In the present study, we tested the spatial learning behavior of four different mouse strains (129/Sv, BALB/c, C57BL and Swiss) in a newly developed circular maze. The maze was based on the circular Barnes maze, which was initially developed for rats. Since mice do not readily enter holes in floor, additional reinforcers (positive and negative) or pretraining procedures have been used to train the animals. Because these methods are not always desirable, we examined whether mice are more willing to enter escape holes (12), which were located in the rim of the apparatus. C57BL mice appeared to improve their performance on three different measures of spatial learning: latency to find escape hole, distance to escape hole and errors (visit to other holes). The other strains also improved their performance although this was only seen for one parameter (i.e. 129/Sv and BALB/c on latency, and Swiss on distance). When the animals were trained to find another location, it was found that only the performance of the C57BL mice was transiently impaired. The C57BL mice were also very efficient in improving their performance in a repeated acquisition paradigm (six trials per day on four successive days). Applying a probe trial procedure, a clear preference for the goal location was found. These findings indicate that these mice used a spatial search strategy. Although this circular maze can be used as an additional tool to assess spatial learning in (genetically modified) mice, it is noted that strain differences in spatial learning seem to be independent of task. Further, our data with different strains indicate that different measures of behavior should be evaluated to assess the spatial learning performance of mice.