The prognostic implications of comorbid depression for outcome from alcohol dependence are unclear. It has been suggested that drinking may represent self-medication of depressive disorders and, alternatively, that the pharmacological properties of alcohol induce episodes of depression. In the present study, these questions were investigated by following 84 alcohol-dependent individuals seeking treatment (34% women) for 3 years in a naturalistic, prospective design. During the follow-up period, depressive and drinking outcomes were significantly related. However, there was no evidence that drinking reliably preceded depressive episodes or that depressive episodes precipitated heavy drinking. Despite their high co-occurrence, this study did not find evidence of a strong, direct causal relationship between these phenomena.