<p>This study examined how the intensity of public physical education shapes working memory and attention control in Chinese university undergraduates, and tested whether self-regulation ability statistically mediates the relationship. Two hundred and fifty second-year students were randomly assigned to five parallel intensity arms — low, moderate-low, moderate, moderate-high, and high — defined by percentage of heart rate reserve (%HRR) and delivered twice weekly across a 12-week course. Cognitive performance was assessed with the 2-back task, Flanker, and Stroop paradigms at baseline, mid-intervention, post-intervention, and follow-up; self-regulation was indexed by the Chinese-adapted Short Self-Regulation Questionnaire. Restricted cubic spline regression revealed an inverted-U dose–response pattern for all four cognitive outcomes, with convergent inflection near 67% HRR (95% bootstrap CI: 63.1–71.2%) and an estimated zone of maximal benefit spanning approximately 60–74% HRR. The performance decline above the inflection was approximately 1.46 times steeper than the rise below it. Bias-corrected bootstrap mediation analysis indicated that self-regulation partially mediated the intensity–cognition association, statistically transmitting roughly 35–40% of the benefit observed around the inflection, while the residual direct path remained robust. Bayesian path estimation produced posterior probabilities exceeding 0.99 for the indirect effect across outcomes. Because no biochemical or neuroimaging indicators were collected, neurobiological language is used only as a theoretical scaffold rather than as direct evidence. The findings tentatively support a statistically distinguishable dual-pathway pattern and offer a candidate intensity band that may inform curriculum design pending multi-site replication.</p>