The brain you keep

Learning to crawl again

In 1959 Pedro Bach-y-Rita, a 65-year-old Catalan poet and scholar who taught at City College of New York, had a stroke that paralyzed half his body and took his speech. Doctors said he would end his days in an institution. Instead his son George, then a medical student in Mexico, took him in and began where a baby begins: on hands and knees, crawling along a wall for months, then kneeling, standing, and walking. Pedro relearned to speak and to type, went back to teaching, and worked until he retired. He died of a heart attack at 72 while hiking at about 9,000 feet (2,700 meters) near Bogotá. His other son, Paul, a scientist, arranged an autopsy. As Paul later described the slides, the lesion was huge and had never healed, and about 97 percent of the fibers running from the motor cortex to the spinal cord were destroyed. A man with that much damage had walked, spoken, and taught.

The lecture tells this story with a Greek professor who had his stroke at 56 and died at 3,600 meters. The accounts we found say a Catalan scholar, 65, and about 9,000 feet. The story comes from the psychiatrist Norman Doidge’s popular book The Brain That Changes Itself (2007) and Paul’s family obituary; the autopsy figures rest on Paul’s memory of the slides, and no pathology report has been published, so treat this as a case, not proof. Paul spent his career on the idea it illustrates. In 1969 he and colleagues described a device that let blind people perceive camera images through patterns of touch on the skin.

Maps that rewrite themselves

The brain holds maps of the body. Between 1928 and 1936 the Montreal surgeon Wilder Penfield touched the exposed brains of 126 awake patients with an electrode and asked what they felt. In 1937 he and Edwin Boldrey published the first map of the body on the brain’s surface, with the hand, lips, and face taking up far more room than the trunk. For decades such maps were assumed to be fixed in adults.

In 1984 Michael Merzenich and colleagues amputated the middle finger of adult owl monkeys and mapped the hand area of the touch cortex two to eight months later. The patch that had served the lost finger had been taken over by the neighboring fingers, whose maps were larger and finer. The change had limits: it stayed within about 500 to 700 micrometers of the old borders, a fraction of a millimeter.

Attention matters too. In owl monkeys trained to tell small differences in touch on one fingertip, the trained skin ended up with a representation 1.5 to more than 3 times larger than comparable skin, while monkeys given the same touches as they attended to sounds changed much less. Practice with attention changes the map; mere exposure changes it less.

In people, the best-known measurement is of string players. In 1995 Thomas Elbert and colleagues found that the area responding to touch on the fingering fingers of the left hand was larger in nine string players, violinists, cellists, and a guitarist, than in non-musicians, with no difference for the right hand, and most of all in those who began before age 12. The lecture says 12 violinists of the Berlin Philharmonic; no source we found has that.

When the senses swap jobs

The visual cortex of people blind from early life does not sit idle. In 1996 Norihiro Sadato and colleagues used PET scans to show that reading Braille activated the visual cortex of blind people, while the same task deactivated it in sighted people. In 1997 Leonardo Cohen and colleagues sent a magnetic pulse through the skull over the visual cortex of early-blind people, and their Braille reading broke down. The same pulse did nothing to touch in sighted people. The “visual” cortex was working for the fingers.

The most surprising study is the shortest. In 2008 Lotfi Merabet and colleagues randomly assigned 47 sighted adults to wear a light-proof blindfold for five days or not, while both groups trained intensively on Braille. By day five the blindfolded group’s visual cortex responded to touch, and a magnetic pulse over it disrupted their Braille reading. A day after the blindfold came off the effect had faded. That speed suggests the connections were already there, switched off, not that new wiring grew in days. And in people blinded after about age 14 the takeover did not appear (Cohen and colleagues, 1999).

Recovery after stroke and injury

After a stroke many people stop using a weak arm because trying fails and the other arm works, and the habit hardens. The psychologist Edward Taub called it “learned non-use” and built a therapy on it: the good hand is restrained in a mitt while the patient practices tasks with the weak one, intensively. In the EXCITE trial, 222 people three to nine months after a first stroke were randomly assigned to the two-week program (106 people) or usual care (116). A year later the time to complete a set of arm tasks had fallen 52 percent in the treated group against 26 percent in the others, and the benefit lasted two years. Taub co-authored the trials, and the patients had some wrist and finger movement left, but the result stands against the idea that recovery ends in the first weeks.

