The alarm that fires first

A woman who could not be afraid

Neuroscientists in Iowa call her S.M. She has a rare inherited illness, Urbach-Wiethe disease, in which calcium slowly builds up in some tissues. In her case it destroyed both amygdalae, the two almond-shaped clusters of cells deep in the temporal lobes that are the best-known part of the brain’s emotional machinery. She laughs, and she feels disgust. What she lacks, as far as anyone can tell, is fear.

In 2011 the team that has studied her for decades, led by Justin Feinstein with Ralph Adolphs, Antonio Damasio, and Daniel Tranel, published a set of tests. S.M. was 44 when tested. In an exotic pet store she held a snake for more than three minutes, rubbing its scales and touching its flicking tongue; her fear never rose above 2 on a scale of 0 to 10. She asked fifteen times to touch a bigger one, and when she tried to touch a tarantula she had to be stopped. Asked why, she said she was curious, though she says she hates snakes. In a haunted house at Waverly Hills, a former tuberculosis hospital in Louisville, Kentucky, she walked with five strangers and ended up leading the group; the costumed monsters made the others scream, while she laughed or tried to talk to them, and her fear rating stayed at zero. Film clips that frighten most viewers did not frighten her, and she laughed at the funny ones and recoiled at the disgusting ones. A diary she kept on a handheld device over three months, at three random moments a day, gave “fearless” as her strongest word.

Her life has not been safe. The paper reports, with police records behind it, that she has been held up at knife point and at gunpoint, was nearly killed in domestic violence, and has been threatened with death more than once. She seems not to have noticed danger the way others do.

It would be easy to conclude that the amygdala is the brain’s fear center. In 2013 the same team did a test that complicates the story. Three women with Urbach-Wiethe disease, S.M. and a pair of identical twins, each breathed a single breath of air with 35 percent carbon dioxide, which often produces a sense of suffocation. All three had panic attacks, against three of twelve healthy women. For S.M., it was the first fear she had felt in any setting since childhood; she scrambled for the air mask and cried for help. The researchers concluded that the amygdala is critical for fear triggered by threats in the outside world, and that fear set off by the body’s own alarms need not depend on it.

Two roads to the alarm

How does the amygdala learn of a threat? Suppose you are hiking and a long, curved shape lies in the grass at the edge of the path. Your eyes send the picture, like all the senses except smell, to a relay station deep in the brain called the thalamus. From there the news travels by two roads. A short one goes straight to the amygdala, which can trigger the body’s defenses at once: a jump, a spike of heart rate and sweat. A longer one goes up to the cortex, which examines the picture in detail and reports back: it is a stick. The first road is quick and crude. The second is slower and careful. The jump is already under way when the all-clear arrives, and then the alarm subsides.

The map of the two roads comes from the laboratory of Joseph LeDoux, now at New York University, who worked on the fear responses of rats. In 1984 LeDoux and colleagues found that rats whose auditory cortex had been removed could still learn to fear a tone that had been paired with a shock, while cutting the route through the thalamus stopped the learning. In 1992 Romanski and LeDoux showed that either road alone was enough for a simple tone; only when both were cut did the learning fail. Recordings from the rat’s amygdala found that the earliest part of its response to a tone arrives in under 15 thousandths of a second, which fits the short road. (Popular books quote precise figures for the two roads, and they do not agree with one another; we could not trace any of them to a paper. The safe statement is only that the direct road is the faster and cruder one.)

Why would the brain be built to jump first? Because nature offers few second chances. A miss and a false alarm do not cost the same. The psychiatrist Randolph Nesse called it the smoke detector principle: a detector that sounds at burnt toast is not broken, because a missed fire costs a life and a false alarm costs a moment. Try the numbers below.

Humans do seem to notice snakes quickly. In experiments by Vanessa LoBue and Judy DeLoache in 2008, preschoolers and adults found a picture of a snake in a grid of other pictures faster than they found a flower, a frog, or a caterpillar. In 2013 a team in Japan recorded from neurons in the pulvinar, part of the thalamus, in two macaques, and found that 37 of the 91 neurons tested responded best to snakes, with the earliest responses at about 55 thousandths of a second. But the explanation is disputed. Studies of babies suggest that attention, color, and learning explain much of the effect, and a result from two monkeys is not a rule for people.

