To determine the efficacy of motor evoked potentials (MEP) and sensory evoked potentials (SEP) in the assessment of severe cervical injury, 17 subjects with severe cervical injury were studied. During the 1st week post-injury and post-surgical treatment, all subjects were submitted to electromyogram (EMG) recordings, dermatomal somatosensory evoked potentials (D.SEP), posterior tibial nerve somatosensory evoked potentials (PTN.SEP), MEP and bilateral cervical electrical stimulations with recording of the diaphragm. For the D.SEP, the latencies of the N9 and N20 responses and the conduction time (N9-N20) were measured in the upper limbs; the latencies of the P40 and P60 responses were measured in the lower limbs. MEP were recorded from distal upper and lower limb muscles following transcranial electrical stimulation of the cortex. (Magnetic stimulation was not indicated because of implanted metallic material in the cervical skull of many patients.) A SEP and MEP grading system was used to improve the assessment of different root neurological levels. In patients with incomplete lesions PTN.SEP, D.SEP and MEP responses could be recorded in territories that were clinically deficient. Patients with complete lesions and absent SEP and MEP responses had a poor outcome. A good correlation was found between the severity of the spinal cord injury and SEP grading. For MEP, the presence or absence of intercostal responses (C4) to cervical and cortical stimulation was the best prognostic indicator. The combined electrophysiological exploration of MEP and SEP proved to be a useful tool for monitoring patients with severe spinal cord injury.