Practice Effects Confound the Emotion–Performance Link in Self-Regulated Learning: A Reanalysis of Lin et al. (2026)

Lin et al. (*Journal of Educational Psychology*, 2026) proposed a computational model of the dynamic interplay between goal setting, performance, and emotions in self-regulated learning. In a repetitive math task (Study 1), they reported that positive emotions predicted lower subsequent performance, interpreting this via the coasting hypothesis. We argue that this finding may instead reflect a temporal confound: in a repetitive task, performance improves due to practice while emotional engagement declines, producing a spurious negative association when time trends are unmodeled. We reanalyze the Study 1 data by introducing a log(trial) learning curve into the best-fitting model. The negative emotion-to-performance effect is no longer credible in this extended model; instead, a strong practice effect emerges. This confound-based interpretation also parsimoniously accounts for why the emotion-to-performance pathway reversed direction in the original Study 2, where task variety and high stakes attenuate practice-driven time trends. Methodologically, we show that the WAIC criterion used in the original study exhibits poor diagnostic reliability and is theoretically biased for autoregressive panel models. We implement exact leave-future-out cross-validation as an unbiased alternative, under which the original model selection conclusions are no longer supported.