The accuracy of self-report of herpes zoster was investigated in the Duke Established Populations for Epidemiological Studies of the Elderly, a longitudinal study of 4162 community-dwelling elderly persons residing in North Carolina, 1986-1993. The authors compared self-reports of zoster with physician diagnosis of zoster and with a zoster verification questionnaire (ZVQ). Compared to physician diagnosis, 3.2% (95% confidence interval 0-61%) of self-reports of zoster (n = 31) were false-positive and no denials of zoster (n = 63) were false-negative. The agreement of self-reports to physician diagnosis was 98.9%, the sensitivity and negative predictive value were 100%, the specificity was 98.4% and the positive predictive value was 96.7%. The ZVQ comparisons were similarly high. These data suggest that the frequency of false-positive and false-negative reports of zoster is low in this elderly population. Zoster self-reports appear to be accurate and suitable for epidemiological investigations.