Motor evoked potential monitoring was tested as an alternative to somatosensory evoked potential monitoring in evaluating spinal cord function during thoracic aortic occlusion in dogs. Twenty-seven animals underwent 60 minutes of cross-clamping of the proximal descending thoracic aorta with (n = 18) or without (n = 9) cerebrospinal fluid drainage. Spinal cord blood flow was measured with microspheres, and neurologic outcome was evaluated at 24 hours with Tarlov's scoring system. Cerebrospinal fluid drainage improved neurologic outcome (p less than 0.05). Motor evoked potentials recorded over the lumbar spinal cord were lost in 9 of 20 dogs with ischemic cord injury and were not lost in any of the 7 dogs that were neurologically normal. Somatosensory evoked potential were lost in 19 of 20 paraplegic/paraparetic dogs and lost in 3 of 7 normal dogs (p less than 0.01). After reperfusion, motor evoked potentials returned in all nine neurologically injured dogs that lost the potentials and were still present at 24 hours. Changes in amplitude, latency, or time until loss or return of motor evoked potentials or somatosensory evoked potentials did not predict neurologic injury. Loss of somatosensory evoked potentials had a high sensitivity (95%) but had low specificity (67%) because of peripheral nerve ischemia. Loss of motor evoked potentials recorded from the spinal cord had high specificity (100%) but a low sensitivity (46%) and was therefore not a reliable predictor of neurologic injury. Return of motor evoked potentials during reperfusion did not correlate with functional recovery. Motor evoked potentials stimulated in the cortex and recorded from the spinal cord had low overall accuracy (59%).(ABSTRACT TRUNCATED AT 250 WORDS)