This paper examines whether golden hamsters can rely on dead reckoning (getting positional information from updated signals generated during locomotion) on an eight-arm maze. Two groups of hamsters were tested: Group L under ordinary room light, Group D in darkness. To enhance the role of dead reckoning, each subject could climb from its own home cage onto the central platform of the maze. In a first experimental phase (15 trials), the L subjects learned to master the maze through developing a locomotor rule (arm chaining) after three to four trials. The D subjects developed arm chaining less readily and fluctuated more in their performance than did the L subjects. In a second experimental phase (15 trials), four arms were blocked at the beginning of each trial. In both experimental groups the performance decreased, yet remained well above chance level. Success and arm chaining were positively correlated in Phase 1 and negatively correlated with success in Phase 2. We assume that in Phase 2 the L subjects switched to the predominant use of visual cues, and the D subjects to dead reckoning.