Aims: Little is known about the heterogeneity in and risk factors associated with trajectories of smoking during adolescence. This study aimed to (1) identify smoking trajectories empirically and (2) identify risk factors for trajectory group membership.

Design: Latent growth mixture models were used to identify population smoking trajectories, and logistic regression was used to estimate risk factors for group membership.

Setting: The participants were drawn from seven middle schools in a Maryland school district.

Participants: Participants consisted of 1320 6th graders who were followed to the 9th grade.

Measurements: Measurements of smoking risk factors were made in the fall of 6th grade and smoking stage was assessed on five different occasions between the fall of 6th and 9th grades.

Findings: Five distinct smoking trajectories were identified. Overall, being female, having friends who smoked, deviance acceptance and outcome expectations were associated with an increased likelihood of being an intender, delayed escalator, early experimenter and early user compared to a never smoker. Additionally, comparisons with never smokers revealed unique identifiers for intenders, early experimenters and early users, but not delayed escalators.

Conclusions: There is much heterogeneity in the manner in which middle-schoolers progress from having no intention of smoking to becoming smokers. Implications for prevention programs are discussed.