The fall in blood pressure that is commonly observed when groups of hypertensive individuals are followed without treatment is usually ascribed to two sources: regression towards the mean and increasing familiarity of the subject with the assessment process. Any effect of the latter process could bias the results of controlled studies in which one group is more frequently assessed than the other, a common situation in community trials. To assess the effect of familiarity, we randomly allocated 116 untreated, mildly hypertensive subjects to three-monthly or yearly assessments. At an independent, blind, year end assessment, both groups showed statistically significant reductions in diastolic blood pressure (P less than 0.001) of 8.4 +/- 1.2 (s.e.m.) and 7.6 +/- 1.6 mmHg respectively, but the difference between the groups was not significant (P = 0.682). We conclude that 'familiarity' does not play an important role in the reduction of blood pressure in long-term follow-up studies of hypertensive subjects.