Three studies of blood pressure measurement were conducted in two primary care clinics. In study I, procedures for taking blood pressure were observed for seven nurses, and their readings were compared with readings taken in a standard manner. The results show that the nurses' procedure needed improvement; the systolic/diastolic discrepancies with the standard averaged 10.2/7.5 mmHg. In study II, the nurses were trained to improve their procedure, and readings were again compared. The results show improvement for moderate, but not for large, discrepancies in diastolic readings. In study III, two physicians' procedure for taking blood pressure indicated that their procedures could also be improved. Overall, there was little correlation between potential biases in providers' procedures and discrepancies with the standard. The discrepancies between the nurse and physician readings led to different treatment indicators in 22-26 per cent of the observations.