Spatial learning and memory deficits in the APP23 transgenic mice have mainly been studied using the Morris water maze (MWM). However learning in the MWM relies on swimming abilities and may be confounded by the stressful nature of this test. We have therefore assessed spatial learning and memory in 12-month-old APP23 using a dry-land maze test developed by Barnes. Mice were given daily learning trials for a total of 41 successive days. After a 12-day interval the mice were re-tested for 4 additional days in order to examine the spatial memory retention. Immediately following this phase, reversal learning was examined for 13 additional days by moving the escape tunnel to the opposite position. During the initial learning phase, APP23 mice showed a significantly longer latency to find the escape tunnel as well as an increased number of errors compared to non-transgenic littermates. These deficits appeared to be due to a delay in switching from a "no strategy" to a spatial strategy. Indeed, this same delay in the use of spatial strategy was observed in the reversal phase of the study. Our results suggest that impairments in APP23 mice in learning and memory maze tests may be due to a specific deficit in the use of spatial strategy.