Rats were given continuous intraventricular infusion of saline or the thiol-proteinase inhibitor leupeptin, via subcutaneously implanted osmotic minipumps, while being trained on a spatial learning water task using spaced trials. Leupeptin caused overnight forgetting during training, but performance eventually reached asymptote in both groups. A retention test conducted 48 h later to assess spatial memory revealed no significant group differences, but did cause, in saline-treated rats only, a disruption of subsequent retraining back to the correct spatial location. The groups showed no differences in Cl-dependent [3H]glutamate receptor binding to hippocampal or entorhinal cortex membranes subsequent to training. In a second experiment, normal rats trained on the same task also showed no differences in Cl-dependent [3H]glutamate binding relative to rats exposed to the water task but given random spatial position training and handled controls. The results are discussed in relation to the hypothesis of Lynch and Baudry (Science (1984) 224, 1057-1063) that a calcium-dependent thiol proteinase is involved in memory formation through its ability to modify glutamate receptor distribution and dendritic spine shape.