Field potentials were recorded in the dentate gyrus of freely-moving rats in a classical conditioning paradigm in which high-frequency stimulation of the perforant path served as a conditioned stimulus. Paired or unpaired perforant path stimulus-footshock presentations were given to animals engaged in a previously acquired food-motivated lever-pressing task. Conditioned suppression of lever-pressing was the behavioural measure of conditioning. Perforant path stimulus trains at an intensity above spike threshold induced long-term potentiation of synaptic transmission in the dentate gyrus. In this condition, animals learned the perforant path stimulus-shock association. Three strategies were employed to block the induction or reduce the magnitude of long-term potentiation induced by the conditioned stimulus: (1) reduction of the intensity of the stimulus below the spike threshold resulted in no long-term potentiation and a failure by the animals to learn the perforant path stimulus-shock association; (2) inhibitory modulation of long-term potentiation by high-frequency activation of commissural input to the dentate gyrus resulted in learning deficits; (3) chronic infusion of DL-2-amino-5-phosphonovalerate, a selective antagonist of the N-methyl-D-aspartate subtype of glutamate receptor, blocked the induction of long-term potentiation and prevented associative learning. A highly significant linear relation emerged from a correlational analysis between the magnitude of the change in synaptic efficacy at the activated synapses and the amount the animals learned about the perforant path stimulus-shock association. The results presented in this paper are consistent with the hypothesis that associative learning depends on the development of lasting changes in synaptic function. We propose that the activation of N-methyl-D-aspartate receptors in the dentate gyrus is involved in this process and that the more change in synaptic efficacy is produced in the activated network, the more the animals learn.