Under etomidate anesthesia, motor evoked potentials produced by magnetic stimulation were successfully recorded from 10 thenar muscles and 10 anterior tibial muscles of eight patients who had undergone surgery on the medulla oblongata and the cervical and thoracic spinal cords. Recordings taken before placing the neural tissue at risk were assessed for variability in amplitude and latency. The lower limit in amplitude was approximately one-third (25-43%) of the baseline. The latencies were more difficult to monitor than were the amplitudes. The latency variations were 2.56 +/- 0.50 milliseconds for the hand and 6.84 +/- 1.37 milliseconds for the leg. During surgery, the unilateral recordings of two patients were transiently lost but partially recovered after a pause in the operation. No obvious postoperative weaknesses in the corresponding limbs occurred. One patient, who showed a permanent loss of unilateral recording, had transient monoplegia with a complete recovery. None of the remaining five patients who had amplitudes larger than one-third of the baseline at the end of the operation had additional motor deficits. Our conclusions are that under etomidate anesthesia, the magnetic motor evoked potentials can be convenient and reliable monitors of motor function, that changes in the amplitude may be superior to those in the latency for intraoperative warning, that the criterion for potential neural damage under magnetic motor evoked potential monitoring might be an amplitude reduction of two-thirds of the control value, and that the magnetic stimulation seems to be more sensitive than the electrical stimulation in the monitoring of motor function and also allows more time and opportunities for the motor function to recover.