68Behave
Robert M. Sapolsky, Behave: The Biology of Humans at Our Best and Worst, Penguin Press, 2017
Linked to 24 chapters: supports 12 · contradicts 6 · mechanism 11 · same claim 3 · pairing 7
Sapolsky takes a single act — someone pulls a trigger, someone touches an arm — and refuses to explain it with one discipline. He proposes a procedure instead. Ask what happened in the brain in the second before. Then ask what sight or smell in the preceding minutes acted on that brain; what hormones in the preceding days changed how it answered; what the preceding years did to its structure, and adolescence, and childhood, and the genes; then the culture that shaped the people who raised the person, and the millions of years behind the genes. The chapters run in that order. What makes the book more than a survey is the use he puts the sequence to: the levels are not alternatives. Name one and you have implicitly named the rest, because each is the end product of the ones before it. No fact of biology causes a behaviour on its own — biology modulates, sensitises, lowers thresholds — and the last chapters argue that a justice system built on blame cannot survive the result.
No buckets
The book opens with a fantasy Sapolsky has had since childhood, in which he captures Hitler and then does things to him that take half a page to describe. He does not believe in souls or in evil, he opposes the death penalty, and he doubts punishment belongs in criminal justice at all — and he also wants Hitler kept alive on a respirator and filled with cancer. The contradiction is the book's starting condition, because it points at what the biology has to explain: we do not object to violence as such, and pay to watch the right kind in a stadium. The same muscles pull a trigger in hideous aggression and in self-sacrificing love. What needs explaining is not the movement but the meaning, which is why he declines to sort behaviours into prosocial and antisocial, too cold for his taste, or good and evil, too frothy, and settles on our best and worst.
His complaint about the usual explanations is that they come in buckets. Ask why the chicken crossed the road and the endocrinologist answers with circulating oestrogen, the bioengineer with the fulcrum formed by the leg bone, the evolutionary biologist with millions of years of selection — each correct, each sealed off. Categories help you remember facts and wreck your ability to think about them, because the boundaries are arbitrary and we forget that they are: two colours that straddle the boundary between the colour words of a language look further apart than they are, and two inside one word look closer together.
He makes the cost concrete with three scientists who each lived in one bucket. John Watson, around 1925, offering to take a dozen healthy infants and train any one of them into a doctor or a beggar-thief regardless of talents or ancestry. Egas Moniz, awarded the Nobel Prize in 1949, writing that it is necessary to alter synaptic adjustments and force thought into new channels, which in practice meant an ice pick. Konrad Lorenz, Nobel laureate and cofounder of ethology, who joined the Nazi Party the moment Austrians were eligible and worked in its Office of Race Policy. Between them they shaped how children are educated, destroyed brains against their owners' will, and helped implement final solutions to problems that did not exist.
One second before
The longest chapter is about the second before, and it is built from three systems. Sapolsky borrows MacLean's three-layer picture of the brain — automatic regulation at the base, emotion above it, cognition on the surface — then spends a page on why it is false as anatomy and keeps it only as a way of arranging the material. There are, he insists, no centres in the brain for particular behaviours.
The first system is the amygdala, and its headline is not the expected one. Its association with aggression is supported by every method available: electrodes recording during attacks, lesions that lower aggression, stimulation that produces rage in patients on the operating table, and people whose amygdalae have been destroyed by disease, who cannot recognise an angry face while recognising every other expression. But ask the specialists what the structure is for and they say fear and anxiety. The region that generates aggression generates dread, and what moves it most is not pain but unpredictable pain: offered two decks with identical odds, one described exactly and one vaguely, people are more anxious about the vague one and their amygdalae say so.
It also learns. In Joseph LeDoux's fear conditioning a tone paired with a shock becomes frightening on its own; when the tone stops predicting shock the fear subsides, but not by erosion, since a second population of neurons learns that the tone is now safe on instructions from the frontal cortex. Pair tone and shock again and the fear returns faster than it was first learned. We do not forget that something was frightening; we learn that it is not.
