BACKGROUND: Several studies that have examined the full range of alcohol consumption have pointed to a possible non-linear association between alcohol use and the common mental disorders. Most of these studies are cross sectional and assessed psychiatric morbidity using non-specific instruments. Our aim was to investigate the longitudinal association between varying levels of alcohol consumption at baseline and the new-onset of depression and generalized anxiety disorder (GAD), in a large international primary care sample. METHODS: The sample consisted of 3201 primary care attenders from 14 countries in the context of WHO Collaborative Study of Psychological Problems in General Health Care. Alcohol use at baseline was assessed using AUDIT and the mental disorders were assessed with the Composite International Diagnostic Interview. RESULTS: Light to moderate alcohol consumption at baseline was associated with a lower incidence of depression and GAD compared to abstinence while excessive alcohol consumption was associated with a higher incidence of depression but not GAD. This non-linear association was not substantially affected after adjustment for a range of possible confounding variables. CONCLUSION: Any causal interpretation of this association is difficult in the context of an observational study and further combined and consistent evidence from different sources is needed.