The Evidence on Teaching

The myth of contextual interference learning benefit in sports practice: a systematic review and meta-analysis

Ammar A, Trabelsi K, Boujelbane MA, Boukhris O, Glenn JM, Chtourou H, Schöllhorn WI · 2023

grade Cmeta-analysisindependentmixed
Sample
36 studies, 183 pooled performance outcomes
Population
Sport-practice participants (athletes, PE students, novices); applied sport skills, not lab tasks
Design
933 records screened, 5 databases; 36 studies, 183 pooled outcomes, sport-practice contexts only. Analysed pre-post and post-retention CHANGE scores, not between-group endpoints. NOT preregistered. Same-corpus companion: Ammar et al. 2024, Educ Psychol Rev 36(2), doi 10.1007/s10648-024-09892-z. Criticised for conflating constant with blocked practice, coding serial/alternating as random, and ignoring dependency.
Key findings
The strongest published challenge to CI as a coaching principle: across 36 sport-practice studies neither acquisition, retention nor transfer differed significantly by schedule, and only 20% of 183 outcomes matched CI predictions. The one place the classic dissociation surfaced was 20-24 year olds (blocked better at acquisition, random better at retention) — exactly the young-adult university population that dominates the lab literature. Methodologically weaker than the Czyż metas (no preregistration, disputed coding of constant/alternating conditions, unhandled dependency), so read its null as 'not demonstrated in sport', not 'disproven'.
Genetic confound
Not addressed. Sport samples are self-selected and pre-filtered by talent, compressing variance in a way that can mask schedule effects among already-selected athletes.
Replication notes
Actively contested. Czyż et al. (2024) report significant effects on overlapping literature; Ammar & Schöllhorn filed a comment, Czyż & Wójcik a rebuttal (SportRxiv 437), plus 2025 Educ Psychol Rev commentaries.

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Retention, high CI vs low CI, overall (sport settings)ES (negative favours random/high CI)-0.159, p=0.141 — non-significantstandardizedretention; delayed; field/applied sportactive-alternativeuncleardomain-skill
Transfer, high CI vs low CI, overall (sport settings)ES-0.243, p=0.071 — non-significantstandardizedtransfer; delayed; field/applied sportactive-alternativeunclearnear-transfer
Age subgroup 20-24y: the classic dissociation does appearESAcquisition 0.282 (p=0.030, blocked better); retention -0.405 (p=0.011, random better); no difference at transferstandardizedacquisition and retention; field/applied sportactive-alternativeuncleardomain-skill
Proportion of outcomes consistent with CI theoryCount of pooled outcomes37 of 183 (20%) agreed with the paradoxical CI prediction on acquisition or relatively permanent gainsstandardizedacquisition+retention; field/applied sportactive-alternativeuncleardomain-skill
Companion reanalysis (Ammar et al. 2024, same corpus)ES by phaseAcquisition (pre-post) blocked 28.9±59.5% vs random 27.9±52.8%, ES=0.1 p=0.35; post-to-retention ES=-0.35 p=0.001 favouring high CI; pre-to-retention long-term gain ES=-0.13 p=0.18standardizedacquisition and retention; field/applied sportactive-alternativeuncleardomain-skill

Cited by