Background: This study tested the hypothesis that data from random digit-dial telephone surveys underestimate the prevalence of cigarette smoking among African-American adults. Method: A novel, community-sampling method was used to obtain a statewide, random sample of N = 2118 California (CA) African-American/Black adults, surveyed door-to-door. This Black community sample was compared to the Blacks in the CA Health Interview Survey (N = 2315), a statewide, random digit-dial telephone survey conducted simultaneously. Results: Smoking prevalence was significantly higher among community (33%) than among telephone survey (19%) Blacks, even after controlling for sample differences in demographics. Conclusion: Telephone surveys underestimate smoking among African-Americans and probably underestimate other health risk behaviors as well. Alternative methods are needed to obtain accurate data on African-American health behaviors and on the magnitude of racial disparities in them.