In this study, we tested to what extent smoking-specific parenting and changes in parenting are related to adolescents' smoking trajectories. Data were used from a four-wave prospective study including 428 adolescents (aged M = 15.2; SD = 0.60). Latent Class Growth Analyses were conducted to identify trajectories. Multinomial Logistic Regression Analyses were executed to examine the relations between parenting and class membership. Longitudinal cross-lagged models were tested to examine causal predominance between parenting and smoking. Four trajectories were found, consisting of Non-smokers, Increasers, Stable smokers and Decreasers. Quality of parental smoking-specific communication was strongly related to the membership in different trajectories. Along with the cross-lagged associations demonstrating that the quality of communication was predominantly related to future smoking rather than vice versa, these findings indicate that parents who talked about smoking in a constructive and respectful manner were more likely to have non-smoking children. In contrast, parents who talked often about smoking-related issues and increased these discussions over time were more likely to have smoking children; cross-lagged associations indicated that these findings could be best explained by children changing their parents. Having a non-smoking agreement was related to a lower risk in becoming a regular smoker.