The Reciprocal Effects Model Is Robust to Alternative Modeling Specifications: A Response to Sorjonen et al. 2025

In a recent commentary, Sorjonen et al. (2025) reanalyzed simulated data based on Marsh et al. (2024), proposing several alternative models to account for reciprocal effects between academic self-concept (ASC) and achievement (ACH). Their contribution is a welcome addition to the ongoing dialogue on longitudinal modeling, expanding the range of theoretically grounded tests of reciprocal processes. While valuable, the models they proposed relied on specifications that, in our view, raised some concerns about model validity or did not fully capture the temporal structure central to Marsh et al.’s (2024) extended reciprocal effects model (REM), which posits contemporaneous skill-development effects (of ACH on ASC) and lagged self-enhancement effects (of ASC on ACH). In addition to responding to Sorjonen et al., we present a generalizable framework for comparing longitudinal models that differ in their assumptions about change processes, temporal structure, and measurement design. Using this comparative framework, we reanalyzed both the simulated data constructed by Sorjonen et al. and the original PALMA dataset used in Marsh et al. (2024). Although Sorjonen et al. questioned the evidence, our reanalyses provide converging support for the extended REM and, in particular, for an effect of academic self-concept on achievement. The extended REM provided the most probable approximation of the data structure and was robust to several alternative modeling specifications. We provide open-access scripts to replicate results and implement the proposed comparative modeling framework in various research settings (see OSF repository at https://osf.io/84bzp/).