Two series of experiments using rats assessed the effects of intraventricular administration of the NMDA antagonist AP5 on performance of non-spatial working memory tasks. The first series used a continuous delayed non-matching to sample (DNMS) design; the second series used a discrete trial delayed matching to sample (DMS) design. Performance was assessed at retention intervals ranging from approximately 5 to 90 sec. The subjects had acquired the behavioural tasks before drug testing commenced. In the DNMS series, minipumps containing vehicle, 5, 10 or 15 nM D-AP5 were implanted. Every 10 days, each rat's minipump was removed and replaced with a fresh pump containing a new drug dose in a counterbalanced design, so that all rats were tested under all four conditions. There were no drug effects on performance at any retention interval. In the DMS series, there were three different basic task variants. Minipumps filled either with 15 mM D-AP5 or vehicle solution were implanted. Vehicle rats performed at approximately pre-operative levels; AP5 rats were impaired only on task variants using repeated stimulus presentations within session. There was no interaction between retention interval and drug treatment. This pattern of results closely resembles that seen following hippocampectomy or fornicotomy, as would be expected if this drug, administered intraventricularly, selectively affected hippocampal function.