The Better Angels of Our Nature: Why Violence Has Declined

by Steven Pinker

  One can imagine, for example, that in a society that was undergoing a Pacification or Civilizing Process, a tendency toward impulsive violence would begin to pay off less than it did in the days of Hobbesian anarchy, because a hair trigger for retaliation would now be harmful rather than helpful. The psychopaths and hotheads would be weeded out by the Leviathan and sent to the dungeons or gallows, while the empathizers and cooler heads would bring up their children in peace. Genes that fortified empathy and self-control would proliferate, while genes that gave free rein to predation, dominance, and revenge would ebb.

  Even a cultural change as simple as a shift from polygyny to monogamy could, in theory, alter the selective landscape. Napoleon Chagnon documented that among the Yanomamö, the men who had killed another man had more wives and more children than the men who had never killed; similar patterns have been found in other tribes, such as the Jivaro (Shuar) of Ecuador.129 This arithmetic, if it persisted over many generations, would favor a genetic tendency to be willing and able to kill. A society that has shifted to monogamy, in contrast, removes this reproductive jackpot, and conceivably could relax the selection for bellicosity.

  Throughout this book I have assumed that human nature, in the sense of the cognitive and emotional inventory of our species, has been constant over the ten-thousand-year window in which declines of violence are visible, and that all differences in behavior among societies have strictly environmental causes. That is a standard assumption in evolutionary psychology, based on the fact that the few centuries and millennia in which societies have separated and changed are a small fraction of the period in which our species has been in existence.130 Since most adaptive evolutionary change is gradual, the bulk of our biological adaptation must be to the foraging lifestyle that prevailed during those tens of millennia, rather than to the specifics of societies that have departed from that lifestyle and diverged from each other only recently. The assumption is supported by evidence for the psychic unity of humankind—that people in every society have all the basic human faculties such as language, causal reasoning, intuitive psychology, sexual jealousy, fear, anger, love, and disgust, and that the recent mixing of human populations has revealed no qualitative innate differences among them.131

  But the assumptions of ancient adaptation and psychic unity are only assumptions. The speed of biological evolution depends on many factors, including the strength of the selection pressure (that is, the average difference in the number of surviving offspring between the carriers of two variants of a gene), the demographics of the population, the number of genes required to implement a change, and the patterns of interaction among the genes.132 Though a complex organ built by a suite of many interacting genes can take aeons to evolve, a quantitative adjustment that can be implemented by one gene, or a small number of independently acting genes, can take place over just a few generations, as long as they have large enough effects on fitness.133 Nothing rules out the possibility that human populations have undergone some degree of biological evolution in recent millennia or even centuries, long after races, ethnic groups, and nations diverged.

  Though people sometimes write that hypotheses about natural selection are just-so stories that can never be verified until someone invents a time machine, in fact natural selection is a distinctive mechanistic process that leaves signs of its handiwork both in the design of organisms’ bodies and in the patterning of their genomes. Since the completion of the first phase of the Human Genome Project in 2000, the search for fingerprints of selection has been one of the most exciting research activities in human genetics.134 One technique juxtaposes the human version of a gene to its counterpart in other species and compares the number of silent changes (which have no effect on the organism and so must have accumulated by random drift) with the number of changes that do have an effect (and thus may have been a target of selection). Another technique looks at the variability of a gene across individuals. A gene that has been subjected to selection should vary less among individual people within the human population than it does between humans as a whole and other mammals. Still other techniques check to see whether a gene lies in the middle of a big chunk of chromosome that is identical across people—the sign of a recent “selective sweep,” which carries a stretch of a chromosome along for the ride with a handy gene, before mutations have had a chance to adulterate it or sexual recombination has had a chance to shuffle it into pieces. At least a dozen of these techniques have been devised, and they are continually being refined. Not only can they be aimed at particular genes, but they can be applied to the entire genome to estimate the fraction of our genes that have been targets of recent natural selection.

