37Peak

Anders Ericsson and Robert Pool, Peak: Secrets from the New Science of Expertise, Houghton Mifflin Harcourt, 2016

Linked to 31 chapters: supports 13 · contradicts 9 · mechanism 19 · same claim 3 · pairing 1

Ericsson spent thirty years asking what separates the best people in a field from everyone else, and this book, written with the science journalist Robert Pool, is his answer for a general reader. It has two halves. The first is that the human brain and body are adaptable: pushed past the point where they can hold themselves steady, they rebuild themselves, and the rebuilding is the skill. The second is that only certain kinds of work push them. Ericsson separates practice into three sorts — naive repetition, purposeful practice, and the deliberate practice that a mature field with teachers and standards makes possible — and argues that what the third sort builds is not speed or reflex but mental representations, the patterns held in long-term memory that let an expert see a chessboard, a patient or a score as a structure rather than a heap of parts. From this he draws the book's most contested conclusion: in every field anyone has examined closely, he found no evidence of an innate gift for it, only a history of practice that nobody had counted. The book also arrives carrying a phrase Ericsson did not coin and does not accept, Malcolm Gladwell's ten-thousand-hour rule, and it gives that phrase a chapter of its own in order to take it apart.

The lesson of perfect pitch

In 1763 a letter to a newspaper in Augsburg reported that the seven-year-old Wolfgang Amadeus Mozart could name any note he heard — on any instrument, from another room, and also the chime of a clock or the sound of a sneeze. Ericsson opens with absolute pitch because it looks like the cleanest case of a gift. About one person in ten thousand has it; Beethoven is thought to have had it and Brahms not, Horowitz had it and Stravinsky not. Two findings had already complicated the picture. Nearly everyone with absolute pitch began musical training between the ages of three and five, which should be irrelevant if the ability is innate. And it is far commoner among speakers of tonal languages, yet no commoner among people of Asian ancestry who grew up speaking English — so the cause lies in the language learned, not in ancestry. The decisive experiment came in 2014, when the psychologist Ayako Sakakibara, at the Ichionkai Music School in Tokyo, took twenty-four children aged two to six and trained them in four or five short sessions a day to identify fourteen major chords by ear. Some finished in under a year, some took eighteen months. Tested afterwards on single notes, every one of them had absolute pitch.

The book's title claim follows. Perfect pitch is not the gift; the capacity to develop perfect pitch is the gift, and nearly everyone is born with it. Leopold Mozart had written the first training manual for children's musical development, had tried it on Wolfgang's older sister Maria Anna, and was working with Wolfgang full time by the age of four — earlier and harder than anything Sakakibara's children went through. What the boy inherited was not an ear but a brain flexible enough to be given one. Ericsson notes the limit as well: the adaptability that produces absolute pitch seems to close around the age of six, and a child who has not been trained by then will probably never acquire it.

Until the 1990s most scientists held that the adult brain's wiring was fixed and that learning could only fill a cup whose size was set at birth. Brain imaging overturned that, and Ericsson takes the overturning as licence to think about potential differently: not a fixed reserve to be drawn down but a vessel whose shape is made by what a person does. His witness for this is Ray Allen, the most prolific three-point shooter in the history of the National Basketball Association, who told a reporter that being described as blessed with a beautiful jump shot infuriated him. “Don't undermine the work I've put in every day,” he said. Allen's high-school coach remembered his jump shot as no better than his teammates' and in fact poor.

The power of purposeful practice

The research programme began in 1978 with an undergraduate at Carnegie Mellon named Steve Faloon, hired to sit in a room while Ericsson read him digits at one a second and repeated them back. Short-term memory holds about seven items, and after four sessions Faloon had confirmed the textbooks, telling Ericsson that there seemed to be a limit for him somewhere around eight or nine digits. Ericsson had found a forgotten 1929 paper by Pauline Martin and Samuel Fernberger reporting that two students had raised their digit spans over four months, and he wanted to watch how. On the fifth session Faloon got eleven digits right. By session sixty he was at twenty, past a hundred sessions he was at forty, which was more than any professional mnemonist had managed, and after more than two hundred sessions he reached eighty-two.

Faloon did not enlarge his short-term memory. He built a way around it. Hearing 907 he heard a two-mile running time of nine minutes seven seconds — he was a distance runner — and the three digits became something already in his long-term memory. To keep the groups in order he fixed a pattern in advance: four groups of four digits, four groups of three, and a six-digit tail that he repeated to himself until he had the sound of it. Ericsson called this a retrieval structure. When Faloon stalled at twenty-two digits, the remedy was not effort but a different arrangement of groups.

