Cognitive Training Does Not Enhance General Cognition
Sala G, Gobet F · 2019
grade Breviewindependentreplicatednumbers spot-checked
Sample
No new statistical synthesis: a narrative Opinion article restating the authors' own meta-analyses (video games k not restated; music; chess; WM training in TD children) alongside independent ones (Melby-Lervag 2016; Simons 2016; Soveri 2017; Dougherty 2016). Every number below is qualitative in the source.
Population
Children, adults and older adults across working-memory, video-game, music, chess and exergame training
Design
Trends in Cognitive Sciences OPINION article (the text calls itself that at p.3), and the anchor the assignment identified by DOI. It reports no pooled effect sizes of its own - the numbers live in the cited meta-analyses, which are separate records here. Read as a position statement backed by meta-analyses, not as a meta-analysis. The paper's own vocabulary is stated explicitly and the record uses it: "near-zero" = |effect| < 0.05, "small" = 0.05-0.20, "moderate" = 0.20-0.50. NOTE ON ATTRIBUTION: doi:10.1016/j.tics.2018.10.004 is Sala & Gobet, not Simons et al. - Simons et al. (2016) is the separate PSPI review, recorded here as simons-2016-brain-training-programs-review. The argument is that apparent inconsistency across the field is not real heterogeneity but design quality plus statistical artefacts.
Key findings
Results are highly consistent across every reviewed domain: minimal effect on domain-general cognitive skills. Between-study variability is fully accounted for by design quality (chiefly active vs passive controls) and statistical artefacts. The programme of research has shown no appreciable benefits and the authors recommend redirecting resources. The positive half is stated just as plainly and is not a concession dragged out of them: cognitive-training programs "do improve the skills that are being trained", and benefits extend "to the trained tasks and, sometimes, similar tasks" - WM training has "a moderate yet reliable positive effect on several memory tests" (Box 3). What fails is generalisation beyond that. Two supporting non-training strands: music-trained twins have the same IQ as their non-music-trained co-twins (Mosing 2016), and musicians' cognitive ability tracks music APTITUDE but not amount of training (Swaminathan 2017) - i.e. selection, not causation. The authors also argue from schooling: if years of school move GCA only marginally, a few hours of training cannot.
Genetic confound
Low. The evidence base is randomised training studies; the failure is of transfer, not of identification.
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| Domain-general cognitive ability (far transfer / g) | pooled effect across domains | minimal to null once design quality and artefacts are modelled | standardized | post-training | unclear | end-of-treatment | g |
| Performance on the trained task itself | qualitative summary across domains | reliably improves - "cognitive-training programs, while they do improve the skills that are being trained, do not enhance GCA"; arithmetic brain-training games do improve simple calculation | treatment-aligned | post-training | unclear | end-of-treatment | domain-skill |
| Tasks similar to the trained task (near transfer), esp. memory tests after WM training | qualitative summary | "moderate yet reliable positive effect on several memory tests" (Box 3); benefits reach the trained tasks "and, sometimes, similar tasks" - attributed to task-specific strategy acquisition, not to expanded capacity | standardized | post-training | unclear | end-of-treatment | near-transfer |
| Music instruction on children's cognitive and academic skills, active controls | qualitative summary of Sala & Gobet 2017 music meta | small overall; "close to zero or null" against an active control - variability across primary studies is mostly control-group type | standardized | post-training | active-alternative | end-of-treatment | far-transfer |
| Video-game training on general cognitive ability | qualitative summary of Sala et al. 2018 (Psych Bull) meta | "close to zero, if not null", regardless of test type, participant age or game genre; competing positive metas judged inflated by publication bias and modelling flaws | standardized | post-training | unclear | end-of-treatment | g |
| Chess instruction on cognitive and academic skills | qualitative summary of Sala & Gobet 2016 chess meta | small to medium overall, but primary-study design quality is poor; only four studies compared chess with an alternative activity and all four found no significant treatment effect | mixed | 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