A series off 3 BP measurements were carried out on 2 groups of people. People not accustomed to BP measurements showed a decrease in pressure between the 1st and 3rd cuff inflation, while people trained to having their BP measured showed no such change. Measurement of the BP of 111 people on 2 successive occasions 8 days apart showed that the mean BP of the whole group decreased between visits. Approximately one third of the people showed a drop in pressure bringing them from borderline hypertension to normotension between the 2 visits. People whose BP dropped substantially between the 1st and 2nd visits achieved higher anxiety scores on a questionnaire than did the remainder of the sample. This may partially explain their labile BP. Single measurements tend to overestimate the BP and the frequency of hypertension. Training experimental subjects by repeated measurement of their BP may be useful in preventing spurious BP changes unrelated to experimental manoeuvres.