Source: Graph from Eisner, 2003.

FIGURE 3-2:. Homicide rates in England, 1200-2000

Before we try to explain this remarkable development, let's be sure it is real. Following the publication of Gurr's graph, several historical criminologists dug more deeply into the history of homicide. The criminologist Manuel Eisner assembled a much larger set of estimates on homicide in England across the centuries, drawing on coroners' inquests, court cases, and local records. Each dot on the graph in figure 3-2 is an estimate from some town or jurisdiction, plotted once again on a logarithmic scale. By the 19th century the British government was keeping annual records of homicide for the entire country, which are plotted on the graph as a gray line. Another historian, J. S. Cockburn, compiled continuous data from the town of Kent between 1560 and 1985, which Eisner superimposed on his own data as the black line.

Once again we see a decline in annual homicide rates, and it is not small: from between 4 and 100 homicides per 100,000 people in the Middle Ages to around 0.8 (eight-tenths of a homicide) per 100,000 in the 1950s. The timing shows that the high medieval murder rates cannot be blamed on the social upheavals that followed the Black Death around 1350, because many of the estimates predated that epidemic.

Eisner has given a lot of thought to how much we should trust these numbers. Homicide is the crime of choice for measurers of violence because regardless of how the people of a distant culture conceptualize crime, a dead body is hard to define away, and it always arouses curiosity about who or what produced it. Records of homicide are therefore a more reliable index of violence than records of robbery, rape, or assault, and they usually (though not always) correlate with them.

Still, it's reasonable to wonder how the people of different eras reacted to these killings. Were they as likely as we are to judge a killing as intentional or accidental, or to prosecute the killing as opposed to letting it pass? Did people in earlier times always kill at the same percentage of the rate that they raped, robbed, and assaulted? How successful were they in saving the lives of victims of assault and thereby preventing them from becoming victims of homicide?

Fortunately, these questions can be addressed. Eisner cites studies showing that when people today are presented with the circumstances of a centuriesold murder and asked whether they think it was intentional, they usually come to the same conclusion as did the people at the time. He has shown that in most periods, the rates of homicide do correlate with the rates of other violent crimes. He notes that any historical advance in forensics or in the reach of the criminal justice system is bound to underestimate the decline in homicide, because a greater proportion of killers are caught, prosecuted, and convicted today than they were centuries ago. As for lifesaving medical care, doctors before the 20th century were quacks who killed as many patients as they saved; yet most of the decline took place between 1300 and 1900. In any case, the sampling noise that gives social scientists such a headache when they are estimating a change of a quarter or a half is not as much of a problem when they are dealing with a change of tenfold or fiftyfold.

Were the English unusual among Europeans in gradually refraining from murder? Eisner looked at other Western European countries for which criminologists had compiled homicide data. Figure 3-3 shows that the results were similar. Scandinavians needed a couple of additional centuries before they thought the better of killing each other, and Italians didn't get serious about it until the 19th century. But by the 20th century the annual homicide rate of every Western European country had fallen into a narrow band centered on 1 per 100,000.

FIGURE 3-3: Homicide rates in five Western European regions, 1300-2000

To put the European decline in perspective, let's compare it to the rates for nonstate societies that we encountered in chapter 2. In figure 3-4 I have extended the vertical axis up to 1,000 on the log scale to accommodate the additional order of magnitude required by the nonstate societies. Even in the late Middle Ages, Western Europe was far less violent than the unpacified nonstate societies and the Inuit, and it was comparable to the thinly settled foragers such as the Semai and the! Kung. And from the 14th century on, the European homicide rate sank steadily, with a tiny bounce in the last third of the 20th century.

Source: Data from Eisner, 2003, table 1.

Another historical change was that homicides in which one man kills another man who is unrelated to him declined far more rapidly than did the killing of children, parents, spouses, and siblings. This is a common pattern in homicide statistics, sometimes called Verkko's Law: rates of male-on-male violence fluctuate more across different times and places than rates of domestic violence involving women or kin. Martin Daly and Margo Wilson's explanation is that family members get on each other's nerves at similar rates in all times and places because of deeply rooted conflicts of interest that are inherent to the patterns of genetic overlap among kin. Macho violence among male acquaintances, in contrast, is fueled by contests of dominance that are more sensitive to circumstances. How violent a man must be to keep his rank in the pecking order in a given milieu depends on his assessment of how violent the other men are, leading to vicious or virtuous circles that can spiral up or down precipitously. I'll explore the psychology of kinship in more detail in chapter 7, and of dominance in chapter 8.

While Europe was becoming less murderous overall, certain patterns in homicide remained constant. Men were responsible for about 92 percent of the killings (other than infanticide), and they were most likely to kill when they were in their twenties. Until the 1960s uptick, cities were generally safer than the countryside. But other patterns changed. In the earlier centuries the upper and lower social classes engaged in homicide at comparable rates. But as the homicide rate fell, it dropped far more precipitously among the upper classes than among the lower ones, an important social change to which we will return.

FIGURE 3-4: Homicide rates in Western Europe, 1300-2000, and in nonstate societies

Sources: Nonstate (geometric mean of 26 societies, not including Semai, Inuit, and! Kung): see figure 2-3. Europe: Eisner, 2003, table 1; geometric mean of five regions; missing data interpolated.