Because of technical error, the behavioral data from six participants were lost; however, the ERP data for these subjects were included in the analyses. This left a total of 273 subjects that could be included in the behavioral analyses and 279 subjects included in the ERP and genetic analyses.

#### Behavioral data

Performance measures in the overall sample, and as a function of *DRD2* and *DAT1* genotype, are presented in Table 1. Consistent with previous work, children were significantly faster on error trials than on correct go trials, *t*(1, 272) = 32.42, *P* < 0.001. Compared to the overall mean of correct trial reaction time, participants were slower to generate a correct response on trials that occurred after an error, *t*(1, 272) = 5.90, *P* < 0.001. However, there were no overall reaction time differences as a function of either *DRD2* or *DAT1* genotype, *F*_{1,269} = 0.02, *P* = 0.89, *F*_{1,269} = 2.90, *P* = 0.09, respectively; neither *DRD2*, *F*_{1,269} = 0.71, *P* = 0.40, nor *DAT1*, *F*_{1,269} = 0.46, *P* = 0.50, interacted with trial type to impact reaction time. However, there was a three-way interaction between trial type, *DRD2*, and *DAT1*, *F*_{1,269} = 4.56, *P* < 0.05.

Table 1. Means (SD) of behavioral measures (milliseconds), ERN, CRN and (ERN–CRN) amplitude (µV) at Cz for the entire sample and the genotypes | All children (*N* = 279) | *DRD2* A1 (*N* = 91) | *DRD2* A2/A2 (*N* = 188) | *DAT1* 10/10 (*N* = 138) | *DAT1* 9 (*N* = 141) |
---|

Errors of commission | 16.01 (7.62) | 15.57 (6.51) | 16.37 (8.10) | 17.05 (8.13) | 1520 (7.01) |

Errors of omission | 10.12 (11.05) | 8.68 (9.58) | 10.83 (11.66) | 11.60 (12.06) | 8.71 (9.83) |

RT errors | 509 (88) | 511 (94) | 507 (85.26) | 514 (92) | 503 (84) |

RT correct | 626 (72) | 624 (77) | 627 (70) | 632 (71) | 621 (72) |

Post-error RT | 655 (119) | 647 (116) | 660 (121) | 655 (112) | 656 (126) |

Post-error slowing | 28 (79) | 22 (81) | 32 (81) | 23 (73) | 35 (89) |

ERN | 0.09 (10.06) | −2.03 (7.37) | 1.12 (10.99) | 0.79 (11.34) | −0.59 (8.61) |

CRN | 9.18 (5.94) | 9.32 (5.55) | 9.11 (6.13) | 8.56 (6.13) | 9.78 (5.70) |

ΔERN | −9.09 (10.26) | −11.35 (8.28) | −7.99 (10.95) | −7.78 (11.99) | −10.37 (8.07) |

As depicted in Fig. 3, *post-hoc t*-tests suggested that within the *DRD2* A1 group, children who were homozygous for *DAT1* 10/10, were slower on correct trials than children with a *DAT1* 9 allele, *t*(1, 88) = −2.65, *P* < 0.01. Within the *DRD2* A1 group, reaction time on error trials did not differ significantly between the two *DAT1* genotypes, *t*(1, 88) = −1.01, *P* = 0.32. Within the *DRD2* A2/A2 group, neither correct, *t*(1, 181) = 0.36, *P* = 0.72, nor error reaction times, *t*(1, 181) = −.59, *P* = 0.56, differed significantly between the *DAT1* groups.

Additionally, post-error slowing did not differ by *DRD2* genotype, *F*_{1,272} = 0.64, *P* = 0.42, or by *DAT1* genotype, *F*_{1,272} = 0.004, *P* = 0.95, and there were no significant two- or three-way interactions involving genotypes and post-error slowing (all *ps >* 0.1).

Overall, participants committed an average of 16.10, SD = 7.62, errors of commission and an average of 10.12, SD = 11.05, errors of omission, out of a total of 240 trials. Children with at least one *DAT1* 9 allele made significantly fewer errors of commission and fewer errors of omission than children who were homozygous for the *DAT1* 10 allele, *F*_{1,272} = 4.08, *P* < 0.05 and *F*_{1,272} = 4.75, *P* < 0.05, respectively. All other effects and interactions did not reach significance (all *ps >* 0.1).

#### ERPs

Means and standard deviations of ERN, CRN and ΔERN as a function of genotype are included in Table 2, response-locked waveforms at Cz for ERN and CRN for each genotype are included in Figs 1 and 2. The ERP response was more negative following errors than correct responses, *F*_{1,275} = 222.25, *P* < 0.001. There was no overall difference in brain activity as a function of the *DRD2* or *DAT1* genotypes, *F*_{1,275} = 2.98, *P* < 0.09, and *F*_{1,275} = 0.51, *P* = 0.48, respectively. However, the effect of trial type was qualified by a significant interaction with *DRD2* genotype, *F*_{1,275} = 6.37, *P* < 0.01. Children with at least one *DRD2* A1 allele had a larger (i.e. more negative) ΔERN than children who were homozygous for the *DRD2* A2 allele, *F*_{1,277} = 6.67, *P* < .01. This effect was driven by the effect of *DRD2* genotype on the ERN, *F*_{1,277} = 6.11, *P* < 0.01 such that children carrying at least one *DRD2* A1 allele had a significantly larger (i.e. more negative) ERN than children carrying the *DRD2* A2 allele (homozygous for A2). Children did not differ in CRN between the two *DRD2* genotypes, *F*_{1,277} = 0.072, *P* = 0.79.