The second system is the frontal cortex, and Sapolsky gives it a definition he uses on every subsequent page: it "makes you do the harder thing when it's the right thing to do". Right, he warns at once, is meant instrumentally and not morally — it takes a good frontal cortex to lie well, and the peasant lying to the thug about the refugees in her attic uses the machinery the financier uses on his investors. The region is the most recently evolved, its patterns of gene expression differ more between two people than the average human brain differs from a chimpanzee's, and it is the last to mature, unfinished until the mid-twenties.
Its two halves do different work. The dorsolateral part is the cold one, most active when someone forgoes an immediate reward for a larger later one; Ernst Fehr silenced it with magnetic stimulation and watched subjects accept insulting offers in an economic game that they still rated as unfair, having not stopped knowing but stopped being able to act on it. The ventromedial part is wired into the limbic system, and Antonio Damasio's patients with damage there are the correction to the idea that emotion is what spoils judgement: their intelligence and working memory are intact and they solve cold frontal puzzles normally, and what they cannot do is decide.
The cost is the hinge. The frontal cortex has a very high metabolic rate when working, and expensive cells, as Sapolsky puts it, are vulnerable cells. Load it with a hard working-memory task and performance on an unrelated frontal task falls immediately afterwards, while people become less charitable and readier to lie. The escape is automaticity: once the harder thing has stopped being hard it costs nothing. Joshua Greene's trolley studies show the halves in contest: between seventy and ninety per cent will pull a lever to kill one and save five, the same proportion refuse to push a man onto the track by hand, and only the second decision engages the amygdala.
The third system is dopamine, and the chapter argues that the word reward is misleading. A monkey pressing a lever for a raisin produces a burst of dopamine; give two raisins and it doubles; keep giving two and it settles back; return to one and it falls below baseline. Wolfram Schultz showed the response is scale-free — a monkey trained to expect two units and one trained to expect twenty produce identical bursts when they unexpectedly get four and forty. What is coded is the discrepancy from expectation, and the expectation comes from the frontal cortex, which is why a species that engineers ever stronger sensations habituates to them at the same pace.
Then the pivot. Once the routine is learned the raisin barely registers, and the large release comes when the light announcing the trial comes on, before any lever has been pressed. Dopamine after learning is less about reward than about anticipation, and it fuels the work: block it and the monkey does not press; stimulate the region instead of flashing the light and it does. Huda Akil found rats that would work simply for the chance to sit where the signal appears. Two things raise the anticipatory release: the size of the expected reward and, remarkably, its uncertainty. Make the reward arrive half the time and far more is released than when it always arrives, and the level climbs through the wait; shift the odds to twenty-five or seventy-five per cent, opposite in likelihood but equally more certain, and that second rise shrinks in both directions.
What follows is an account of willpower with no will in it: contemplating an immediate reward activates the limbic targets of the dopamine system and contemplating the delayed one the frontal targets, and the relative strength predicts the choice. No warthog restricts its calories to look good next summer; we run this circuitry across decades and past our own deaths, and how we manage it is unknown.
Seconds to days before
Widen the field and the question becomes what reached the brain from outside, and the answer is that most of it was never noticed. People who read a paragraph containing the word "ocean" name Tide as their favourite detergent more often, then explain what makes it clean so well.
The case he dwells on is race. Flash a face for less than a tenth of a second, too briefly for anyone to be sure they saw anything, and people guess its race better than chance; by a hundred milliseconds the amygdala has responded, and more so the higher the viewer's implicit bias. Elizabeth Phelps found that a face paired with a shock is learned as a threat faster when it is of another race, and that a neutral other-race face is rated as angrier; John Gabrieli found the fusiform region, which tells faces apart, least engaged by other-race faces in the most biased subjects, and not from unfamiliarity, since a purple-skinned face gets an ordinary response. Electrical recordings give the timing: within two hundred milliseconds the threat component is larger for black faces in white subjects, armed or not, while the later frontal component that inhibits it is smaller, and the larger the ratio of the first to the second, the likelier the subject is to shoot an unarmed black man in Joshua Correll's laboratory task. Sapolsky's case from outside the laboratory is the forty-one shots fired at Amadou Diallo in 1999 as he reached for his wallet.
