records have failed to uncover a single case of a state's being formed in that ethereal atmosphere of dispassionate farsightedness. Smaller units do not voluntarily abandon their sovereignty and merge into larger units. They do so only by conquest, or under external duress. A third theory, still popular with some historians and economists, sets out from the undoubted fact that, in both Mesopotamia and North China and Mexico, large-scale irrigation systems began to be constructed around the time that states started to emerge. The theory also notes that any big, complex system for irrigation or hydraulic management requires a centralized bureaucracy to construct and maintain it. The theory then turns an observed rough correlation in time into a postulated chain of cause and effect. Supposedly, Mesopotamians and North Chinese and Mexicans foresaw the advantages that a large-scale irrigation system would bring them, even though there was at the time no such system within thousands of miles (or anywhere on Earth) to illustrate for them those advantages. Those farsighted people chose to merge their inefficient little chiefdoms into a larger state capable of blessing them with large-scale irrigation. However, this 'hydraulic theory' of state formation is subject to the same objections leveled against social contract theories in general. More specifically, it addresses only the final stage in the evolution of complex societies. It says nothing about what drove the progression from bands to tribes to chiefdoms during all the millennia before the prospect of large-scale irrigation loomed up on the horizon. When historical or archaeological dates are examined in detail, they fail to support the view of irrigation as the driving force for state formation. In Mesopotamia, North China, Mexico, and Madagascar, small-scale irrigation systems already existed e ore tne 'se of states. Construction of large-scale irrigation systems did not accompany the emergence of states but came only significantly later in ac of those areas. In most of the states formed over the Maya area of 2. 8 4 'GUNS,GERMS, AND STEEL Mesoamerica and the Andes, irrigation systems always remained small-scale ones that local communities could build and maintain themselves. Thus, even in those areas where complex systems of hydraulic management did emerge, they were a secondary consequence of states that must have formed for other reasons. What seems to me to point to a fundamentally correct view of state formation is an undoubted fact of much wider validity than the correlation between irrigation and the formation of some states—namely, that the size of the regional population is the strongest single predictor of societal complexity. As we have seen, bands number a few dozen individuals, tribes a few hundred, chiefdoms a few thousand to a few tens of thousands, and states generally over about 50,000. In addition to that coarse correlation between regional population size and type of society (band, tribe, and so on), there is a finer trend, within each of those categories, between population and societal complexity: for instance, that chiefdoms with large populations prove to be the most centralized, stratified, and complex ones. These correlations suggest strongly that regional population size or population density or population pressure has something to do with the formation of complex societies. But the correlations do not tell us precisely how population variables function in a chain of cause and effect whose outcome is a complex society. To trace out that chain, let us now remind ourselves how large dense populations themselves arise. Then we can examine why a large but simple society could not maintain itself. With that as background, we shall finally return to the question of how a simpler society actually becomes more complex as the regional population increases. we have seen that large or dense populations arise only under conditions of food production, or at least under exceptionally productive conditions for hunting-gathering. Some productive hunter-gatherer societies reached the organizational level of chiefdoms, but none reached the level of states: all states nourish their citizens by food production. These considerations, along with the just mentioned correlation between regional population size and societal complexity, have led to a protracted chicken-or-egg debate about the causal relations between food production, population variables, and societal complexity. Is it intensive food production that is the cause, triggering population growth and somehow leading to a com- FROMegal1tarianism TO KLEPTOCRACY • Z 8 5 lex society? Or are large populations and complex societies instead the cause, somehow leading to intensification of food production? Posing the question in that either-or form misses the point. Intensified food production and societal complexity stimulate each other, by autoca-talvsis. That is, population growth leads to societal complexity, by mechanisms that we shall discuss, while societal complexity in turn leads to intensified food production and thereby to population growth. Complex centralized societies are uniquely capable of organizing public works (including irrigation systems), long-distance trade (including the importation of m: rals to make better agricultural tools), and activities of different groups of economic specialists (such as feeding herders with farmers' cereal, and transferring the herders' livestock to farmers for use as plow animals}. Al! of these capabilities of centralized societies have fostered intens'fied food production and hence population growth throughout history. In addition, food production contributes in at least three ways to specific features of complex societies. First, it involves seasonally pulsed inputs of labor. When the harvest has been stored, the farmers' labor becomes available for a centralized political authority to harness—in order to build public works advertising state power (such as the Egyptian pyramids), or to build public works that could feed more mouths (such as Polynesian Hawaii's irrigation systems or fishponds), or to undertake wars of conquest to form larger political entities. Second, food production may be organized so as to generate stored food surpluses, which permit economic specialization and social stratification. The surpluses can be used to feed all tiers of a complex society: the chiefs, bureaucrats, and other members of the elite; the scribes, craftspeople, and other non-food-producing specialists; and the farmers themselves, during times that they are drafted to construct public works. Finally, food production permits or requires people to adopt sedentary living, which is a prerequisite for accumulating substantial possessions, developing elaborate technology and crafts, and constructing public works. The importance of fixed residence to a complex society explains wny missionaries and governments, whenever they make first contact with Previously uncontacted nomadic tribes or bands in New Guinea or the mazon, universally have two immediate goals. One goal, of course, is the , . VIOus one o* 'pacifying' the nomads: that is, dissuading them from lng missionaries, bureaucrats, or each other. The other goal is to induce 1 8 6 •GUNS,GERMS, AND STEEL the nomads to settle in villages, so that the missionaries and bureaucrats can find the nomads, bring them services such as medical care and schools, and proselytize and control them. Thus, food production, which increases population size, also acts in many ways to make features of complex societies possible. But that doesn't prove that food production and large populations make complex societies inevitable. How can we account for the empirical observation that band or tribal organization just does not work for societies of hundreds of thousands of people, and that all existing large societies have complex centralized organization? We can cite at least four obvious reasons. One reason is the problem of conflict between unrelated strangers. That problem grows astronomically as the number of people making up the society increases. Relationships within a band of 20 people involve only 190 two-person interactions (20 people times 19 divided by 2), but a band of 2,000 would have 1,999,000 dyads. Each of those dyads represents a potential time bomb that could
