On Model Interpretability and Time-Reversal Testing in Analyses of Reciprocal Effects: Clarifying Modeling Issues in Sorjonen et al.’s (2025a, 2025b) Commentaries

Journal article : Comment

Núñez-Regueiro et al. (2025) recently demonstrated that the alternative models used by Sorjonen et al. (2025a) to reject the reciprocal effects model (REM) contained misspecifications that, when addressed, yielded results consistent with the REM. In a reply to our critique, Sorjonen et al. (2025b) defended their original models as valid modeling approaches. They further proposed time-reversal tests, which they interpreted as contradicting the REM. In this new comment, we adopt a more pedagogical approach to resolve these misunderstandings. First, we explain in details why the models in Sorjonen et al. (2025a) are difficult to reconcile with several principles of structural equation modeling, by yielding improper solutions, impossible or confounded model parameters, untested temporal structures, and omitted adjustments that limit the interpretability of reciprocal effects. Second, we further explain why the renewed claims of validity in Sorjonen et al. (2025b) extend these issues by mobilizing simulation studies that do not align with the models at stake. Third, we demonstrate that their time-reversal testing repeated some of these issues and that a rectified version yields results consistent with predictions based on the REM (https://osf.io/ry6p3/overview?view_only=2af2bc8cd4694433b9024eb2577451b6). Our objective is to advance the debate on reciprocal effects by clarifying the modeling issues associated with Sorjonen et al.’s commentaries and providing additional empirical evidence relevant to evaluating the REM.

Authors: Núñez-Regueiro, F; Marsh, HW; Pekrun, R; Lüdtke, O; Guo, J

• Publication status: Published
• Publisher: Springer
• Journal: Educational Psychology Review
• Volume: 38
• Issue: 1
• Article number: 80
• Publication date: 2026-05-26
• Acceptance date: 2026-05-18
• DOI: 10.1007/s10648-026-10179-8
• EISSN: 1573336X
• ISSN: 1040726X
• Language: English
• Keywords: Reciprocal effects model; Causal inference; Structural equation modeling; Model admissibility; Scientific debate
• Subtype: Comment