The study hypothesized that sociodemographic factors, characteristics of family and friends, attitudes toward drugs, and personal risk perceptions would be significantly related to alcohol and other drug use. It was also hypothesized that various explanatory models could be developed for children and adolescents to predict alcohol and other drug use. Data were collected from children and adolescents between 9 and 17 years of age in 100 central locations across the United States. The overall research design included a repeated sampling approach using four waves of data and involved the use of age-specific, self-administered questionnaires. The analysis, which primarily employed stepwise multiple regression for model estimation, reinforced the importance of marijuana use by friends, perceived risk, and tobacco and alcohol use in predicting marijuana use. An exposure model which included antidrug advertisements was a particularly powerful model in predicting alcohol and drug use by young people. Nonetheless, no single model worked equally well in identifying correlates of alcohol and other drug use by children and adolescents. Implications of the findings for drug prevention and intervention are discussed. An appendix lists sociodemographic and attitudinal variables used in the study. 25 references, 7 notes, 6 tables, and 1 figure