He then refuses the conclusion the reader is reaching for. Are we built to fear another race's face? No. Not every amygdala responds, and the response is undone by manipulations that take seconds: mixed-race faces are remembered better once described as belonging to one's own group, and people of mixed ancestry arbitrarily assigned to one race show the weakened response to the race they have just been told is not theirs.
The point widens past race: gender is read from a face within 150 milliseconds and social status within 40, sitting near smelly rubbish makes people more socially conservative without shifting their views on economics, and the same Prisoner's Dilemma produces cooperation when called the Community Game and defection when called the Wall Street Game.
Some signals come from inside the body. Interoception, the continuous report the viscera send to the insula and the frontal cortex, shifts judgement the same way a face does. Pain does not cause aggression, Sapolsky notes; it amplifies what was there. Blood glucose falls during frontally demanding work, frontal performance improves after a sugary drink and not after a sweetener, and hungry people choose harsher punishments. Then the chapter reverses the arrow. Damasio's patient with a selectively destroyed amygdala was poor at recognising fear on a face and spent half as much time as controls looking at the eyes — and improved the moment she was told to look at them. The amygdala does not only detect fear; it decides where the eyes go to find it.
The hormone chapter is written against its own reputation. Sapolsky builds the conventional case for testosterone and aggression — the correlation, the castrations that lower it, the replacement that restores it — and then takes it apart. Castration never drives aggression to zero, and the more fighting a male did beforehand the more survives, which means the residue is learning. The prison studies cannot say which way the arrow points, because behaving aggressively raises testosterone. And individual differences within the normal range do not predict individual aggression in any species examined; John Archer's 2006 review found the association in adults weak and inconsistent.
What the hormone does instead is amplify. It raises confidence, lowers anxiety, makes strangers' faces less trustworthy, and weakens the link between prefrontal cortex and amygdala while strengthening the fast route in. At the level of a single cell the mechanism is almost embarrassingly plain: testosterone shortens the refractory period of amygdaloid and hypothalamic neurons. It does not make them fire. It makes them fire faster when something else does. Give it to the middle-ranking male in a group of talapoin monkeys and his aggression rises, aimed at the two below him and never at the two above.
The positive form is the challenge hypothesis, proposed by John Wingfield and colleagues in 1990: testosterone rises at a challenge to status, and what it prompts is not aggression but whatever preserves status. It rises in a spectator whose team wins, which rules out exertion and leaves identification. Where status depends on being fair, the hormone makes you fair: Christoph Eisenegger and Ernst Fehr used the Ultimatum Game, where rejection leaves both sides with nothing and is felt as an insult, and testosterone made offers more generous, while subjects who merely believed they had been given it made meaner ones. The problem is not the hormone but how often we award status for aggression.
Oxytocin gets the same treatment from the other direction. Infused into a virgin rat's brain it produces maternal behaviour; inhaled, it makes people trust more in economic games, but not against a computer. Then the correction: its gifts stop at the edge of the group. Carsten de Dreu had men contribute to a team pot, which oxytocin made them do more generously, then had them play against the other team, where, with money at stake, oxytocin made them likelier to strike first; in a second study, Dutch students under oxytocin were less willing to sacrifice one man to save five when he was called Dirk than when he was called Helmut or Ahmed. It is a parochial hormone, not a universal one.
Stress bears most directly on judgement, and it is Sapolsky's own subject. Over seconds to minutes the stress response is superb: energy into the blood, growth and repair postponed, cognition and the senses sharpened. The primate innovation is to run the machinery on the thought of a stressor rather than the presence of one, and to keep running it. The relation between stress and function is an inverted U. No stress at all is aversively boring; we pay for moderate transient stress and call it being challenged or at play, and rats will press levers for glucocorticoids in exactly that range. The core of psychological stress, the loss of control and predictability, is what we buy a ticket for on a roller coaster. Past some point the curve turns down, and where it turns differs enormously between people.
