I postulate that multiple sclerosis is an autoimmune disease that involves genetically determined failure of activation-induced apoptosis of autoreactive T cells in the central nervous system. Activation of central-nervous-system-reactive T cells in peripheral lymphoid organs by exposure to crossreacting antigens or superantigens derived from common infectious agents may trigger attacks of multiple sclerosis. In normal individuals these activated T cells are deleted by activation-induced apoptosis, but in individuals predisposed to multiple sclerosis they survive, proliferate, and damage the central nervous system. The clinical course of multiple sclerosis may vary according to the antigens in the central nervous system being targeted: targeting of myelin antigens leads to a relapsing-remitting course of clinical recovery due to remyelination or other mechanisms; targeting of axonal antigens leads to a progressive course from onset because axonal regeneration is limited in the central nervous system. This hypothesis can account for many characteristics of multiple sclerosis and has predictions that can be tested.