Objective: To examine externalizing behavior problems and cigarette smoking as predictors of subsequent cannabis use. Method: Dutch adolescents (N = 1,606; 854 girls and 752 boys) from the TRacking Adolescents' Individual Lives Survey (TRAILS) ongoing longitudinal study were examined at baseline (ages 10-12 [T1]) and at two follow-up assessments (ages 12-15 [T2] and 15-18 [T3]). The analysis focused on DSM-IV externalizing behavior (conduct, attention deficit hyperactivity, and oppositional) problems at T1, assessed by the Youth Self Report and the Child Behavior Check List, on self-reported ever smoking at T2, and on cannabis use at T3. Results: All associations of parent-rated externalizing behavior problems with cannabis were mediated by earlier smoking. Considering self-reported problems, none of these associations with cannabis were mediated by smoking, except the influence of self-reported conduct problems in girls. Interestingly, even after adjusting for externalizing problems, earlier smoking independently and consistently predicted cannabis use. The adjusted odds ratios for smoking varied in boys from 4.8 to 5.2 (ever) from 10 to 12 (daily) and from 22 to 23 (early-onset) whereas in girls from 4.9 to 5.0, 5.6 to 6.1, and 27 to 28, respectively (p less than 0.001 for all). Conclusions: Our findings challenge the view that externalizing behavior problems directly predict cannabis initiation. Such associations were inconsistent across informants and sexes and were often mediated by earlier smoking. Early smoking onset is a powerful predictor of later cannabis initiation independent of preceding externalizing behavior problems. Although externalizing behavior problems are important as a starting point for substance use trajectories, early-onset smoking should be identified as an important marker of cannabis use risk. (Contains 4 tables.)