Effectiveness of exercise for improving cognition, memory and executive function: a systematic umbrella review and meta-meta-analysis
Singh B, Bennett H, Miatke A, Dumuid D, Curtis R, Ferguson T, Brinsley J, Szeto K, Petersen JM, Gough C, Eglitis E, Simpson CE, Ekegren CL, Smith AE, Erickson KI, Maher C · 2025
grade Creviewindependentfailed
Sample
133 systematic reviews covering 2,724 RCTs and 258,279 participants
Population
All populations and ages, including children and adolescents
Design
Umbrella review of systematic reviews of RCTs, AMSTAR-2 risk-of-bias scoring, random-effects pooling of review-level effect sizes. Graded C despite the enormous corpus because the analysis pools at REVIEW level rather than study level, does not apply modern selection-model publication-bias adjustment, and does not report prediction intervals — the three failures that Bartos et al. 2025 corrected by rebuilding the same 2,239 estimates at study level. Overlapping reviews double-count the same primary RCTs. The stated sensitivity check (excluding low-quality reviews) does not address selective reporting within the primary literature.
Key findings
This is the strongest recent statement of the pro-exercise position and the reason the claim remains in circulation: exercise "significantly improved" general cognition (SMD 0.42), memory (SMD 0.26) and executive function (SMD 0.24), with the explicit claim that memory and executive-function gains are LARGER for children and adolescents than for adults, and that effects are larger for low- and moderate-intensity and for short (1-3 month) interventions. It concludes there is "strong evidence ... across all populations". Bartos et al. 2025 rebuilt this dataset at study level and found publication-bias-adjusted executive function of 0.012 and moderate-to-strong evidence AGAINST any effect in healthy children and adolescents. The pattern of larger effects from shorter, lower-intensity interventions is itself a small-study-bias signature, not a dose-response finding.
Genetic confound
Underlying corpus is RCTs, so the estimates are not genetically confounded; the problem here is publication bias and heterogeneity, not heredity.
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| General cognition (pooled at review level, uncorrected) | SMD | 0.42 | standardized | far transfer; end of intervention | business-as-usual | end-of-treatment | far-transfer |
| Memory (pooled, uncorrected) | SMD | 0.26 | standardized | far transfer; end of intervention | business-as-usual | end-of-treatment | far-transfer |
| Executive function (pooled, uncorrected) | SMD | 0.24 — reduces to 0.012 (95% CrI 0.000 to 0.147) on study-level bias-adjusted re-analysis (Bartos 2025) | standardized | far transfer; end of intervention | business-as-usual | end-of-treatment | far-transfer |
| Moderator claim on age | subgroup comparison | memory and executive-function gains reported LARGER for children and adolescents than adults | standardized | end of intervention | business-as-usual | end-of-treatment | far-transfer |
| Moderator claim on intensity and duration | subgroup comparison | effects larger for low/moderate intensity and for SHORTER (1-3 month) interventions — a small-study-bias signature | standardized | end of intervention | business-as-usual | end-of-treatment | far-transfer |
Cited by
- Physical activity as an input — dose, fitness, and school outcomesno effectconf: mediumgc: low