DEEP READING · READ 82

The Better Angels of Our Nature

The Better Angels of Our Nature · Steven Pinker · 2011

中文 →

In One Sentence

Measured as a rate rather than a raw body count, violence has fallen across the long run of human history — tribal raiding, medieval street killing, judicial torture, war between great powers, nearly every scale you can measure; and it fell not because people became kinder, but because states, trade, literacy and reason rewrote the arithmetic of whether violence pays. Pinker asks you to accept something counterintuitive: you live in the least violent era of human existence, however hard the news works to convince you otherwise.

Coordinates

Pinker is a cognitive scientist at MIT and Harvard, and the most readable — and most combative — public advocate evolutionary psychology has. He made his name on language (The Language Instinct) and on human nature (The Blank Slate). This 2011 doorstop of nearly eight hundred pages is his attempt to test that position at historical scale; the title comes from the last line of Lincoln's 1861 inaugural address. It detonated in the academy and in public argument at once: Bill Gates has called it one of the most important books he has read, and it drew the fiercest sustained fire — from statisticians, anthropologists and historians — of anything Pinker has written.

Central Claims

• The long decline of violence is an empirical fact you can check against data, not a progressive article of faith. Homicide rates, war death rates, the scope of torture and capital punishment, violence against women, children and minorities — each curve slopes down on its own timescale.
• Human nature did not change. We carry five "inner demons" and four "better angels" at once; what changed is the environment — Leviathan, commerce, literacy, the standing of women, reason — which decides which side gets switched on.
• Most violence is not the absence of morality but a product of it. Perpetrators overwhelmingly believe they are righting a wrong, defending honour, or building a better world. Humanity's problem has never been too little moralizing; it is too much of it, too cheaply available, too quick to reach for force.

The Core Concepts, One by One

The denominator: why it has to be a rate, not a body count

This is the book's foundation and its single most contested move, so let us settle it first. Pinker measures violence not in deaths but in violent deaths per 100,000 people per year — total divided by population, then divided by time. Why? Because population exploded. One killing a year in a band of fifty and a thousand killings a year in a nation of fifty million differ by a factor of 1,000 in absolute terms, but the danger of living there differs by two orders of magnitude the other way: 2% chance of being killed this year versus 0.002%. Asking "how violent is the world" is really asking "pick a person at random — what is the chance they are killed this year?" And that question requires a denominator.

Install that conversion and the historical league table inverts. The two world wars killed roughly seventy million people, which sounds unsurpassable. But using Matthew White's compilation of atrocity death tolls, Pinker rescales each catastrophe by the world population at the time, expressing it in mid-twentieth-century equivalents — and the worst thing people ever did to each other turns out not to be Hitler but the An Lushan revolt in eighth-century China, whose recorded loss of thirty-six million scales to something over four hundred million. Which is to say: the twentieth century looks like the bloodiest in history partly because it was the most populous, the best documented, and the first to be photographed.

This denominator habit is the first real tool the book hands you, and it reaches far past violence: whenever a frightening absolute number lands in front of you, ask what the denominator is and over what span. Fair warning, though — this is also precisely where the critics aim hardest (see below): "share of all deaths that were violent" in a small-scale society and "annual rate per 100,000" in a modern state are not the same quantity, and the assumptions used to bridge them do not survive close inspection.

Six declining curves — not one trend but six

Pinker does not tell one story about falling violence. He tells six, on timescales from five millennia to five decades, each with its own machinery. That matters: because the curves have different drivers and different starting centuries, their pointing the same way is hard to write off as coincidence.

The Hobbesian trap and the Leviathan: why people who don't want war fight anyway

Start with the trap. The Hobbesian trap — the security dilemma — says that where no third party enforces anything, two neighbours will slide into attacking each other even if neither wants to. I fear you will strike first, so I had better strike first; you know I reason that way, which gives you all the more reason to strike first. Fear alone manufactures war; malice is optional. That is why anarchic conditions carry such staggering rates of violent death: not because the people there are worse, but because nothing there can make "I won't attack you" a credible promise.

