Data from population-based studies are important for the study of comorbidity. Cross-sectional research shows a consistent positive association of diabetes and depression. Prospective population-based studies are reviewed, showing that the temporal order may be from diabetes to depression, or from depression to diabetes, depending to some extent on the type of diabetes. The size of the effects is fairly consistent among the small number of studies and not trivial. Possibilities for future epidemiologic research on comorbidity are discussed.