To improve assessment of the DSM-IV alcohol tolerance criterion in adolescents, this study tested the performance of a minimum percentage increase in drinking quantity, and a proxy measure of tolerance (i.e., average heavy-drinking quantity per occasion) in identifying adolescents with alcohol dependence. Two adolescent samples were examined. In one sample (N = 415, 58% male, 79% white, 57% clinical), a modified version of the SCID was used to determine DSM-IV alcohol diagnoses, and lifetime drinking history data were collected by interview. In the second sample (N = 470, 60% male, 76% white, 100% clinical), the Adolescent Diagnostic Interview was used to determine DSM-IV alcohol diagnoses and to collect data on initial- and current-drinking quantities needed to become intoxicated. The performance of a percentage increase and average heavy-drinking quantity in identifying those with dependence was evaluated using receiver operating characteristic analysis. The utility of a percentage increase definition was limited by the high degree of variability in initial-drinking quantities. Percentage increase may underassign the tolerance symptom when initial-drinking quantities are high and overassign the symptom when initial-drinking quantities are low. Average heavy-drinking quantity per occasion, combined with a minimum frequency of drinking, demonstrated better performance than any percentage increase definition. Alternatives to a change-based (e.g., percentage increase) definition of tolerance warrant study due to limits of change-based definitions when initial-drinking quantity shows a high degree of variability. The variability in initial-drinking quantity may reflect individual differences in initial sensitivity that need to be considered in tolerance assessment.