The evidence for a fast road in humans is the most contested part of the story. A brain-scan study in 1999 by John Morris, Arne Öhman, and Raymond Dolan flashed angry faces for 30 thousandths of a second and masked them so that ten men did not see them; the right amygdala’s activity moved with that of the pulvinar, in the thalamus, and the superior colliculus, in the midbrain. A patient known as G.Y., who lost half his visual field to damage to his visual cortex at age eight, could guess the expressions of faces shown in the blind half at better than chance, and his amygdala responded differently to fearful faces there, though his brain may have rewired itself after the childhood injury. But in 2010 Luiz Pessoa and Ralph Adolphs reviewed the anatomy and the timing in monkeys and humans and argued that there is no clean, fast road, but many roads, among them very quick sweeps through the cortex itself. In 2016 recordings from electrodes placed inside the brains of patients found that the human amygdala responds to fearful faces from about 74 thousandths of a second, earlier than the fear-related activity in visual cortex, which restored some confidence in a fast route. The honest status: a fast threat route is well established in rats, and a fast threat response is established in humans, but whether it is a subcortical shortcut is still argued.

Smell, the oldest road

Smell is the one sense that skips the thalamus. The nerve cells of the nose connect to the olfactory bulb, and from there to the olfactory cortex and the amygdala, without the usual relay; the thalamus joins in later, along a second route to the front of the brain. That anatomy is often said to explain why a smell can bring back a scene from childhood in an instant. Marcel Proust’s madeleine is the classic example, and the psychologist Rachel Herz has tested it. In 2004 she gave people cues to memories such as a campfire or cut grass as a smell, a picture, or a sound. The memories brought by smell were more emotional, though no more vivid. A 2006 study of 93 older adults by Johan Willander and Maria Larsson found that smells cued memories from the first ten years of life, while words and pictures cued memories from young adulthood.

One corollary is a myth. People say that humans have a poor sense of smell, a leftover from the days before we walked upright, and the claim has a long history: the nineteenth-century anatomist Paul Broca linked the human olfactory bulb’s small size to our capacity for free will. In 2017 the neuroscientist John McGann reviewed the evidence and called this a myth of the nineteenth century. The human bulb is a tiny fraction of brain volume, but in absolute terms it is large, with about 60 million neurons, and humans have about 390 working odor-receptor genes. We outperform mice at some smells and lose to them at others, and people can follow a scent trail across a field.

The claim that scent can steer a crowd’s mood is harder to support. A story goes around that the New York subway once sprayed the smell of chocolate chip cookies and vanilla ice cream at rush hour to calm passengers. We could find no record that it happened. What studies do show is smaller: a pleasant ambient scent can lift mood a little, and in one study baking-cookie and roasting-coffee smells in a shopping mall made passers-by more likely to help a stranger, by way of improved mood. Effects like these are small, depend on context, and have not been tested in large preregistered replications.

Feeling stamps memory

The amygdala sits next to the hippocampus, the structure the brain uses to lay down memories of events, and the two talk. When something frightening or thrilling happens, the body releases stress hormones, adrenaline and cortisol, and they act through the amygdala to strengthen the memory being stored. James McGaugh at the University of California, Irvine, spent a career on this. In 1994 he and Larry Cahill gave volunteers a drug, propranolol, that blocks the action of adrenaline-like hormones at certain receptors, and then told them a story with slides. People on the drug remembered the emotional part of the story less well. They remembered a similar but unemotional story as well as anyone, and they were not sedated.

The effect of emotion is on vividness and confidence more than on accuracy, though. Roger Brown and James Kulik in 1977 coined the term “flashbulb memory” for the sense of a photograph of the moment you heard shocking news. Jennifer Talarico and David Rubin tested it on the attacks of September 11, 2001. They asked 54 Duke University students to write down, on 12 September 2001, how they had heard the news, and also to describe an ordinary recent event. They tested the students again after 1, 6, or 32 weeks. Details of both kinds of memory faded at the same rate. What differed was that the students’ belief that they remembered 9/11 accurately stayed high, while confidence in the ordinary memory fell. Emotion made the memory feel like a photograph. It did not make it one.

Built by care

If the alarm is learned as well as built in, then early experience should matter, and the best evidence comes from animals. At McGill University in Montreal, Michael Meaney’s group noticed that laboratory rat mothers differ in how much they lick and groom their pups. In 1997 they reported that pups raised by mothers who licked more became adults with smaller stress-hormone responses and more of the receptor that helps shut the stress response off. In 1999 the group swapped pups at birth between the two kinds of mothers, which showed that the mothering, not the genes, passed the trait on. These are ordinary lab mothers, not neglectful ones. In 2004 Ian Weaver and colleagues, working with small groups of about five animals, found a chemical change in the DNA near the gene for that receptor, in the hippocampus, that differed between the two groups, appeared in the first week of life, and lasted. The idea that experience can leave a chemical mark on a gene, without changing its sequence, is called epigenetics.

