Settled at three, or made by what comes after?

A CD for every newborn

On 13 January 1998 the governor of Georgia, Zell Miller, proposed to spend $105,000 a year so that every baby born in the state would go home with a classical-music recording. He had been reading about the brain. The CDs were eventually handed out, according to the state’s encyclopedia, with private funding. Florida passed a law about classical music in state-funded child care, and Tennessee had a newborn-CD program too. They were the most visible result of a wave that had been building for four years.

In 1994 the Carnegie Corporation published Starting Points, a report on the needs of children from birth to age three. In April 1997 the White House held a conference on what new brain research said about young children, and the film director Rob Reiner’s “I Am Your Child” campaign was launched alongside it. The message that many people took away was that the first three years decide how a child’s brain, and so a life, will turn out. A conference summary published that year said that early experiences, “particularly in the first three years of life,” have a dramatic impact on brain development.

The music came from a one-page 1993 letter in Nature by Frances Rauscher, Gordon Shaw, and Katherine Ky. Thirty-six college students listened to about ten minutes of a Mozart sonata for two pianos (K. 448), or to a relaxation tape, or to silence, and then took spatial-reasoning puzzles of the paper-folding kind. Scores were higher after the Mozart, for about ten to fifteen minutes. The effect was quickly called “the Mozart effect,” and a 1997 book of that name made it a brand. Nothing in the original work involved babies, long-term intelligence, or even other kinds of tests.

The evidence since has not been kind to the idea. In 2001 William Thompson, Glenn Schellenberg, and Gul Husain found that the boost was matched by an upbeat piece and vanished once the listeners’ mood and arousal were accounted for. A 2010 meta-analysis of nearly 40 studies by Jakob Pietschnig and colleagues found a small overall effect (d = 0.37), a negligible difference between Mozart and other music people enjoy (d = 0.15), signs of publication bias, and “little evidence” for a specific Mozart effect. Listening to something you like can lift your mood for a few minutes, and that is all.

What the brain science says

The campaign did not invent its science. It stretched three real findings. The first is that the infant brain overproduces connections and then prunes them. In 1979 the neurologist Peter Huttenlocher counted synapses in the middle frontal gyrus in 21 human brains from newborn to age 90. Density rose during infancy to a maximum at about one to two years, about 50 percent above the adult level, and then fell between the ages of two and sixteen. Later work showed no single peak: synapse density in auditory cortex peaks around three months, in prefrontal cortex after fifteen months, and in the prefrontal cortex pruning continues into mid-adolescence and beyond (Petanjek and colleagues, 2011). So the three-year mark is an invention. The ages differ for each region, and the pruning does not stop at three.

The second thing is that more synapses do not mean a smarter child. The educational psychologist John Bruer, who wrote the best-known critique, The Myth of the First Three Years (1999), put it this way: there is no linear connection between the number of synapses and brainpower. Pruning is how the brain tunes itself to the world it finds, and a brain that has been trimmed is not a lesser brain. The lecture on which this course is based compares the process to Michelangelo carving the David, a statue made by removing marble, and the image is good if you keep two processes apart: before birth roughly half of the neurons in a region die, and after birth the connections are made in excess and then cut back.

The third finding is the critical period. In the early 1960s David Hubel and Torsten Wiesel closed one eye of kittens and recorded from the visual cortex, and by 1970 they had shown that the kitten’s visual cortex is vulnerable to the loss of use of an eye during a window that begins in about the fourth week of life, is highest until about the sixth to eighth week, and fades by about the third month. Closing one eye for three months from birth cut the share of cortical cells that the eye could influence from 85 percent to about 7 percent, and in an adult cat even a year of eye closure did nothing. They shared the 1981 Nobel Prize in Physiology or Medicine. Critical periods are real. The question is what they look like in people, and whether the window is “three years.”

Windows, not doors

Human windows are not all the same size. Newborns with dense cataracts in both eyes need surgery within the first weeks: in a 2009 study of 37 infants by Eileen Birch and colleagues, visual outcome got worse by about one line on the eye chart for every three weeks of delay up to about 14 weeks, and after that did not depend on age at surgery. But different features of vision have different windows, and some deficits from early deprivation only show up years later.

For language, the window is wider. In 1989 Jacqueline Johnson and Elissa Newport tested the English grammar of 46 native Chinese and Korean speakers who had arrived in the United States at ages from 3 to 39. Among those who arrived before puberty, the later they arrived, the lower their score, in a nearly straight line; among those who arrived later, scores were low and varied. A far larger study in 2018 by Joshua Hartshorne, Joshua Tenenbaum, and Steven Pinker analyzed the grammar tests of 669,498 English speakers and found that the ability to learn grammar is preserved until about age 17.4 and then declines. That is a window much later than three, and it is not a cliff. For first languages the evidence is sadder: Rachel Mayberry and colleagues found that deaf people who had little language in early life, signed or spoken, did poorly with later language, whereas those who had a first language from infancy learned a second one well. In animals, the “brakes” that close a window can sometimes be lifted to restore plasticity (Takao Hensch and others), and whether that helps children with a lazy eye is still being tested.