Children show the deepest reorganization, and its limits. When severe epilepsy comes from one side of the brain, surgeons sometimes disconnect or remove that hemisphere. Among 111 patients at Johns Hopkins from 1975 to 2001, 65 percent became free of seizures and 89 percent could walk without help, though in a related series of 71 children cognition was mostly well below average, predicted mainly by the underlying disease. Language can move to the other side: in five people with early left-brain injury, the right hemisphere showed activity much like the left in healthy controls (Staudt and colleagues, 2002). It can also crowd: in young adults with early left injuries, the more their language lay on the right, the worse their visuospatial skills (Lidzba and colleagues, 2006).

The lecture’s most striking example is a girl of four who lost part of her left brain and then learned to read. In the published case, what was removed was a visual region, including the usual letter-recognition spot, not a language area. By 11 she read, only a little more slowly than usual, and a scan showed a letter-recognition spot in her right hemisphere (Cohen and colleagues, 2004).

Taxi drivers and jugglers

To get a license, a London taxi driver must learn “the Knowledge”: about 25,000 streets within six miles of Charing Cross and thousands of places of interest, a three-to-four-year process that, by the licensing body’s account, half or more of trainees fail. In 2000 Eleanor Maguire and colleagues scanned 16 licensed drivers and 50 other men. The back of the drivers’ hippocampus was larger, the front was smaller, and the size tracked years on the job. Bus drivers, who drive all day on fixed routes, did not show it (2006), and the taxi drivers were worse at acquiring new visual and spatial information, a hint of a trade-off.

Skeptics could say that people with big hippocampi become taxi drivers. In 2011 Katherine Woollett and Maguire scanned 79 trainees at the start of training and 3 to 4 years later, with 31 controls. The groups looked alike at the start. Afterward, only the 39 who qualified had gained tissue in the back of the hippocampus; the 20 who failed, and the controls, had not. The qualifiers were also worse than controls at recalling a complex drawing after a delay.

Juggling gave the first scan-based sign that adult learning changes structure. In 2004 Bogdan Draganski and colleagues taught 12 young volunteers a three-ball routine over three months, until each could keep it going for a minute. Gray matter in a visual-motion region increased, and then shrank back when practice stopped. In 2008 Janina Boyke and colleagues taught people aged 50 to 67. Only 10 of 44, 23 percent, reached the one-minute mark, but they too showed transient increases, in the motion area, the hippocampus, and a reward-related region.

Read these studies with care. A scan measures volume or density, not neurons. The authors of the second juggling study list cell size, supporting cells, new connections, and blood flow among the possible causes, and a 2013 critical review argued that design and statistical problems may underlie many reported effects.

New neurons?

The idea that the brain adds neurons in adulthood is old. In 1962 Joseph Altman reported dividing cells in adult rat brains, and in the early 1980s Fernando Nottebohm found that two song-control regions of the male canary brain are much larger in spring than in autumn, in birds that learn new songs each year. In 1998 Peter Eriksson and colleagues reported dividing cells in the hippocampus of five cancer patients, aged 57 to 72, who had been given a labeling chemical before they died. The lecture says the patient was 89; the oldest was 72. In 2013 Kirsty Spalding and colleagues used carbon-14 from Cold War bomb tests to date hippocampal neurons and estimated about 700 new neurons a day in each hippocampus.

Then, in March and April 2018, two large studies gave opposite answers. Shawn Sorrells and colleagues found no young neurons in the adult hippocampus; Maura Boldrini and colleagues found thousands of immature neurons in people up to 79. Later work kept splitting: studies in 2019, 2022, and 2025 reported immature neurons or progenitor cells in adults, while a 2022 gene-activity study found no clear sign of them. In mice, blocking neurogenesis impaired telling nearby places apart. In people no one has shown that new neurons matter for memory.

Does brain training work?

Merzenich went on to help found and advise companies that sell brain-training programs. Does practicing a computer game improve anything beyond the game? In 2010 Adrian Owen and colleagues enrolled 11,430 people in six weeks of online training and found improvement on the practiced tasks and no evidence of transfer to untrained ones. In 2014 more than 70 scientists signed a statement that brain-game claims were overstated, and 133 signed a reply. A 2016 review by Daniel Simons and colleagues found extensive evidence of improvement on trained tasks, less for closely related ones, and little for daily life. In January 2016 the US Federal Trade Commission announced a $50 million judgment against Lumos Labs, maker of Lumosity, suspended to $2 million, over claims that its games helped work and school performance and delayed age-related decline.

The strongest hint comes from the ACTIVE trial. In 2,832 older adults, ten sessions of memory, reasoning, or speed training were compared with no training, and each group reported less difficulty with daily activities ten years later. In a follow-up analysis, the speed-training group had about a 29 percent lower risk of dementia, right at the edge of significance (hazard ratio 0.71, 95 percent interval 0.50 to 0.998), while the memory and reasoning groups did not differ. The trial was not designed to test dementia, so this is a lead to confirm, not a treatment.