  These analyses have delivered a surprise. As the geneticist Joshua Akey concluded in a stringent 2009 review, “The number of strong selective events thought to exist in the human genome today is considerably more than that imagined less than a decade ago.... [Approximately] 8% of the genome has been influenced by positive selection, and an even larger fraction may have been subject to more modest selective pressure.”135 Many of the selected genes involve the functioning of the nervous system, so they could, in theory, affect cognition or emotion. The pattern of selection, moreover, differs among populations.

  Some journalists have uncomprehendingly lauded these results as a refutation of evolutionary psychology and what they see as its politically dangerous implication of a human nature shaped by adaptation to a hunter-gatherer lifestyle. In fact the evidence for recent selection, if it applies to genes with effects on cognition and emotion, would license a far more radical form of evolutionary psychology, one in which minds have been biologically shaped by recent environments in addition to ancient ones. And it could have the incendiary implication that aboriginal and immigrant populations are less biologically adapted to the demands of modern life than populations that have lived in literate state societies for millennia. The fact that a hypothesis is politically uncomfortable does not mean that it is false, but it does mean that we should consider the evidence very carefully before concluding that it is true. Is there any reason to believe that declines in violence in particular societies can be attributed to genetic changes in their members?

  The neurobiology of violence is a target-rich area for natural selection. Selective breeding of mice for four or five generations can produce a strain that is markedly more or less aggressive than an off-the-shelf lab mouse.136 Violence in humans, of course, is fantastically more complicated than violence in mice, but if variations among people in their inclinations toward or away from violence are heritable, selection could certainly favor whichever variants result in more surviving offspring, which would change the concentration of the inciting and pacifying genes over time. So first we must establish whether any portion of the variation in aggression among people is caused by variation in their genes, that is, whether aggression is heritable.

  Heritability may be measured in at least three ways.137 One is to look at correlations in traits between identical twins who have been separated at birth and reared apart; they share their genes but not their family environments (within the range of environments in the sample). A second is to see whether there is a higher correlation between identical twins (who share all their genes and most of their family environments) than between fraternal twins (who share only half their variable genes and most of their family environments). The third is to see whether there is a greater correlation between biological siblings (who share half their genes and most of their family environments) than between adopted siblings (who share none of their variable genes and most of their family environments). Each of the methods has strengths and weaknesses (for example, identical twins may become partners in crime more often than fraternal twins), but the strengths and weaknesses are different among the three, so if the methods converge, there are good grounds for believing that the trait in question is heritable.

  These methods have shown that having an antisocial personality and getting into trouble with the law hav
e a substantial heritable component, though their effects sometimes depend on features of the environment. In a large 1984 study of Danish adoptees, among the adolescents and young adults who had been brought up in families in which an adoptive parent had been convicted of a crime, about 25 percent of the biological offspring of criminals had been convicted of a crime, whereas only 15 percent of the biological offspring of noncriminals had been convicted.138 In that study, the effect of biological relatedness was seen only in nonviolent crimes such as auto theft, so in the 1980s many textbooks said that only nonviolent criminality was heritable, not a tendency to violence itself. But the conclusion was premature. There are far fewer criminal convictions of violent than nonviolent crimes, so the sample size for the violent ones is smaller and the ability to detect heritability weaker. Also, conviction rates are buffeted by vicissitudes of the criminal justice system that may overwhelm the effects of an offender’s violent tendencies.