Around this experiment Ericsson builds a taxonomy. Naive practice is repetition in the hope that repetition is enough; his example is a music teacher's exchange with a pupil who practised an hour a day, played the test piece ten or twenty times, got it right once or twice, and when asked how he had practised it answered that he had just played it. Purposeful practice has a specific goal, demands full attention, generates feedback, and takes place just outside what the learner can already do. Most people, Ericsson argues, reach an acceptable level at anything — driving, tennis, teaching, medicine — let it become automatic, and stop improving from that day. He cites research showing that doctors twenty or thirty years into practice do worse on objective measures than doctors three years out, and he likes the case of Benjamin Franklin, who played chess for fifty years, sometimes against the strongest player in Europe, and remained above average.

The limits of purposeful practice appear in the two people who followed Faloon. Renée Elio improved at his rate to about twenty digits and then hit a wall she never crossed; she had invented mnemonics on the fly from dates and times instead of fixing a structure first. Dario Donatelli, taught Faloon's method at the start, moved faster, stalled at thirty, then built a retrieval structure of his own and eventually passed a hundred digits. Trying hard was not the difference. Knowing what to try was.

Harnessing adaptability

The body's adaptability is easy to see; the brain's had to be found. Ericsson's best case is the London taxi driver. To be licensed for the whole city a driver must learn the Knowledge: every street, building and landmark within six miles of Charing Cross, some twenty-five thousand streets, tested by examiners who name two points and require the best route named turn by turn. Candidates spend years riding the 320 runs in the official guide and the quarter-mile around each end of them. Half drop out.

Eleanor Maguire, at University College London, scanned sixteen drivers in 2000 and found the posterior hippocampus — the part of the brain engaged by spatial navigation — larger than in fifty controls, and larger the longer a man had driven. Comparing taxi drivers with bus drivers, who cover the same city on fixed routes, she found the same difference, which ruled out driving itself as the cause. The objection that remained was selection: perhaps men with large hippocampi were the ones who passed. So she followed seventy-nine trainees and thirty-one controls from the start of training. At the outset the groups did not differ. Four years later forty-one had qualified and thirty-eight had not, and only in the forty-one had the posterior hippocampus grown.

Why the brain and body should behave this way is, Ericsson argues, a consequence of their preference for stability. Cells work to hold temperature, acidity and energy supply steady; exercise that stays inside what those mechanisms can absorb changes nothing. Push past them and the cell's chemistry goes out of range, genes switch on that were dormant — one study counted 112 of them in an overloaded rat muscle — and the tissue rebuilds itself until the new load is comfortable again. Then the change stops. This is why improvement requires continually raising the demand, and why it must be raised only a little: too far outside the comfort zone is injury, not adaptation.

The same pattern shows up wherever anyone has looked. In a 1995 study in Science, Edward Taub and four German colleagues mapped the brains of six violinists, two cellists and a guitarist: the region controlling the fingers of the left hand had expanded into territory normally given to the palm, and the earlier a musician had started, the larger the expansion. The right hand showed no difference. Mathematicians have more grey matter in the inferior parietal lobule, and more of it the longer they have worked. Ericsson adds two qualifications that are easy to miss. Development in one direction costs something in another: Maguire's licensed cabbies, tested on remembering a complex figure after a delay, did worse than people who had never trained, though all three groups had scored alike four years earlier. And the changes need upkeep — retired taxi drivers had less grey matter in the posterior hippocampus than working ones.

Mental representations

On 27 April 1924 the Russian grandmaster Alexander Alekhine sat in a chair at the Hotel Alamac in New York with his back to twenty-six chessboards and played all of them from memory. Runners called out the moves. Twelve hours later he had won sixteen, lost five and drawn five. Alekhine had never trained at blindfold chess in particular; like most grandmasters who can do it, he acquired the ability as a by-product of learning chess.

What makes it possible was worked out in the early 1970s by Ericsson's mentor Herbert Simon with Bill Chase. Shown a position from a real game for five seconds, a master could replace about two-thirds of the twenty-odd pieces and a novice about four of them. Shown the same number of pieces scattered at random, the master managed two or three — the same as the novice. The masters were not remembering pieces. They were remembering patterns, which Chase and Simon called chunks and which are held in long-term memory; Simon estimated that a master has accumulated about fifty thousand of them. The same thing happens with words: read twenty words in random order and most people recall six, but arrange them into a sentence and most people recall nearly all.

