The population prevalence of multiple sclerosis is 0.1 %¹; however, the risk of the disease in the siblings of affected individuals is very much higher at 3–5%^2,3. The importance² of genetic factors in accounting for this increased risk is confirmed by the results of twin and adoption studies^4–7. Despite the evidence for a strong genetic effect, a weak major histocompatibility complex (MHC) association is the only consistently observed feature in the genetics of multiple sclerosis^8–11. Other candidates have been proposed, including genes encoding the immunoglobulin heavy chain^12–14, T cell receptor β chain^15,16 and APOC2¹⁷, but none has yet been confirmed^18,19. Evidence for linkage and association to the myelin basic protein gene has been reported in a genetically isolated Finnish population²⁰, but it has not been possible to reproduce these results in other populations^21,22. We used a two-stage approach to search the human genome for the genes causing susceptibility to multiple sclerosis. Two principal regions of linkage are identified, chromosomes 17q22 and 6p21 (MHC). Our results are compatible with genetic models involving epistatic interaction between these and several additional genes.