Spatial cognition is influenced by the external environment, which interacts with an animal’s internal motivation. However, the underlying phylogenetic effects on spatial cognition are poorly studied. Our study considers the geographic and phylogenetic correlates of spatial cognition in four populations of the striped mouse Rhabdomys, a genus that is widespread throughout southern Africa. We tested four populations from South Africa: R. pumilio from the arid western parts, R. dilectus from the moist eastern grasslands, and one population of each species occurring in sympatry in the central grasslands. The spatial cognition of individuals of both sexes was studied in a modified 5-hole Barnes maze, and we measured distance travelled, error rate and latency to enter the correct hole during the initial training phase ( i) maze un-rotated, and ii) rotated and external cues removed) and a later memory test phase. Populations did not differ in the initial phase in un-rotated and rotated mazes, suggesting the use of internal cues in the maze. In the memory testing phase, arid R. pumilio and sympatric R. dilectus both spent the most time in the correct quintile and were most active in the Barnes maze. The sympatric and arid populations of R. pumilio made fewer errors followed by allopatric R. dilectus, whereas sympatric R. dilectus made the most errors. There were no population differences in distance travelled or latency to find the correct hole, and no sex differences in any test. We conclude that R. pumilio has better memory and made better decisions, despite differences in activity between the arid and sympatric populations. In contrast, the sympatric R. dilectus was active, but did not make good decisions, whereas the allopatric R. dilectus showed intermediate behaviours. Thus, spatial ability is phylogenetically determined and is modulated by population differences in activity and geographic origin.