In the context of a collaborative study on the epidemiology of hypertension in populations of West African origin procedures for standardization of the measurement of blood pressure were evaluated. Comparisons of mean levels of blood pressure, which in large part determine prevalence rates, are highly sensitive to differences in technique. While rotating a single field team may be the ideal approach to multisite studies, it is not practical in international collaborative research. Appropriate techniques to standardize multiple teams over a long period of time have not been well developed, however. In the present study 8981 individuals were examined in eight sites in six countries with the standard mercury sphygmomanometer. An evaluation of the effectiveness of central training, site visits, monitoring of digit preference, and the use of an electronic device for internal standardization is described. In all but one of the sites reliability was high and comparable to the observers at the Coordinating Center. Digit preference for the entire set of measurements was limited (frequency of terminal zero = 23.5% for systolic and 28.9% for diastolic readings) and could be shown to have virtually no effect on prevalence rates or correlation estimates. Mean differences among observers within a given site and between sites were small (+/- 0-5 mmHg). While logistically complex, these methods can provide the basis for standardization in international comparative blood pressure surveys.