69Determined

Robert M. Sapolsky, Determined: A Science of Life Without Free Will, Penguin Press, 2023

Linked to 23 chapters: supports 12 · contradicts 8 · mechanism 7 · same claim 2 · pairing 5

Sapolsky's claim is that there is no point in the causal chain at which an uncaused intent could enter, and that almost everything we do about blame and credit assumes there is one. A behaviour happened because of the state of a brain a second before, which was the product of that morning's hormones, of a decade of neural remodelling, of an adolescence during which the frontal cortex was still under construction, of a childhood that set the regulation of the genes, of a fetal environment, of a genome, and of the centuries of ecology that produced the culture which raised the person. Each link runs without a seam into the one before it. What he demands of anyone who disagrees is a single neuron whose firing was not the outcome of that history, and his claim is that nobody can produce one. No one discipline falsifies free will, he insists, and the disciplines are not separable anyway: to speak of a neurotransmitter is already to speak of the gene that builds it and of the evolution of that gene. What follows occupies the second half of the book. No blame, no credit, no desert and no hatred, with dangerous people restrained the way one keeps a car with failed brakes off the road.

Turtles all the way down

The title comes from a story he and his college friends told about William James, lecturing on the nature of the universe, and the old woman who informed him that the world rests on the back of a turtle. Asked what the turtle stands on, she answers: another turtle. Asked again, she answers that it is no use, it is turtles all the way down. They told it as mockery. He now tells it the other way round. An infinite regress of causes is not the absurd part; the absurd part is believing that somewhere down there one turtle floats in the air.

His definition of determinism is Laplace's: a demon knowing the position of every particle could predict every future moment, and re-creating any past arrangement would reproduce this present exactly. His own illustration is a graduation ceremony, with the families and the glowing immigrant parents, and at the perimeter the member of the grounds crew emptying the bins. Trade every factor neither of them controlled — the genes, the womb, the piano lessons or the threat of eviction — and the two switch places. We know this, he says, and congratulate the graduate anyway. Stated broadly, "we are nothing more or less than the cumulative biological and environmental luck" that has brought anyone to any moment.

His definition of free will is a demand rather than a description. A man pulls a trigger; his finger moved because a muscle contracted because a neuron fired because the neuron upstream fired. Find the neuron that started it, one that fired for no reason, and show that its functioning was untouched by whether the man was tired or hungry or in pain, by the hormones of the preceding days, by his childhood, by his genes, by the culture his ancestors invented. Alfred Mele, one of the compatibilists he argues with, calls this setting the bar absurdly high. Sapolsky answers that it is the only bar there can be, because all of that is what came to constitute the man.

He sorts the field into four positions and takes the sparsest: the world is deterministic and free will therefore cannot exist, a hard incompatibilism he shares with a handful of scholars he names — Gregg Caruso, Sam Harris, Derk Pereboom. Roughly ninety per cent of philosophers and legal scholars are compatibilists, holding that a world of atoms obeying laws is nonetheless a world containing free will and just deserts. Those are his opponents throughout, and he narrows them with a thought experiment. Raise a compatibilist in a sealed room where they learn nothing about brains, then tell them about Phineas Gage, whose frontal cortex was destroyed by a tamping iron in 1848 and who was afterwards no longer himself. If the immediate reply is that free will survives anyway, he is not interested; the compatibilist he wants is the one who first asks whether they have been wrong about everything. His stated ambition is modest for a book this absolute: not that every reader abandon free will, but that they conclude there is so much less of it than they assumed that they must change their minds about things that matter.

The final three minutes of a movie

The neuroscience that dominates the free-will debate began with Benjamin Libet at the University of California, San Francisco, in 1983. Subjects watched a fast clock, pushed a button whenever they felt like it, and reported the moment they became aware of deciding — about two hundred milliseconds before the finger moved. But the readiness potential, an electrical ramp over the supplementary motor area, began some three hundred milliseconds before that. John-Dylan Haynes at Humboldt University repeated the design with functional imaging twenty-five years later and found the choice decodable in the prefrontal cortex up to ten seconds before awareness; Itzhak Fried at UCLA, recording single neurons in patients awaiting epilepsy surgery, found individual cells committing seconds ahead. Three methods, three scales, one headline: the die is cast before you believe you are casting it.

