Predictors of tobacco consumption were investigated with path analytic techniques using household survey data from 1,093 teenagers aged 13-17. When current smokers are compared to past or never smokers, pro-tobacco attitudes are always the most important predictor variable followed by friends' use of marijuana and own use of marijuana, suggesting that pro-tobacco attitudes are responsible for maintaining tobacco consumption. We infer that peer illicit drug use has the greatest effect in taking up smoking because it is the most important predictor of past smoker status compared to never smoking and it does not appear as such an important predictor of current use compared to never smoker status. Pro-tobacco attitudes are a consequence of smoking rather than an influence upon taking up smoking.