Purpose: There is substantial research linking tobacco and alcohol use to subsequent cannabis use, yet the specificity of this relationship is still under debate. The aim of this study was to examine which substance use model-the gateway hypothesis, the common liability (CL) model and/or the route of administration model-best explains the relationship between early onset of tobacco and alcohol use and subsequent cannabis use initiation. Methods: We used data from 2,113 (51% female) Dutch adolescents who participated in three consecutive assessment waves (mean age: 11.09, 13.56, and 16.27 years, respectively) of the TRacking Adolescents' Individual Lives Survey study. (Pre) adolescent cannabis, tobacco and alcohol use was assessed using the Youth Self-Report and a TRacking Adolescents' Individual Lives Survey developed questionnaire. Results: We found that, during adolescence, early onset of tobacco use does not pose a significantly higher risk of initiating cannabis use than early onset alcohol use. Therefore, we can rule out the route of administration model. Moreover, we found that adolescents who reported early onset comorbid use of both tobacco and alcohol have a higher likelihood to initiate cannabis use than adolescents who have tried either tobacco or alcohol. The gateway hypothesis is not broad enough to explain this finding. Therefore, the CL model best predicts our findings. Conclusion: Future research on adolescent cannabis initiation should focus on testing the robustness of the CL model. Furthermore, identifying adolescents who use both tobacco and alcohol, before the age of 13, may help to curtail the onset of cannabis use. (C) 2011 Society for Adolescent Health and Medicine. All rights reserved.