On the far side the damage lands on the frontal cortex. Prolonged high doses of glucocorticoids given to healthy volunteers impaired working memory into the range seen after frontal damage. Stress weakens the frontal cortex's links to the hippocampus, where the news that should prompt a change of plan arrives, and strengthens its links to habit circuits, so the stressed person does the same thing again, harder and faster. In the amygdala the arithmetic runs the other way: glucocorticoids raise excitability where fear is learned, and they do not invent that excitation, they amplify it. Fear is learned more easily and unlearned less easily, since extinction is relearning and the frontal cortex has to drive it. Sustained stress, stated broadly, impairs risk assessment.
It also narrows the circle of concern. Jeffrey Mogil showed in 2006 that a mouse's pain threshold drops near another mouse in pain, but only if that mouse is a cagemate, and Sapolsky's collaboration with Mogil's group found that briefly blocking glucocorticoid secretion makes a mouse show a stranger the empathy it would show a friend. Hormones, the chapter ends, do not determine, command, cause or invent behaviours. They make us more sensitive to social triggers and exaggerate what was already there.
Days to millennia before
The middle chapters keep widening. Neuroplasticity, he insists, is value free: the London taxi driver's hippocampus is the same phenomenon as the trauma that enlarges an amygdala and shrinks a hippocampus. It is also finite, and its limits are ordinary ones.
Adolescence supplies the fact the rest of the book leans on. The frontal cortex is unfinished until the mid-twenties, and it finishes by subtraction: grey matter is overproduced and then pruned, and in a longitudinal study the longer the thickening lasted before pruning began, the higher the adult intelligence score. Adolescents who manage adult-level control show more frontal activation doing it, because a well-pruned adult cortex does the same work for less. Why the delay? Not because the region is hard to build; it uses the same neurons as the rest of the cortex. Because being last to mature is the same thing as being least constrained by genes and most shaped by experience. The genetic programme of human brain development appears to have evolved, as far as it can, to free the frontal cortex from genes.
Childhood follows with a warning attached: epigenetic findings have been overinterpreted and most epigenetic changes are transient. Little in childhood determines an adult behaviour and virtually everything in childhood shifts the propensities. Ask not what a gene does, he writes, but what it does in a particular environment and in a particular network of other genes: genes are about potentials and vulnerabilities rather than inevitabilities. And culture runs both ways, the strongest link between brains and cultures being the delayed frontal cortex, which is how a culture gets written into a person.
Us versus them
The Us and Them chapter opens with the detail that during the filming of Planet of the Apes the actors playing chimps and the actors playing gorillas ate lunch in separate groups. Everything after it is about how little it takes.
A fifty-millisecond exposure to a face of another race activates the amygdala and fails to engage the face-recognition region as a same-race face would, and the brain sorts faces by gender and status at about the same speed. None of it is available to introspection; the Implicit Association Test detects it only as a few milliseconds of hesitation when the pairings run against the grain. Henri Tajfel's minimal group experiments in the 1970s showed that a division as thin as whether someone over- or underestimated the dots in a picture produces favouritism, including toward anonymous in-group members, which makes it identification rather than liking. Merely grouping people activates parochial biases, he writes, however tenuous the grouping — and it works by inflating our opinion of Us rather than lowering our opinion of Them.
Thems come in kinds: threatening, ridiculous, or held to be simple and interchangeable, which is essentialism — the belief that while we are individuals, they have a single unchanging unpleasant essence. And some are disgusting, which is where the insula comes in. It handles gustatory disgust in most animals, and in humans it has been extended to moral disgust, so the region that fires when you bite something rotten fires when you think about a group you find loathsome. Disgust, in Paul Rozin's formulation, serves as an out-group marker: people least willing to sit in a seat just vacated hold the most negative views of immigrants.