The orphans of Bucharest

In 1966 Romania’s government issued Decree 770, which banned abortion and restricted contraception, and births nearly doubled the following year. When the Ceaușescu regime fell in December 1989, tens of thousands of children, by different estimates 100,000 to 170,000, were living in state institutions. In 2000 the psychologists Charles Nelson, Charles Zeanah, and Nathan Fox began the Bucharest Early Intervention Project. They screened 187 children under thirty months in six institutions and chose 136, aged 5 to 31 months. Half were randomly assigned to foster families that the project built from nothing, and half stayed in the institutions. A third group, 72 children who had never been institutionalized, was followed for comparison. Foster care barely existed in Romania, so the team had to create it, and the ethics of the trial were debated.

The results were clear. At 42 and 54 months, children who stayed in institutions scored much lower on cognitive tests, and the benefit of foster care was greatest in those placed youngest. Secure attachment at 42 months was found in 17.5 percent of the institutional group, 49 percent of the foster-care group, and 65 percent of the never-institutionalized group. Language scores in children placed before 15 months were similar to those of their peers. At age 18 the gap between the foster and institutional groups was still about 9 points in IQ (73.5 versus 64.7), sixteen years after randomization, and the gap with other young people was large.

A second study, in Britain, followed Romanian children adopted into UK families. Those who had spent less than six months in an institution were, as young adults, much like UK-born adoptees. Those who had spent longer showed persistently higher rates of autism-like patterns, difficulty with social boundaries, and inattention, though their lower IQs had caught up by adulthood. The two studies cannot be compared directly, since no BEIP child was placed before six months and the British adoptive families were selected.

What the evidence supports is a sensitive period for family care, with different windows for different abilities (language by about 15 months, a physical growth window about 12), and some effects, such as psychiatric symptoms, where timing did not matter but family care did. Recovery was substantial and also partial. The data do not say that a child is lost at three. They say that early good care matters, that it can help even when late, and that it is not a vaccine.

What enrichment shows

If deprivation hurts, does enrichment help, and do the benefits last? The results are mixed in a way that matters. The Carolina Abecedarian Project, begun in 1972, enrolled 111 mostly African American infants from low-income families and randomly gave half of them full-day, year-round, high-quality child care from the first months until age five. At age 30 the treated group averaged 13.5 years of schooling against 12.3, and 23 percent had a bachelor’s degree against 6 percent; the researchers saw strong educational benefits, mixed economic benefits, and little effect on social adjustment. The Perry Preschool Project in Michigan, which began in 1962 with 123 children, showed adult gains in education and employment and fewer arrests, although its IQ gains faded within a few years.

Programs run at scale have done less. In the national Head Start Impact Study, which randomly assigned 4,667 children in 2002, modest gains in language and reading during the Head Start year were gone by third grade. In a Tennessee trial of a state pre-kindergarten program with 2,990 children, the children who had been offered places did worse than the others on state tests in third through sixth grades, with the sharpest difference in sixth grade, and the authors said they did not know why. Non-randomized studies of Head Start’s long-term effects, such as one comparing siblings, have been more positive.

A gene is not a verdict

The lecture’s story of genes begins with a family. In 1993 the Dutch geneticist Han Brunner and colleagues described a large family in which several men had borderline intellectual disability and bursts of impulsive aggression, including arson and attempted rape. The men they tested lacked a working copy of an enzyme, monoamine oxidase A, or MAOA, that breaks down serotonin, norepinephrine, and dopamine. The lecture says that every male in the family was awaiting execution. That is wrong: no source says so, and the Netherlands had no death penalty for ordinary crimes after 1870. Brunner himself later cautioned that the findings did not make MAOA an “aggression gene.” The family is one of only a handful known in the world.

The story then moved to the common version of the gene. MAOA sits on the X chromosome, so a man has one copy, and one common variant, the “low-activity” form, is carried by about a third of white men and a larger share in some other groups. In 2002 Avshalom Caspi, Terrie Moffitt, and colleagues reported that among 442 New Zealand boys followed from birth, those who had both been badly maltreated and carried the low-activity form were far more likely to be convicted of violence. About one boy in eight was in that group, and they accounted for 44 percent of the cohort’s violent convictions. Other studies found a smaller effect, and a study of more than 4,000 young Americans (Haberstick and colleagues, 2014) could not confirm it. Because the variant is so common and its effect is small and depends on severe maltreatment, it cannot be used to single out a dangerous person.

