with a ruler, measuring the number of column inches devoted to each century.24 The range was so great that he had to plot the data on a logarithmic scale (on which an exponential fade looks like a straight line). His graph, reproduced in figure 5–4, shows that as you go back into the past, historical coverage hurtles exponentially downward for two and a half centuries, then falls with a gentler but still exponential decline for the three millennia before.
If it were only a matter of missing a few small wars that escaped the notice of ancient chroniclers, one might be reassured that the body counts were not underestimated, because most of the deaths would be in big wars that no one could fail to notice. But the undercounting may introduce a bias, not just a fuzziness, in the estimates. Keegan writes of a “military horizon.”25 Beneath it are the raids, ambushes, skirmishes, turf battles, feuds, and depredations that historians dismiss as “primitive” warfare. Above it are the organized campaigns for conquest and occupation, including the set-piece battles that war buffs reenact in costume or display with toy soldiers. Remember Tuchman’s “private wars” of the 14th century, the ones that knights fought with furious gusto and a single strategy, namely killing as many of another knight’s peasants as possible? Many of these massacres were never dubbed The War of Such-and-Such and immortalized in the history books. An undercounting of conflicts below the military horizon could, in theory, throw off the body count for the period as a whole. If more conflicts fell beneath the military horizon in the anarchic feudal societies, frontiers, and tribal lands of the early periods than in the clashes between Leviathans of the later ones, then the earlier periods would appear less violent to us than they really were.
FIGURE 5–4. Historical myopia: Centimeters of text per century in a historical almanac
Source: Data from Taagepera & Colby, 1979, p. 911.
So when one adjusts for population size, the availability bias, and historical myopia, it is far from clear that the 20th century was the bloodiest in history. Sweeping that dogma out of the way is the first step in understanding the historical trajectory of war. The next is to zoom in for a closer look at the distribution of wars over time—which holds even more surprises.
THE STATISTICS OF DEADLY QUARRELS, PART 1: THE TIMING OF WARS
Lewis Richardson wrote that his quest to analyze peace with numbers sprang from two prejudices. As a Quaker, he believed that “the moral evil in war outweighs the moral good, although the latter is conspicuous.”26 As a scientist, he thought there was too much moralizing about war and not enough knowledge: “For indignation is so easy and satisfying a mood that it is apt to prevent one from attending to any facts that oppose it. If the reader should object that I have abandoned ethics for the false doctrine that ‘tout comprendre c’est tout pardonner’ [to understand all is to forgive all], I can reply that it is only a temporary suspense of ethical judgment, made because ‘beaucoup condamner c’est peu comprendre’ [to condemn much is to understand little].”27
After poring through encyclopedias and histories of different regions of the world, Richardson compiled data on 315 “deadly quarrels” that ended between 1820 and 1952. He faced some daunting problems. One is that most histories are sketchy when it comes to numbers. Another is that it isn’t always clear how to count wars, since they tend to split, coalesce, and flicker on and off. Is World War II a single war or two wars, one in Europe and the other in the Pacific? If it’s a single war, should we not say that it began in 1937, with Japan’s full-scale invasion of China, or even in 1931, when it occupied Manchuria, rather than the conventional starting date of 1939? “The concept of a war as a discrete thing does not fit the facts,” he observed. “Thinginess fails.”28
Thinginess failures are familiar to physicists, and Richardson handled them with two techniques of mathematical estimation. Rather than seeking an elusive “precise definition” of a war, he gave the average priority over the individual case: as he considered each unclear conflict in turn, he systematically flipped back and forth between lumping them into one quarrel and splitting them into two, figuring that the errors would cancel out in the long run. (It’s the same principle that underlies the practice of rounding a number ending in 5 to the closest even