Indirect blood pressures recorded with a random zero sphygmomanometer were compared with simultaneous intra-arterial blood pressures recorded with the Oxford system. Twenty five patients undertook a graded bicycle exercise test, cycling at workloads increasing from 41, 65, 114, and 163 W (250 to 400, 700, and 1000 kpm per min) with each grade being maintained for three minutes unless the exercise test was terminated earlier at the point of fatigue. Intra-arterial pressures were recorded continuously and indirect measurements made at steady state levels in the 30 seconds before each change in grade and immediately after the termination of the exercise protocol. The mean difference in systolic blood pressure at 5.5 minutes of exercise showed that the indirect measurement underestimated the direct measurement. Immediately after the termination of exercise the blood pressure fell precipitiously to a highly significant degree. For both systolic and diastolic pressures there was considerably individual variability. These data confirm that indirect methods of blood pressure measurement during dynamic exercise testing are inaccurate and may provide misleading information.