Kuller, L H. (Johns Hopkins Univ. School of Hygiene and Public Health, Baltimore, Md. 21205), A. Bolker, M. S. Saslaw, B. L Paegei, C. Sisk, N. Borhani, J. A. Wray, H. Anderson, D. Peterson, W. Winkelstein, Jr., J. Cassel, P. Spiers, A. G. Robinson, H. Curry, A. M. LJIienfeld and R. Seltser. Nationwide cerebrovascular disease mortality study. II. Comparison of clinical records and death certificates. Amer. J. Epid., 1969, 90: 545–555.—The large differences in cerebrovascular disease mortality among geographic areas of the United States can not be explained by variations in certification practices such as the choice of the underlying cause of death. There may be differences in the frequency that clinical stroke diagnoses on hospital records, physician's and medical examiner's reports or from a family interview were listed on the death certificate. In order to measure the relationship between clinical stroke diagnosis and information on the death certificate, pertinent clinical records were reviewed for every death certificate included in the nationwide cerebrovascular disease mortality study. Clinical information or a family interview was obtained for 96.5% of the 16, 956 death certificates in the adjusted sample. When all sources of stroke diagnosis were pooled together, it was found that 18.9% of the deaths had a stroke diagnosis. The frequency was highest in the high cerebrovascular disease death rate areas. Approximately 60% of all stroke diagnoses were listed on the death certificate including 42.2% as the underlying cause of death. Differences among the areas were comparatively small. The sensitivity of the death certificate was highest for hospital deaths and lowest for those certified by the medical examiner, while the specificity was the same irrespective of the place of death. Differences in the frequency in which clinical stroke diagnoses were listed on the certificate did not explain the geographic differences in death rates.