The nickname “the warrior gene” was popularized in New Zealand in 2006 and 2007, in media reports and a medical-journal paper about Māori men that other scholars criticized as unsupported and stigmatizing. In courts, the evidence has been heard in about a dozen cases. In Tennessee in 2009 a jury convicted Bradley Waldroup of voluntary manslaughter and attempted second-degree murder instead of murder, after hearing that he had been badly abused as a child and carried the low-activity variant; he was sentenced to 32 years, and one juror later said the science had helped. In Italy in 2009 an appeals court cut a man’s sentence from nine years to eight, citing his genetic profile. Researchers who reviewed the cases said it is impossible to say how much the gene mattered. A 2019 test of 18 famous candidate genes for depression, in samples of 62,000 to 443,000 people, found no support for any of them.

Chemical messengers, with caution

Brain chemicals attract more popular claims than any other part of this field. Dopamine is the best known. Arvid Carlsson showed in the 1950s that it is a transmitter in its own right, and in 1960 Oleh Hornykiewicz and Herbert Ehringer found it depleted in the movement regions of the brains of people with Parkinson’s disease, which led within a year to the first treatment with L-DOPA. In 1997 Wolfram Schultz, Peter Dayan, and Read Montague showed that dopamine neurons in monkeys signal errors in the prediction of reward. The label “the pleasure molecule” is disputed: Kent Berridge and Terry Robinson argued that dopamine drives wanting, not liking.

The lecture tells a story of a doctor whose eight-year-old son’s brain scan showed that attention-deficit disorder is “under-arousal,” which led to stimulants. That story is not documented. What we can trace is that in 1937 the psychiatrist Charles Bradley, in Rhode Island, gave Benzedrine to 30 children to relieve headaches and noticed that about half did dramatically better at school and became calmer, and that methylphenidate was made in 1944 and named Ritalin after its inventor’s wife, Rita. The idea that stimulants calm only children with ADHD was tested in 1978 by Judith Rapoport and colleagues, who gave a stimulant to 14 normal boys: they too became less active and performed better on tests.

The gut gets “second brain” headlines. About 90 to 95 percent of the body’s serotonin is made by cells in the gut lining, and the gut’s own nervous system holds some 200 to 600 million neurons. But serotonin made in the gut does not cross into the brain, which makes its own, and the gut’s nervous system runs digestion and does not think. In 2022 Joanna Moncrieff and colleagues reviewed 17 studies and found no consistent support for the idea that depression is caused by low serotonin; other psychiatrists replied that serotonin still matters, though nobody now defends a simple chemical imbalance.

Oxytocin is the third. In A Midsummer Night’s Dream, the flower called love-in-idleness makes anyone whose eyes it touches fall in love with the first living creature they see, and the queen Titania ends up adoring a man with a donkey’s head. It is a plot device, and oxytocin is not a love potion. It is a nine-amino-acid hormone that makes the womb contract in labor and triggers milk let-down. In 2005 Michael Kosfeld and colleagues reported that a nasal spray of oxytocin made people more trusting, and the finding made headlines. Later, larger studies, including a registered replication in 2020 that involved one of the original authors, did not find the effect, and we found no study showing that handshakes or kisses work through oxytocin. “The love hormone” is a headline, not a finding.

Touch and early care

The lecture’s most vivid animal story is one where the facts hold. In the 1990s Michael Meaney’s group in Montreal found that rat mothers differ in how much they lick and groom their pups, and that pups who were licked more grew up less reactive to stress (Liu and colleagues, 1997). In 1999 Darlene Francis and colleagues swapped pups between mothers. The calmness, and also the mothering style, followed the mother who raised the pups, not the one who gave birth, so the trait passed from mother to daughter by rearing and not by genes. In 2004 Ian Weaver and colleagues found that the licking changed chemical tags on the DNA of a stress-hormone-receptor gene in the pups’ hippocampus, differences that lasted into adulthood and were reversed in adults by a drug. A 2009 study by Patrick McGowan and colleagues found a similar pattern in the brains of 12 people who had died by suicide after childhood abuse, compared with 12 who died by suicide without abuse and 12 controls.

The lecture adds that brushing pups with a soft paintbrush restores the effect. What is documented is narrower. Brush stroking restored growth hormone and growth-related markers in rat pups separated from their mothers (Schanberg and Field, 1987). In 1986 Field’s team reported that premature human infants who had stroking and limb movements three times a day for ten days gained 47 percent more weight than controls; a later Cochrane review found a smaller gain and weak evidence for developmental benefit. Skin-to-skin “kangaroo care” has stronger support: a Cochrane review of 21 studies found that it reduces deaths and infections in low-birth-weight infants. It is much less clear that it changes brain development, and the proposal that early care leaves a chemical trace on genes in people rests on small studies and is far from settled.