Sapolsky then spends most of the chapter dismantling that headline, using the compatibilists' own objections. Manuel Vargas points out that the experiments insist on a purely immediate, impulsive act, which is precisely what free will is not for; Uri Maoz distinguishes picking from choosing, and Libet studies which box of cereal you take off the shelf. One experiment gave subjects two buttons, each sending a thousand dollars to a charity, against a version in which both charities got five hundred whichever button was pressed. The arbitrary version produced the usual readiness potential; the consequential one did not. Prediction accuracy, the number that would settle it, is poor — about sixty per cent in the imaging studies, near enough to chance, and around eighty per cent with single neurons. Libet's own rescue, a conscious veto in the final fraction of a second, Sapolsky refuses on grounds of symmetry: circuits routinely achieve an effect either by exciting a neuron or by inhibiting the neuron that inhibits it, so a free-won't neuron would have to be as uncaused as a free-will one. His verdict is that free will survives Libet, and that this does not matter in the least, because the whole literature declines to ask where the intent came from — which is like judging a film by its last three minutes.

What fills the rest of the film is the longest chapter in the first half, and it is organised as a walk backwards in time. In the seconds before, sensory input is already at work: a foul smell in the room lowers people's reported warmth towards gay men and makes them more punitive about violations of purity, because the insula, which evolved to make mammals spit out rancid food, was handed moral disgust as a second job too recently for anything new to have evolved. Beauty and goodness are assessed by the same patch of orbitofrontal cortex, which is part of why attractive defendants are convicted less often.

In the minutes to days before, hormones. Testosterone does not create aggression but lowers the threshold for aggression already present. Oxytocin does the opposite, and only for an in-group: in one Dutch study it made subjects readier to sacrifice a person with a German or Middle Eastern name while leaving Dutch names untouched. Glucocorticoids, secreted under stress, make the amygdala more excitable and the frontal cortex less so. Over weeks to years the brain itself is rebuilt — trauma enlarging the amygdala, years of depression shrinking the hippocampus. Adolescence is the last stretch of construction of the frontal cortex, which matures a decade after the rest of the brain, and he argues the delay was selected for: no gene can specify what counts as the right thing in a particular culture, so that region had to be left to experience. The genetic programme, as he puts it, evolved to free the frontal cortex from genes.

Further back, childhood. Adverse Childhood Experience scores put the variability on one axis, and each step up that scale raises the likelihood of adult antisocial behaviour, impaired impulse control, substance abuse and early death by roughly thirty-five per cent. Before that, the womb: maternal stress hormones, androgens, starvation. Before that, the genome — sixteen thousand genes active in the brain, with variation at something like four million spots — and the rule that a variant does nothing on its own, only something in an environment. One variant of a serotonin-metabolising gene raises the risk of antisocial behaviour, but only in those severely abused as children. Before that, centuries. Collectivist and individualist cultures show up in where people place themselves in a drawing of their social world and in which facial expressions activate their dopamine systems; the standard explanation is the collective labour of rice terraces against the solitary labour of wheat, and Chinese from the wheat-growing north behave like Westerners.

At each step he refuses the argument he could have made. Sensory effects are mild and have exceptions; hormones change likelihoods rather than causing behaviours; genes are about vulnerability rather than inevitability. None of this disproves free will on its own, and he says so each time. What disproves it is that there is no seam: discussing mothering style is already discussing the epigenetics of the child's frontal cortex, and discussing evolution is already discussing which gene variants exist to be regulated. Neil Levy, whose Hard Luck of 2011 supplies the book's philosophical spine, calls this constitutive luck and concludes that what rules out free will is not ontology but luck.

This is where Daniel Dennett enters, and where Sapolsky is at his least detached. Dennett's image for why a bad start does not excuse is a marathon rather than a hundred-yard dash: a small initial disadvantage counts for nothing, because luck averages out in the long run. No it does not, Sapolsky answers. Bad luck is amplified rather than cancelled, because a childhood of adversity both reduces the opportunities that arrive and reduces the capacity to use the few that do.

