Part of the evidence used to corroborate school motivation theories relies on modeling methods that estimate cross-lagged effects between constructs, that is, reciprocal effects from one occasion to another. Yet, the reliability of cross-lagged models rests on the assumption that students do not differ in their trajectories of growth over time (e.g., no high- or low-achievers). The present review explains why deviations from this assumption produce unreliable findings by confounding between- and within-person processes of change. To relax this assumption, next-generation cross-lagged models are presented and illustrated using panel data on high school students (N = 944). These issues and solutions are discussed using, as a case study, the pervading theory that motivation develops as a function of reciprocal effects between beliefs about the self (e.g., academic self-concept) and school achievement. Implications regarding the use of cross-lagged models and knowledge building in school motivation research are discussed. Online supplementary materials containing technical notes on cross-lagged models, as well as open-source data and scripts for R and Mplus, are provided to aid educational researchers use and compare these alternative models.