Genetic influences on behavior are complex and, as such, the effect of any single gene is likely to be modest. Neuroimaging measures may serve as a biological intermediate phenotype to investigate the effect of genes on human behavior. In particular, it is possible to constrain investigations by prior knowledge of gene characteristics and by including samples of subjects where the distribution of phenotypic variance is both wide and under heritable influences. Here, we use this approach to show a dissociation between the effects of two dopamine genes that are differentially expressed in the brain. We show that the DAT1 gene, a gene expressed predominantly in the basal ganglia, preferentially influences caudate volume, whereas the DRD4 gene, a gene expressed predominantly in the prefrontal cortex, preferentially influences prefrontal gray matter volume in a sample of subjects including subjects with ADHD, their unaffected siblings, and healthy controls. This demonstrates that, by constraining our investigations by prior knowledge of gene expression, including samples in which the distribution of phenotypic variance is wide and under heritable influences, and by using intermediate phenotypes, such as neuroimaging, we may begin to map out the pathways by which genes influence behavior.