Willing willpower

The concession most people will make, and the one the willpower chapter exists to remove, is that the hand you were dealt is not your doing but what you do with it is. He first clears away two evasions. One puts free will in the past: you have had years to take responsibility for your dispositions, so bad constitutive luck stops excusing, and Dennett's version, in an exchange with Caruso, is that autonomy is something one grows into. Sapolsky's reply is that was was once now — if a neuron's function today is embedded in its history, you cannot show that the same neuron's function last week was not. The other evasion moves free will elsewhere: to a brain region not yet studied, or, in Michael Gazzaniga's version, out of the brain entirely and into the social level. Sociality, he answers, is as much a product of biology interacting with environment as the shape of a nose.

Then the substance. The chapter opens with a baboon he watched in Kenya who, closing on a gazelle his ally would have shared with him, could not tolerate eating second, jumped his partner, tripped, lost the gazelle and lost the alliance, and was demoted twice within months. The word for the human version is squander, and against it stands a catalogue of grit. The dichotomy, formalised, looks like this:

The clearest statement of the split he finds in a psychologist writing after the conviction of the Penn State coach Jerry Sandusky: one cannot choose not to be a paedophile, but one can choose not to be a child molester. The right-hand column is made of the same tissue as the left. Willpower is the prefrontal cortex, the last part of the brain to mature, the most expensive to run, and the part whose job is doing the harder thing when the harder thing is right.

The evidence that this column has a history is the same in kind as before. Josh Greene's experiment gave people a chance to cheat through a rigged computer glitch; in those who cheated, temptation produced heavy frontal activation, the picture of a person wrestling. In those who never cheated the frontal cortex did not stir at all. Greene's word for the second state is grace rather than will: for those people, doing the right thing was not the harder thing, and nothing was resisted. Depletion is measurable. Fifteen minutes of resisting a bowl of sweets makes people worse at a trivial switching task, and vice versa; emergency room doctors show more implicit racial bias at the end of a heavy shift than at the start.

Longer legacies follow the same path. Roughly half the people incarcerated for violent crime have a history of concussive head trauma, against about eight per cent of the general population. The socioeconomic status of a child's family predicts the volume and grey matter of the frontal cortex in kindergarteners, in toddlers, in six-month-olds and in four-week-olds; by the age of three a well-off child has heard some thirty million more words at home than a poor one. His favourite finding is the one that closes on itself: activity in the ventromedial frontal cortex tracks how strongly a person prefers a task that rewards self-discipline, which makes it the machinery that encodes a belief that there is no machinery. The chapter's conclusion is stated three times in three registers, and the shortest is that we cannot successfully will ourselves to have more willpower.

Is your free will chaotic?

Six chapters then answer the three twentieth-century sciences that are supposed to have opened cracks in determinism. Each is given a primer of its own before it is turned down, and the primers are the most generous writing in the book.

Chaos arrives through Edward Lorenz at MIT, restarting a twelve-variable weather simulation after lunch, typing 0.506 where the machine had been carrying 0.506127, and watching the two runs diverge until they had nothing in common. His 1963 paper declared perfect weather prediction impossible; by 1972 the butterfly had become the field's emblem. Sapolsky builds the intuition by hand with a cellular automaton, a row of boxes and a rule about which fill in the next row: complexity erupts from nothing, and there is no way to know what row one thousand looks like except by computing every row between. The argument some have drawn from this — that we live in a world of clouds rather than watches, and so a world with room for freedom — he answers in one sentence: determinism and predictability are very different things. Every one of the hundred thousand boxes in his automaton was filled by the rule, and at none of them could anything else have happened. Otherwise free will becomes the name for the biology we cannot yet predict, which shrinks whenever anyone publishes.