Ericsson generalises this into the central mechanism of the book. A mental representation is a pattern of information held in long-term memory that lets a person handle far more at once than short-term memory allows. Everyone has them — the word dog is one — and what separates experts is their quality and number. They are also strictly local. Faloon's structures for digits did nothing for his memory for letters; a diver's are useless in basketball. So there is no such thing as training memory, or training to be an athlete, or training to be a doctor: one trains for digit strings, for the marathon, for diagnosis.

The representations do more than store. They generate candidate moves, and they tell the expert when something has gone wrong. A baseball hitter has no better eyesight or reflexes than anyone else; he has learned to read the pitcher's arm before the ball is out of the hand. A neuro-ophthalmologist presented with a police officer who has an unbearable earache and one pupil smaller than the other recognises Horner's syndrome, then generates the possibilities that could produce both symptoms at once — stroke, shingles, a tear in the carotid artery — asks about weightlifting, and gets the diagnosis that two hundred newspaper readers missed. Writing shows the same shift. A schoolchild describing his method says he sets down ideas until his supply runs out, which researchers call knowledge telling; expert writers build a representation of the whole first and revise it when the writing shows it to be wrong.

Because representations are what practice builds, they also govern how well practice works. Gary McPherson and James Renwick filmed children of seven to nine practising at home. The weakest, a first-year cornet player, made eleven mistakes a minute on a first run through and corrected three in ten of them on the second; the strongest, a first-year saxophonist, made 1.4 a minute and corrected eight in ten. Neither wanted it more. The saxophonist could hear what the piece was supposed to sound like and therefore knew when he had missed it. Skill and representation improve together, each raising the ceiling on the other — a staircase, in Ericsson's image, that you climb as you build it.

The gold standard

In 1987 Ericsson moved to the Max Planck Institute for Human Development in Berlin, near the Universität der Künste, whose music academy produces a steady supply of international soloists. With Ralf Krampe and Clemens Tesch-Römer he studied thirty violin students in three groups of ten, matched for age and sex: those the professors named as possible international soloists, those who were very good but not that, and students in the music-education department who would become teachers. Ten middle-aged violinists from the Berlin Philharmonic and the Radio-Symphonie-Orchester were added as a picture of where the best students were heading. Everyone was interviewed at length about their history and kept a diary of a week in fifteen-minute blocks.

The groups agreed about almost everything. All of them named solitary practice as the thing that improved them most, all of them said it was hard and not enjoyable, and all of them stressed sleep. One thing separated them: accumulated hours alone with the instrument. By the age of eighteen the future teachers had practised an average of 3,420 hours, the very good 5,301, and the best 7,410. The orchestral players averaged 7,336 by the same age, almost exactly the figure for the best students.

From this study Ericsson named deliberate practice, and the definition is narrower than most readers expect. It requires a field already developed enough that the best performers are clearly separated from beginners and that a body of teaching technique exists — musical performance, ballet, chess, the individually scored sports. It requires a teacher who knows what expert performers do and can assign exercises that build towards it. Within those conditions it is purposeful practice with a direction: defined goals, near-maximal effort just past current ability, full conscious attention, feedback that the learner gradually takes over from the teacher, and, as both product and instrument, better mental representations. “Deliberate practice is purposeful practice that knows where it is going and how to get there.”

Most fields do not meet the preconditions — management, teaching, engineering and consulting have no agreed measure of superior performance — and Ericsson's advice there is to get as close as possible: find who is genuinely better by some objective outcome, work out what they do differently, and build exercises from that. He is blunt about how badly the first step usually goes. Judges at the California State Fair wine competition, given three glasses of the same wine in a flight of thirty, scored them 91, 87 and 83, and judges who were consistent one year were inconsistent the next.

Ericsson also disposes of the number the book is best known for. Gladwell's Outliers took ten thousand hours from Ericsson's paper, where it was the average accumulated by the best violinists at twenty; the figure at eighteen, 7,410 hours, was equally available and less memorable. Half of those ten violinists had not reached ten thousand hours at twenty, and none of them was yet a master — pianists who win international competitions tend to do it near thirty, by which time they have twenty to twenty-five thousand hours behind them. Faloon became the best person in the world at his task after about two hundred hours. Gladwell's estimate that the Beatles played ten thousand hours in Hamburg was revised to about eleven hundred by Mark Lewisohn's 2013 biography, and in any case performing a set to a crowd is not practice designed to fix a weakness; nor would either activity explain Lennon and McCartney's songwriting, which is what the band is remembered for. What Gladwell got right, Ericsson says, is that in any field with a long history of people trying to be the best, it takes many years of hard work — and the reason the number is large is that everyone you are competing with has done the same.

