<p>Training of both domain-general and domain-specific skills has shown potential to benefit the acquisition of children’s math skills. In particular, training working memory and numerical knowledge have both shown to improve math performance in kindergarteners. However, it is unclear whether engaging in both, domain-general and domain-specific training would lead to greater learning outcomes as compared with completing only one type of training. Using regression and latent linear growth modeling, we analyzed a combined dataset from three studies involving 468 diverse kindergarteners to determine the relative benefits of three types of cognitive interventions – targeting working memory only, numerical knowledge only, or working memory followed by numerical knowledge training - to improve math and working memory skills in the short- and long run. In addition, we tested the extent to which the improvements are related to individual differences at baseline, and whether and how those individual differences might interact with training type. Overall, our results indicated that the additive training that targeted both, working memory and numerical knowledge, led to the most pronounced long-term gains in math compared to the other treatments. Furthermore, participants with relatively higher initial math skills tended to show the most substantial long-term gains in both working memory and math skills from the additive training. In contrast, for those with lower initial math skills, single domain training seemed to be the most effective approach. In sum, our data highlight the importance of taking into account individual strengths and needs when designing and implementing an intervention to promote math learning in early childhood.</p>