Hobbes supplied the solution: a Leviathan, a third party holding a monopoly on legitimate force. Once a state exists, it avenges and punishes on your behalf, so you need not; and it guarantees that whoever strikes first ends up worse off, which drains the incentive to strike at all. The Leviathan does not make people good. It makes restraint the option that each person can see, on their own arithmetic, is the better deal. Pinker treats this as the book's strongest explanatory variable: wherever state authority withdraws — medieval borderlands, the American frontier, failed states, gang-run neighbourhoods, prisons — violence floods back, and it comes back in a strikingly consistent form: private vengeance and duels over honour.

The payoff for you is not in international relations but in reading any situation where nobody defers to anybody: when conflict is unusually frequent somewhere, don't start by asking why the people are so difficult — ask whether there is a referee everyone accepts.

The Civilizing Process: manners are not about manners

The German sociologist Norbert Elias published The Civilizing Process in 1939 to near-total silence; Pinker retrieved it and made it a load-bearing beam. Elias had noticed something odd: from the late Middle Ages, Europe generated a flood of etiquette manuals. Erasmus, in his 1530 guide for boys, advises with a straight face: don't blow your nose into the tablecloth, don't spear the shared meat with the knife you have been licking, don't break wind at table, don't relieve yourself in the hallway. It reads as comedy today — but Elias's insight is that a society only writes such things down when it is moving from doing as one pleases toward monitoring oneself, continuously.

And that monitoring capacity is general-purpose. The same apparatus — think one step ahead, damp down the impulse, anticipate how others will react — keeps a knife out of the shared platter at dinner and out of a man's ribs in the street. Elias argued two forces installed it: royal power consolidating, so that violence was confiscated upward and the knight mutated from a man who settled things himself into a courtier who survived by reading faces; and commercial networks thickening, so that you dealt daily with strangers you needed to trade with again tomorrow, which demanded foresight and restraint.

The data cooperate handsomely. The long-run European homicide series assembled by the criminologist Manuel Eisner shows medieval rates of tens of killings per 100,000 per year in many places — over a hundred in some towns — falling to around 1 by the twentieth century: one to two orders of magnitude, with England, the Netherlands, Italy and Germany each tracing the same curve independently.

The most persuasive evidence is the counter-example. In Culture of Honor, the psychologists Richard Nisbett and Dov Cohen walked American students from Northern and Southern backgrounds down a narrow corridor, where a confederate bumped them and muttered an insult. Northerners mostly laughed it off. Southerners showed sharp spikes in cortisol (the stress hormone) and testosterone, and behaved more aggressively and less yieldingly in subsequent confrontations. The explanation is not Southern temperament but Southern economics: herding country in the uplands, livestock easy to steal, law far away — a man who failed to establish that crossing him carried a price did not keep his herd. Honour culture is what fills in for a Leviathan that never arrived. The institutions changed long ago; the psychology is still being handed down.

Gentle commerce: your neighbour is worth more alive

Doux commerce is an Enlightenment idea — Montesquieu: "the natural effect of commerce is to lead to peace" — and Pinker gives it a game-theoretic spine. Plunder is a zero-sum game: what I take is exactly what you lose. Voluntary trade is a positive-sum game: each of us hands over what we value less for what we value more, and both of us end up ahead. Once your neighbour is your customer, supplier or creditor, he is worth more to you alive than dead, and arithmetic starts doing the work that virtue could not.

This is a change in incentives, not a moral conversion. Historically, as roads, coinage, credit and market towns spread through Europe, the returns to robbery fell while the returns to exchange rose, and violence receded with them. Run it backwards and you get the resource curse: where wealth sits in immovable, seizable form — diamonds, oil, coca — violence proves stubborn, because taking the mine beats operating it. A fast diagnostic for any conflict: is wealth here created by exchange, or acquired by occupation?

The Humanitarian Revolution: how novels helped abolish torture

Something very strange happened in the seventeenth and eighteenth centuries. Judicial torture, burning at the stake, breaking on the wheel, witch trials, debtors' prisons, the slave trade — practices that had been unremarkable for millennia — were dismantled in under a century. Beccaria's On Crimes and Punishments (1764) demolished the legitimacy of judicial torture more or less single-handedly, in pamphlet form. Why then?

