Effect of Exercise on Cognition, Memory, and Executive Function: A Study-Level Meta-Meta-Analysis Across Populations and Exercise Categories

2,239 effect-size estimates from 215 meta-analyses of RCTs; the same evidence corpus as Singh et al. 2025 (BJSM), re-assembled at study level

All populations and ages, with separate subgroup estimates for healthy children and adolescents

Study-level reconstruction of the pooled exercise-cognition literature, then robust Bayesian model-averaging (RoBMA) with selection-model and PET-PEESE-type publication-bias adjustment, plus explicit between-study heterogeneity and prediction intervals. This is methodologically the strongest correction applied to this literature to date: it avoids trim-and-fill, which Carter et al. showed has high false-positive rates and undercorrects. It is a preprint (PsyArXiv), not yet peer reviewed, which is the only reason it is graded B rather than A.

Re-analysing the exact data behind the 2025 BJSM umbrella review that claimed "strong evidence" of benefit across all populations, publication-bias-adjusted effects collapse: general cognition SMD 0.227 (95% CrI 0.116 to 0.330), memory 0.027 (0.000 to 0.227), executive function 0.012 (0.000 to 0.147). Prediction intervals span negative and positive effects. The finding that matters most for schools: in HEALTHY CHILDREN AND ADOLESCENTS the analysis reports moderate-to-strong evidence AGAINST an effect of any exercise category on executive function — Bayesian evidence for the null, not merely absence of evidence. Specifically, exergaming and mind-body exercise show evidence of NO benefit in healthy youth. The only youth signal is exergaming for executive function in NON-healthy children and adolescents (a clinical population, off-scope for general schooling).

Not addressed and not needed — the pooled corpus is RCTs, so randomization controls genetic confounding. That is precisely what makes the near-null informative: the much larger observational fitness-cognition correlations are the ones requiring explanation, and passive gene-environment correlation supplies it.