The 10,000-Hour Rule: The Researchers Behind It Say the Number Was an Average and Nothing Is Magic About It

There are two useful things to know about the ten-thousand-hour rule. The first is that the researchers whose work produced the number have said in print that it is not a rule. The second is that when somebody pooled every study of practice and performance they could find, practice explained a good deal less of the difference between people than the rule implies.

What the original study actually was

The number comes from Ericsson, Krampe and Tesch-Römer (1993) in Psychological Review, which studied violinists at a Berlin music academy and found that the most accomplished had accumulated more solitary practice than the less accomplished.

In 2016, Anders Ericsson and Robert Pool published a piece adapted from their book Peak setting out, in their own words, what Malcolm Gladwell’s Outliers had done to that finding. Their objections are specific.

The number was an average, not a threshold. “the number of ten thousand hours at age twenty for the best violinists was only an average.” Roughly half the top group had accumulated less than ten thousand hours. A rule that half of your exemplars fail to meet is not a rule.

There is nothing special about the figure. “there is nothing special or magical about ten thousand hours”, and it was seized on, they write, because it was “a nice round number.”

The subjects were not masters. “they were very good, promising students who were likely headed to the top of their field, but they still had a long way to go.” The study is about promising students at twenty, not about world experts.

Not all hours count the same. Gladwell’s version conflated time spent performing with deliberate practice, which is a specific, effortful, feedback-driven activity. As they put it, “an hour of playing in front of a crowd… is not the same as an hour of focused, goal-driven practice.”

What the 1993 study showed, on their own account, is that among students already admitted to a competitive academy, the better ones had logged more solitary practice. That is a correlation within an already highly selected group. It is not a demonstration that ten thousand hours produces mastery.

How much does practice actually explain?

Macnamara, Hambrick and Oswald (2014), in Psychological Science 25(8), meta-analysed the deliberate-practice literature across every major domain in which it had been studied. Their framing question is exactly the one a reader has: “This view is a frequent topic of popular-science writing-but is it supported by empirical evidence?”

There is a complication in reporting their numbers, and it is worth being open about it. A corrigendum was issued in 2018. We could not retrieve its content. So rather than quote the 2014 abstract’s figures, which may have been superseded, we are using the numbers the same three authors state in their own 2020 review of the field, Hambrick, Macnamara and Oswald (2020), Frontiers in Psychology 11, 1134:

Domain	Variance in performance explained by deliberate practice
Games	24%
Music	23%
Sports	20%
Education	5%
Professions	1%
Overall	14%

Read the bottom rows. In professions - medicine, law, the work most people actually do - accumulated deliberate practice accounts for about one per cent of the difference between people. In education, five.

That does not mean practice is irrelevant. Fourteen per cent of the variance is a real and substantial contribution, and nobody in this argument claims you can play the violin without practising. It means the difference between people who have practised comparably is mostly made of something else - starting age, working memory, physical attributes, coaching quality, motivation, opportunity, luck - and the ten-thousand-hour framing tells you none of that.

The same authors’ 2020 verdict on the wider theory:

“It seems difficult to deny that there are serious problems with the deliberate practice view, as Ericsson and colleagues have presented it over the past 25 years.”

Their specific charge is that the theory has shifted its definitions and applied inconsistent evidentiary standards, which risks making it unfalsifiable. Their bottom line is measured: deliberate practice “is an important predictor of individual differences in expertise” but likely “not as important as Ericsson and his colleagues have argued.”

What the argument is really about

It is easy to read this as debunking, and that would be the wrong reading.

Nobody disputes that expert performance requires enormous amounts of practice. The dispute is about whether practice is sufficient, and whether the differences that remain among heavily practised people are explained by more practice.

Ericsson’s position was that with the right kind of practice, almost anyone could reach expert performance in most domains. Macnamara and colleagues’ position is that the data do not support that, and that the domains where practice explains most - games, music, sport - are precisely the domains with the most structured training regimes and the clearest performance measures, while the domains where it explains almost nothing are the messy ones.

For a reader, the practical consequence is small but real. Counting hours is a poor measure of progress, because the studies that count hours explain a minority of the difference between people. What the same literature does support is that the kind of practice matters more than the quantity: effortful, feedback-driven, deliberately targeted at what you cannot yet do.

That is the same shape as findings elsewhere on this site: spacing your practice is well supported while the specific scheduling rules attached to it are not, and commercial products that promise transferable cognitive gains have consistently failed to deliver them. The pattern is that the mechanism is usually real and the number attached to it usually is not.

Limitations and disputes

• A 2018 corrigendum to the central meta-analysis could not be read. The figures printed here are the authors’ own 2020 restatement, which differs from the 2014 abstract. We have not printed the 2014 abstract’s percentages.
• A direct replication exists that we could not read. Macnamara and Maitra (2019), Royal Society Open Science 6, 190327, revisited Ericsson, Krampe and Tesch-Römer (1993). We could verify the citation but not the findings, so this article does not rely on it and does not characterise what it found.
• The 1993 paper itself was not read in the original. Its design and claims are reported as its own lead author described them in 2016 and as the meta-analysts describe them.
• The critics are one research group. Macnamara, Hambrick and Oswald are co-authors on the meta-analysis, its corrigendum and the 2020 review. That is a coherent programme of work, and it is also a single perspective, and Ericsson disputed it in print until his death in 2020.
• Variance explained is not the same as importance. A variable can explain little variance in a selected sample precisely because everyone in it has a lot of the variable. This is a real objection to the professions figure.
• Nothing here tells anyone not to practise.