This longitudinal study aimed at comparing aggregate measures of heavy or problem drinking and their variations across time among the same subjects. We examined middle-aged men participating in a health survey over a 5-year interval. Of the 133 consecutive men in the whole age group interviewed as 40-year-olds in 1989, 114 were reached and re-interviewed in 1994. Alcohol consumption was measured by self-report, Malmo-modified Michigan Alcoholism Screening Test (Mm-MAST), and serum carbohydrate-deficient transferrin (CDT). Self-reported alcohol consumption decreased with years (142 vs 105 g/week, P = 0.01), as did CDT (16.9 vs 14.4 U/l, P = 0.02), but there was no change in the Mm-MAST results. There was no significant difference in the number of heavy drinkers (either Mm-MAST score > or = 3, or by self-reported alcohol consumption > or = 280 g/week, or by CDT > or = 20 U/l) at 40 and 45 years of age (37 and 47% respectively). At the individual level, alcohol consumption both increased and decreased with age. At 45 years of age 5/114 (4%) of the men reported that they had increased their alcohol consumption by more than 80 g/week and 25/114 (22%) said that they had reduced their drinking by the same amount. The remaining 84 (74%) reported drinking the same amount as 5 years earlier (+/- 80 g/week). This indicates that alcohol drinking habits are not stable in middle age. Most heavy drinkers in both age groups were detected by Mm-MAST and this proportion increased with age while the proportion of positive self-reports and CDTs decreased. Thus, the social consequences, measured here by the Mm-MAST, may be more readily experienced with years even at smaller consumption levels.