Recent studies of the spinal cord and cerebellum have highlighted the importance of atrophy in the development of neurological impairment in multiple sclerosis. We have therefore developed a technique to quantify the volume of another area commonly involved pathologically in multiple sclerosis: the cerebral white matter. The technique we describe extracts the brain from the skull on four contiguous 5 mm periventricular slices using an algorithm integrated in an image analysis package, and quantifies their volume. Intra-observer scan-rescan reproducibility was 0.56%. We have applied this technique serially to 29 patients with multiple sclerosis selected for an 18-month treatment trial with a monoclonal antibody against CD4+ lymphocytes (deemed clinically ineffective). A decrease in volume beyond the 95% confidence limits for measurement variation was seen in 16 patients by the end of the 18-month period. The rate of development of atrophy was significantly higher in those who had a sustained deterioration in their Kurtzke expanded disability status scale (EDSS) score compared with those who did not (respective means: -6.4 ml year-1 and -1.8 ml year-1, P < 0.05) but in both groups these changes differed significantly from baseline (P < 0.05). Baseline T2 lesion load, change in T2 lesion load over 18 months and the volume of new gadolinium enhancing lesions on monthly scans for the first 10 months showed no correlation with the development of atrophy. This study demonstrates that progressive cerebral atrophy can be detected in individual patients with multiple sclerosis, correlates with worsening disability and gives additional information to that obtained with conventional MRI. The effect of putative therapies aimed at preventing disability could be objectively assessed by this measure.