In absolute numbers, of course, civilized societies are matchless in the destruction they have wreaked. But should we look at absolute numbers, or at relative numbers, calculated as a proportion of the populations? The choice confronts us with the moral imponderable of whether it is worse for 50 percent of a population of one hundred to be killed or 1 percent of a population of one billion. In one frame of mind, one could say that a person who is tortured or killed suffers to the same degree regardless of how many other people meet such a fate, so it is the sum of these sufferings that should engage our sympathy and our analytic attention. But in another frame of mind, one could reason that part of the bargain of being alive is that one takes a chance at dying a premature or painful death, be it from violence, accident, or disease. So the number of people in a given time and place who enjoy full lives has to be counted as a moral good, against which we calibrate the moral bad of the number who are victims of violence. Another way of expressing this frame of mind is to ask, "If I were one of the people who were alive in a particular era, what would be the chances that I would be a victim of violence?" The reasoning in this second frame of mind, whether it appeals to the proportion of a population or the risk to an individual, ends in the conclusion that in comparing the harmfulness of violence across societies, we should focus on the rate, rather than the number, of violent acts.

Though descriptions of violence in nonstate societies demolish the stereotype that foraging peoples are inherently peaceful, they don't tell us whether the level of violence is higher or lower than in so-called civilized societies. The annals of modern states have no shortage of gruesome massacres and atrocities, not least against native peoples of every continent, and their wars have death tolls that reach eight digits. Only by looking at numbers can we get a sense as to whether civilization has increased violence or decreased it.

How can one establish the cause of death when the victim perished hundreds or thousands of years ago? Some prehistoric skeletons are accompanied by the stone-age equivalent of a smoking gun: a spearhead or arrowhead embedded in a bone, like the ones found in Kennewick Man and Ötzi. But circumstantial evidence can be almost as damning. Archaeologists can check prehistoric skeletons for the kinds of damage known to be left by assaults in humans today. The stigmata include bashed-in skulls, cut marks from stone tools on skulls or limbs, and parry fractures on ulnar bones (the injury that a person gets when he defends himself against an assailant by holding up his arm). Injuries sustained by a skeleton when it was inside a living body can be distinguished in several ways from the damage it sustained when it was exposed to the world. Living bones fracture like glass, with sharp, angled edges, whereas dead bones fracture like chalk, at clean right angles. And if a bone has a different pattern of weathering on its fractured surface than on its intact surface, it was probably broken after the surrounding flesh had rotted away. Other incriminating signs from nearby surroundings include fortifications, shields, shock weapons such as tomahawks (which are useless in hunting), and depictions of human combat on the walls of caves (some of them more than six thousand years old). Even with all this evidence, archaeological death counts are usually underestimates, because some causes of death -- a poisoned arrow, a septic wound, or a ruptured organ or artery -- leave no trace on the victim's bones.

What happens, then, when we use the emergence of states as the dividing line and put hunter-gatherers, hunter-horticulturalists, and other tribal peoples (from any era) on one side, and settled states (also from any era) on the other? Several scholars have recently scoured the anthropological and historical literature for every good body count from nonstate societies that they could find. Two kinds of estimates are available. One comes from ethnographers who record demographic data, including deaths, in the people they study over long stretches of time. The other comes from forensic archaeologists, who sift through burial sites or museum collections with an eye for signs of foul play.

The topmost cluster shows the rate of violent death for skeletons dug out of archaeological sites. They are the remains of hunter-gatherers and hunter-horticulturalists from Asia, Africa, Europe, and the Americas and date from 14,000 BCE to 1770 CE, in every case well before the emergence of state societies or the first sustained contact with them. The death rates range from 0 to 60 percent, with an average of 15 percent.

Once researchers have tallied a raw count of violent deaths, they can convert it to a rate in either of two ways. The first is to calculate the percentage of all deaths that are caused by violence. This rate is an answer to the question, "What are the chances that a person died at the hands of another person rather than passing away of natural causes?" The graph in figure 2-2 presents this statistic for three samples of nonstate people -- skeletons from prehistoric sites, hunter-gatherers, and hunter-horticulturalists -- and for a variety of state societies. Let's walk through it.

Sources: Prehistoric archaeological sites: Bowles, 2009; Keeley, 1996. Hunter-gatherers: Bowles, 2009. Hunter-horticulturalists and other tribal groups: Gat, 2006; Keeley, 1996. Ancient Mexico: Keeley, 1996. World, 20th-century wars & genocides (includes man-made famines): White, 2011. Europe, 1900-60: Keeley, 1996, from Wright, 1942, 1942/1964, 1942/ 1965; see note 52. Europe, 17th-century: Keeley, 1996. Europe and United States, 20th century: Keeley, 1996, from Harris, 1975. World, 20th-century battle deaths: Lacina & Gleditsch, 2005; Sarkees, 2000; see note 54. United States, 2005 war deaths: see text and note 57. World, 2005 battle deaths: see text and note 58.

