We investigated whether rats can learn mazes by a procedure in which rats were left in the maze (residential maze) in groups for 1 h a day. Water and food locations, which served as the start and goal boxes respectively in the test trial, were at the opposite ends of the maze. On the test trial conducted everyday before the residence period, animals put in the start box showed a significant decrease of the error response into the blind alleys and running time to reach the goal box. Systemic administration of scopolamine (0.25, 0.5 mg/kg), a muscarinic receptor antagonist, dose-dependently increased the number of errors, but mecamylamine, a nicotinic receptor antagonist, had no effect. Bilateral hippocampal lesions retarded both the acquisition and retention of this maze learning. The results suggest that this residential maze procedure is useful for testing maze learning ability in rats.