Three experiments were performed to explore the rat's ability to count objects in its environment. Previous research demonstrating counting behavior in animals has typically involved overtraining or extreme motivational conditions in which food or safety were at a premium. In the present experiments, a simulated natural environment was employed in which rats were required to enter a particular tunnel in an array of six to obtain food. Spatial, olfactory and visual cues were controlled so that selection of the correct tunnel was based solely upon the ordinal position of the tunnel. Experiments 1 and 2 demonstrated that this behavior could be learned relatively rapidly and maintained despite change in the spatial configuration of the test area. In Experiment 3, retention of the tunnel entry discrimination was demonstrated following an inactive period of one year as well as six months thereafter. Collectively, these data extend the range of situations in which rats have performed numerical discriminations, and suggest that the environmental support necessary to demonstrate such abilities may include relatively natural test situations, rather than the use of extreme motivational states induced under arbitrary environments.