A pretest-training-posttest design assessed whether training to improve spatial skills also improved mathematics performance in elementary-aged children. First grade students (mean age = 7 years, n = 134) and sixth grade students (mean age = 12 years, n = 124) completed training in 1 of 2 spatial skills-spatial visualization or form perception/VSWM--or in a nonspatial control condition that featured language arts training. Spatial training led to better overall mathematics performance in both grades, and the gains were significantly greater than for language arts training. The same effects were found regardless of spatial training type, or the type of mathematics tested. [This paper will be published in "Journal of Educational Psychology."]