Movement, sleep, and learning

The best-known study of exercise and the brain is by Kirk Erickson and colleagues (2011). They assigned 120 older adults to a year of walking three days a week or to stretching. The walkers’ hippocampus grew about 2 percent while the stretchers’ shrank about 1.4 percent. Both groups improved on a spatial-memory test, though, so walking did not beat stretching on memory. Larger reviews are mixed: a Cochrane review of 12 small trials found no cognitive benefit in healthy people over 55, a meta-analysis of 39 trials found a small to moderate one, and the LIFE trial of 1,635 sedentary adults found none after two years. In the Whitehall II study of 10,308 people followed about 27 years, midlife activity did not predict dementia, and activity fell in the years before diagnosis, which suggests early dementia reduces activity, not the reverse. Exercise is good for the heart and the body; its effect on thinking is small, if real.

Sleep has older evidence. In 1924 John Jenkins and Karl Dallenbach had two students learn nonsense syllables and tested them after hours awake or asleep; they forgot much less after sleep, which the authors read as less interference. In 2013 Lulu Xie and colleagues reported that in sleeping mice the space between brain cells was over 60 percent larger and waste protein was cleared twice as fast, but a 2024 study in mice found the opposite, so this is open. In 20 people, one night without sleep raised a brain marker of the Alzheimer’s protein by about 5 percent in two regions. The guideline from the major sleep societies for adults is seven or more hours.

Aging and memory

A popular claim is that memory declines after 25. In 2015 Joshua Hartshorne and Laura Germine combined 48,537 online participants with standard test norms and found that abilities peak at different ages: processing speed in the late teens, short-term memory for digits in the early twenties, vocabulary around 50 to 65, and reading emotion in people’s eyes held steady from about 40 to 60. A Swedish study that followed 829 people for five years found no decline in memory for personal events before 60. Averages hide that some older adults keep their memory well.

The lecture’s interference idea holds for everyday forgetting: where you parked today is blurred by all the other days you parked in that lot. That is a teaching illustration, not an aging finding. The data say that memory for specific events declines on average from midlife, while vocabulary and general knowledge keep growing into the sixties.

Brain myths

The claim that we use only 10 percent of our brains is traced to William James, who wrote in 1907 that we use “only a small part of our possible mental and physical resources.” His essay has no percentage. “About ten percent” appears in a 1929 almanac, and in a 1936 foreword Lowell Thomas credits James with it. The 2014 film Lucy made it a plot, though its director said he meant it as a metaphor. Brain imaging shows activity across the whole brain over a day.

The “left-brained” or “right-brained” person is also a misreading. Roger Sperry shared the 1981 Nobel Prize for split-brain research, which showed that the two halves specialize, not that people come in two types. In 2013 Jeffrey Nielsen and colleagues analyzed scans of 1,011 people aged 7 to 29 in 7,266 regions and found that connections are lateralized locally, not as a whole-brain left or right style. And “learning styles,” the belief that people learn better when taught in their preferred mode, were reviewed by Harold Pashler and colleagues, who found “virtually no evidence” that matching teaching to a style helps. In a 2012 survey of 242 teachers in the UK and the Netherlands, more than 80 percent believed in learning styles and in hemispheric dominance.

Protecting the hardware

Protecting the brain starts with its inputs and its casing. The US National Institute for Occupational Safety and Health recommends keeping noise to 85 A-weighted decibels averaged over an eight-hour workday, and halving the time for every 3 decibels above that. A 10-decibel rise is ten times the sound energy. By that rule, 94 decibels allows an hour. The calculator applies the rule to workplace noise; a phone’s volume slider is not calibrated in decibels.

Hearing is also a brain issue. In 2020 the Lancet Commission on dementia estimated that 12 risk factors account for around 40 percent of dementia, which “could theoretically be prevented or delayed” if the links are causal and the factors could be removed. Hearing loss was the largest, at about 8 percent. Whether fixing hearing helps is not settled: a 2023 trial of 977 adults aged 70 to 84 found that hearing aids and counseling did not significantly slow three-year cognitive decline overall.

On helmets the lecture is right. A Cochrane review of 61 studies found that helmets cut a motorcyclist’s risk of death by 42 percent and of head injury by 69 percent, and a 2017 analysis of 40 studies of more than 64,000 injured cyclists found that helmets halved the odds of head injury. These are observational data, consistent across many studies.