The peptide rAc1–11 represents the dominant T cell epitope of rat myelin basic protein (MBP) in mice of the H-2^u haplotype. Residue 4 has been shown previously to govern binding of the peptide to the class II molecule, I-A^u. We have constructed peptide analogues bearing amino acid substitutions at position 4 and have assessed their ability to stimulate an antigen-specific T cell hybridoma when presented by viable antigen presenting cells (APC). Complexes between I-A^u and one such analogue, rAc1–11[4A], were rapidly lost from the surface of live APC displaying a half-life (t½) of ∼10 min. Neither shedding of intact complexes from the cell surface, nor their Internallzatlon and recycling through an acidic intracellular compartment were found to account for their loss. The possible dissociation of rAc1–11[4A] from the peptide binding cleft was therefore addressed by comparing the t½, of complexes between I-A^u and peptide analogues of higher affinity. The tyroslne-substltuted analogue, rAc1–11[4Y], remained stably bound to I-A^u for at least 4 h, thereby displaying a t½, far in excess of that evident for rAc1–11[4A]. Significantly, the wild type peptide, rAc1–11 , bound so transiently that functional complexes could not be detected on the surface of peptlde-pulsed APC. The physiological relevance of these findings was confirmed by extending our studies to an analysis of the homologous epitope of murine MBP; evidence that this epitope likewise displays minimal affinity for I-A^u suggests a novel strategy for the escape from tolerance induction by encephalitogenlc T cells.