This cross-sectional study examined the validity of the conduct disorder hypothesis of adolescent drug involvement, largely tested to date on normal populations, on a large drug clinic-referred sample (N = 2582). All subjects met DSM-III-R criteria for either abuse or dependence of at least one psychoactive substance. Based on self-report data from the Personal Experience Inventory (PEI) (Winters & Henly, 1989), multiple regression analyses were conducted to predict prior twelve months drug use frequency from psychosocial risk variables. It was hypothesized that the Delinquency factor would account for the most variance in drug use compared to three competing factors (Psychological Distress, Nonconventional Values, and Family Distress). The analyses provided support for the hypothesis across gender, age and ethnic groups. The Delinquency factor uniquely accounted for about 50%-60% of the variance in drug use severity. Backward deletion regression analysis of individual scales indicated that Peer Chemical Environment and Deviant Behavior (both part of the Delinquency block) and, to a lesser degree, Psychological Disturbance (part of the Psychological Distress block), were consistently the most predictive of drug use. The findings are seen as consistent with the viewpoint that delinquency behaviors are important mediators of adolescent drug abuse and, thus, they merit central attention in prevention programs. The continuing need to develop and test theories about the pathways to drug involvement pertinent to severe-end, clinic-referred adolescents is discussed.