Thirty-one cases (26 patients) of clinically definite multiple sclerosis with predominant spinal cord involvement were studied retrospectively, focusing in particular on the results of sensory testings which were compared with the findings of spinal cord magnetic resonance imaging (MRI) and somatosensory evoked potentials (SEPs). Factor analysis of the sensory impairments showed two factors corresponding to superficial (pain and light-touch) and deep (position and 'thumb/big toe localizing') sensations. Vibratory sense depended on both factors, but more on deep sensations. Multiple regression analysis of the SEP abnormalities and sensory deficits showed that impaired deep sensations contributed to abnormal SEPs, but the results of clinical sensory testings and SEPs were dissociated in 31% of the extremities examined. The MRI sagittal images correlated well with the levels of sensory impairments, whereas the axial images showed a poor relationship to the extent of sensory impairments. We concluded that the use of SEPs as well as MRI is not always superior to such clinical sensory tests as the 'thumb/big toe localizing tests'.