Emergence gets the same treatment. Ants find near-optimal solutions to a routing problem with more possibilities than they have members; a slime mould placed on a map of Tokyo with oat flakes at the suburban stations grows a tubule network resembling the rail system. Nobody drew a plan. But the free-will versions fail structurally. Christian List generates indeterminism at the higher level by coarse-graining the state space, which is to say by rounding, by ruling that five similar-but-different starting states are the same state; Sapolsky's biological version is that oestrogen and testosterone look nearly identical on a chemistry exam. Others treat emergent levels as orphaned from cause and effect. And downward causation is real — a conformity experiment in 2005 showed the visual cortex itself firing differently when people privately came round to the group's view — but it changes the context in which the bricks do their simple things, not what the bricks are. There is still gravity.

The quantum chapters he writes with visible reluctance, since the field attracts cranks and he says he does not fully understand it. He grants the physics, and grants that quantum effects on ion channels are plausible, even likely. What he denies is that they scale. The brain is warm, wet and enormous — tens of trillions of glutamate receptors in the hippocampus alone — and quantum states decohere long before anything behavioural happens. Roger Penrose and Stuart Hameroff's microtubules he compares to deciding that a library's knowledge lives in the carts rather than the books, noting Max Tegmark's calculation that the relevant states are orders of magnitude too short-lived. Even Bernard Katz's spontaneous release of neurotransmitter, which really does occur without an antecedent cause, turns out to be regulated: a separate molecular pathway evolved for it, and hormones and experience tune how often it happens. Something mechanistic determines when it is time to be indeterministic. The deeper objection needs none of this arithmetic. Indeterminate behaviour is random behaviour, and John Searle's question stands unanswered: how do we get from randomness to rationality? The proposals for harnessing randomness — Dennett's generator of considerations filtered by your values, Kevin Mitchell's elbow room, Robert Kane's stirring up of chaos at life's forks — end in the same place, because whatever does the filtering has values, and those came from one second before and one century before.

Will we run amok?

The second half opens with the objection everyone raises. If people stop believing in free will, will they behave worse? There is a literature. Prime subjects with a passage from Francis Crick to the effect that they are nothing but a pack of neurons, and their readiness potentials shrink, their error-related brain responses flatten, they cheat a little more and dispense nearly twice as much punitive hot sauce. Sapolsky, who could have used this, takes it apart instead. The effects are small — the outcomes, he writes, were more amok-ish than amok — and most of the studies fail to replicate, sometimes in larger samples. A 2022 meta-analysis of a hundred and forty-five experiments, ninety-five of them unpublished, found that the manipulations do mildly weaken belief in free will without any consistent effect on ethical behaviour.

He then reaches for a better-powered natural experiment: atheism, where a comparable belief really has collapsed within living memory. The naive literature says the irreligious are less honest and less charitable. Nearly all of it is self-report, and asking people only tells you how charitable they wish to appear; watch behaviour instead and the differences in blood donation, tipping, honesty and forgiveness vanish. Control for sex, age, income, marriage and community involvement and most of what remains disappears too, since charitable giving tracks attendance at a house of worship rather than frequency of prayer. Scandinavia, the most secular region on earth after a century of collapsing belief, does better than the religious United States on life expectancy, infant mortality, inequality, homicide and corruption. What survives of the religious advantage is directed at the in-group; towards out-groups it is the atheists who are more prosocial.

The reframing that follows is the sharpest move in the second half. The priming studies necessarily sample people whose belief can be shifted, and it is exactly those people who then cheat; nobody compared settled sceptics with settled believers. Damien Crone and Neil Levy did, and found the two groups identical in ethical behaviour, with those who most defined themselves by moral identity the most honest and generous whatever they believed. The religion literature has the same shape: in obedience studies the moderates complied most, while committed believers and committed non-believers resisted equally, and those who hid Jews in occupied Europe came disproportionately from both ends. The divide that matters is conviction against indifference.

