We report here data suggesting that reactivation of a well-established memory by a retention test triggers cellular events which depend upon N-methyl-D-aspartate (NMDA) receptors for up to 2 h after reactivation. Rats were overtrained on a maze task requiring integration of distal spatial information contained in cues strategically placed around the maze. Previous experiments showed that pretrial injection of the noncompetitive NMDA receptor antagonist, MK-801, at a dose which had no effect on overt behavior (0.05 mg/kg), markedly disrupted the well-trained performance of the task. Surprisingly, the behavioral deficit persisted on subsequent, drug-free trials, 24 h later. The present experiments showed that post-trial injections produced the same effects on performance on one or two subsequent daily trials. A temporal gradient for this amnestic effect of the drug treatment was established by injecting rats at 5, 30, 60, 90, 120 or 180 min after the performance trial. Only those rats whose MK-801 treatment was delayed for 120 min or more after the trial were able to perform the task normally 24 h later. All other treatment times induced significant amnesia for the task, when the rats were tested 24 h later. A subsequent experiment, using a more difficult version of the task, showed a longer amnesia gradient, but the predrug performance level could be reinstated within one multiple trial test session. Thus, it appears that activation of a well-established memory circuit renders the trace labile, requiring its reconsolidation. To what extent the entire post-acquisition cascade of NMDA receptor-dependent intracellular events is recapitulated each time a memory is activated and reorganised is probably a function of the age and complexity of the memory and the amount of new information to be integrated into the circuit. These results provide physiological evidence for the notion that memory is a dynamic process undergoing continual reorganization as a function of the ongoing experience of the organism.