OBJECTIVE: The purpose of the study was to explore the predictive value of women's alcohol habits in relation to incidence of diabetes and all-cause mortality. Special attention was paid to potential confounding factors such as age, heredity, education, socioeconomic group, physical inactivity, smoking, blood pressure, serum lipids, and, in particular, obesity. RESEARCH DESIGN AND METHODS: A longitudinal population study consisting of a representative sample of 1,462 women aged 38-60 started in Göteborg, Sweden, in 1968-1969 monitoring for diabetes and mortality over 32 years. RESULTS: Alcohol intake, expressed as intake of wine, hard liquor, or total grams of alcohol, was significantly negatively associated to 32-year diabetes incidence independent of age. However, the apparently protective effect of the alcohol variables was attenuated when BMI was included as a covariate. The inverse relationship between wine intake and diabetes did not remain after adjustment for physical activity or socioeconomic group. Beer and wine intake were significantly negatively associated to mortality. Increase of alcohol intake between the examination in 1968-1969 and 1980-1981 was significantly inversely related to the mortality between 1980-1981 and 2000-2001 and independent of all covariates. No relationship was observed between an increase in alcohol intake and diabetes incidence. However, after adjustment for age, family history, and basal alcohol consumption altogether, a significant inverse relationship was observed between increase of alcohol and diabetes incidence. CONCLUSIONS: The initially significant inverse associations observed between alcohol and diabetes as well as mortality were dependent on a number of confounding factors, of which BMI seems to be the most important.