A vasopressin metabolite, AVP4-9, was injected 1 h prior to retention tests at 1, 6, 11, and 16 days after learning to test the hypothesis that the peptide exerts qualitatively different effects on memory retrieval depending on the accessibility of the memory. The findings provided strong support for this hypothesis: At a retention interval associated with excellent recall in control animals, pretest administration of AVP4-9 (3.0 μg/kg) significantly impaired memory, while this same treatment significantly improved recall at an interval associated with poor retention in controls. At retention tests associated with intermediate recall in controls, retrieval was not significantly affected by the peptide treatment. This pattern of results indicates that the peptide treatment is interacting with endogenous changes that correspond to the accessibility of the memory. © 1990 Academic Press, Inc.