OBJECTIVE To evaluate the response rates when random digit dialing was used as a substitute for geographic area sampling and household interviews to recruit 2100 African Americans for a blood pressure measurement and hypertension-related knowledge and attitudes survey. METHODS Random digit dialing was used to identify African American adults living in 12 low-income ZIP code areas of Houston, Texas. A brief survey of hypertension awareness and treatment was administered to all respondents. Those who self-identified as African American were invited to a community location for blood pressure measurement and an extended personal interview. An incentive of $10 was offered for the completed clinic visit. A substudy of nonrespondents was carried out to test the effectiveness of a $25 incentive in increasing the response rate. Data from the initial random telephone interview were used to identify differences between those who did and did not attend the measurement session. RESULTS Ninety-four percent of eligible persons contacted completed the telephone survey, and 65% agreed to visit a central community site for blood pressure measurement. In spite of the financial incentive and multiple attempts to reschedule missed appointments, only 26% of the 65% who agreed to attend completed the scheduled visit. In the substudy of the higher financial incentive, all of those who missed the original appointment agreed to another appointment, and 85% of this subgroup kept it. Not being employed full-time and a history of hypertension were consistently associated with agreement to be measured and keeping an appointment. In spite of the low response rate for scheduled appointments, differences--other than in employment status and a history of hypertension--between responders and nonresponders were small and consistent with what is usually observed in health surveys. CONCLUSIONS The use of random digit dialing as a substitute for area sampling and household screening resulted in unacceptably low response rates in the study population and should not be undertaken without further research on ways to increase response rates.