Pinker adopts the historian Lynn Hunt's explanation and extends it: the explosion of literacy and printing let people inhabit other minds at scale for the first time. The epistolary novels of the eighteenth century — Richardson's Pamela and Clarissa, Rousseau's Julie — had country gentlemen staying up all night inside a servant girl's interior monologue and weeping over her predicament. That was genuinely new in human experience: sympathy had always required a face; now it could travel on paper, across class, sex and border. Add the Republic of Letters — the transnational correspondence network in which Enlightenment writers argued with one another — and "our customs are obviously right" became a claim that had to be defended.

Connect this to the philosopher Peter Singer's expanding circle: people are built to extend moral concern only to their own — family, then clan, village, tribe, nation, race, and outward to all humans and eventually other species. The direction of history is that circle widening. The crucial point is that it widens not because kindness grows but because the capacity to imagine being you keeps getting amplified by technology — writing, print, travel, photography, film. Which is also what drives the rights revolutions: once a group's first-person account can be read and seen by the majority, violence against it becomes hard to say out loud.

The five inner demons: the deadliest is not sadism but idealism

Pinker refuses to treat violence as a pathology. He sorts its motives into five kinds, each of them ordinary psychology doing ordinary work:

• Predatory or instrumental violence: force as a means to an end, no emotion required — robbery, silencing a witness, seizing land. The coolest and the commonest.
• Dominance: the contest for standing, face and glory. At the individual level it is the brawl that starts with "what are you looking at" (young men's homicides typically ignite over something trivial); at the group level it is nationalism and racial supremacy — face, collectivized.
• Revenge: the moralized urge to punish. The brain's reward circuitry participates in retribution — vengeance genuinely feels good, which is exactly what makes it dangerous, since blood feuds can hold two families in a loop for generations.
• Sadism: pleasure taken in suffering. The rarest, and it requires long habituation to develop — Pinker notes that history repeatedly institutionalized that habituation, with public executions, torture and blood sports as family entertainment.
• Ideology: a shared belief system promising an unboundedly good end state, against which any price paid to reach it looks like a bargain. This is the deadliest category by body count, because it recruits idealists rather than brutes, and because an infinite good cancels any finite evil in the moral arithmetic.

Which leads to the book's most counterintuitive claim, taken from the psychologist Roy Baumeister: the myth of pure evil. We imagine perpetrators as people who know they are doing wrong and relish it. Real perpetrators almost always see themselves as the wronged party, settling a score, or acting under duress. Victims and perpetrators remember the same event with systematic asymmetry: the victim recalls enormous, unprovoked, still-unhealed harm; the perpetrator recalls something provoked, limited, and long overdue for letting go. This is why conflicts are so nearly unresolvable: both sides are morally sincere. Hence Pinker's hard line — most of the world's violence is not a shortfall of morality but its output.

The four better angels and the escalator of reason

On the other side stand: empathy (properly, sympathetic concern — and Pinker rates it lowest of the four, since empathy works like a spotlight, illuminating one appealing face while remaining unmoved by a statistic of millions, and it is easily steered by kinship and resemblance); self-control (delay of gratification — the Elias thread); the moral sense (double-edged: it restrains violence or launches it, depending on whether its content is fairness and reciprocity or purity, loyalty and obedience); and reason.

Reason is the one Pinker rates highest, under the name the escalator of reason. Once you start handling human affairs through abstract, communicable argument, you are on a moving staircase that only goes one way — because reasoning has a built-in symmetry. To persuade you, I must offer a reason that does not privilege me in particular; and once I concede that your suffering counts in principle as much as mine, I can no longer justify counting only my own. Violence stops being a contest to win and becomes a problem to solve.

His empirical support is the Flynn effect: across the twentieth century, raw IQ scores rose by roughly three points a decade in country after country, and the gains were concentrated in the most abstract subtests — analogies, matrix reasoning, classification — not in general knowledge or arithmetic. Modern schooling and symbol-saturated environments, in other words, train people to abstract a concrete situation into a general rule and follow out its consequences: the same muscle moral reasoning uses. Pinker invokes it to explain why an ordinary person today reaches almost automatically for "what if I were him," a move that two centuries ago was a philosopher's manoeuvre. Honesty demands the caveat: going from "abstract reasoning scores rose" to "moral judgement improved" is a correlation, and Pinker travels that distance faster than his evidence does.

