Although studies have demonstrated a relationship between working memory and achievement in math and reading, it is still unclear why working memory might be related to these abilities. In the present article, we examined the viability of several possible theories in 2 separate analyses of math and reading. In each case, we contrasted the predictions of a cognitive filter model, a transactional model, and a positive manifold model using data from the 2011 Early Childhood Longitudinal Study Kindergarten (ECLS-K). Results of path analyses in a structural equation modeling (SEM) framework indicated an excellent fit for the transactional model, while a poor fit was shown for the other 2 models for both math and reading. Findings across these analyses suggested that working memory and achievement interact in a reciprocal, recursive manner over time. Findings are discussed in terms of their implications for theory, practice, and future research.