The substantial prevalence of noncompliance with drugs in populations of patients with treated hypertension suggests that a reliable, clinically applicable screening test for this behavior is needed. Among employed white males, patient response to a nonjudgmental clinician inquiry has been reported to be highly predictive of noncompliance when positive but relatively insensitive (40 per cent). We assessed the performance of patient self-report in a demographically different population and confirmed the generalizability of prior observations. A compound decision rule combining blood pressure and verbal inquiry observations, however, had higher sensitivity (83 per cent) for noncompliance in our population than self-report alone (55 per cent) and could be considered for use when the prevalence of noncompliance among uncontrolled hypertensives is sufficiently high. Because 40 per cent of well-controlled hypertensives in our test population were noncompliant by pill-count, a question is raised regarding the need for "stepping down" drug therapy in some individuals under treatment for high blood pressure.