The behavioral data () showed a characteristic relationship between previous
outcomes and current
choices. The analysis of the standard outcomes by means of a three-way ANOVA with treatment (placebo vs. pramipexole), type of previous outcome (gain vs. loss) and magnitude of outcome (25 or 5) showed a significant effect of type of previous outcome [F(1,12)
12.4; p<0.01], with losses in the previous trial leading to riskier choices (i.e., higher probability of selecting 25) than gains. Although, the overall number of times that subjects made a risky choice (25 cents) did not differ between treatments: 47% after placebo and 49% after pramipexole (t(12)
0.33); the analysis of the infrequent boost trials by means of a two-way ANOVA with treatment and type of boost trial (“+50” vs. “+10”) as factors showed a main effect of treatment [F(1,12)
5.0, p<0.05]. Pairwise comparisons by means of t-tests showed a significant effect for the “+50” trials only; t(12)
−2.28, p<0.05. Thus, pramipexole abolished the tendency to be conservative after a boosted win of 50 that was present under placebo. This behavioral pattern is consistent with clinical observations of an increase of risk behavior in Parkinson's patients treated with dopaminergic agonists.
Functional MRI results showed that monetary gains (standard gain plus boost trials vs. losses) in the placebo condition robustly increased activity in the rostral basal ganglia, in an area encompasing the globus pallidus and parts of the ventral striatum (). At the same threshold, this activation was greatly diminished under pramipexole. To assess the response pattern in more detail, the time-course of the blood oxygen level dependent (BOLD) response in rostral basal ganglia was examined by defining a region of interest (ROI) around the peak of activation (x
: −15,−3,−3) in the globus pallidus. In the placebo condition, win trials (green; solid: 25 cent wins, dotted: 5 cent wins) showed an increase in BOLD relative to loss trials (red). In particular the 50 cent boosted wins (blue thick line) gave rise to a pronounced increase of activity. By contrast, this increased response to unexpected high gains was absent in the pramipexole condition. Percentage bold increase to the boosted +50 condition under pramipexole was lower (0.58%) than after placebo (0.80%), t(12)
−2.44, p<0.05, potentially reflecting decreased phasic dopamine release 
Figure 3 (a) fMRI results (rostral basal ganglia). Axial slices at z=−3 showing voxels of greater activation for win (normal gain+boost) than loss trials in the ventral striatum after placebo and pramipexole. The time (more ...)
Additionally, under placebo, unexpected high gains (boost +50) were associated with higher midbrain activity relative to normal gains (+5, +25) and +10 boost trials (see ), in line with data from animal studies (e.g., 
). Interestingly, this increase in neural activity was decreased under pramipexole. In addition to the voxel-based analysis, the BOLD response for a ROI defined around the peak of activation (x
:−4,−23,−8) was also assessed. Similarly to the results found in the rostral basal ganglia, the BOLD increase to the boost +50 trials (dark blue line) was lower under pramipexole (0.44%) than in the placebo condition (0.58%), t(12)
While our main focus was on the responses in the ventral striatum and the midbrain, we also found significant activations in the contrast (standard gain plus boost trials vs. losses) in the anterior cingulate gyrus (x,y,z coordinates: −6, 43, 7) and the posterior cingulate gyrus (x,y,z coordinates: −6, −52, 16, see ) in both, placebo and pramipexole conditions (see 
for similar activations).