A cross-sectional survey was conducted to consider whether there is agreement between self-report and clinical diagnosis in the prevalence of selected common chronic diseases among elderly Taiwanese. Between July 1992 and June 1993, both face-to-face interview and clinical evaluation were applied to a sample of 228 Taiwanese aged 65 y and older to estimate the prevalence of heart diseases, hypertension and diabetes. The results showed that a self-reported history of diabetes had the highest sensitivity (66. 7%) and specificity (95.2%). The self-report of heart diseases was the least sensitive (20.5%), while self-reports of hypertension was the least specific (82.8%). Multivariate analyses showed that age, education and number of self-reported diseases suffered from, appeared to influence the accuracy of the self reported data. Our data suggest a notable lack of agreement abetween self-report and clinical diagnosis for medical conditions, even for those with clear and unambiguous diagnostic criteria. The elderly in Taiwan tended to under report (with a consistently higher false negative rate than false positive rate) the existence of the three selected medical conditions. This was especially so for heart diseases. Our findings suggest that, if the level of disease prevalence in the population needs to be known for health planning estimates for hypertension and diabetes would not necessarily be a problem. However, the prevalence for heart disease would be seriously under-estimated. We also argue that information obtained by self-report should be validated before use, especially in the case of assessment of the prevalence of medical conditions in the elderly.