The Fullerton Longitudinal Study offers a unique opportunity to model the stability of intelligence and achievement and their relations from elementary through secondary school. Using latent variable modeling, we fit a cross-lagged panel model to examine the relations between intelligence and achievement in two academic domains: mathematics and reading. Findings revealed that students' achievement is highly stable across the school years. Childhood intelligence is a strong predictor of initial mathematics and reading achievement. After age 7-years, intelligence is not predictive of either mathematics or reading achievement after accounting for prior achievement. Students who enter school with strong academic skills tend to maintain their academic advantage throughout their elementary and secondary education. We discuss the implications of these results for talent development.