Three groups of water-deprived rats collected water from the ends of the 8-arm radial maze. Sighted subjects, and subjects blinded either with or without pre-enucleation experience on the radial maze, all retrieved the water efficiently. Most of the subjects exhibited the same response stereotypy, regularly choosing 8 adjacent arms of the maze, then stopping in the center of the maze. The strategies underlying this performance were analysed by interrupting trials and rotating the maze 180 degrees after the subject had made 3 choices. Sighted subjects depended on extramaze stimuli, naive-blind subjects depended on intramaze stimuli and experienced-blind subjects ignored their initial 3 choices after the trial was interrupted. Choice accuracy was equally good whether the subject was returned to the position from which it had been removed, or returned to the opposite side of the central platform. All 3 groups of subjects maintained their stereotyped adjacent-arm responding only as long as such responding was consistent with high choice accuracy. Response stereotypy was prevalent on the radial maze, but response strategies were secondary to memory strategies.