During the period 1960-1972, the medical record linkage index method provided an estimated average annual age- and sex-adjusted rate for first transient ischemic attack (TIA) of 134 per 100,000 population for those age 50 years and older. The cohort method estimate was 237 per 100,000 population (95% confidence limits (CL), 165-310). The discrepancy was due to 18 cases of TIA noted at the time of patient visits in the cohort analysis that were not identified from the medical record analysis. The cohort analysis for all strokes at age 50 years and older was not significantly different from the medical record linkage analysis. Although the cohort follow-up method provided better case ascertainment in this study, current indexing procedures provide case ascertainment equal to that of the cohort method and at less cost. No trend was found to suggest that incidence rates for TIA were decreasing in the period 1955-1979, in contrast to stroke in Rochester, Minn.