Susan Fiske's model sorts Thems on two independent axes, warmth and competence, and the imaging follows the sorting.
| Competent | Incompetent | |
|---|---|---|
| Warm | Pride: Us, and those we take for Us | Pity: the elderly, the disabled |
| Cold | Envy: the market-dominant minority, the ex-colonial ruler | Disgust: the homeless, addicts, the very poor |
The last corner is the one that matters. Shown pictures of people in it, subjects activate the amygdala and the insula and do not activate the face region or the ventromedial prefrontal cortex — the profile evoked by a disgusting object rather than a person — while the other three corners do engage the ventromedial cortex. Atrocity comes from the transitions: persecution of a resented, competent group characteristically begins by degrading it into the disgusting corner, which is what the dunce caps of the Cultural Revolution were for.
The second half is about how easily the sort is redone. Mary Wheeler and Susan Fiske asked three different questions about the same photographs. Asked to find a dot in the picture, subjects showed no amygdala response, because no face was processed. Asked whether the person looked older than a given age, the response grew, because thinking in one category strengthened thinking in another. Asked whether the person would like a particular vegetable, the response vanished, because the question cannot be answered without picturing one person shopping and sitting down to dinner. Automatic, Sapolsky concludes, does not mean inevitable.
The reasons we give ourselves arrive afterwards; cognition runs to catch up with the affective self. Two conservatively dressed young Mexicans chatting quietly in Spanish on a suburban commuter platform for two weeks made the regular commuters more hostile to Mexican immigration, and to no other kind. Sapolsky's closing advice is a short list, offered without any promise that the impulse can be removed: distrust essentialism, treat rationality as often being rationalisation, work on goals both sides care about, practise perspective taking, and individuate. The goal is not to be cured of Us and Them, which he does not want: some of the happiest moments of his life came from feeling like an Us.
Biology, the criminal justice system, and free will
The last argument begins with a proposal Sapolsky sent to a foundation asking for ideas too provocative to submit anywhere else. His was titled "Should the Criminal Justice System Be Abolished?", and he meant it.
He clears away the reform agenda first. In 1563 the Dutch physician Johann Weyer urged reform of witch trials, accepting that witches exist and should be punished but noting that elderly women can suffer atrophy of the tear ducts, so the test that identified a witch by her failure to weep at the crucifixion story would condemn the innocent. That is what reform by science looks like: sound thinking imposed on one corner of an irrational structure.
His target is the middle position, which almost everybody holds: that biology can constrain behaviour but something else remains in charge of the rest. Nobody now holds the extreme opposite. A person convulsing in a seizure who strikes someone is not charged with assault, and five hundred years ago they would have been; in 1457 a sow was convicted and executed for eating a child while her piglets were spared as too young to be responsible. The sentence that marks the distance travelled is: it's not him, it's his disease.
The M'Naghten rule is where the middle position entered law. In 1842 a Scotsman named Daniel M'Naghten, in the grip of a persecutory delusion, tried to kill the Prime Minister and shot his private secretary instead; he was found not criminally responsible, and the rule that emerged holds a person innocent when a defect of reason from disease of the mind prevents them telling right from wrong. Behind it, Sapolsky argues, sits a picture nobody would defend if stated plainly: separate from all the neurons and hormones and childhoods and genes, in a concrete bunker somewhere in the brain, a small person sits at a control panel, made of parchment and vacuum tubes and rivets of gumption rather than, as he puts it, squishy biological brain yuck.
The reply he anticipates is that nobody believes anything so crude, that free will lurks only around the edges, in the choice of socks. His answer is that frequency does not matter: if ninety-nine point nine nine per cent of what you do is biologically determined and once a decade you freely choose which direction to floss in, you have invoked something operating outside the rules of science.
He then walks the lines people draw and shows each bending. The Supreme Court barred the execution of juveniles in 2005 on frontal-cortex reasoning, which is the reasoning that spared the piglets, and the dissents objected that nothing magical happens on an eighteenth birthday. Everyone accepts that enough brain damage removes responsibility, nobody can say what enough is, and a large proportion of death row inmates have a history of early frontal injury. Stephen Morse, the most formidable defender of responsibility, holds that everything is caused but only some things compel, and Sapolsky cannot draw that line without something standing outside the causal universe.
