferred in the same context (MNI 0, 50, 19, t = 4.01, z = not 3.34). Accordingly, when we searched for regions that contained a conjunction of voxels responding to both types of executed, or choice-relevant, value differences (p < 0.05) we recovered a signal in vmPFC. At the same threshold, a region within dmPFC contained voxels representing modeled, or choice-irrelevant, value difference, be it those of the subject or those of the partner (Figure 3C). The data presented in Figures 3A and 3B are not multiple comparison corrected and therefore do not constitute formal tests. We present these data to illustrate the effects in the individual conditions that underlie the formal tests of interaction. In order to provide a formal test that the regions switched agents between conditions, we designed a test that selected peaks exclusively from one choice condition and extracted data from these peaks in the alternative choice condition.