58, p = 0.035). The absence of an RPE response was likewise observed

in a ROI defined in the dorsal striatum and ventral putamen Obeticholic Acid in vitro (see Figures S4F–S4K and Supplemental Experimental Procedures). As with the VTA, we next assessed the extent to which ventral striatal BOLD fluctuations to unexpected rewards depended upon a group-specific temporal hazard function. However, in the case of the VS, we could also assess the degree to which unexpected rewards elicited a larger response than unexpected zero outcomes on average across all variable timing trials. We could not perform this analysis for the VTA because this very contrast had been used to define the VTA ROI, and so would be subject to selection bias. In GroupU, where unexpected rewards also carry unexpected timing information, unexpected rewards led to an increase in VS activity (t27 = 3.69, p = 0.001, response learn more to 40p versus 0p in 50:50 trials; Figure 4D). By contrast in groupS, where all events carry the same timing information, there was no difference in the average VS responses between rewarded and unrewarded

variable timing trials (t27 < 1, p > 0.3; Figure 4D). Direct comparison between the effects observed in the two groups showed larger differences between responses to 40p versus 0p in groupU compared to groupS (2-way interaction: group × 40p-versus-0p response: F1,52 = 5.18, p = 0.026). Again, whereas the VTA responded to unexpected rewards, the VS responded to unexpected information about event timing. Furthermore, unlike in VTA, the BOLD signal to unpredictable rewards in variable timing trials did not conform with the group-relevant temporal hazard function (Figure 4C and Figure S4E) (ANOVA group × hazard function, F1,52 = 1.68, p = 0.28). Formal comparison with the VTA data revealed a three-way interaction (ROI × group × hazard function, F1,104 = 4.72, p = 0.032). The absence of an effect of the temporal hazard function was also true for dorsal striatum and ventral putamen (see Figures S4F–S4K and Supplemental Experimental

Procedures). In summary, at US time in variable timing trials, the only event that elicited a significant increase in VS activity was an unexpected reward in groupU—the only event that revealed unexpected timing information. To examine whether this response to unexpected timing information at through US time was related to subject behavior, we performed two further analyses on BOLD responses to unexpected rewards in groupU. We assumed that, in order to perform well on test trials, subjects would covertly time the outcome in each trial. It is therefore conceivable that the VS response to the US in classical conditioning trials might reflect the accuracy of subjects’ internal timing estimates and drive behavioral change. If the VS signal is monitoring task-performance then trials where the subject’s prediction is more accurate than expected should elicit a large BOLD response at US time.