Development of early math skill depends on a prerequisite level of cognitive development. Identification of specific cognitive skills that are important for math development may not only inform instructional approaches but also inform assessment approaches to identifying children with specific learning problems in math. This study investigated the specific cognitive correlates of math problem solving across early grade levels (1-4) while controlling for basic calculation skills. As expected, basic calculation skill was a significant predictor of math problem solving across the entire sample. However, the addition of cognitive measures almost doubled the variance explained (R[superscript 2] = 0.61). Additionally, only select cognitive variables contributed to the prediction of math problem solving, and these variables change in importance as children develop higher-level math skills. Results are discussed within a developmental model, which emphasizes the increasing importance of abstract code representations required in higher levels of math performance.