Supplementary Figure 1. Reinforcement altered Training phase RTs. To investigate if there were any effects of reinforcement on Simon task performance, the slope of RTs within each block was quantified for each stimulus (the six separate Training blocks were then averaged together). All stimuli experienced a decline in RT over time, but the B vs. C stimuli differed in the speed of decline: a 2 (B vs. C) * 2 (congru vs. incongru) ANOVA revealed a significant interaction where conditions associated with reward (B congruent and C incongruent) had steeper slopes than those associated with punishment (F 1,82 =6.12, p=.015). These findings suggest that reward associations hastened response times as a function of reinforcement history. No such effects were found for Study II, possibly due to floor effects of much faster RTs (placebo: 343 ms; cabergoline: 340 ms) than in Study I (574 ms). Error bars are S.E.M. 1
Supplementary Figure 2. Feedback-locked time-frequency representations from the FCz electrode. A) Same as Figure 3a for comparison here. An additional delta band ROI is highlighted below the theta band ROI. B) Phase consistency was greater in the theta band for punishments (see similarities between power and phase ROIs), but it was greater in the delta band for rewards, consistent with suggestions of a rewardrelated positivity. 2
Supplementary Figure 3. Delta-band EEG reveals additional evidence for cortical systems affected by the cost of conflict leading to diminished reward value. The topography of correlation coefficients (delta power and choice bias) is shown for each empirical contrast, as well as a scatterplot from the CPz electrode. A) In the case of rewarding feedback, conflict was hypothesized to diminish the relationship between salience of reward and future action selection identical to the prediction for FMθ in the main text. B) Individuals with greater reward-related delta on the B condition had a stronger bias to seek B>C. C) Individuals with greater reward-related delta on the C condition (following conflict) had no relationship between feedback-related activities and action selection. D) The difference between reward-related correlation coefficients was significant in posterior central areas, demonstrating how conflict diminished the relationship between reward-related delta band signals and action valuation. E) In the case of punishing feedback, there were no predictions for delta band activities as this signal has only been related to better-than-expected outcomes. These tests are shown here for completeness. F-G) Null effects. H) No delta band differences in punishment Correlations were Spearman s rho tests (N=83), z indicates rho-to-z test of differences between coefficients. **p<.01 3
Supplementary Figure 4. Delta plots show training phase conflict effects (incongruent minus congruent) as a function of the speed of response quantile. A) In Study I, there was a peak in conflict-related effects at about 450ms. B) In Study II, the peak in conflict-related effects was ~300-400ms in each condition, and this peak extended over more quantiles than in Study I. Error bars are mean +/- S.E.M. 4
Supplementary Figure 5. Study II: time-frequency plots from the FCz electrode for the cabergoline placebo contrast. A-B) While there were some varied significant differences between cabergoline and placebo sessions, there were no differences in the pre-response or post-feedback time-frequency regions described in the main text. 5
Supplementary Figure 6. An alternative behavioral measure of the cost of conflict yielded similar significant genetic, pharmacological, and eye blink effects as the Conflict Cost contrast described in the main text. A) An alternative measure of the cost of conflict was constructed by comparing choices between B and C when they were paired with A and D, termed the Relative contrast (this contrast also had a constant of.5 added to it to facilitate direct comparison with the other contrasts). The Conflict Cost and Relative measures were highly correlated with each other (rho 81 =.57, p=1.31 e-8 ). B) A one-way ANOVA for distributions of the DARPP-32 geneotype revealed a significant step-wise effect for the Relative contrast (F 2,80 =3.91, p=.02). C) A t-test revealed a significant effect of cabergoline on the Relative contrast (t 26 =-2.43, p=.02). D) Placebo blink rate predicted the cabergolineinduced change in the Relative contrast (Spearman s rho 25 =-.37, p=.05). E) The cabergolineinduced change in blink-rate was also linearly related to the shift in the Relative effect (although non-significant: rho 25 =.31, p=.12). Error bars are mean +/- SEM. *p<.05 6