Train-of-four (TOF) monitoring of neuromuscular block at submaximal current stimulation has been recommended because it is less painful than monitoring at supramaximal stimulation. The measurement error, however, when using submaximal stimulation has not as yet been fully elucidated. The authors therefore quantified the measurement error of TOF monitoring at low currents near the initial threshold for stimulation (ITS) by assessing precision (the difference between duplicated TOF ratios measured at the same current) and accuracy (the difference between TOF ratios at lesser currents and the TOF ratio at 58 mA, which served as the reference standard method). A stable neuromuscular block at a TOF ratio of 0.40 and subsequently 0.70 was established in 20 patients using a continuous infusion of atracurium. The ulnar nerve was stimulated at the wrist using TOF stimulation, and the evoked twitch response of the adductor pollicis muscle was measured with a force-displacement transducer. The stimulating current was varied in 5-mA steps between ITS and 55 mA. Whereas there were no statistical significant differences between the mean TOF ratios at the different currents, the standard deviation increased with decreasing currents (P < 0.01 at currents < 40 mA). The precision was acceptable except at ITS, where the limits of precision (mean intracurrent difference in TOF ratio ± two standard deviations) exceeded 0.05, which was the chosen acceptable difference. The standard deviation of the accuracy was significantly increased at currents ≤ ITS + 25 mA (P < 0.01). At currents < ITS + 25 mA, the limits of accuracy (mean intercurrent difference in TOF ratio ± two standard deviations) exceeded the chosen acceptable difference. It is concluded that the accuracy of TOF ratio monitoring is unacceptable at currents < ITS + 25 mA.