Near and Far Transfer in Cognitive Training: A Second-Order Meta-Analysis
Sala G, Aksayli ND, Tatlidil KS, Tatsumi T, Gondo Y, Gobet F · 2019
grade Bmeta-analysisindependentunreplicated
Sample
Three models pooling meta-analyses: k = 99 (near), k = 119 (far, WM training), k = 233 (far, all training types)
Population
Children, adults and older adults; working-memory, video-game, music, chess and exergame training
Design
Meta-analysis of meta-analyses, which is the right tool for a field whose apparent disagreement comes from how individual reviews handle controls. Explicitly models placebo effects and publication bias, and tests whether population type (children vs adults vs older adults) moderates transfer.
Key findings
Working-memory training does induce near transfer, moderated by population. Far-transfer effects are small or null in every model. Crucially, once placebo effects and publication bias are controlled, the overall far-transfer effect size AND the true between-study variance both equal ZERO - regardless of population or training programme. The authors describe the lack of generalisation as an invariant of human cognition.
Genetic confound
Low. Randomised training contrasts throughout.
DOI / URL
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| Near transfer (memory measures) from WM training | second-order pooled effect | present; moderated by population type | standardized | post-training | active-alternative | end-of-treatment | near-transfer |
| Far transfer across all cognitive-training programmes | second-order pooled effect, placebo and publication-bias corrected | zero, with zero true variance | standardized | post-training | active-alternative | end-of-treatment | far-transfer |
Cited by
- Does working-memory or brain training transfer to intelligence and achievement?no effectconf: highgc: low