The problems and issues associated with the use of pharmacological antagonists in studies on learning and memory are considered in a review of the role of N-methyl-D-aspartate (NMDA) receptors, NMDA receptor-mediated long-term potentiation (LTP), and muscarinic receptors in spatial learning in the water maze. The evidence indicates that neither NMDA nor muscarinic receptors, nor NMDA receptor-mediated LTP, are required for spatial learning, although they might normally contribute to it. Detailed behavioral analyses have indicated that the water maze task is more complex than generally has been appreciated, and has a number of dissociable components. Naive rats trained under NMDA or muscarinic antagonism display sensorimotor disturbances that interfere with their ability to acquire the task. Rats made familiar with the general requirements of the task can learn the location of a hidden platform readily under NMDA or muscarinic antagonism. The ability of a rat to acquire the water maze task depends on its ability to apply instinctive behaviors to performance of the task in an adaptive manner. The instinctive behaviors undergo modification as the rat learns the general strategies required in the task. The evidence suggests that at least some of the plastic changes involved in acquiring the task occur in existing neural circuits situated in widespread areas of the brain, including sensory and motor structures in the cortex and elsewhere, and are therefore difficult to distinguish from existing sensorimotor mechanisms. More generally, the findings indicate the difficulty of inferring the occurrence or nonoccurrence of learning from behavior, and the difficulty of causally linking the action of particular receptor populations with the formation of specific memories.