The vestibular system was activated by galvanic electrical stimulation in 22 patients with spinal cord injury. Three patients were studied standing and all were studied sitting. Electromyographic responses recorded in soleus (standing patients) and the erectores spinae (all patients) were compared with data from 18 control subjects. The vestibular stimulus polarity and head position were arranged so as to produce excitatory medium latency muscle responses in the controls. Responses in the patient group were present bilaterally, present unilaterally or absent below the level of injury. The amplitude of response recorded in erectores spinae at lumbar levels below the lesion in 21 patients (left and right side responses summed) and five control subjects was positively correlated with American Spinal Injuries Association (ASIA) grade: the smallest amplitudes were found in patients with the most severe impairment (Spearman rank correlation coefficient rs=0.59; P=0.002, two-tailed). The latency of response (averaged for both sides) was negatively correlated with ASIA grade in 21 patients: the longest latencies were found in patients with the most severe impairment (rs=–0.57; P < 0.01, two-tailed). Amplitude and latency were negatively correlated (rs=–0.72, P < 0.002, two-tailed). The latencies of responses recorded in the erectores spinae at different vertebral levels were linearly related to the vertical distance from the inion to the recording site in both patient and control groups. The conduction velocities of the spinal pathways activated by vestibular stimulation were 4.6 and 10.4 m/s in patient (recording below lesion) and control groups, respectively. Both clinical status (patients recording below lesion, patients recording above lesion and controls) and distance were significant predictors of latency (general linear model, P < 0.0005). It is concluded that measurement of vestibular-evoked responses could provide information on the level and density of spinal cord lesions.