The positive relationship between spatial ability and mathematical skills is a classical result in developmental and cognitive psychology. In this study, we analysed the effects of two types of spatial training and handedness on primary school children’s arithmetical ability. The participants were pre-tested on a test of arithmetic and assigned to one of three groups: (a) one hour of mental rotation and translation training, (b) one hour of mental translation training only, or (c) a no-contact group. The results showed no significant difference between training groups and a significant interaction between training group and category of handedness. Interestingly, only extremely right-handed children seemed to benefit from the training. These outcomes suggest that any spatial training needs to include mental rotation activities to be effective, and that the relationship between spatial training and achievement mathematics appears to be moderated by handedness.