AIMS: Longitudinal full cross-lagged models are essential to test causal relationships. This study used such a model to test the predictive value of internal (enhancement and coping) and external (conformity and social) drinking motives for changes in alcohol use over time, and tested possible reversed causality (i.e. alcohol use explains later drinking motives). DESIGN: Longitudinal data consisting of two waves (separated by 1 year) were used to estimate cross-lagged structural equation models. SETTING: Three comparable (regarding urbanization and social stratification) Dutch communities. PARTICIPANTS: A total of 454 alcohol-using adolescents aged 13-16 years (mean = 14.8 years, SD = 0.78) at wave 1. MEASUREMENTS: Standardized questionnaires including the Drinking Motive Questionnaire-revised, and items on total weekly consumption and frequency of heavy episodic drinking. FINDINGS: In adolescence, drinking motive preferences are already relatively stable over time. Also, only social motives significantly predicted increases in total weekly consumption and frequency of heavy episodic drinking. No feedback mechanisms by which alcohol consumption explains later drinking motives scores were found. CONCLUSIONS: Among drinking adolescents in a wet drinking culture, such as the Dutch drinking culture, social drinking motives, rather than enhancement or coping motives for drinking, appear to predict overall consumption and frequency of heavy episodic use a year later. Parents and other important social actors have an active role in reducing alcohol availability and monitoring adolescents' drinking.