We report on the first South African two-stage community prevalence study of psychiatric morbidity, conducted in Mamre, a rural "coloured' village, 50 km from Cape Town. Randomly selected adults (N = 481) were assessed using the Self-Reporting Questionnaire (SRQ) as a first-stage screen and the Present State Examination (PSE-9) was administered to a proportion of the sample (N = 121) as the second-stage criterion. Demographic, health care utilization, and substance abuse data were also collected. Using the PSE-9 CATEGO Index of Definition of 5, the weighted prevalence of psychiatric morbidity was 27.1% (confidence interval of 19.5-34.7%), the majority of cases being given a tentative diagnosis of depressive or anxiety disorder. The CATEGO algorithm may not be fully appropriate in this cultural context as there was an apparent over-diagnosis of paranoid states. The SRQ's weighted sensitivity and specificity were 0.49 and 0.82 respectively. Overall, the SRQ correctly identified 67% of cases and non-cases. No demographic variables predicted psychiatric morbidity, but there was an indirect link between morbidity and primary care utilization. Further South African studies of the validity of both the SRQ and of criterion instruments are needed. These may contribute to knowledge regarding cultural factors affecting psychiatric diagnosis.