The effect of contextual interference on transfer in motor learning: a systematic review and meta-analysis
Czyż SH, Wójcik AM, Solarská P · 2024
grade Cmeta-analysisindependentmixednumbers spot-checked
Sample
1,421 participants across 34 studies
Population
Mixed lifespan (<18 through >=60); lab motor tasks and applied sport skills
Design
Companion to the retention meta, same PROSPERO registration (CRD42021228267). 34 studies meta-analysed, 86 effect sizes, 1,421 participants (42 studies in the narrative review). Delayed transfer = >24h post-acquisition. Total I²=86% (level3 47%, level2 39%). Sensitivity analysis removing 5 outliers drops pooled SMD from 0.55 to 0.40. Egger's/rank-correlation could not be run because of heterogeneity; no study rated 'strong', only 2 'moderate' (26 weak on selection bias, 30 weak on withdrawals). CRITICALLY, unlike the retention paper, the SETTING moderator test here is NON-SIGNIFICANT: F(1,84)=1.34, p=0.25, and the authors write it "suggests the subgroup analyses are not necessary, though, we decided to perform them analogically to Brady (2004)". Every lab-vs-applied number below is therefore a post-hoc split under a null omnibus test.
Key findings
Transfer LOOKS like it shows the same lab-vs-field gradient as retention — medium and significant in the lab (0.75), small and non-significant in applied settings (0.37) — but the setting moderator test is non-significant here (p=0.25), so unlike the retention paper this gradient is not statistically supported. The overall 0.55 is fragile: removing 5 outliers cuts it to 0.40. Children under 18 again show essentially nothing (0.12, ns), while the large older-adult estimate (1.28) is itself non-significant on 8 effect sizes — the paper's own abstract nonetheless claims the age analysis "turned out to be significant only in adults and older adults", which its results section contradicts. Transfer is measured on near variants within the same movement class, so this is not evidence that varied practice builds general athleticism.
Genetic confound
None modelled. Transfer tasks are near-transfer variants within the same movement class, so this speaks to narrow transfer, not to raising general motor capacity.
Replication notes
Consistent with Brady (2004) on the lab/applied split. Contradicted by Ammar et al. (2023) sport-only meta; the two groups have exchanged published comment and rebuttal.
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| Delayed transfer, high CI (random) vs low CI (blocked), overall | SMD | 0.55 (95% CI 0.25-0.86, p<0.001); 34 studies, 86 effect sizes, 1,421 participants; drops to 0.40 (0.18-0.62, p<0.001) after removing 5 outliers | mixed | transfer; delayed >24h; lab+field pooled | active-alternative | end-of-treatment | near-transfer |
| Delayed transfer, laboratory studies | SMD | 0.75 (95% CI 0.26-1.25, p=0.004); 17 studies, 46 effect sizes, 738 participants; falls to 0.62 (0.24-1.01, p=0.002) after removing 2 outliers. Setting moderator test was non-significant (p=0.25) | researcher-designed | transfer; delayed; lab | active-alternative | end-of-treatment | near-transfer |
| Delayed transfer, applied/field studies | SMD | 0.37 (95% CI -0.02 to 0.69, p=0.06), non-significant; 17 studies, 40 effect sizes, 683 participants; falls to 0.11 (-0.13 to 0.34, p=0.36) after removing 2 outliers. (The paper's abstract reports this subgroup as 0.34; its results section says 0.37 — an internal inconsistency in the paper.) | researcher-designed | transfer; delayed; field/applied | active-alternative | end-of-treatment | near-transfer |
| Delayed transfer in young participants (<18y) | SMD | 0.12 (95% CI -0.28 to 0.53, p=0.54), negligible; 21 effect sizes, 468 participants; falls to 0.02 (-0.36 to 0.39, p=0.36) after removing 1 outlier | mixed | transfer; delayed; lab+field | active-alternative | end-of-treatment | near-transfer |
| Delayed transfer in adults (18-59) and older adults (>=60) | SMD | Adults 0.54 (0.16-0.92); 55 effect sizes, 812 participants; falls to 0.38 (0.13-0.62, p=0.003) after removing 5 outliers. Older adults 1.28 (-0.34 to 2.90, p=0.10, ns; only 8 effect sizes, 121 participants) | mixed | transfer; delayed; lab+field | active-alternative | end-of-treatment | near-transfer |
Cited by
- Practice scheduling for motor skills — varied vs blocked, spaced vs massedmixedconf: highgc: low