The human version rests on much thinner evidence. In 2009 Patrick McGowan and colleagues examined hippocampus tissue from 36 French-Canadian men held in the brain bank at the Douglas Institute in Montreal: twelve who had died by suicide after severe abuse in childhood, twelve who had died by suicide without abuse, and twelve who died of other causes. The abused group had more of the chemical mark near the receptor gene, and less of the receptor’s expression. That is a striking finding, from a small group, in a single region, with childhood abuse reconstructed from relatives’ accounts, and with no abused group who did not die by suicide. It does not show that abuse caused suicide through methylation. What is solid is the broader pattern: a pooled analysis in 2020 of 79 studies covering 337,185 young people found that those who had been sexually abused had about 3.4 times the odds of a suicide attempt, with physical and emotional abuse about 2.2 times, a link the authors stress is not proof of cause.

A reliable but modest finding is that groups of people who were maltreated or traumatized as children tend, on average, to have slightly smaller hippocampi and amygdalae. In 1997 Douglas Bremner and colleagues found the left hippocampus 12 percent smaller in 17 adults with abuse-related post-traumatic stress, compared with 17 controls. A study of identical twins complicated the picture: in 2002, Mark Gilbertson and colleagues, studying 40 pairs of identical twins, found that the twin brothers of veterans who never saw combat also had small hippocampi if the veteran had severe PTSD, suggesting that a smaller hippocampus may sometimes come before trauma, as a risk, rather than after. And a pooled analysis of 1,868 people in 2018 found a difference in hippocampal volume of only about 0.17 standard deviations, which means the groups overlap so much that a scan cannot tell you whose life was hard.

The rodent studies suggest how stress might change the cells. Rats restrained for six hours a day for 21 days, in a 1992 study by Yoshifumi Watanabe, Elizabeth Gould, and Bruce McEwen, lost branches from the dendrites, the receiving branches of nerve cells, in one part of the hippocampus (eight rats in each group). In 2002 Ajai Vyas, Sumantra Chattarji, and colleagues found that a related stress, two hours of immobilization a day for ten days, made dendrites in the amygdala grow, by a quarter in median length, as the hippocampal ones shrank. The same stress could thus strengthen a threat alarm while weakening the structure that helps put threat in context. Retracted branches are not dead cells, and the studies used male rats. Monkey studies show the same pattern of early experience. In a colony of rhesus macaques, a study by Dario Maestripieri in 2005 found that 9 of 16 females raised by abusive mothers became abusive themselves, against none of 15 raised by non-abusive mothers, and fostering confirmed it was upbringing, not genes.

The largest natural experiment involved children. In 1966 Romania’s government issued Decree 770, banning almost all abortion and restricting contraception. The birth rate nearly doubled within a year, and by the mid-1980s, according to standard histories, a woman was exempt only if she was over 45 or already had five children. Many families could not feed the children, and by 1990 more than 100,000 children lived in state institutions, where the routine was regimented and little attention went to any one child. After 1989 the institutions were opened to the world. Harry Chugani and colleagues at the Children’s Hospital of Michigan scanned ten adopted Romanian children, around age nine, and found low metabolic activity in the orbital part of the frontal lobe, the amygdala, the head of the hippocampus, and the lateral temporal cortex.

The most rigorous study is the Bucharest Early Intervention Project. Charles Nelson, Charles Zeanah, and Nathan Fox randomly assigned 136 institutionalized children, aged between about 6 and 31 months, to foster care or to staying in the institution, 68 in each group, and followed them for years, alongside 72 children who had never been in an institution. At age 8 the average IQ was 76 for the children who stayed, 82 for those in foster care, and 107 for the never-institutionalized group. The earlier a child was placed, the better the outcome on several measures, and the strongest evidence is for placement in the first two years. But foster care did little for attention problems, and a larger British study of Romanian adoptees found that the children who had endured more than six months of deprivation were more likely, as adults, to show autism-like traits, indiscriminate friendliness, and inattention, though their thinking deficits faded. The popular line that a “critical period” for emotional development ends at age five is not supported. Both studies placed children before the age of about three and a half, so neither could test it.

On 12 May 2008 a magnitude 7.9 earthquake struck Wenchuan in Sichuan province, killing more than 69,000 people. Within about three weeks, Su Lui and colleagues scanned 44 survivors, none injured, who had witnessed death or collapse, and 32 comparison people. The survivors’ brains showed more resting activity in parts of the frontal lobe, insula, and striatum and weaker connections in networks that include the amygdala and hippocampus. There was no scan from before the quake, and a later study that rescanned survivors found no change in the volume of any region, though changes in connection persisted for two years after the anxiety had cleared. The right statement is that a single severe event can change how brain networks work within weeks. It is not that emotion is the “strongest” way to change a brain.