- Social Sciences
- African Studies
- American Studies
- Asian Studies
- Communication Sciences
- Ethnic Studies
- European Studies
- Gender Studies
- Physical Sciences
- Life Sciences
- Animal Communications
- Cell Biology
- Evolutionary Biology
- Food Science and Technology
- Human Anatomy
From Warre to Tyranny: Lethal Conflict and the State
by Mark Cooney
From Warre to Tyranny: Lethal Conflict and the State
American Sociological Review
Updated: February 19th, 2013
University of Georgia
Following Hobbes, many social theorists have claimed that the state reduces the amount of violence in human societies. Are they right? I review the cross- cultural and cross-national evidence on the impact of the state on the most conzmon form of extreme violence-lethal conflict (i,e., war, rebellion, homi- cide, and execution). Drawing on the sociology of conflict management (Black 1993),I argue that the relationship between the state and lethal conflict is not negative as Hobbesian theory predicts. Rather, it appears to be U-shaped. A combination of materials from anthropology, criminolog,y, and political sci- ence suggests that rates of lethal conflict tend to be high when state authority is absent and also when it is extremely strong or centralized. Between these extremes, in less centralized states, lethal conflict typically declines.
Theorists have long argued that the state reduces violence among those subject to its jurisdiction. Hobbes ([I6511 1909) pro- vided an early and eloquent statement of this idea, contending that in the absence of a strong central authority, violence pervades social life. Weber's ([I9221 1968:54) influ- ential definition of the state emphasizes its ability to successfully monopolize the legiti- mate use of violence, a capacity that would seem to depend partly on the state's ability to restrict violent behavior. Elias ([I9391 1982) nominates the state as a primary source of the long-term civilizing process that decreases violence in everyday life. And Koch's (1974) experience as an ethnographer in New Guinea leads him to stress the importance of third- party modes of conflict management, like those provided by state legal systems, as vio- lence-controlling mechanisms in human so- cieties. But are these authors correct? Does the state diminish violence in human affairs? If
" Direct all correspondence to Mark Cooney, Department of Sociology, University of Georgia, Athens, GA 30602-161 1 (mcooney @uga.cc.uga. edu). An early version of this paper was presented at the annual meeting of the American Society of Criminology, Miami, November 1994. For com- ments and advice I thank M. P. Baumgartner, E.
M. Beck, Donald Black, William Finlay, John Herrmann, Ivy Kennelly, James Tucker, and sev- eral anonymous ASR reviewers.
the state did not exist, would life be more violent? Neither the literature on the state nor that on violence contains a sustained empiri- cal analysis of these issues, even though the "many-sided problem of order" remains a prominent topic of theoretical discussion (Wrong 1994). In this paper, I seek to fill the void, reviewing the available information in light of ideas developed in the sociology of conflict management, a broad field dedicated to describing and explaining the handling of human conflict (Black 1976, 1984, 1993; also see Horwitz 1990).
The argument I advance is that the Hobbe-
sian thesis is only partially correct. Anthro- pological evidence suggests that the state tends to reduce the amount of violent conflict in human societies. But political science data indicate that the state may, under cer- tain conditions. also increase violence. Con- sequently, while the form of violent conflict may change from the "warre of every man against every man" (Hobbes [I65 11 1909:98) to state tyranny directed against citizens, the sheer volume of violence may remain ap- proximately the same. I therefore propose that the overall relationship between the state and violent conflict is U-shaped: High levels of violent conflict are found when state au- thority is weak or absent and when it is ex- tremely strong or centralized. Between these extremes, in less centralized states, low and intermediate amounts of violence are found.
American Sociological Review, 1997, Vol. 62 (April:3 16-338)
The information on violence is most abun- dant for lethal violence, whether committed by individuals, groups, or governments. Most lethal violence arises out of conflict. Hence, in using data on lethal violence I am, in ef- fect, measuring lethal conflict. My argument encompasses two hypotheses. First, rates of lethal conflict are higher in stateless than in noncentralized state societies. Second, rates of lethal conflict are higher in centralized state societies than noncentralized state soci- eties.
