Ocular movements were studied in 108 patients with established or suspected multiple sclerosis using an on-line computer-based electro-oculographic technique. In one group of patients peak eye movement velocities alone were measured during horizontal refixation saccades. In a second group saccade reaction times and accuracies were measured in addition to velocities, while in a subgroup a quantitative analysis of horizontal pursuit eye movements was also carried out. With the saccade velocity test abnormalities were present in 44 per cent of cases studied and were subclinical in 18 per cent. Abnormalities were found in 57 per cent of cases in whom the detailed saccade analysis was performed, including 48 per cent of patients with clinically normal eye movements. Saccade reaction time and accuracy were more sensitive parameters than saccade velocity, and the highest yield of abnormalities was obtained when all three were taken into consideration. Abnormalities of pursuit movements were found in 71 per cent of cases studied and were frequently subclinical. Abnormalities of saccadic and pursuit movements were not always present together in the same patient, and the overall yield of abnormalities was higher when the results of both types of study were taken into account. The yield of abnormalities with the eye movement studies was somewhat lower than with the pattern-reversal VEP in the clinically definite multiple sclerosis group, but was higher in patients in the other categories. Subclinical abnormalities of eye movement were found in a significant number of patients with normal VEPs. The finding of such an abnormality in patients with spinal cord syndromes allowed reclassification of 14 patients to a category with a higher degree of diagnostic certainty. It is concluded that quantitative electro-oculography is a valuable adjunct to the clinical evaluation of eye movements and has an important role in the investigation of patients suspected of multiple sclerosis.