Our aim is to examine how peer smoking affects smoking in adolescents. We would also like to look at the relationship between parental smoking, parenting and child smoking. We will use longitudinal latent class analysis in mplus to generate trajectories for adolescent and peer smoking. Adolescent smoking trajectories will be based on both self report smoking data and cotinine measures. We hypothesise that adolescents who smoke have more peers who smoke and that trajectories of adolscent smoking closely map the trajectories of peer smoking. The association between child and peer smoking may be due to influence, selection or both. Children with parents who smoke are expected to be more likely to smoke, however parental monitoring is expected to be associated with lower smoking. We also hypothesise that the association between peer and adolescent smoking may differ by level of parental monitoring. Smoking behaviour may also be associated with gender, SES (housing tenure, crowding status and maternal educational attainment) and other substance use. These covariates will also be included in our model.