ISSUES AND PERSPECTIVES
Scholarship on the state has analyzed vari- ous aspects of violence. These include the link between war-making and state-forma- tion (Tilly 1975, 1990; Porter 1994), the re- lationship between internal and external state violence (Rasler 1986; Starr 1994), the growth of coercive states (Gurr 1988), and the trend away from interstate war toward more amorphous forms of conflict within and across national boundaries (van Creveld 1991; Tilly 1995). Anthropologists have pur- sued similar questions in the context of struc- turally simple societies, investigating, for ex- ample, how the development of political in- stitutions affects the incidence of different forms of violence (Otterbein and Otterbein 1965; Otterbein 1968).
Important though these questions are, I ad- dress a different topic here: the effect of state authority on the overall quantity of lethal conflict. Consider two scenarios. In scenario A, the people of a region are divided into nu- merous small, autonomous, political commu- nities. In scenario B, these same people are now organized into two sovereign states. Is there more lethal conflict in scenario A or B? In other words, compared to their counterpart in B, does the average person in A have a greater, lesser, or the same chance of being killed in a violent conflict-whether the vio- lence is warfare, revolution, execution, or homicide?
Theorists take three positions. The first posi- tion-that the state reduces lethal conflict- finds its classic expression in Hobbes's ([I 65 11 1909) Leviathan:
[I]t is manifest that during the time men live without a common power to keep them all in awe, they are in that condition which is called Warre; and such a warre, as is of every man, against every man. (P. 96)
The consequences of warre are
. . . continuall feare, and danger of violent death; And the life of man, solitary, poore, nasty, brutish, and short. (P. 97)
The solution is sovereign power
. . . as great, as possibly men can be imagined to make it. And though of so unlimited a Power, men may fancy many evil1 conse- quences, yet the consequences of the want of it, which is perpetual1 warre of every man against his neighbor, are much worse. (P. 160)
Although the Hobbesian view is undoubt- edly the dominant one, a second position holds that the state has no impact on rates of lethal conflict. The anarchist theorist, Kro- potkin ([I8861 1975), expressed one version of this idea:
[Tlhe severity of punishment does not dimin- ish the amount of crime. Hang, and if you like, quarter murderers, and the number of murders will not decrease by one. On the other hand, abolish the penalty of death, and there will not be one murder more; there will be fewer. Sta- tistics prove it. But if the harvest is good, and bread cheap, the weather fine, the number of murders immediately decreases. This again is proved by statistics. The amount of crime al- ways augments and diminishes in proportion to the price of provisions and the state of the weather. (P. 42)
A third position is that the state increases the incidence of violent death. One way it does this is by increasing the scale of war- fare (Reyna 1994). This argument has been stated most forcefully by the anthropologist Lee (1979) in his study of the !Kung San (now called the Jul'hoansi; Lee 1993:ix; cf. Thomas 1994), a much-studied group of hunter-gatherers in the Kalahari desert. Re- ferring to the Hobbesian thesis, as expressed by Sahlins (1968), Lee (1979) remarks:
An alternative but complementary view is to regard the process of social evolution leading to the state as one of externalizing violence rather than controlling or eliminating it. . . . As human societies have evolved from bands (like the !Kung) to tribes and chiefdoms, each step up in the level of sociocultural integration has
reduced the problems of violence at the previ- ous level of integration, but has opened up new forms of violence at the new level. So, for ex- ample, in the nineteenth century the Batswana chiefdom imposed its order on the band-level San hunters in Eastern Botswana, only to wage intertribal warfare on a much larger scale against neighboring chiefdoms such as the Matebele and the Kalanga-Shona. Then at the end of the nineteenth century, the British in- dustrial state brought a Pax Britannica to the warring chiefdoms of Southern Africa. But a generation later, the British mobilized thou- sands of Tswana warriors' sons to fight in the Mediterranean theater against the German and Italian national states. At each new level of in- tegration, the scale on which violence is prac- ticed becomes greater in terms of the numbers involved, the degree of organization, the length and intensity of the conflict, and the technologi- cal sophistication. (P. 399, emphases omitted)'
I argue that none of these positions fully captures the complex relationship between the state and the amount of lethal conflict that occurs. The state does affect rates of le- thal conflict, contrary to Kropotkin's view, but the relationship is neither positive, as Lee holds, nor negative, as Hobbes and others believe. Rather, the relationship appears to be U-shaped, an argument that draws on re- cent theoretical developments in the sociol- ogy of conflict management.2
The of Conflict Management
The sociology of conflict management seeks to isolate the social conditions underlying variation in the way people handle conflict. Early work by Marx (Cain and Hunt 1979), Maine ([I 86 11 1963), Durkheim ([I 8931 1964, 1899-1 900), Weber (1954) and others focused on law, but the subject now extends to the entire range of formal and informal tactics that individuals and groups use in pur- suing grievances (Black 1984).