The Long Peace and the New Peace: why great powers stopped fighting

Since 1945 no great power has fought another directly, and Western Europe — a region that spent roughly half of the preceding five centuries at war with itself — has become a place where war between members is not even imaginable. The historian John Lewis Gaddis named it the Long Peace. Pinker's preferred explanation is the "Kantian triangle" documented by the political scientists Bruce Russett and John Oneal: democracies rarely fight each other, trading partners rarely fight each other, and states embedded in the same international organizations rarely fight each other — with the three reinforcing one another.

Notably, he does not give the credit to nuclear weapons. Following the political scientist John Mueller, he argues that what really changed was the idea: before 1914, war was still widely described in Europe as noble, purifying, a tonic for national character. The two world wars bankrupted that rhetoric, and war was demoted in developed-world discourse from glory to failure. Once something moves from "an honourable last resort" to "humiliating under any circumstances," its frequency collapses — the same obsolescence that finished off duelling, foot-binding and public execution.

The post-Cold War second wave — the New Peace — refers to further declines in civil war, genocide and coups. Pinker is at his most tentative here, and rightly so; see below.

Power laws: why "no big war for decades" proves less than it seems

This section shows Pinker's honesty and also plants his largest vulnerability. Lewis Fry Richardson, meteorologist and peace researcher, spent a career counting "deadly quarrels" and found two things. First, the onset of wars approximates a memoryless random process (like radioactive decay: when the last one ended tells you nothing about when the next begins). Second, the magnitude of wars follows a power law — enormous numbers of small conflicts, vanishingly few large ones, and those few large enough to swallow most of the deaths.

What makes a power law dangerous is that it has no typical size and a very heavy tail: the mean can be dominated by a single extreme event and never settles down however much data you collect. By analogy: height is bell-shaped, so you will never meet a three-metre person; wealth is power-law, so one person can hold a million times the median. War deaths belong to the second family. The uncomfortable corollary follows: in a power-law world, "seventy years without a great war" is not by itself sufficient to show the underlying machinery changed, because long quiet stretches are exactly what such a process produces — until it doesn't. Pinker discusses this, and still argues the Long Peace is a real change of mechanism rather than a run of luck. That is the book's most fragile joint, and precisely where the Taleb line of criticism goes in.

The Distilled Skeleton

Eight hundred pages compressed into a single line of argument:

• Step one, establish the fact. Skeletal trauma from archaeology, ethnographic records, parish and court archives, war datasets — six independent declining curves, everything expressed per capita. Only one thing needs proving here: the decline is data, not mood.
• Step two, explain why nobody believes it. Availability bias (the easier something is to recall, the likelier we judge it), plus news that reports what happened and never what didn't, plus the moral reflex that treats "the world is improving" as a defence of the status quo. Our intuitions are calibrated to the visibility of violence, and visibility has been exploding.
• Step three, install the psychological engine. Five demons and four angels, all of them normal cognition, none of them a disease — so the change cannot have come from human nature.
• Step four, find the external variables. Five historical forces — Leviathan, gentle commerce, feminization (women's interests entering the calculus, fewer unattached young men), cosmopolitanism, the escalator of reason — each suppressing certain demons and amplifying certain angels. The structure of the argument is: constant nature × changing environment = changed behaviour.
• Step five, end on a conclusion that permits no complacency. The decline is not a law of history, not providence, and not guaranteed. It is the output of specific institutions, and it reverses when they fail.

