In order to provide a strong test of this prediction, subjects were divided into two groups. In both groups, 40p outcomes were signaled at the time of delivery. In groupS (signaled group), 0p outcomes were also signaled. Hence, each successive time-step after the CS was more

likely to contain an outcome (and thus a reward) as the subject knew that the outcome had not yet been delivered. The hazard function thus increased monotonically through the trial (Figure 3B; inverted function shown in green). In groupU (unsignaled group), 0p outcomes were unsignaled. In this group, the passage of time initially increased the chances of imminent reward (as the peak delivery time approached), and then decreased these chances BMN 673 clinical trial as it became increasingly likely that the crucial time had passed, resulting in a hazard function that was approximately quadratic and peaking at 6 s (Figure 3B; inverted function shown in red). Because of these group differences in hazard functions, we predicted different BOLD responses to an unexpected reward in the two groups (Figure 3B). www.selleckchem.com/products/sch772984.html We tested

the two hazard functions on the BOLD response to unexpected rewards (for details regarding the general linear model [GLM] see Experimental Procedures). Parameter estimates for both hazard functions were extracted from the VTA ROI. In both groups, VTA data conformed to predictions: the monotonic hazard function predicted data from groupS (t13 = 2.60, p = 0.022), and the quadratic hazard function predicted data from ZD1839 purchase groupU (t13 = 4.22, p = 0.001), but not vice versa (both p > 0.05; Figure 3C). Furthermore, this difference survived the stringency of a formal

between-group comparison (ANOVA group × hazard function, F1,52 = 5.18, p = 0.027). Hence, in both groups an unexpected reward delivered early leads to a stronger response than an unexpected reward delivered at an expected time; however, an unexpected late reward only leads to a strong response in groupU, where the temporal hazard function decreases late in the trial. This effect can be seen in the raw BOLD time courses extracted from the VTA, plotted separately for short, middle and long CS-US intervals (Figure S3). Although we found that the BOLD response to the CS increased in proportion to the expected reward for fixed timing trials, there was no such effect for variable timing trials. There was a general increase in BOLD signal in response to variable cue onset (p < 0.001, t27 > 4.0) but this increase did not distinguish between the three reward conditions (p > 0.3; Figure S3). Overall, effects of variable timing cues showed a trend toward being smaller than those of fixed timing cues (t27 = 1.99, p = 0.057, comparing responses to any fixed timing CS to those evoked by any variable timing CS), rendering it possible that any effects were too small for such a scaling to be detectable.