As the United States experiences greater income inequality, more and more students experience an early science achievement gap. This study tested several competing theoretical models of early science achievement with a longitudinal sample of 14,624 children who were followed from kindergarten entry to the end of 1st grade. To understand why and when differences emerge among children, we adopted a theoretical approach to study early science achievement because theories, if they are accurate, appeal to the factors responsible for achievement differences. Results from a structural equation modeling analysis revealed a mediation-only model provided an excellent fit to the data that was minimally changed when some of the indirect pathways were dropped and several direct pathways were added. This finding suggests that children who are more advanced in their understanding of science topics, have more expertise, or have more working-memory capacity are likely to process the same classroom material more effectively and completely than their classmates who have a less advanced understanding, less expertise, or less working-memory capacity. From a theoretical perspective, these findings provide strong support for individual theories of conceptual knowledge, expertise, working memory, and motivational development, and they also demonstrate the value of a large-scale, longitudinal, and multivariate approach for understanding science achievement. Implications for developmental theory and educational policy are discussed.