The effect of stereotactic lesional surgery for the treatment of tremor in multiple sclerosis was examined in a prospective case-controlled study. Surgery was not undertaken in 33 patients (72% of 46 cases referred for stereotactic surgery), two of whom died within 4 months of referral. Twenty-four multiple sclerosis patients were included in the study; 13 underwent surgery and were matched against 11 controls on the basis of age, sex, expanded disability system scores (EDSS) and disease duration. Assessments were carried out at baseline/preoperatively, and then 3 and 12 months later; these included accelerometric and clinical ratings of tremor, spirography, handwriting, a finger-tapping test, nine-hole peg test, tremor-related disability, general neurological examination, Barthel Activities of Daily Living (ADL) Index of general disability, EDSS, a 0-4 ataxia scale, Mini-Mental State (MMS) examination, speech and swallowing assessments and grip strength. Postoperative MRI scans demonstrated that tremor could be attenuated by lesions centred on the thalamus in seven cases, on the zona incerta in five cases and in the subthalamic nucleus in one case. Two patients developed hemiparesis and in two cases epilepsy recurred. Two surgical patients and one control patient died between the 3 and 6 months assessments. Both groups had a significant deterioration in EDSS but not Barthel ADL Index scores at 1 year, but the difference between the groups was not significant. Similarly, no differences between the groups' rates of deterioration of speech or swallowing or MMS were found. Significant improvements in contralateral upper limb postural (P2) and kinetic tremors, spiral scores and head tremor were detected at 3 and 12 months after surgery (but not handwriting or nine-hole peg test performance). Tremor-related disability and finger-tapping speed were also significantly better 12 months after surgery, the latter having significantly worsened for the control group. A 3 Hz 'filter' for postural (P2) upper limb tremor was detected by accelerometry/spectral analysis above which tremor was always abolished and at or below which some residual tremor invariably remained. Criteria for selecting multiple sclerosis patients for this form of surgery are discussed.