We tested a model of the process of becoming involved with drugs during junior high. The sample included 698 students who were not using alcohol, cigarettes, or marijuana during Grade 7 (T1); the follow-up data points were 12 and 15 months later (T2 and T3). The final model, which predicted 72% of the variance in drug use at T3, provides support for hypotheses drawn from both social and cognitive theories. Weak familial and school attachments fostered use by increasing the likelihood of exposure to pro-drug social influences (drug use offers); weak bonds with school also directly affected cognitive motivations (lower resistance self-efficacy [RSE] and more positive outcome expectancies). In turn, social influences at T1 played a dominant role in initial use at T2, but cognitive motivations were also significant. At T3, prior use assumed the most prominent position. Drug-specific measures of RSE and expected use directly affected later use of that substance. The results indicate that both generic and drug-specific effects are needed to explain adolescent drug use. Implications for prevention programs are discussed.