Taking the method outside the conservatory produces the book's most striking single result. In 1968 United States Navy pilots over Vietnam were losing roughly one aircraft for every one they shot down. The Navy's answer was the school that became Top Gun: its best pilots flew Soviet tactics against students day after day, cameras and radar recording every encounter, and the real work happened on the ground afterwards, where instructors made students account for what they had seen, what they had done and why. Between 1970 and 1973 Navy pilots shot down 12.5 enemy aircraft for each one lost, while the Air Force, which had not changed its training, stayed near two to one. Ericsson's reading is that training in most professions teaches knowledge when what is wanted is skill; a 2005 Harvard review of sixty-two studies found that doctors' quality of care mostly declined with years of experience, and that continuing medical education, which is overwhelmingly lectures, barely moves it.

For individuals the advice is a teacher first, then engagement — group classes make it too easy to go through the motions — then a method for plateaus. When Faloon stalled, Ericsson slowed the reading rate slightly, and the extra digits Faloon then held proved the problem was encoding speed rather than capacity; another time he gave strings ten digits longer than any Faloon had managed, and Faloon recalled more digits than ever before while still getting the string wrong, which located the limit in a single faulty group. Work out which component fails and attack that. On motivation Ericsson is unsentimental: his study of 274 competitors at the 2006 Scripps National Spelling Bee found that the best spellers had done the most solitary study and that none of them enjoyed it, and he doubts a general faculty of willpower exists at all, since the evidence suggests it is specific to the activity.

The road to extraordinary

In the late 1960s the Hungarian psychologist László Polgár, having studied hundreds of people called geniuses, concluded that any child could be made into one and looked for a wife who would help him test it. Klara agreed. They chose chess over languages and mathematics for a practical reason she later gave a reporter: chess is objective and easy to measure. Their three daughters were home-schooled around the game. Susan won the Budapest girls' under-eleven championship at four with ten wins and no losses, and became the first woman to earn the grandmaster title by the route men take. Sofia, at fourteen, won eight of nine games against grandmasters in Rome. Judit became a grandmaster at fifteen years and five months, younger than anyone before her, held the top place among women for twenty-five years, and reached eighth in the world overall.

Ericsson reads the sisters as an extreme instance of an ordinary sequence. Benjamin Bloom's study at Chicago in the early 1980s examined the childhoods of 120 experts in six fields — concert pianists, Olympic swimmers, tennis champions, research mathematicians, research neurologists and sculptors — and found three stages, to which Ericsson adds a fourth. First the thing is a toy: Susan Polgár remembered finding the chess set in a cupboard and liking the shapes, Tiger Woods was given a club at nine months. Parents turn the play towards the real purpose of the toy, praise what the child achieves, and pass on their own interests; older siblings, who appear in case after case, do part of the teaching. Second comes a teacher, and with the teacher the first real practice, which is work. Third, usually in the mid-teens, comes commitment — the best available coach rather than the nearest, families moving house, and costs Bloom found few families could bear for more than one child.

Starting young matters, but for specific reasons. Ballet turnout has to be trained before the hip and knee joints finish forming, between about eight and twelve. A tennis player's serving motion and the 20 per cent extra bone thickness in the racket arm come from early training, though bone keeps adapting into adulthood. The corpus callosum is larger in musicians only if they began before seven, while the cerebellum is larger whenever they started. Against the hard version of the age limit Ericsson sets Paul Brady, a Bell Labs researcher who at thirty-two set a computer to generate pure tones and practised half an hour a day; after two months he could name all twelve notes, and over fifty-seven days of testing on a piano he was exactly right thirty-seven times and half a tone out eighteen, which meets the technical definition of absolute pitch.

The fourth stage, Ericsson argues, is not a different faculty. Innovators are almost always expert performers first — Picasso painted in a near-classical style before he left it — and what their teachers gave them was the ability to improve without a teacher, representations good enough to diagnose their own performance. From there the work is the same, one rung at a time. There are no big leaps, only leaps that look big to people who did not see the steps. Nigel Richards, three times world Scrabble champion, won the French championship in 2015 without speaking French, having spent nine weeks memorising the dictionary; he will not discuss his methods, but his results told other players what was possible, which was enough to move the field.