  Today’s studies use more sensitive measures of violence, including confidential self-reports, validated scales of aggression and antisocial behavior, and ratings by teachers, friends, and parents (for example, whether the person being rated “hurts others for his or her own advantage” or “deliberately frightens and causes discomfort in others”). All of the measures correlate with the probability of being convicted of a violent crime, while providing far more plentiful data.139 When they are analyzed with the tools of behavioral genetics, all three methods reveal substantial heritability of aggressive tendencies.140

  The analysis of twins separated at birth is the rarest method in behavioral genetics, because nowadays few twins are separated at birth. But the largest of the studies, based at the University of Minnesota, looked at aggression in identical twins reared apart and found a heritability coefficient of 0.38 (which means that about 38 percent of the variation in aggression in the sample can be explained by variation in their genes).141 Adoption studies are somewhat more common, and one of the better ones estimated the heritability of aggressive behavior in its sample as 0.70.142 Studies that compare identical and fraternal twins on aggressive tendencies such as arguing, fighting, threatening, destroying property, and disobeying parents and teachers tend to yield estimates of heritability between 0.4 and 0.6, particularly in childhood and adulthood. (In adolescence, the influence of peers often overshadows the influence of genes.)143

  The behavioral geneticists Soo Hyun Rhee and Irwin Waldman recently reviewed the entire research literature on the genetics of aggression, which includes more than a hundred twin and adoption studies.144 They selected nineteen that met stringent criteria of research quality and that specifically zeroed in on aggressive actions (such as physical fighting, cruelty to animals, and bullying) rather than the broader category of antisocial tendencies. They also examined all the published twin and adoption studies of criminal arrests and convictions. They estimated the heritability of aggressive behavior at around 0.44, and the heritability of criminality at around 0.75 (of which 0.33 consists of additive heritability, that is, variation that breeds true, and 0.42 consists of nonadditive heritability, variation caused by interactions among genes). Though their dataset on criminality did not distinguish violent from nonviolent crimes, they cited a Danish twin study that separated the two and that yielded a heritability estimate of 0.50 for the violent ones.145 As in most studies in behavioral genetics, the effects of being brought up in a given family were tiny to nonexistent, though other aspects of the environment that are not easily measurable by these techniques, such as effects of the neighborhood, subculture, or idiosyncratic personal experiences, undoubtedly do have effects. The exact numbers should not be taken too seriously, but the fact that they are all substantially above zero should be. Behavioral genetics confirms that aggressive tendencies can be inherited, and that gives natural selection material to work with in shifting the average violent tendencies of a population.

  Heritability is a necessary condition for evolutionary change, but it measures a hodgepodge of diverse contributors to behavior. When we tease them apart, we find many specific pathways by which natural selection could adjust our inclinations toward or away from violence. Let’s consider a few.

  Self-domestication and pedomorphy. Richard Wrangham has noted that the domestication of animals usually tames them by slowing down components of the developmental timetable to retain juvenile traits into adulthood, a process called pedomorphy or neoteny. 146 Domesticated strains and species tend to have more childlike skulls and faces, to show fewer sex differences, to be more playful, and to be less aggressive. These changes can be seen in farm animals that have been deliberately domesticated, such as horses, cattle, goats, and foxes, and in one species of wolf that was self-domesticated after it started to hang around human campsites thousands of years ago scrounging leftover food and eventually evolved into dogs. And in chapter 2 we saw that bonobos evolved from a chimpanzee-like ancestor via a process of pedomorphy after their foraging ecology had reduced the payoff for aggression in males. Based on pedomorphic changes in the fossils of Paleolithic humans, Wrangham has suggested that a similar process has been taking place in human evolution during the past thirty to fifty thousand years, and may still be taking place.

  Brain structure. The neuroscientist Paul Thompson has shown that the distribution of gray matter in the cerebral cortex, including the dorsolateral prefrontal region, is highly heritable: it is almost identical in identical twins and considerably less similar in fraternal twins.147 So is the distribution of white matter connecting the frontal cortex to other regions of the brain.148 It is possible, then, that the frontal lobe circuitry that implements self-control varies genetically among individuals, making it eligible for recent natural selection.

