In other analyses, we tested for an association of ERP variables with alcohol measures via their correlations. First, in between-family analysis of twins as individuals, we correlated individual alcohol measures with ERP variables for all twins in which both co-twins had usable data, the same samples used for HMR analysis. Then, in 2 within-family analyses, we exploited the paired co-twin structure of our data. These within-family analyses control for unmeasured—and often unknown—between family confounds (e.g., known confounds include parental drinking history, family structure and status, household environment) that could mediate associations of brain function and drinking history among adolescents. In the first approach, we compared ERP measures in co-twins from twin pairs discordant for DSM-III-R diagnosis of alcohol disorder; discordant co-twins, especially discordant MZ co-twins, offer an incisive analytic tool (Dick et al., 2000), although, often, when both predictor (here, drinking measures) and outcome (here, ERP measures) are moderately heritable, discordant co-twins will be limited in number, and power will be constrained. In a second within-family analysis, we used all twin pairs with usable data from both co-twins on both