I Lee's view is consistent with Hobbes's in im- plying that a worldwide state would be the most effective way to reduce violence. It diverges from Hobbes's position in its assessment of the overall amount of violence found under a system of mul- tiple states.
Although my argument is confined to states, the same relationship may hold for other forms of authority (e.g., domestic, occupational, local).
AMERICAN SOCIOLOGICAL REVIEW
Black (1976, 1993) has developed a body of theory that explains variation in conflict management regardless of time, place, or structural level. Theory of that generality must necessarily incorporate findings from several disciplines, including anthropology, history, political science, and criminology (Black 1995). An interdisciplinary focus also characterizes much work among the growing number of scholars who use Black's ideas to study a variety of forms of conflict manage- ment (or "social controln-Black uses the two terms interchangeably). These include studies of collective violence (Senechal de la Roche 1996), international discord (Borg 1992), avoidance and nonconfrontation in suburbia (Baumgartner 1988), disputing among corporate executives (Morrill 1995), cross-cultural patterns of domestic violence (Baumgartner 1993), the treatment of mental illness (Horwitz 1982), and the handling of grievances in nonhierarchical corporations (Tucker forthcoming).
The literature on conflict management cur- rently does not specify an overall relation- ship between the state and violent conflict, but it suggests a question for analyzing the issue: In what ways could the state promote or inhibit the use of violence among people in conflict? The literature also provides some preliminary answers.
One possible effect of the state on conflict. for example, is to reduce the incidence of le- thal violeke by providing disputants with a peaceful me& for resolving their differ- ences-law. Not that statelessness should au- tomatically lead to violence. There are many ways of handling conflict nonviolently besides invoking law. These include negotia- tion, mediation, arbitration, avoidance, and toleration (Black 1990). Nevertheless, the presence of law should, in the aggregate, make violence less likely because it creates another peaceful alternati~e.~
"he state suffers from an important limitation as an explanatory variable: It lacks generality. For example, it is unable to explain variation in vio- lence within and across stateless societies. As such, it is a poor candidate for inclusion in a gen- eral sociological theory of violence (Black 1990: 4349). Consequently, the U-shaped relationship I propose cannot be regarded as a fundamental theoretical proposition.
On the other hand, if state authority grows extremely strong, violence can be expected to increase again (Black 1993, chap. 8). Strong or centralized states do not merely provide institutions and personnel to resolve disputes, they become directly involved in conflict themselves, as principals. Highly centralized regimes tend to be intolerant and severe toward their own citizens and other states. Given to persecution at home and ag- gression abroad, they create, find, and pun- ish enemies to a degree unknown in other states.
I combine these ideas into a single formu- lation. I argue that lethal conflict is most pro- nounced where state authority is weak or ab- sent and where it is extremely strong.4 Be- fore turning to my argument, however, I must clarify some key terms.
I define a state conventionally, as "an autono- mous political unit, encompassing many communities within its territory and having a centralized government with the power to collect taxes, draft men for work or war, and decree and enforce laws" (Carneiro 1970:733). States differ in many ways, but one key distinction identified by political sci- entists and sociologists is the strength of their authority or how centralized they are.5 State centralization, in the sense used here, has at least two principal dimensions (Rummel 1995). The first is autocracy-de- mocracy, or the extent to which political de- cision-making is concentrated in the hands of a single individual or small group. Measures of autocracy include how few people make decisions on matters of state importance
A reviewer suggested that the empirical pat- terns described here are better explained by the presence or absence of the rule of law. But the rule of law is not a satisfactory explanatory vari- able because law responds to its social environ- ment (Cooney 1995). Hence, even if lethal vio- lence is negatively correlated with the rule of law, the question still remains: What explains the rule of law? One factor discussed here is the strength of state authority.