The remaining worry is that determinism makes change impossible, and the answer is that we do not change our minds, we are changed. He takes it down to Aplysia, the sea hare whose gill withdrawal Eric Kandel spent decades on. One shock to the tail, and a later touch produces twice the retraction; four shocks, three times; a cluster over days, ten times and lasting weeks. Each step has its mechanism — more stored neurotransmitter, then a gene whose protein blocks that neurotransmitter's breakdown, then a chain of genes that builds an additional synapse. The same molecules do the same work in mammals learning to blink before a puff of air, and in the human amygdala learning which group of people to fear. Change is real, thorough and available; nobody authors it.

We really have done this before

The strongest chapters are the historical ones, and their argument is that the subtraction he is asking for has been performed before without the roof caving in. His case is epilepsy. Assyrian physicians described seizures clinically four thousand years ago and called the condition the hand of sin. Physicians blamed the phases of the moon for sixteen centuries before anyone tested the claim, and Galen had phlegm plugging a ventricle, with the fit as the brain's attempt to dislodge it. Christianity read the possession of the boy in Mark's gospel into the disorder, and in 1487 the Malleus maleficarum named seizures among the reliable marks of a witch; it caught the new printing press and went through thirty editions in the following century. After the Enlightenment the blame migrated rather than lifting: to masturbation, to unseemly sexual appetite, and, in a proposal of Thomas Beddoes in 1800, to reading too many novels instead of gardening. Because it was thought heritable, men were castrated in parts of Europe into the nineteenth century; Buck v. Bell upheld Virginia's sterilisation of the feeble-minded and epileptic in 1927, and the statute was not repealed until 1974.

Then it went. Autopsies disposed of the phlegm, statistics disposed of the moon, Hughlings Jackson localised the focus, and in 1808 a person who killed during a seizure was acquitted. By 1860 the psychiatrist Jules Falret could write that such a killer bore not the slightest responsibility, striking mechanically and without motive — and then, unable to follow his own sentence, grant some of them the privilege of moral responsibility after all, which is where Sapolsky dates compatibilism's characteristic move. What happens now when a driver has a first seizure at the wheel and kills a child is grief at monstrous luck, not a mob; and the dangerousness problem, supposedly the fatal objection, was solved quietly and minimally, by suspending a licence until the person has been free of seizures for about six months.

He runs the same arc through schizophrenia within living memory. The schizophrenogenic mother was coined by Frieda Fromm-Reichmann in 1935 and became orthodoxy when analysts held the psychiatry chair of every major American medical school; there was never any evidence for it. What ended it was not an argument but advocacy: families organising in the seventies, the founding of what became the National Alliance on Mental Illness, and E. Fuller Torrey on daytime television in December 1988 holding up scans of identical twins discordant for the illness, where the enlarged ventricles were visible to anyone. No leader of the old school apologised, though Leo Kanner, who had blamed refrigerator mothers for autism, stood up at a parents' association in 1969 and said that he acquitted them. The list continues: shell shock as cowardice until it was post-traumatic stress, Gulf War syndrome dismissed as psychosomatic until 2008, dyslexia as laziness until the cortical malformations were found.

Punishment gets the hardest chapter, and he does not pretend the science is on his side. Punishing a cheat activates dopamine reward circuitry — the region roused by cocaine — and more punishment brings more of it. In Tania Singer's experiment six-year-olds and chimpanzees both paid, in tokens or in shoving a heavy door, to watch someone be beaten, and paid far more often when the person had earlier been unkind to them. Third-party punishment, the willingness to pay to punish a wrong done to somebody else, may be uniquely human and is what makes large cooperative societies possible. So the objection is not that retribution is useless. It is that its justification is gone. Derk Pereboom's quarantine model supplies the replacement: a malady can make someone dangerous, it is not their fault, restraining them is collective self-defence, and the restraint must be the minimum required and not an inch more. Retribution runs on backward-looking proportionality, more harm earning more punishment; quarantine runs forward, more danger earning more constraint. Caruso extends it on a public-health model, arguing that the conditions producing criminals are more dangerous than any criminal. Saul Smilansky objects that constraint for something not your fault demands compensation, which would make crime rational; Caruso's answer is Norway, with a murder rate an eighth of the American, an incarceration rate an eleventh, a recidivism rate a quarter, and prisons with kitchens and workshops. The test case is Anders Breivik, who killed seventy-seven people in 2011, received Norway's maximum of twenty-one years, and was admitted by the University of Oslo to study remotely. Sapolsky notes, fairly, that the court reached that sentence by deciding Breivik was sane and therefore could have done otherwise — the reasoning he has spent four chapters attacking.

