Cognitive mapping implies the development of an internal representation of the spatial relationships among objects in the environment. One can assess the development of a cognitive map by demonstrating that an animal can select, when appropriate, a novel path to reach a goal in the absence of landmarks and when path integration does not provide an adequate account. Rats were trained to find reinforcement in two of three goal boxes in a three-arm maze. In test, the rats were given a choice between two novel paths, one that led to the goal box that had been baited during training, the other that led to the goal box that had been unbaited during training. When they were trained in the absence of distinctive intramaze cues (Experiment 1), no preference was found; however, when distinctive intramaze alley cues were present during training but were unavailable as directional cues during testing (Experiment 2), the rats demonstrated a significant preference for the correct novel path. These results suggest that under appropriate conditions rats are able to form simple cognitive maps of their environment.