High contextual interference improves retention in motor learning: systematic review and meta-analysis
Czyż SH, Wójcik AM, Solarská P, Kiper P · 2024
grade Cmeta-analysisindependentmixednumbers spot-checked
Sample
2,068 participants across 54 studies
Population
Mixed lifespan, ages ~5.5-82; laboratory tasks and applied sport/clinical settings
Design
PROSPERO-preregistered (CRD42021228267). 54 studies, 194 effect sizes, 2,068 participants; three-level mixed model handling within-study dependency, plus random-effects model on averaged SMDs. Delayed retention only (>24h). Total I²=90% (I²level3=66%, I²level2=24%). EPHPP quality: only 3 of 54 studies strong/moderate; 44 weak on selection bias, 50 weak on withdrawals — by the tool's own rule only 2-3 studies should have been meta-analysed at all, and the authors say so and pool anyway. No funnel/Egger test reported. Setting moderator IS significant (F(1,192)=4.50, p=0.03); the AGE moderator is NOT (F(2,189)=1.69, p=0.19) — the authors ran the age subgroups anyway "because the differences between age groups were quite substantial", so every age-specific number below is a post-hoc split under a null omnibus test.
Key findings
54 studies, n=2,068. Random-vs-blocked retention SMD 0.63 (0.33-0.93) — but carried by laboratory tasks (0.92) and small/non-significant in applied field settings (0.23, p=.24). SENSITIVITY ANALYSES CUT IT HARD: after outlier removal overall falls to 0.43 (0.19-0.67), applied to -0.01 (-0.35-0.32, p=.94), and under-18s to -0.39 (-1.30-0.51, p=.39). Age and setting are CONFOUNDED in this corpus — the authors state that ALL experiments with older adults were run in a laboratory and only one applied study included young participants — so the age moderation and the setting moderation are one confound, not two findings. The age omnibus moderator test was non-significant (p=0.19) and the subgroups were run regardless.
Genetic confound
No heritability or trainability modelling; individual differences in schedule responsiveness treated as error variance. Age moderation read developmentally, not genetically.
Replication notes
Directly contested by Ammar et al. (2023, 2024), who find no CI effect in sport. Czyż & Wójcik published a point-by-point rebuttal (SportRxiv 437) noting their review was preregistered and Ammar's was not.
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| Delayed retention, high CI (random) vs low CI (blocked), overall | SMD (three-level mixed model) | 0.63 (95% CI 0.33-0.93, p<0.001); 54 studies, 194 effect sizes, 2,068 participants; random-effects model on averaged SMDs 0.71 (0.41-1.01, p<0.001) | mixed | retention; delayed >24h; lab+field pooled | active-alternative | end-of-treatment | domain-skill |
| Delayed retention, overall, AFTER OUTLIER REMOVAL (the honest headline) | SMD (three-level mixed model, 10 outcomes beyond Cook's D 4/n removed) | 0.43 (95% CI 0.19-0.67, p<0.001) — down from 0.63; the averaged-SMD model likewise falls 0.71 to 0.56 (0.32-0.80) after removing 2 outliers | mixed | retention; delayed >24h; lab+field pooled | active-alternative | end-of-treatment | domain-skill |
| Delayed retention, laboratory studies only | SMD | 0.92 (95% CI 0.48-1.36, p<0.001); 30 studies, 1,210 participants | researcher-designed | retention; delayed; lab | active-alternative | end-of-treatment | domain-skill |
| Delayed retention, applied/field studies only | SMD | 0.23 (95% CI -0.16 to 0.62, p=0.24), non-significant; 24 studies, 105 effect sizes, 858 participants. After removing 3 outliers it reverses to -0.01 (-0.35 to 0.32, p=0.94), favouring BLOCKED practice | researcher-designed | retention; delayed; field/applied | active-alternative | end-of-treatment | domain-skill |
| Delayed retention in young participants (<18y) | SMD | 0.02 (95% CI -0.90 to 0.94, p=0.97), negligible; 418 participants, 49 effect sizes. After removing 2 outliers it goes NEGATIVE, -0.39 (-1.30 to 0.51, p=0.39) — i.e. the point estimate for children favours blocked practice | mixed | retention; delayed; lab+field | active-alternative | end-of-treatment | domain-skill |
| Delayed retention in older adults (>=60y) | SMD | 1.45 (95% CI 0.55-2.35, p=0.003); 205 participants, 24 effect sizes; averaged-SMD model 1.58 (0.62-2.54, p=0.001). ALL older-adult experiments in this corpus were laboratory studies, so this is the lab effect wearing an age label | mixed | retention; delayed; lab+field | active-alternative | end-of-treatment | domain-skill |
| Delayed retention in adults (18-59) | SMD (three-level mixed model) | 0.63 (95% CI 0.30-0.96); 1,425 participants, 119 effect sizes | mixed | retention; delayed; lab+field | active-alternative | end-of-treatment | domain-skill |
Cited by
- Practice scheduling for motor skills — varied vs blocked, spaced vs massedmixedconf: highgc: low