The Evidence on Teaching

The Role of Deliberate Practice in Expert Performance: Revisiting Ericsson, Krampe & Tesch-Römer (1993)

Macnamara, B. N., & Maitra, M. · 2019

grade Breplicationindependentfailed
Population
Conservatory student violinists, three skill strata (replicating the original 1993 design, n=30)
Design
Preregistered twice (OSF khjs7; results-blind Registered Report jyn5w) and DOUBLE-BLIND, unlike the original: experimenters unaware of skill group, violinists unaware there were multiple groups. N=39 Cleveland Institute of Music violinists, three faculty-nominated groups (13/13/13), sex- and age-matched. Added teacher-designed practice (Ericsson's theoretical DP) alongside practice-alone (his operational DP).
Key findings
The foundational study of deliberate-practice theory fails to replicate under double-blind, preregistered conditions. The best violinists had accumulated FEWER hours (8,224) than the merely good (9,844) — 'complete correspondence between skill level and accumulated practice' is gone. Ericsson's stricter theoretical construct (teacher-designed practice) explained no more variance than his loose operational one, removing the definitional escape hatch. Both top groups passed 10,000 h by age 20, so the number is a correlate of conservatory enrollment, not a threshold for eminence.
Genetic confound
No twin/sibling design. Retrospective self-report remains open to recall bias, though the double-blind procedure removes experimenter-expectancy inflation — likely why the effect shrank versus 1993.
Replication notes
This IS the replication: n=39 vs original n=30, preregistered and results-blind accepted at Royal Society Open Science. Direction REVERSED on the key best-vs-good comparison. No counter-replication exists.
DOI / URL
10.1098/rsos.190327

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Accumulated practice alone to age 18: best vs good violinistsmean hours; t; dBest M=8,224 h [6,400,10,048] vs Good M=9,844 h [6,937,12,751] — best practised NUMERICALLY LESS; t24=-0.93, p=.364, d=-0.38. Core 'complete correspondence' claim NOT replicatedresearcher-designedattained expertise; retrospective; field (conservatory)nonenot-applicabledomain-skill
Accumulated practice alone: good vs less accomplishedmean hours; t; dGood M=9,844 vs Less accomplished M=4,558 h [3,264,5,851]; t16.57=3.26, p=.005, d=1.33 — link survives only across the wide skill gapresearcher-designedattained expertise; retrospective; fieldnonenot-applicabledomain-skill
Overall group effect of accumulated practicechi-square; eta-squaredchi2(2)=13.90, p=.001, eta2=0.26 [.03,.44] — substantial but far below the original's implied r=.70, which exceeded the upper CI of the music meta-analytic r=.48 [.38,.56]researcher-designedattained expertise; retrospective; fieldnonenot-applicabledomain-skill
Teacher-designed practice (theoretical DP) vs practice aloneeta-squared; correlationTeacher-designed eta2=0.23 [.02,.41] vs practice-alone eta2=0.26 — no advantage; highly correlated r37=.72, p<.001. Best M=6,251 h vs Good M=6,821 h, t24=-0.37, p=.717researcher-designedattained expertise; retrospective; fieldnonenot-applicabledomain-skill
10,000-hour benchmarkdescriptiveBy age 20 BOTH the best and the good violinists exceeded 10,000 h of practice alone on average — the threshold did not discriminate the top groupresearcher-designedattained expertise; fieldnonenot-applicabledomain-skill

Cited by