The last chapter takes the cost seriously. If there is no free will there is also no meaning, no answer to why beyond what came before, and he allows that we may be better off believing the falsehood. The evidence is uncomfortable: in the study where an illusory sense of will can be induced, the people who resist the illusion are the clinically depressed, who turn out not to be distorting but to be accurate while everyone else overestimates. Against that he sets what the belief costs the unlucky. An advocate who discovered that she carried a mutation in the gene for leptin, the hormone that signals fullness, described it as the day she stopped being her own worst bully. Sherman James named John Henryism for the folk hero who beat the machine and died of it: a conviction that enough effort overcomes anything, which raises the risk of cardiovascular disease in Black American labourers, because the system they strain against is built to hold. His conclusion is stated plainly: "There is no justifiable 'deserve.'" Hating a person for what they have done, on this account, makes as little sense as hating the sky for storming or the earth for quaking. He admits he cannot live it, failing ninety-nine per cent of the time, and admits one indefensible foothold, the certainty that it is good when people feel less pain and more happiness, which is meaningless to say about a machine. He is under no illusion about how the future will see us. They will find us as ignorant as the peasants who thought Satan caused fits. He would prefer that they not also find us heartless.

Since 2023

The philosophical replies came quickly and largely agreed on the charge. John Martin Fischer, reviewing the book for Notre Dame Philosophical Reviews in November 2023, argued that Sapolsky defines free will as requiring an uncaused cause and thereby settles the question against compatibilism before any evidence arrives; that he never distinguishes freedom as the ability to do otherwise from freedom as ownership of the actual sequence, which is where most contemporary compatibilism lives; and that generalising from epilepsy and schizophrenia to all human action is a leap, since what makes those cases exculpating is that their causes are atypical. Kieran Setiya made a similar case in The Atlantic that month, praising the science and faulting the engagement with philosophy, and Jessica Riskin, in the New York Review of Books in February 2025, held that no experiment could settle a question of this kind. Uri Maoz, whose distinction between picking and choosing the book uses, reviewed it in Science, called the treatment of compatibilism dismissive, and recommended it anyway. Sapolsky anticipated the charge: his position is that compatibilists resist for reasons that are emotional and dress them as intellectual, which is itself part of what his critics object to.

The opposing book appeared within weeks. Kevin Mitchell's Free Agents argued from evolution and neuroscience that organisms are genuine causes of their own behaviour, and the two men debated publicly through 2024. Read together they disagree less about biology than about the word: Mitchell rejects a definition of freedom that requires escape from all prior causes, calling that randomness rather than agency, and means by free will the capacity for conscious control that evolution built by degrees. Sapolsky calls the same gradations differences in self-regulation and declines to call them freedom. Dennett, whose marathon the book attacks by name, died in April 2024 without replying at length.

Some of the empirical floor has moved, mostly beneath the chapter on willpower. The ego-depletion effect, which underwrites the bowl of sweets, failed a preregistered replication across twenty-three laboratories and more than two thousand subjects in 2016 and has not recovered. The parole study drawn in figure 69.5 has been challenged on the ground that cases are not randomly ordered, since represented prisoners are heard first, and simulations published in 2016 suggest the effect, if real, is much smaller than reported. The Adverse Childhood Experiences literature holds up as population statistics, but a 2021 study in JAMA Pediatrics found such scores too imprecise to predict outcomes for an individual, which is awkward for the claim that we already know enough. The readiness potential has meanwhile been reinterpreted as an artefact of averaging backwards from the actions that happened, a case set out in Seth Being You; that one cuts Sapolsky's way, since he had already refused to rest anything on Libet. What has not been contested is the part of the book that is a catalogue: the chain from a smell to a genome to a culture summarises settled findings, and the historical chapters describe a subtraction of blame that demonstrably occurred.