The highly encephalitogenic guinea pig peptide 68-88 has been used to develop an effective and reproducible model of protection in the Lewis rat. Doses as low as 0.1 nmol of peptide protected 70% of rats when injected intraperitoneally six and four weeks prior to challenge with 0.05 nmol of the peptide in complete Freund's adjuvant. Fragments derived from guinea pig peptide 68-88 by selective enzyme cleavage were then tested for their capacity to provide protection in this model system. These fragments had previously been well characterized both biochemically and immunologically. The protection provided by each fragment closely paralleled its capacity to induce disease. This suggests that the region of peptide 68-88 required for protection is similar to that needed for induction of experimental allergic encephalomyelitis and the other T-cell functions of the peptide. B-cells did not appear to participate; peptide 68-85, which has no capacity to produce antibody against peptide 68-88, gave full protection, while peptide 79-88, which contains the major B-cell determinant of the peptide, afforded no protection. Rat peptide 68-88 did not protect against challenge with the guinea pig peptide, demonstrating a critical role for serine 79. These studies support the concept that nonencephalitogenic agents do not protect against experimental allergic encephalomyelitis at doses comparable to those of encephalitogenic agents.