Table 2. Overall and Incremental results from hierarchical regression analysis of genotype predicting ERPs at Cz | ERN | CRN | ΔERN |
---|

*R*^{2} | *F* | *R*^{2} | *F* | *R*^{2} | *F* |
---|

Overall results |

Step 1: *DRD2* | 0.022 | 6.11** | 0.00 | 0.072 | 0.024 | 6.67** |

Step 2: *DAT1* | 0.025 | 3.64^{*} | 0.01 | 1.49 | 0.038 | 5.52** |

Incremental results |

Step 1: *DRD2* | 0.02 | 6.11** | 0.00 | 0.072 | 0.024 | 6.67** |

Step 2: *DAT1* | 0.004 | 1.17 | 0.01 | 2.91 | 0.015 | 4.28^{*} |

In addition, the difference between ERN and CRN also varied as a function of *DAT1* genotype, *F*_{1,275} = 3.88, *P* < 0.05. Although neither the ERN nor the CRN differed between the two genotypes alone, *F*_{1,277} = 1.32, *P* = 0.25 and *F*_{1,277} = 2.95, *P* = 0.09, respectively, children with the *DAT1* 10 allele (homozygous for *DAT1* 10) had a significantly smaller (i.e. less negative) ΔERN than children with the *DAT1* 9 allele (with at least one *DAT1* 9 allele), *F*_{1,277} = 4.53, *P* < 0.05.

Neither the *DAT1* by *DRD2* two-way interaction, *F*_{1,275} = 3.62, *P* = 0.06, nor the three-way interaction between trial type, *DAT1* genotype, and *DRD2* genotype reached significance, *F*_{1,275} = 0.01, *P* = 0.92, suggesting two independent effects on error-related brain response related to the *DAT1* and *DRD2* genes. To investigate the possibility that the genotypes related differently to frontal/posterior electrode sites, a repeated-measures anova was conducted that suggested that the effect of trial type at Pz was also qualified by a significant interaction with the *DRD2*, *F*(1, 275) = 6.053, *P* < 0.01, and the *DAT1* genotype, *F*(1, 275) = 6.46, *P* < 0.01. An additional repeated-measures anova suggested that the effect of trial type at Fz was qualified by a significant interaction with the *DRD2* genotype, *F*(1,275) = 5.74, *P* < .05, but not the *DAT1* genotype, *F*(1,275) = 2.51, *P* = 0.12.).

#### Hierarchical multiple regression analyses

To test for unique contribution of each genotype on the ERN, CRN and ΔERN we conducted separate hierarchical multiple regression analyses in which each of the ERPs were the dependent variables and potential predictor variable were the *DRD2* and *DAT1* genotypes. Results are shown in Table 2. As can be seen from the table, the additional variance accounted for in the ERPs by adding *DAT1* as a predictor was significant in the case of ΔERN, R^{2} = 0.015, *P* < 0.05, although not in any of the other ERP measures. The variance in the difference score ΔERN accounted for by *DRD2* alone was 2.4% and after *DAT1* was added, the variance accounted for increased to a total of 3.8%, a significant increment. Thus, *DAT1* significantly predicts the difference score ΔERN even after controlling for *DRD2*. This suggests an additive effect of the two genotypes. Overall and incremental results from the hierarchical regression analyses are included in Table 2.

#### Mediation analysis

A follow-up repeated measure anova suggested that when overall accuracy, reaction time (on error and correct trials), and age are added as covariates, the interaction between trial type and *DRD2* remained significant, *F*_{1,265} = 6.07, *P* < 0.01, however, the effect of trial type was no longer qualified by the interaction with *DAT1*, *F*_{1,265} = 1.94, *P* = 0.17. Statistical analyses were conducted to test the potential mediation of accuracy on the relationship between *DAT1* and ΔERN. The original beta for the relationship between *DAT1* and ERN was 2.60, *t*(278) = 2.13, *P* < 0.05 and the beta for the relationship between *DAT1* and accuracy was −4.75, *t*(272) = −2.73, *P* < 0.01. In the second regression analysis, the beta for accuracy predicting ERN was -.12, *t*(272) = −2.51, *P* < 0.01 and the beta for *DAT1* predicting ERN was reduced to 2.03, *t*(272) = 1.63, *P* = 0.12. This reduction was significant; Sobel's test *Z* = 1.85, *P* < 0.05 (Fig. 3).

#### Gender

A follow-up repeated measure anova suggested that when gender was added as a covariate, it did not result in a significant interaction with ERN, *F*_{1,274} = 0.16, *P* = 0.69, and the interaction of *DRD2* and ERN, *F*_{1,274} = 6.35, *P* < 0.01, and *DAT1* and ERN, *F*_{1,274} = 3.99, *P* < 0.05, remained significant. However, a previous study has suggested that males and females differ in *DRD2* binding in the striatum (Pohjalainen *et al.* 1998b) so further *post hoc* analyses were completed by dividing the sample into males and females. In males, the effect of trial type was not significantly qualified by an interaction with *DRD2* genotype, *F*_{1,155} = 0.88, *P* = 0.35. However, in females there was a significant interaction of trial type with *DRD2* genotype, *F*_{1,123} = 7.20, *P* < 0.01, such that females with the *DRD2* A1 allele had a significantly more negative ERN, M = −3.67, SD = 7.53, than females without the *DRD2* A1 allele, M = 1.01, SD = 11.04.