The reproducibility and sources of variation in ankle and brachial systolic pressures were determined for a reactive hyperaemia test used to assess peripheral arterial disease in the lower limbs. Thirty-six subjects had repeat measurements taken by four observers on two separate days using a Doppler probe and random zero sphygmomanometer. Ankle pressures were measured before and after 4 min of arterial occlusion above the knee. The 95 per cent confidence limits for a postocclusion reduction in pressure was approximately +/- 25 mmHg and for the postocclusion ankle-brachial ratio was +/- 20 per cent. The corresponding figures for the mean of four measurements taken by two observers on 2 days was approximately +/- 17 mmHg and +/- 13 per cent. In patients with vascular disease analysis of variance showed that the greatest sources of variation in the postocclusion ankle-brachial ratio were the 'biological' variations between subjects and between legs. But for postocclusion reductions in pressure, variability due to observers and repeat readings was as important as the 'biological' differences between subjects and legs. Thus, in respect of variability, the ratio may be a better parameter to detect differences between subjects.