The random zero sphygmomanometer is widely used in studies involving blood pressure measurement because it is believed to eliminate digit preference and reduce measurement error. We performed blood pressure measurements sequentially using random zero and standard sphygmomanometers in random order in 1,356 participants in the Cardiovascular Health Study. Despite adherence to the manufacturer's instructions, we observed a substantially nonuniform distribution of zero levels generated by the random zero sphygmomanometer and a disturbing correlation between the zero level and blood pressures taken with the standard sphygmomanometer. With the random zero device, the pooled estimated slopes for the regression of standard systolic and diastolic pressures on the zero level were −0.71 and −0.17, respectively (bothp<0.0001). The only plausible explanation for this relation between the random zero device and the standard device is that by some unknown mechanism the subject's blood pressure is influencing the zero level. Systolic and diastolic blood pressures measured with the random zero device were, respectively, 1.65 and 1.84 mm Hg lower (bothp<0.0001) than standard blood pressures. Digit preference was detectable in the uncorrected blood pressure and zero level measured with the random zero device but was eliminated after calculation of the corrected blood pressure. For most epidemiological studies, the random zero sphygmomanometer offers no significant advantage over the standard sphygmomanometer. It may still be useful in those epidemiological studies and clinical trials where blinding is important.