Researchers studied more than 34,000 elite performers across sport, science, music and chess and found something striking: the children who excelled fastest tended to specialize early, while the adults who eventually reached the very top usually developed more slowly and explored more widely first.

Talent systems are built to notice speed. The child winning early is selected, funded and given more specialised practice. The slower child is easier to overlook, even when that slower development may be part of the path that eventually carries someone further.

A review published in Science in December 2025 found a recurring split across sport, science, music and chess. Young high performers tended to progress quickly through intensive practice in one discipline. Adults who later reached the highest international levels more often developed gradually and accumulated experience across several disciplines before narrowing their focus.

The finding is worth taking seriously, but it should not be read as the final word. This was a synthesis of existing studies, not an experiment that assigned children to specialise or explore. It shows a consistent association across unusual careers. It cannot promise that breadth will create excellence in any individual child.

The 34,839 figure comes from 19 different datasets

The Science review was written by Arne Güllich of RPTU University Kaiserslautern-Landau, Michael Barth of the University of Innsbruck, David Hambrick of Michigan State University and Brooke Macnamara of Purdue University. Its 19 datasets contained 34,839 adult international top performers. The authors also considered 66 representative studies of young and sub-elite performers.

Those 34,839 people were not one cohort followed from childhood under a shared protocol. The review combined evidence about international athletes, Nobel laureates and nominees, top chess players, classical composers, scientists, artists and film directors. Each field defined performance differently, using measures such as medals, rankings, citations, awards or the reception of major work.

The paper’s indexed abstract describes three cross-domain patterns: exceptional young performers and world-leading adults were largely different populations; early success was associated with more specialised practice and faster progress; adult world-class performance was associated with more multidisciplinary practice and slower early progress.

That is more precise than saying researchers “studied 34,839 children”. They assembled a large retrospective picture of adult excellence, then compared it with evidence about earlier and lower levels of performance.

Early specialisation can do exactly what it is designed to do

Among young and sub-elite performers, intensive practice in the main discipline was associated with better current performance. Stronger junior athletes started their main sport younger, entered talent programmes sooner and reached national or international milestones earlier than lower-performing peers. They accumulated more practice in that sport and less in others.

If the outcome is who wins at 14, this makes sense. A child who has spent years rehearsing the exact event being measured holds an advantage over one dividing time among several activities. Selection can then amplify the difference by directing stronger coaching, facilities and competition towards the current winner.

The classic deliberate-practice account of expert performance established why sustained, effortful work on specific weaknesses matters. The new review does not overturn the need for practice. Its contribution is to separate two questions that are often collapsed: what predicts superior performance during development, and what characterised the people who eventually reached the very top at peak age?

Those questions produced opposite answers often enough to challenge early rank as a simple measure of potential.

The senior comparison reversed the junior pattern

In sport, eventual world-class adults had generally started their main discipline later, entered formal talent programmes later and reached performance milestones later than senior athletes who peaked at national level. During childhood and adolescence, they had accumulated less practice in their main sport but more practice in other sports.

World-class athletes in the review averaged involvement in two additional sports across nine years. That result agrees with a 2022 systematic review and meta-analysis of 9,241 athletes, which found that early specialised, coach-led practice was associated with junior success, while more multisport practice and gradual early progress were associated with senior international achievement.

The continuity between junior and senior elites was also low. In the athletic datasets, roughly 13 per cent of international-level junior and senior performers were the same people. This does not mean 87 per cent of talented children failed. It means that being in the very small international junior group and being in the very small international senior group were mostly occupied by different individuals.

Chess supplied a smaller but unusually clear comparison. Players who later ranked in the world top three had, on average, 62 fewer Elo points at age 14 than those whose adult peak was fourth to tenth. At adult peak, the top-three players were 48 points ahead. An earlier grandmaster title distinguished fast developers, but not the group that eventually reached the highest rank.

Breadth meant something different in every field

Multidisciplinary experience cannot be reduced to a single activity. In sport it meant practising other sports. In science it included work across scientific disciplines and experience in other professions, arts, music or craftsmanship. Nobel laureates averaged two additional avocations and showed more multidisciplinary experience than recipients of major national science awards who had not won a Nobel.

The Nobel comparison also complicates the familiar picture of the obvious young star. Laureates in physics and chemistry had slower early growth in citation impact than nominees who never received the prize. In another dataset, German Nobel laureates were less likely than national award winners to have held a student scholarship and took longer to reach their first professorship.

For music, the review included an analysis of 911 operas by 59 highly regarded composers. An opera’s success was positively associated with the composer’s previous work across genres and negatively associated with the number of earlier operas in the same genre. Research on scientists, artists and film directors similarly found that periods of unusually high-impact work were often preceded by exploration across subjects, genres or styles.

An earlier meta-analysis of deliberate practice in music, games, sport, education and professions found that specialised practice explained part, but not all, of the variation in performance. Depth still matters. The live question is how the timing of depth interacts with the range of earlier experience.

The review offers explanations, not a causal test

The authors propose three reasons breadth might help. A search-and-match process gives a young person more opportunities to find a discipline that fits their interests and capabilities. Learning across different activities may build methods of adapting that remain useful after specialisation. A broader path may also limit the chance of an early route ending through loss of interest, an unbalanced workload or injury in physical fields.

None of those mechanisms was directly tested across all 34,839 performers. The evidence is largely observational and draws on studies with different designs. Retrospective accounts can be incomplete. Families and institutions able to provide several activities may differ in wealth, time and access from those that cannot. Definitions of “world-class” are also clearer in ranked competition than in science or art.

The sample begins at the far end of the outcome distribution, which introduces another caution. Studying people who survived long enough to become eminent can reveal common histories, but it does not tell us how many equally broad young people never reached that level. Comparison groups reduce that problem; they do not remove it.

Nor does the paper say that future top performers were ordinary beginners who avoided serious work. Many were already above most peers when young, and many accumulated considerable discipline-specific practice. They developed more slowly relative to other exceptional performers. Breadth preceded depth, but did not replace it.

The strongest implication concerns selection systems

The institutional problem is clearer than any parenting rule. A programme that admits children according to current junior rank and evaluates coaches by near-term results creates pressure for earlier selection and narrower training. It may become highly efficient at producing youth winners while excluding many of the slower developers who will make up the adult elite.

The University of Innsbruck’s summary recommends giving young people opportunities to pursue two or three disciplines, even when the activities are not closely related. That is the authors’ practical interpretation. The datasets cannot specify an optimal menu or number of weekly hours for an individual child.

The harder policy question is when outcomes are counted. If funding depends on this season’s medals, examination scores or promotions, a coach or teacher has little incentive to preserve time for exploration whose value might appear a decade later. Short measurement windows encourage short development windows.

Silicon Canals recently examined another attractive claim about predicting adult achievement from a child’s mindset. The common error is to take something that predicts the next stage and promote it into a forecast of an adult life. Early performance contains information, but less than a talent system may assume.

A child who excels early has achieved something real. A child developing more slowly has not supplied evidence of a lower ceiling. The review asks institutions to stop treating those two statements as contradictions.