FIGURE 2-2: Percentage of deaths in warfare in nonstate and state societies

Next are figures from eight contemporary or recent societies that make their living primarily from hunting and gathering. They come from the Americas, the Philippines, and Australia. The average of the rates of death by warfare is within a whisker of the average estimated from the bones: 14 percent, with a range from 4 percent to 30 percent.

Finally we get to some figures for states. The earliest are from the cities and empires of pre-Columbian Mexico, in which 5 percent of the dead were killed by other people. That was undoubtedly a dangerous place, but it was a third to a fifth as violent as an average pre-state society. When it comes to modern states, we are faced with hundreds of political units, dozens of centuries, and many subcategories of violence to choose from (wars, homicides, genocides, and so on), so there is no single "correct" estimate. But we can make the comparison as fair as possible by choosing the most violent countries and centuries, together with some estimates of violence in the world today. As we shall see in chapter 5, the two most violent centuries in the past half millennium of European history were the 17th, with its bloody Wars of Religion, and the 20th, with its two world wars. The historian Quincy Wright has estimated the rate of death in the wars of the 17th century at 2 percent, and the rate of death in war for the first half of the 20th at 3 percent. If one were to include the last four decades of the 20th century, the percentage would be even lower. One estimate, which includes American war deaths as well, comes in at less than 1 percent.

In the next cluster I've lumped pre-state societies that engage in some mixture of hunting, gathering, and horticulture. All are from New Guinea or the Amazon rain forest, except Europe's last tribal society, the Montenegrins, whose rate of violent death is close to the average for the group as a whole, 24.5 percent.

Now let's turn to the present. According to the most recent edition of the Statistical Abstract of the United States, 2,448,017 Americans died in 2005. It was one of the country's worst years for war deaths in decades, with the armed forces embroiled in conflicts in Iraq and Afghanistan. Together the two wars killed 945 Americans, amounting to 0.0004 (four-hundredths of a percent) of American deaths that year. Even if we throw in the 18,124 domestic homicides, the total rate of violent death adds up to 0.008, or eight-tenths of a percentage point. In other Western countries, the rates were even lower. And in the world as a whole, the Human Security Report Project counted 17,400 deaths that year that were directly caused by political violence (war, terrorism, genocide, and killings by warlords and militias), for a rate of 0.0003 (three-hundredths of a percent). It's a conservative estimate, comprising only identifiable deaths, but even if we generously multiplied it by twenty to estimate undocumented battle deaths and indirect deaths from famine and disease, it would not reach the 1 percent mark.

Recently the study of war has been made more precise by the release of two quantitative datasets, which I will explain in chapter 5. They conservatively list about 40 million battle deaths during the 20th century.("Battle deaths" refer to soldiers and civilians who were directly killed in combat.) If we consider that a bit more than 6 billion people died during the 20th century, and put aside some demographic subtleties, we may estimate that around 0.7 percent of the world's population died in battles during that century. Even if we tripled or quadrupled the estimate to include indirect deaths from war-caused famine and disease, it would barely narrow the gap between state and nonstate societies. What if we added the deaths from genocides, purges, and other man-made disasters? Matthew White, the atrocitologist we met in chapter 1, estimates that around 180 million deaths can be blamed on all of these human causes put together. That still amounts to only 3 percent of the deaths in the 20th century.

The major cleft in the graph, then, separates the anarchical bands and tribes from the governed states. But we have been comparing a motley collection of archaeological digs, ethnographic tallies, and modern estimates, some of them calculated on the proverbial back of an envelope. Is there some way to juxtapose two datasets directly, one from hunter-gatherers, the other from settled civilizations, matching the people, era, and methods as closely as possible? The economists Richard Steckel and John Wallis recently looked at data on nine hundred skeletons of Native Americans, distributed from southern Canada to South America, all of whom died before the arrival of Columbus. They divided the skeletons into hunter-gatherers and city dwellers, the latter from the civilizations in the Andes and Mesoamerica such as the Incas, Aztecs, and Mayans. The proportion of hunter-gatherers that showed signs of violent trauma was 13.4 percent, which is close to the average for the hunter-gatherers in figure 2-2. The proportion of city dwellers that showed signs of violent trauma was 2.7 percent, which is close to the figures for state societies before the present century. So holding many factors constant, we find that living in a civilization reduces one's chances of being a victim of violence fivefold.

