The incorrect stimulus resulted in the reverse (30:70) ratio. Thus, on 30% of trials subjects received “misleading” feedback. After 40 trials

the reinforcement contingencies reversed, so that the frequently rewarded stimulus now became frequently punished and vice versa. Each subject completed a pseudorandom fixed sequence of 80 trials. Subjects were instructed that the identity of the correct stimulus could change, but received no information as to how often such a change might occur (for details see Supplemental Experimental Procedures). Details of DNA extraction from the saliva samples and genotyping are described in the Supplemental Experimental Procedures. For DAT1, two alleles of interest were analyzed: the common 10R allele and the rarer 9R allele. The insertion/deletion polymorphism in the SERT promoter region (5HTTLPR) was genotyped BIBW2992 cost for the long (S) or short (L) alleles in combination with the single nucleotide polymorphism rs25531 A/G substitution in the same region. For the behavioral analysis,

we used a biallelic model, where the S allele was grouped with the rare LG allele (indicated as S′), given that the G-substitution in the L allele results in reduced expression more similar to the S allele ( Hu et al., 2006 and Praschak-Rieder et al., 2007). LA alleles were indicated as L′. Given the large sample size, all genotypes could be analyzed separately, which enabled testing for dose-dependent gene effects. In all analyses, GSK1349572 sex, age, and education level were included as covariates of no interest. The statistical significance threshold for all tests was p = 0.05, using a Bonferroni correction where appropriate.

To increase sensitivity, we did not use a Bonferroni correction for any of the control analyses. Using the χ2 test, we assessed whether there were any differences between genotype groups in the proportion of subjects passing the acquisition learning however criterion of eight consecutive correct responses, which we report in the Supplemental Experimental Procedures, where we also report baseline effects of task engagement/learning for both the pass and fail groups. Effects of reinforcement on subsequent choice were operationalized as the probability of repeating responses after reward (“win-stay”) and shifting responses after punishment (“lose-shift”) (Figure 1A). Errors during the reversal phase were divided into two types. Perseverative errors were defined as two or more consecutive incorrect choices of the previously rewarded stimulus. Thus, perseverative errors required subjects to erroneously stay with the previously correct stimulus, despite punishment. The remaining errors during the reversal phase were defined as “chance errors.