The deepest line runs between aptitude and effort: we assign talent and impulse to biology, and grit and restraint to the will. Against it he lists what the book has shown moves restraint — blood glucose, sleep, a concussion, glucocorticoid levels, pain, childhood abuse, lead in the tap water, the culture you were raised in, how much cognitive load you have carried in the last few minutes, whether you have been smelling the sweat of someone frightened. Working hard, he concludes, is as much a property of the physical universe as being smart.
He is honest about the weakness. Ask a neuroscientist on the stand why this brain-damaged defendant became a murderer when most do not, and there is no answer. Explaining a fractured leg predicts the limp almost perfectly; explaining a social behaviour predicts very little. His answer is that this is not less biology but a different kind. Take a woman who is depressed today: her serotonin transporter variant alone might account for a tenth of it, a parent's death in childhood for a quarter, poverty and isolation for two fifths, today's glucocorticoid level and her culture for more than half — and most of the factors have not been discovered. Not different amounts of causation, but different kinds of it.
Marvin Minsky once defined free will as "internal forces I do not understand", and Sapolsky's amendment is one word: yet. If you believe science stops tonight, the small person in the bunker is safe. If you believe anything more will be learned, he is being squeezed, and you have also accepted that people will look back at us as we look at the authors of the Malleus Maleficarum, the 1487 manual that identified a witch by her convulsions. Sapolsky declines to call them monsters; he imagines an inquisitor coming home tired, describing the woman he condemned, remembering her bewildered children, and concluding that it is what it is.
What he proposes is smaller than abolition and harder. Protection stays: nobody who rejects free will thinks dangerous people should walk free, any more than a car with failing brakes should be driven, and deterrence is a problem he does not solve. What must go is the idea that punishment can be deserved and that punishing can be virtuous. He takes that impulse seriously as biology: judging culpability engages the cold dorsolateral cortex, while deciding how much to punish engages the ventromedial cortex, the amygdala and the insula, and the more they fire the heavier the sentence. Punishment that feels just feels good, which is how we bear its expense. His evidence that the mind-set can change is the case that has already run to completion: epileptics were once punished for their intimacy with the devil, and now, if the seizures are uncontrolled, they may not drive — and nobody thinks the driving ban is virtuous.
He ends by admitting he cannot live by any of it. He and his wife once praised a host for choosing a good pineapple, and he can deconstruct the praise — the olfactory receptors the man did not build, the culture that taught him ripeness, the income that put him in that shop — and they praised him anyway. The most that can be managed, he concludes, is to keep our myths benign and save the hard rational thinking for the moments when we judge other people harshly.
Since 2017
Some of the book's most quotable findings sit where the replication crisis pressed hardest. Ego depletion, which stands behind the account of the frontal cortex as a finite resource and behind the glucose results, failed a large multi-site replication in 2016 and has been contested since. Intranasal oxytocin has fared no better: the doses reach the brain unreliably, effect sizes are small, and several social findings have not replicated cleanly, though the animal work on receptor distribution has held. Priming results of the kind the subliminal-cue chapter is built from have thinned under the same scrutiny. Schultz's prediction-error work has not been touched by any of this.
The framing has aged well. That biology modulates rather than causes, that the same hormone produces opposite behaviours depending on context and prior disposition, that automatic and inevitable are different words — these have become the default position rather than the corrective one. The stress work is his own and has not been seriously challenged, and the developmental account of the frontal cortex has been sharpened rather than overturned by the large longitudinal imaging consortia.
The free-will chapter became a book. Determined, published in 2023 and read here as Sapolsky Determined, gives these last pages three hundred of their own, adding chaos theory, emergence and quantum indeterminacy as the escape routes it has to close, and replacing the small person in the bunker with a better image, which is also its first chapter's title: turtles all the way down. It drew a sharper response from philosophers, most of whom are compatibilists and few of whom accept that he has engaged their position rather than the folk version of it — an objection already available in 2017, since he concedes here that only a subset of the views he attacks are compatibilism in the philosophers' sense.