"Centralization" should not be confused with network "centrality" (Hage and Harary 1983:30- 39). Centralization could also be labeled "authoritarianism" or "totalitarianism."
(e.g., finance, security), the absence of checks on leaders' decisions, and the lack of free and regular elections. The second is to- talitarianism-libertarianism, or the degree to which the state penetrates and controls social life. Measures of totalitarianism include the extent of state ownership, the size and scope of the bureaucracy, and the number and in- fluence of state officials in neighborhoods, workplaces, schools, and other institutions.
In highly centralized states both autoc- racy-democracy and totalitarianism-liber- tarianism are strongly developed: The state controls most social activity, places its offi- cials throughout the society, and political de- cision-making is in the tenacious hands of a single person or a small group that is not an- swerable to a popular electorate or subject to review by other branches of government. Ex- amples are ancient empires, Communist re- gimes, and totalitarian polities (e.g., Nazi Germany). The least centralized states are those in which all citizens participate equally in decision-making and in which the state exerts only minimal control over social insti- tutions. Because there are (at least yet) no states with such minimal centralization, states can be considered noncentralized if they have a low overall score on both dimen- sions of centralization. One example is tradi- tional states with political institutions based on limited representation that levy taxes and maintain order but otherwise do not penetrate the lives of their citizens. Another is modern democracies, which regulate many institu- tional spheres but do not typically dominate them by subordinating their interests to those of the polity, and which distribute public de- cision-making among a variety of officehold- ers elected by popular and periodic vote. Be- cause of the great availability of data on modern democracies, I use them as the ex- emplar of noncentralized states.
Although centralization is a continuum, the preliminary nature of the present discus- sion requires analyzing state authority in terms of a simple trichotomous distinction between stateless societies, noncentralized states, and centralized states. This means that at least four important questions must be de- ferred: (1) how lethal conflict fares under in- termediate forms of pre-state authority, such as chiefdoms; (2) the extent to which the state indirectly affects the frequency of lethal conflict (e.g., by weakening the strong kin- ship groups and local groups associated with feuding in stateless societies; Thoden van Velzen and van Wetering 1960); (3) the de- gree of centralization states must attain be-
fore rates of lethal conflict begin to increase; and (4) which dimension of political central- ization, autocracy-democracy or totalitarian- ism-libertarianism, has the greatest impact on the incidence of lethal conflict.
I define "lethal conflict" as the deliberate or careless killing of another through physical means in the course of a conflict. A conflict is a disagreement over "right" and "wrong." Excluded are killings believed to be caused by supernatural intervention (e.g., witch- craft), but included are killings by neglect, such as the calculated starvation of an en- emy, individual or collective. The killer or victim may be an individual, an organization, a sovereign polity (whether state, chieftain- ship, or some other entity) or one of its agents. This yields four forms of lethal con- flict (Table 1).
Broad though it is, the typology in Table 1 excludes lethal violence arising out of preda- tory behavior such as rape, robbery, banditry, and piracy as well as other killings not re- lated to conflict, such as infanticide and se- rial killing.
Since official statistics rarely distinguish clearly between conflict-related killings and other forms of killing, it is uncertain exactly what percentage of lethal violence originates in conflict. Criminologists have long estab- lished that most modern homicides arise out of disputes, altercations, disagreements, ven- dettas, and the like (Wolfgang 1958:190-99). In the contemporary United States, about 70 percent of criminal homicides appear to be- gin as conflicts (Block 1986:3-4; Maxfield 1989). In other countries, the percentage may well be higher (Home Office Criminal Sta- tistics 1993:81-82 [England and Wales]; Silverman and Kennedy 1993:55 [Canada]; Strang 1993: 14 [Australia]). Qualitative eth- nographic evidence reveals that the great ma- jority of killings in preindustrial societies are committed in the course of disputes over honor, sexual fidelity, prior homicides, and the like (e.g., Hasluck 1954; Koch 1974;
Table I. Four Forms of Lethal Conflict Killer
Victim IndividualiGrouD Age~~~?~olitv
Polity1 Agent of Polity Rebellion
Knauft 1990).6 At larger structural levels, conflict also dominates, although numerical estimates are again lacking. Political killing by and against states typically arises out of disputes over the imposition of and resis- tance to state authority (Rummel 1994). Similarly, most warfare, between tribes and modern nation-states alike, is preceded by nonlethal skirmishes such as diplomatic ne- gotiation, the seizure of hostages, the occu- pation of territory, and the assassination of political leaders. In both types of society, wars tend to arise between physically contiguous groups, a good indicator that prior conflict lies behind the outbreak of hostili- ties (Meggitt 1977:42; Bremer 1992). In short, the great majority of violence is rooted in conflict, regardless of the structural level at which it occurs.
