Several cross-sectional studies have shown a decline in alcohol intake with increasing age. Longitudinal studies have failed to confirm this trend, which suggests that cohort effects may account for the reported decline. To address this, both cross-sectional and longitudinal analyses of alcohol use in 270 healthy elderly persons over a seven-year period (1980-1987) were performed. Alcohol consumption was assessed by three-day diet records. One hundred sixty-five subjects (61.1%) remained in the study until 1987; 143 (53%) completed diet records for every year. Longitudinal analysis showed a statistically significant decline in the percent of subjects consuming any alcohol over time (slope = -2% per year; 95% confidence interval -2.8, -1.1%). A cross-sectional analysis of the 1980 data revealed a similar decline in percent drinkers with increasing age (slope = -2.7% per year; 95% confidence interval -4.4, -1.1%). Mean alcohol intake for those who continued to drink did not change over time except among heavy drinkers (consumption of greater than 30 g per day in 1980), who did show a significant decline in mean alcohol intake (P = .02). Thus, in our population the decline in percent of drinkers with age found by a cross-sectional analysis was confirmed in longitudinal analyses, suggesting that this represents a true age-related decline rather than a cohort effect.