OBJECTIVE To report our experience with neurophysiological monitoring of corticospinal function using compound muscle action potentials (CMAPs) produced by repetitive transcranial electrical stimulation in a large series of patients, after defining optimal stimulus parameters in a small group of patients. METHODS In 100 patients undergoing spinal surgery, corticospinal volleys were recorded using epidural electrodes, or CMAPs were recorded from innervated muscles, or both techniques were used to monitor spinal cord function. In subsets of patients, stimulus parameters were varied to determine the optimal parameters for CMAP recordings, using the corticospinal volleys to guide the initial choice. RESULTS Recordings of corticospinal volleys indicated that less energy was delivered to the cortex if the duration of each stimulus in the stimulus train was brief (e.g. 50 micros) and that there was attenuation of D and I waves in the corticospinal volley when the interstimulus interval in the train was <5 ms. An interstimulus interval of 5 ms proved significantly more effective than an interstimulus interval of 2 ms in evoking CMAPs, but resulted in a more complex, dispersed electromyographic (EMG) potential. The superiority of the 5 ms interval did not depend on stimulus intensity or the existence of pre-existing neurological deficit. Using trains of 5 pulses of duration 50 micros, interstimulus interval 5 ms and intensity 500 V, satisfactory CMAPs could be recorded in 55 of 82 patients, significantly less often in neurologically impaired patients than in neurologically normal subjects. Epidural recordings of the corticospinal volley were obtained in 61 of 69 patients, again more often in neurologically normal subjects. CONCLUSIONS When epidural recordings can be made, direct recordings of corticospinal activity are probably more reliable than recordings of CMAPs. However, epidural recordings are not suitable under all circumstances, and the ability to record CMAPs reliably represents an advance in intraoperative monitoring. Under the anaesthetic conditions used in the present study, the optimal stimulus parameters consist of a train of 5 stimuli of 50 micros duration at an interstimulus interval of 5 ms and an intensity of 500 V.