Procedures are described to deal with problems of sampling variability and of changes in sampling design in cohort studies that utilize national survey sample data. Such studies with existing data are inexpensive and may vastly increase knowledge of the attitudinal and behavioral correlates of aging. For inferring the effects of aging, these studies are superior to both crosssectional and panel studies. An illustrative study with United States data shows a substantial increase in the percentage of teetotalers in the cohorts that matured beyond middle age from 1945 to 1960, and this increase cannot be attributed to influences that would have increased teetotalism in the total population.