Let's turn to the second way of quantifying violence, in which the rate of killing is calculated as a proportion of living people rather than dead ones. This statistic is harder to compute from boneyards but easier to compute from most other sources, because it requires only a body count and a population size, not an inventory of deaths from other sources. The number of deaths per 100,000 people per year is the standard measure of homicide rates, and I will use it as the yardstick of violence throughout the book. To get a feel for what these numbers mean, keep in mind that the safest place in human history, Western Europe at the turn of the 21st century, has a homicide rate in the neighborhood of 1 per 100,000 per year. Even the gentlest society will have the occasional young man who gets carried away in a barroom brawl or an old woman who puts arsenic in her husband's tea, so that is pretty much as low as homicide rates ever go. Among modern Western countries, the United States lies at the dangerous end of the range. In the worst years of the 1970s and 1980s, it had a homicide rate of around 10 per 100,000, and its notoriously violent cities, like Detroit, had a rate of around 45 per 100,000. If you were living in a society with a homicide rate in that range, you would notice the danger in everyday life, and as the rate climbed to 100 per 100,000, the violence would start to affect you personally: assuming you have a hundred relatives, friends, and close acquaintances, then over the course of a decade one of them would probably be killed. If the rate soared to 1,000 per 100,000 (1 percent), you'd lose about one acquaintance a year, and would have a better-than-even lifetime chance of being murdered yourself.

Figure 2-3 shows war death rates for twenty-seven nonstate societies (combining hunter-gatherers and hunter-horticulturalists) and nine that are ruled by states. The average annual rate of death in warfare for the nonstate societies is 524 per 100,000, about half of 1 percent. Among states, the Aztec empire of central Mexico, which was often at war, had a rate about half that. Below that bar we find the rates for four state societies during the centuries in which they waged their most destructive wars. Nineteenth-century France fought the Revolutionary, Napoleonic, and Franco-Prussian Wars and lost an average of 70 people per 100,000 per year. The 20th century was blackened by two world wars that inflicted most of their military damage on Germany, Japan, and Russia/USSR, which also had a civil war and other military adventures. Their annual rates of death work out to 144, 27, and 135 per 100,000, respectively. During the 20th century the United States acquired a reputation as a warmonger, fighting in two world wars and in the Philippines, Korea, Vietnam, and Iraq. But the annual cost in American lives was even smaller than those of the other great powers of the century, about 3.7 per 100,000. Even if we add up all the deaths from organized violence for the entire world for the entire century -- wars, genocides, purges, and man-made famines -- we get an annual rate of around 60 per 100,000. For the year 2005, the bars representing the United States and the entire world are paint-thin and invisible in the graph.

Though war is common among foraging groups, it is certainly not universal. Nor should we expect it to be if the violent inclinations in human nature are a strategic response to the circumstances rather than a hydraulic response to an inner urge. According to two ethnographic surveys, 65 to 70 percent of hunter-gatherer groups are at war at least every two years, 90 percent engage in war at least once a generation, and virtually all the rest report a cultural memory of war in the past. That means that hunter-gatherers often fight, but they can avoid war for long stretches of time. Figure 2-3 reveals two tribes, the Andamanese and the Semai, with low rates of death in warfare. But even they have interesting stories.

So by this measure too, states are far less violent than traditional bands and tribes. Modern Western countries, even in their most war-torn centuries, suffered no more than around a quarter of the average death rate of nonstate societies, and less than a tenth of that for the most violent one.

The Andaman Islanders of the Indian Ocean are recorded as having an annual death rate of 20 per 100,000, well below the average for nonstate peoples (which exceeds 500 per 100,000). But they are known to be among the fiercest hunter-gatherer groups left on earth. Following the 2004 Indian Ocean earthquake and tsunami, a worried humanitarian group flew over to the islands in a helicopter and were relieved to be met with a fusillade of arrows and spears, signs that the Andamanese had not been wiped out. Two years later a pair of Indian fishers fell into a drunken sleep, and their boat drifted ashore on one of the islands. They were immediately slain, and the helicopter sent to retrieve their bodies was also met with a shower of arrows.

