Rats injected with either drug or saline were trained in a food-rewarded lever-pressing task until they could complete an FR10 requirement within the first 120s of the session, and were then tested for the retention of this response under various treatment conditions. In this procedure, state-dependency occurred with ketamine, phencyclidine, MK-801 and CGS 19755; rats trained with any of these NMDA antagonists failed to show response transfer when tested with saline. Also, in rats trained with saline the response failed to transfer in tests with any of these drugs. The doses at which the failure to transfer occurred in drug-to-saline tests, were approximately 3-fold lower than those at which the failure occurred in saline-to-drug tests. Even higher doses of these compounds were required to inhibit acquisition. The state-dependency which NMDA antagonists appear to produce in tests for saline-to-drug transfer, might constitute the mechanism whereby these compounds apparently disorganize the behavior of animals and, perhaps also, of humans. The data do not support the widely held notion that NMDA antagonists produce deficits in memory or retention; instead, at doses considerably lower than those impairing acquisition, NMDA antagonists produce a state upon which the recall of the newly learned response is dependent.