Misreadings & Honest Criticism

• Taleb's statistical counterattack (the one to know). Nassim Taleb and Pasquale Cirillo argue that war deaths are fat-tailed — variance enormous or undefined — so the sample mean carries almost no information. Under such a distribution, seven relatively quiet post-war decades are nowhere near enough to reject the null hypothesis that nothing about the mechanism changed: long calm spells are what fat tails routinely produce. Pinker replies that the trend rests on many curves at once and on identifiable mechanisms. Fairly stated: his mechanism argument is strong, but the part that treats "no great war lately" as evidence of trend does not stand up statistically.
• Prehistoric rates may be systematically inflated. The anthropologist R. Brian Ferguson and others note that the archaeological samples Pinker cites skew toward sites that were studied and published because they showed signs of violence, while sites without such traces never enter the tally. And the contemporary hunter-gatherer societies used as proxies for prehistory had mostly already been disturbed by state expansion, colonization and the gun trade — much of the tribal warfare observed is a product of contact, not a baseline.
• The denominators are not commensurable. Comparing "share of all deaths that were violent" in small-scale societies with "homicides per 100,000 per year" in modern states requires a chain of assumptions about lifespan, age structure and record-keeping; the real error bars are far wider than the charts suggest. Historical death tolls are shaky in their own right — the thirty-six million of the An Lushan revolt most likely reflects the collapse of the census and the loss of people from the tax rolls rather than actual deaths.
• "Violence" is defined narrowly. John Gray and other critics argue the frame excludes structural harm: mass incarceration, killing at a distance by drone, torture outsourced to allied states, deaths caused by sanctions, the long consequences of ecological damage. The objection lands: if violence has partly been displaced, outsourced and reshaped, then part of the decline is an artefact of the accounting. (Pinker's reply: even so, one must first grant the fact that far fewer people are being killed.)
• The Enlightenment's ledger is kept too clean. The peak of the transatlantic slave trade, colonial conquest, and the destruction of Indigenous America ran concurrently with the Enlightenment. Crediting the Humanitarian Revolution entirely to reason and literacy discounts the other half of the same era.
• Reality has pushed back since 2011. After publication came Syria, Yemen, Tigray, the war in Ukraine from 2022 and the war in Gaza, driving global battle deaths back to their highest levels since the end of the Cold War. The New Peace looks considerably more fragile than it did in 2011. This does not necessarily overturn the long trend — the level remains far below the mid-twentieth century — but it does establish that the curve was never guaranteed to keep falling.
• Some of the psychology has aged badly. Several experimental findings the book leans on were damaged by the subsequent replication crisis, most conspicuously ego depletion (the idea that willpower drains like a battery), which is one of the supports under the self-control chapter. Discount the individual experiments; the long-run archival series deserve more confidence.
• Do not read it as permission to relax. Pinker denies this reading repeatedly. His conclusion is conditional: violence fell because a particular set of institutions is doing work, and those institutions were neither inevitable nor indestructible. A reader who takes the book to mean history improves on its own is exactly the reader it was written against.

The Pinker Outside This Book

Read only this one and you would take Pinker for a historian of violence. In fact the book is an application of a much larger claim, and missing the background means misreading it.

He began with language. The Language Instinct (1994), building on Chomsky, argues that language is not a cultural invention but an evolved instinct: children everywhere grow grammar on the same schedule without systematic instruction, and communities of deaf children spontaneously generate sign languages with full grammatical structure. The general programme follows: the mind is not a blank page but a set of special-purpose organs assembled by natural selection.

The Blank Slate (2002) is that position on the offensive, and it is the key to Better Angels. He attacks three modern doctrines: the blank slate (we are born empty and written on by environment), the noble savage (we are naturally benign and corrupted by civilization), and the ghost in the machine (some soul beyond biology does the deciding). His claim is that acknowledging a structured human nature entails neither fatalism nor an apology for inequality — on the contrary, only once you admit that dominance and vengeance really do live in the human mind can you set about designing institutions to hold them down. Better Angels is that sentence in historical form: violence fell enormously while human nature stayed put, which refutes "people are naturally good" and "people are beyond saving" in the same stroke.

Enlightenment Now (2018) extends the method from violence to everything — lifespan, health, wealth, literacy, safety, each with its rising curve, all credited to reason, science and humanism — and magnifies the controversy along with it; critics found its treatment of climate change, inequality and democratic decay too easy, and its Enlightenment too tidy a single story. Rationality (2021) returns to the psychology, arguing among other things that public irrationality is often not ignorance but team loyalty.

Taken together the system is one sentence: human nature is fixed and scientifically describable, and it is precisely because it is fixed that institutions and knowledge can reliably improve our condition. You may disagree — but to refute him you have to refute both halves.

Ten Sentences