In previous studies the estimation of the incidence of intellectual changes in multiple sclerosis has varied from as low as 2 per cent. (Cottrell and Wilson, 1926) to as high as 72 per cent. (Ombredane, 1929). In order to establish the incidence accurately, the most satisfactory method would clearly be a long-term follow-up (covering lifetime) of a sample of multiple sclerosis patients. Any cross-sectional study would include patients in varying degrees of advancement of the disease, and patients found to be free of intellectual deficits at the time of investigation would not necessarily remain so. Also, the indirect psychometric assessment of intellectual loss is notoriously difficult, and the direct method of follow-up would give much more accurate results. The only direct study up to date has been that of Canter (1951), who found a highly significant (i.e. 13·48 points) loss on re-testing multiple sclerosis patients on the Army General Classification Test after a four-year period. Even after such a short period as six months he found slight losses on most Wechsler-Bellevue subtests, in contrast to an average gain of six full IQ points of the control group.