Deliberate Practice and Performance in Music, Games, Sports, Education, and Professions: A Meta-Analysis (with 2018 Corrigendum)
Macnamara, B. N., Hambrick, D. Z., & Oswald, F. L. · 2014
grade Cmeta-analysisindependentreplicatednumbers spot-checked
Population
Mixed ages and skill levels across five domains: music, chess/Scrabble, sports, academic courses, professional work (aviation, medicine, sales)
Design
157 effect sizes, N=11,135 (music 28, games 11, sports 60, education 51, professions 7; 29 unpublished). Mixed-effects meta-analysis of CROSS-SECTIONAL correlations between retrospective accumulated-practice estimates and attained skill. No randomization or within-family control. I2=84.9. Funnel symmetric; trim-and-fill imputed 0 missing. 2018 corrigendum fixed a dependent-effect formula (+~2 points).
Key findings
The deflation of the 10,000-hour rule. Deliberate practice explains 12% of performance variance overall (14% per the 2018 corrigendum) — 86-88% unexplained. By domain: games 24%, music 23%, sports 20%, education 5%, professions <1% (n.s.). Tracks domain predictability (24% high / 12% moderate / 6% low). CORRIGENDUM APPLIED: the performance-measurement moderator is 12% lab / 11% expert-rated / 10% standardized (Q(3)=9.75, p=.021) — so within THIS corpus the usual 'researcher-designed measures inflate ~2x' rule barely holds and should not be cited from here.
Genetic confound
No twin/sibling/within-family design. Authors nominate heritable g and working-memory capacity for the unexplained variance. Practice is itself heritable (Mosing 2014), so 12-14% is an upper bound on any causal share.
Replication notes
Corrected 2018 (Psych Sci 29:1202-1204), no change in substance. Pattern replicated in the 2016 sports meta and Hambrick et al. (2014). Contested by Ericsson (2016) and Ericsson & Harwell (2019) on definitional grounds, not arithmetic.
DOI / URL
10.1177/0956797614535810 (corrigendum: 10.1177/0956797618769891)
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| Overall performance (all domains) | meta-analytic r; % variance (r^2) | r=.35 [.30,.39]; 12% variance [9%,15%] — 88% unexplained. Corrigendum: r=.38 [.33,.42], 14% [11%,18%] | mixed | attained expertise, not an acquisition/retention/transfer manipulation; cross-sectional; mostly field/real-skill | none | not-applicable | domain-skill |
| Performance by domain | % variance (r) | games 26% (r=.51); music 21% (.46); sports 18% (.42); education 4% (.21); professions <1% (.05, p=.62). Corrigendum: 24/23/20/5/1% | mixed | attained expertise; cross-sectional; field | none | not-applicable | domain-skill |
| Moderator: how practice was MEASURED | % variance (r) | retrospective interview 20% (r=.45); questionnaire 12% (.34); objective LOG 5% (.22). Q(2)=16.19, p<.001. Corrigendum: questionnaire 15%, log 4% | mixed | attained expertise; cross-sectional; field | none | not-applicable | domain-skill |
| Moderator: predictability of task environment | % variance (r) | high predictability 24% (r=.49); moderate 12% (.35); low 4% (.21). Q(1)=20.49, p<.001 | mixed | attained expertise; cross-sectional; field | none | not-applicable | domain-skill |
| Moderator: how PERFORMANCE was measured | % variance (r) | group membership 26% (r=.51); lab task 14% (.37); expert rating 9% (.30); standardized objective score 8% (.28). Q(3)=14.41, p=.002 | mixed | attained expertise; lab vs field contrast inside one meta | none | not-applicable | domain-skill |
Cited by
- Deliberate practice and the 10,000-hour rulemixedconf: highgc: medium