The use of proxy respondents in surveys designed to provide population estimates of smoking prevalence offers an inexpensive way to obtain these data. The accuracy of this information is examined in analyzing data from tobacco use surveys of adults conducted in 22 North American communities as part of the National Cancer Institute's Community Intervention Trial for Smoking Cessation. Proxy-reported smoking status was obtained in a cross-sectional telephone survey conducted from August 1993 to January 1994 (n = 99,682). Self-reported smoking status was obtained from an in-depth interview of a sample of the respondents aged 25-64 years enumerated from the telephone survey (n = 31,417). Discrepancy rates were calculated by comparing the proxy-reported and self-reported smoking statuses of a given individual (n = 10,226). In both surveys, respondents were categorized as current smokers (those who currently smoke and have smoked at least 100 cigarettes in their lifetime), recent quitters ( 8 years since cessation), and never smokers. The overall discrepancy rate between the self-report and the proxy report was 5.4%. Self-respondents who were black, Hispanic, Asian, recent quitters, or aged 25-34 years were more likely to have inconsistent proxy reports. The authors estimate that the screener interview underestimated the true smoking prevalence by 0.1% when they corrected for smoking status discrepancies. These results confirm that proxy-reported smoking status is an accurate and effective means to monitor populationwide smoking prevalence of adults.