Abstract
Objective
A recent meta-meta-analysis concluded that the average effect of exercise on memory is near zero after correcting for publication bias (SMD = 0.027). We tested whether this null finding reflects excessive aggregation rather than a genuine absence of effect.
Methods
We conducted a focused reanalysis a publicly available meta-meta-analytic database, applying a theoretically motivated outcome reclassification protocol to isolate verbal declarative memory (VDM). From 2,239 effect-size estimates spanning 215 meta-analyses,
k
= 37 VDM-specific estimates were synthesized using a random-effects model. Publication bias was assessed via Egger’s test, PET-PEESE, and trim-and-fill.
Results
The pooled effect was
g
= 0.216 [95% CI (0.075, 0.357),
p
= 0.003], approximately eight times larger than the aggregate estimate of 0.027. Moderate heterogeneity was observed (
I
2
= 56.1%), with no significant funnel plot asymmetry (
p
= 0.458). Aerobic exercise produced the most robust benefit (
g
= 0.229,
k
= 27).
Conclusion
The null aggregate memory effect masks a meaningful signal for verbal declarative memory. The findings support domain-specific exercise prescriptions: for hippocampus-dependent verbal memory, moderate-intensity aerobic exercise represents an evidence-supported strategy for enhancing retention.