  Oxytocin, the so-called cuddle hormone which encourages sympathy and trust, acts on receptors in several parts of the brain, and the number and distribution of those receptors can have dramatic effects on behavior. In a famous experiment, biologists inserted a gene for the receptor for vasopressin (a hormone similar to oxytocin that operates in the brains of males) into meadow voles, an aggressive and promiscuous species that lacks it. Higamous, hogamous, the voles were monogamous, just like their evolutionary cousins the prairie voles, which come with the receptors preinstalled.149 The experiment suggests that simple genetic changes in the oxytocin-vasopressin system can have profound effects on sympathy, bonding, and by extension the inhibition of aggression.

  Testosterone. A person’s response to a challenge of dominance depends in part on the amount of testosterone released into the bloodstream and on the distribution of receptors for the hormone in his or her brain.150 The gene for the testosterone receptor varies across individuals, so a given concentration of testosterone can have a stronger effect on the brains of some people than others. Men with genes that code for more sensitive versions of the receptor have a greater surge of testosterone when conversing with an attractive woman (which can lead to reduced fear and greater risk-taking), and in one study were overrepresented in a sample of convicted rapists and murderers.151 The genetic pathways that regulate testosterone are complicated, but they offer a target by which natural selection could alter people’s willingness to take up aggressive challenges.

  Neurotransmitters are the molecules that are released from a neuron, seep across a microscopic gap, and lock onto a receptor in the surface of another neuron, changing its activity and thereby allowing patterns of neural firing to propagate through the brain. One major class of neurotransmitters are the catecholamines, which include dopamine, serotonin, and norepinephrine (also called noradrenaline, and related to the adrenaline that triggers the fight-or-flight response). The catecholamines are used in several motivational and emotional systems of the brain, and their concentration is regulated by proteins that break them down or recycle them. One of those enzymes is monoamine oxidase-A, MAO-A for short, which helps to break down these neurotransmitters, preventing them from building up in the brain. When they do build up, the organism can become h
yperreactive to threats and more likely to engage in aggression.

  The first sign that MAO-A can affect violence in humans was the discovery of a Dutch family that carried a rare mutation that left half the men without a working version of the gene.152 (The gene is found on the X chromosome, which men have only one of, so if a man’s MAO-A gene is defective, he has no backup copy to compensate.) Over at least five generations, the affected men in the family were prone to aggressive outbursts. One, for example, forced his sisters to disrobe at knifepoint; another tried to run over his boss with his car.

  A more common kind of variation is found in the part of the gene that determines how much MAO-A is produced. People with a low-activity version of the gene build up higher levels of dopamine, serotonin, and norepinephrine in the brain. They also are more likely to have symptoms of antisocial personality disorder, to report that they have committed acts of violence, to be convicted of a violent crime, to have amygdalas that react more strongly and an orbital cortex that reacts less strongly to angry and fearful faces, and, in the psychology lab, to force a fellow participant to drink hot sauce if they think he has exploited them.153 Unlike many other genes that affect behavior, the low-activity version of the MAO-A gene seems to be fairly specific to aggression; it does not correlate well with any other personality trait.154

  The low-activity version of the MAO-A gene makes people more prone to aggression primarily when they have grown up with stressful experiences, such as having been abused or neglected by their parents or having been held back in school.155 It’s hard to pinpoint the exact stressors that have this effect, because stressful lives are often stressful in many ways at once. In fact, the modulating factor may consist of other genes which are shared with an abusive parent, predisposing both parent and child to aggression, and which may also elicit negative reactions from the people around him.156 But whatever the modifying factor is, it does not turn the effects of the low-activity version of the gene upside down. In all the studies, the gene has an aggregate or main effect in the population that could make it a target of selection. Indeed, Moffitt and Caspi (who first discovered that the effect of the gene depends on stressful experiences) suggest that rather than thinking of the low-activity version of the gene as a contributor to violence, we should think of the high-activity version as an inhibitor of violence: it protects people from overreacting to a stressfilled life. Geneticists have discovered statistical evidence of selection for the MAO-A gene in humans, though the evidence does not single out the low- or high-activity variant; nor does it prove that the gene was selected for its effects on aggression.157