There are, to be sure, hunter-gatherers and hunter-horticulturalists such as the Semai who have never been known to engage in the protracted, collective killings that can be called warfare. Anthropologists of peace have made much of these groups, suggesting that they could have been the norm in human evolutionary history, and that it is only the newer and wealthier horticulturalists and pastoralists who engage in systematic violence. The hypothesis is not directly relevant to this chapter, which compares people living in anarchy with those living under states rather than hunter-gatherers with everyone else. But there are reasons to doubt the hypothesis of hunter-gatherer innocence anyway. Figure 2-3 shows that the rates of death in warfare in these societies, though lower than those of horticulturalists and tribesmen, overlap with them considerably. And as I have mentioned, the hunter-gatherer groups we observe today may be historically unrepresentative. We find them in parched deserts or frozen wastelands where no one else wants to live, and they may have ended up there because they can keep a low profile and vote with their feet whenever they get on each other's nerves. As Van der Dennen comments, "Most contemporary 'peaceful' foragers… have solved the perennial problem of being left in peace by splendid isolation, by severing all contacts with other peoples, by fleeing and hiding, or else by being beaten into submission, by being tamed by defeat, by being pacified by force." For example, the! Kung San of the Kalahari Desert, who in the 1960s were extolled as a paradigm of hunter-gatherer harmony, in earlier centuries had engaged in frequent warfare with European colonists, their Bantu neighbors, and one another, including several all-out massacres.

FIGURE 2-3: Rate of death in warfare in nonstate and state societies

The low rates of death in warfare in selected small-scale societies can be misleading in another way. Though they may avoid war, they do commit the occasional murder, and their homicide rates can be compared to those of modern state societies. I've plotted them in figure 2-4 on a scale that is fifteen times larger than that of figure 2-3. Let's begin with the right-most gray bar in the nonstate cluster. The Semai are a hunting and horticulturalist tribe who were described in a book called The Semai: A Nonviolent People of Malaya and who go out of their way to avoid the use of force. While there aren't many Semai homicides, there aren't many Semai. When the anthropologist Bruce Knauft did the arithmetic, he found that their homicide rate was 30 per 100,000 per year, which puts it in the range of the infamously dangerous American cities in their most violent years and at three times the rate of the United States as a whole in its most violent decade. The same kind of long division has deflated the peaceful reputation of the! Kung, the subject of a book called The Harmless People, and of the Central Arctic Inuit (Eskimos), who inspired a book called Never in Anger. Not only do these harmless, nonviolent, anger-free people murder each other at rates far greater than Americans or Europeans do, but the murder rate among the! Kung went down by a third after their territory had been brought under the control of the Botswana government, as the Leviathan theory would predict.

The reduction of homicide by government control is so obvious to anthropologists that they seldom document it with numbers. The various "paxes" that one reads about in history books -- the Pax Romana, Islamica, Mongolica, Hispanica, Ottomana, Sinica, Britannica, Australiana (in New Guinea), Canadiana (in the Pacific Northwest), and Praetoriana (in South Africa)-- refer to the reduction in raiding, feuding, and warfare in the territories brought under the control of an effective government. Though imperial conquest and rule can themselves be brutal, they do reduce endemic violence among the conquered. The Pacification Process is so pervasive that anthropologists often treat it as a methodological nuisance. It goes without saying that peoples that have been brought under the jurisdiction of a government will not fight as much, so they are simply excluded from studies of violence in indigenous societies. The effect is also noticeable to the people themselves. As an Auyana man living in New Guinea under the Pax Australiana put it, "Life was better since the government came" because "a man could now eat without looking over his shoulder and could leave his house in the morning to urinate without fear of being shot."

Sources:! Kung and Central Arctic Inuit: Gat, 2006; Lee, 1982. Semai: Knauft, 1987. Ten largest U. S. cities: Zimring, 2007, p. 140. United States: FBI Uniform Crime Reports; see note 73. Western Europe (approximation): World Health Organization; see note 66 to chap. 3, p. 701.

FIGURE 2-4: Homicide rates in the least violent nonstate societies compared to state societies

The anthropologists Karen Ericksen and Heather Horton have quantified the way that the presence of government can move a society away from lethal vengeance. In a survey of 192 traditional studies, they found that one-on-one revenge was common in foraging societies, and kin-against-kin blood feuds were common in tribal societies that had not been pacified by a colonial or national government, particularly if they had an exaggerated culture of manly honor. Adjudication by tribunals and courts, in contrast, was common in societies that had fallen under the control of a centralized government, or that had resource bases and inheritance patterns that gave people more of a stake in social stability.

One of the tragic ironies of the second half of the 20th century is that when colonies in the developing world freed themselves from European rule, they often slid back into warfare, this time intensified by modern weaponry, organized militias, and the freedom of young men to defy tribal elders. As we shall see in the next chapter, this development is a countercurrent to the historical decline of violence, but it is also a demonstration of the role of Leviathans in propelling the decline.