Place navigation in the Morris water maze can be directed by memory of the target coordinates relative to remote landmarks (allocentric) or by the memory of the start-goal route (egocentric). When the start and goal positions remain constant and visual cues are eliminated by darkness, memory of the route may become decisive. This assumption was tested in 10 male hooded rats using an infrared television tracking system allowing navigation training in the dark. In Expt. 1, these animals were trained to swim in the dark from the start at the S rim of the pool to the goal position in the center of the NW quadrant of the pool. Mean escape latencies decreased from 47 s initially to 16 s during the 24 daily sessions. Another group of 10 male hooded rats learned the same task in the light. Mean escape latencies decreased from 20 s initially to 5 s during 4 daily sessions. In Expt. 2, possible allocentric location of the target was tested in the same rats by rotating both the start and goal positions by 90 degrees counterclockwise (i.e., to E-SW and later to N-SE). Mean escape latency during 5 days after the first rotation increased to 24 s, but returned back to the asymptotic level of 18 s after the second rotation. The same change of the start and goal position (from S-NW to E-SW) in the light only increased escape latency in the first session. In Expt. 3, both the goal position and route direction were changed to N-SW. Surprisingly, the animals rapidly acquired a new heading angle at the start and mean escape latencies were not significantly changed. It is concluded that overtrained place navigation in darkness can be easily changed to a new direction.