Although the onset of illicit substance use during adolescence can hit parents abruptly like a raging flood, its origins likely start as a trickle in early childhood. Understanding antecedent factors and how they grow into a stream that leads to adolescent drug use is important for theories of social development as well as policy formulations to prevent onset. Based on a review of the extant literature, we posited a dynamic cascade model of the development of adolescent substance-use onset, specifying that (1) temporally distinct domains of biological factors, social ecology, early parenting, early conduct problems, early peer relations, adolescent parenting, and adolescent peer relations would predict early substance-use onset; (2) each domain would predict the temporally next domain; (3) each domain would mediate the impact of the immediately preceding domain on substance use; and (4) each domain would increment the previous domain in predicting substance use. The model was tested with a longitudinal sample of 585 boys and girls from the Child Development Project, who were followed from prekindergarten through Grade 12. Multiple variables in each of the seven predictor domains were assessed annually through direct observations, testing, peer nominations, school records, and parent-, teacher-, and self-report. Partial least-squares analyses tested hypotheses. Of the sample, 5.2% had engaged in substance use by Grade 7, and 51.3% of the sample had engaged in substance use by Grade 12. Five major empirical findings emerged: (1) Most variables significantly predicted early substance-use onset; (2) predictor variables were significantly related to each other in a web of correlations; (3) variables in each domain were significantly predicted by variables in the temporally prior domain; (4) each domain's variables significantly mediated the impact of the variables in the temporally prior domain on substance-use outcomes; and (5) variables in each domain significantly incremented variables in the previous domain in predicting substance-use onset. A dynamic cascade represented the most parsimonious model of how substance use develops. The findings are consistent with six features of social development theories: (1) multiple modest effects; (2) primacy of early influences; (3) continuity in adaptation; (4) reciprocal transactional development; (5) nonlinear growth in problem behaviors during sensitive periods; and (6) opportunities for change with each new domain. The findings suggest points for interventions, public policies, and economics of substance-use and future inquiry.

Kenneth A. Dodge is the William McDougall Professor of Public Policy and Professor of Psychology and Neuroscience at Duke University, where he directs the Center for Child and Family Policy. His research is aimed at understanding how problem behaviors develop and how they can be prevented.

Patrick S. Malone is Associate Professor of Psychology at the University of South Carolina. As a quantitative social psychologist, he utilizes advanced methodological techniques to address substantive questions of social risk forsubstance use outcomes.

Jennifer E. Lansford is Associate Research Professor at the Duke UniversityCenter for Child and Family Policy. Her research focuses on how family, peer, and cultural contexts affect the development of aggression and other behavior problems in youth.