The development of math reasoning and 3-d mental rotation skills are intertwined. However, it is currently not understood how these cognitive processes develop and interact longitudinally at the within-person level -- either within or across genders. In this study, 553 students (52% girls) were assessed from fifth to seventh grades on 3-d mental rotation spatial skills (assessed each fall) and numerical and algebraic math reasoning skills (assessed each spring). Boys outperformed girls on mental rotation tests across all three grades, and on fifth and seventh grade math reasoning tests. Consistent with the literature on between-person comparisons, there was a positive correlation between mental rotation and math reasoning skills in the full sample and for both genders. A random intercept cross-lagged panel model was used to control for these confounding group-level differences in order to isolate within-person associations between earlier and later performance. Initially in fifth grade, math reasoning predicted subsequent sixth grade mental rotation skills. By seventh grade, more advanced mental rotation skills were associated with subsequent math reasoning skills while math reasoning skills were no longer predictive of mental rotation skills. An examination of gender differences revealed that this pattern was driven by boys while girls experienced less within-person change. These findings suggest that boys may initially rely in part on their math reasoning skills to solve 3-d mental rotation tasks. However, as their 3-d mental rotation skills mature, they begin to primarily depend upon these developing spatial skills to solve math reasoning problems rather than the reverse.