Deliberate practice and the 10,000-hour rule

Verdict

Ericsson's framework — that expertise is "largely" explained by accumulated structured, effortful, feedback-rich practice — was popularised by Gladwell into a threshold rule that the original paper never claimed (10,000 hours was the best violinists' mean by age 20, not a cutoff). The honest number is far smaller than either version implies, and the deflation runs deeper than the headline meta-analysis because of a genetic confound the meta cannot address.

What the evidence shows

Three findings kill the threshold reading. Chess masters' practice hours span a nearly 30-fold range; a third of masters had less practice than the mean of the tier below; and in a preregistered, double-blind replication of Ericsson's own 1993 study the best violinists had fewer accumulated hours than the merely good ones, with both groups passing 10,000 hours by age 20.

The variance-explained figure tracks artifacts as much as reality. It rises with domain predictability (24% high / 12% moderate / 6% low) and falls sharply with measurement quality — interview 20% > questionnaire 12% > objective logs 5%. The headline numbers rest substantially on retrospective recall contaminated by current skill.

Ericsson's rebuttal is testable and fails. Every definitional restriction he demanded still leaves the sports estimate at 17–22%; his own best-case reanalysis retains 14 of 88 effects and assumes reliabilities to reach 61% — and even there, teacher-designed practice (r=.56) is statistically indistinguishable from self-directed practice (r=.51).

Hereditarian-lens assessment

Risk: medium, and this is the most important thing on the page. The R² figures are correlations between practice and attainment, and the adversarial pass established that they cannot be read as practice's causal contribution:

• Practice is itself heritable (h² ≈ .69 men / .41 women), so people who practise more are genetically non-random.
• The practice–ability covariance is predominantly genetic (rA .33–.57) with a nonshared- environment correlation of ~0.
• In the decisive test, MZ co-twins discordant by up to 20,228 practice hours did not differ in ability (all p > .2).

So 14% is an upper bound containing gene-environment correlation; the true causal environmental effect is smaller. Note the confound runs in the deflationary direction — it makes the verdict more secure, not less, which is why confidence stays high at medium risk.

Two honest counterweights. The elite r ≈ .11 is computed after selection on both practice and performance, so range restriction inflates the collapse — the artifact licenses neither "practice stops mattering at the top" nor a hereditarian reading. And Mosing's outcome is auditory discrimination (a capacity), not instrument performance, which the authors concede likely does improve with practice.

Boundaries & what critics say

• This is not "talent is everything." Practice is necessary, non-trivial, and the largest single identified predictor in several domains.
• Domain matters enormously: 24% in games vs <1% in professions. Highly predictable, well-structured activities are where practice buys most.
• Retrospective measurement inflates the estimates; objective logs give 5%.

Practical guidance

• Provide structured, feedback-rich practice with a coach — it is necessary, and the alternative is worse.
• Do not promise thresholds. No hour count guarantees expertise; the range is enormous at every level.
• Expect and plan for large differences in rate, rather than treating slow progress as insufficient effort — the strongest causal test finds no within-genotype effect of extra hours.
• Prefer domains and framings where practice actually pays (structured, predictable skills).

Open questions

• What the causal environmental effect of practice actually is, once gene-environment correlation is removed, is unestimated — only bounded above.
• Mosing's design needs replication on performance outcomes (instrument playing, sport skill) rather than discrimination capacity.