OBJECTIVE Median and ulnar somatosensory evoked potentials (SEP) in the assessment of cervical spinal cord injury (SCI) in relation to the outcome of hand function. METHODS Cohort analytical study. METHODS Spinal cord injury center, university hospital. METHODS Consecutively sampled, 23 acute and 46 chronic tetraplegics. METHODS (1) Median and ulnar SEP recordings graded in 5 categories, from normal to absent; (2) clinical examination of algesia/aesthesia in dermatomes C5/6 and C8/Th1 graded in categories of normal, impaired, and loss of sensation; (3) clinical examination of hand function graded in 4 categories, from active to inactive hand function. RESULTS Pathological median and ulnar SEP were found in 46% and 76% of patients, respectively. A differentiation between upper (C3-C6) and lower (C6/C7-Th1) cervical SCI was possible by comparing the SEP N20 latencies (but not amplitudes) of both nerves (Mann-Whitney U test, p < .01). In follow-up examinations over 6 months the N20 amplitude increased significantly (ANCOVA, p < .001), while the N20 latency did not change. By both ulnar SEP recordings and assessment of the sensory deficit in dermatome C8/Th1 the outcome of hand function, i.e., the development of active or inactive hand function, could be predicted (Mann-Whitney U test, p < .01). CONCLUSIONS Median and ulnar SEP are valuable to indicate the level of injury, the degree of sensory impairment and to predict the outcome of hand function even in unconscious patients. They can improve the diagnostic assessment of cervical SCI.