Then Ericsson turns to the objection he is always given. His method with any reported prodigy is two questions: what exactly is the ability, and what training could have produced it. In thirty years he says he never found a case that survived both. Paganini's feat of finishing a piece after three strings broke was an act he designed: he had written pieces for one string, and he broke the others himself. The Mozart compositions dated to age six survive in Leopold's hand; the keyboard concertos credited to the eleven-year-old turn out to be reworkings of other men's sonatas, plausibly set as exercises; the first works unambiguously his own come at fifteen or sixteen, after a decade of training. Donald Thomas, who cleared seven feet in basketball shoes and won the world championship a year later, had jumped about six feet four in high school, used a competent Fosbury Flop that nobody acquires without practice, and had spent years training the single-leg take-off that dunking and high jumping share; in the nine years after his discovery he never beat the 2.35 metres he jumped in 2007, which is not the trajectory of untapped talent.

The strongest evidence comes from chess. In 2006 Merim Bilalić, Peter McLeod and Fernand Gobet measured intelligence and practice together in fifty-seven schoolchildren from chess clubs. Across the whole group practice explained most of the difference in playing strength, and intelligence — memory and processing speed rather than visuospatial ability — explained some. Among the twenty-three who played in tournaments, practice still mattered and intelligence did not: within that group the players with lower measured intelligence were on average slightly the better players, because they had practised more, having had to work harder to keep up at the start. The pattern recurs elsewhere. Piano students with higher scores do better after six months and no better by the time they are professionals; dental students' visuospatial scores predict performance on a simulated jaw and predict nothing once they are residents; Korean Go masters have been measured with an average intelligence quotient around 93. Among working scientists the measure does not predict productivity at all.

What Ericsson thinks genes most likely do, if they do anything beyond height and body size, is act on practice rather than on ability — making one child find drawing or music more pleasurable, or another able to concentrate for longer. The reason he presses the point is the self-fulfilling prophecy. Canadian professional hockey players are heavily drawn from those born in January, February and March, because the age cut-off makes them the oldest in their childhood cohorts; coaches read the resulting advantage as talent, give those children more attention and more games, and the reading comes true. In one study more than 90 per cent of thirteen-year-old soccer players nominated as the best were born in the first half of the year. If the same selection operates in school mathematics, and Ericsson thinks it does, the children labelled ungifted at seven are lost to engineering and physics for no reason at all. His last illustration is a week of first-year physics at the University of British Columbia in which Carl Wieman's group replaced lectures with small-group problems and immediate feedback: the redesigned section scored 74 per cent against 41, an effect larger than any teaching intervention previously measured. He ends by proposing a new name for the species — Homo exercens, practising man, in place of knowing man.

Since 2016

The neuroscience has held. Maguire's four-year study of the taxi drivers, Taub's map of the violinists' cortex and the general finding that training reshapes the regions it taxes are not in dispute, and the applied programme Ericsson argued for has spread: simulation with structured debriefing is now normal in surgery, aviation and emergency medicine, much as Top Gun did it. What has been fought over is the size of the claim. In 2014 Brooke Macnamara, David Hambrick and Frederick Oswald pooled eighty-eight studies and reported that accumulated practice accounted for 26 per cent of the variation in performance at games, 21 per cent in music, 18 per cent in sports, 4 per cent in education and under 1 per cent in the professions. Ericsson answered that their measure counted any practice at all, including the group rehearsal and match play he had spent his career distinguishing from deliberate practice, and that retrospective questionnaires cannot do the work his interviews did.

The more uncomfortable result came in August 2019, when Macnamara and Megha Maitra published a replication of the 1993 Berlin study in Royal Society Open Science. They recruited thirty-nine violinists, sorted them into three groups of thirteen by teacher and expert rating, and ran the estimates and diaries under a double-blind procedure the original had not used. The least accomplished group had practised markedly less, as before. But the two upper groups did not separate: the violinists rated merely good had logged as many solitary hours as the best, or more. Macnamara and Maitra also noted that ten thousand hours had never been a threshold in the original paper, only one of several averages. Ericsson replied that neither the samples nor the ratings were comparable with his; the reply appeared after his death.

Ericsson died at his home in Tallahassee on 17 June 2020, aged seventy-two. The argument he left is now usually described as settled in the middle. Nobody disputes that expert performance requires an enormous amount of the right sort of work, or that the amount of it in most people's lives is the first thing worth fixing. Hambrick and others hold that practice is one factor among several, including starting age, working memory and whatever disposes a person to practise in the first place — which is close to the position Ericsson himself took on genes, minus his insistence that the remaining variance would evaporate under better measurement. The ten-thousand-hour rule, which Ericsson spent a chapter of Peak refuting and the rest of his life refuting in public, remains the one thing most people know about his work.