Is the aggregate amount of lethal conflict found in human societies affected by the state? Anthropologists, criminologists, politi- cal scientkts, and others have conducted a substantial amount of research that helps to answer that question. However, their infor- mation is far from complete, particularly for stateless societies. Data on lethal conflict is available only for societies that have been studied by anthropologists who happen to be interested in the subject. Moreover, only in some of those cases are there reasonably ac- curate counts of the number of people killed. Many of the older ethnographies report their information in relatively vague terms (e.g., feuding is "frequent" or "very frequent"). Consequently, there are few numerical esti- mates of the victims of lethal conflict, and
"use the convention known as "the ethno- graphic present tense" to refer to stateless societ- ies as described in the anthropological literature. Most have changed substantially since they were first studied.
even those present some problems of inter- pretation.
Yet while the amount of lethal conflict that occurs in the typical stateless society cannot be quantified, the quality and quantity of the available information is sufficient to make the general comparisons across the three cat- egories of society required by the issue un- der discussion. The procedure I adopt is to support the argument using different types of data wherever possible. Thus, by drawing on synchronic and diachronic analyses, case studies and cross-cultural surveys, personal narratives and professional investigations, it is possible to piece together a coherent, con- sistent, and triangulated account of the rela- tionship between the state and lethal conflict.
In its strongest form, Hobbesian theory is easily refuted. Even a single stateless soci- ety with low rates of violence falsifies its ciaims. Two well-documented examples of stateless societies in which violent conflict is extremely rare are the Mbuti Pygmies (Turn- bull 1961, 1965, 1978) and the Semai (Robarchek 1977; Dentan 1978, 1979, 1988; Robarchek and Dentan 1987).7 Both societ- ies negate the idea that stateless societies are necessarily violent (also see Howell and Willis 1989).
A weaker version of Hobbesian theory is more resilient, however. In this version, the state is but one factor influencing killing. Rates of lethal conflict in stateless societies should, therefore, be higher on average than in state societies, even though particular stateless societies may seldom experience killing: "It may peradventure be thought, that there was never such a time, nor condition of warre as this; and I believe it was never gen- erally so, over all the world" (Hobbes [I6511 1909:77).~
Both the Mbuti (Zaire) and the Semai (Ma- laysia) are nominally incorporated into nation states, but the geographical isolation of both groups means that the state exercises no control over the way they handle conflict.
In places, Hobbes implies that the "state of nature" is a pre-social state not to be equated with actual stateless societies. But elsewhere (for ex- ample, in the passage cited above) he suggests otherwise (Wrong 1994, chaps. 2 and 4).
From a conflict management perspective this weaker version of Hobbesianism is con- siderably more plausible because violence is only one way of pursuing a grievance in the absence of law. Although a dearth of precise statistical data precludes a multivariate test, weak Hobbesianism receives strong support from two comparative analyses: a synchronic comparison of the incidence of lethal conflict in selected stateless and state settings, and a diachronic comparison of the effect on lethal conflict of the introduction of the state.
Stateless and Noncentralized State Societies
Several synchronic cross-cultural studies re- port that as political institutions become more compiex and state-like, violence de- clines (Koch and Sodergen 1976; Masamura 1977; Rosenfeld and Messner 1991).9 Be- cause they are not based on firsthand obser- vational data, a criticism sometimes made of these studies is that they reflect more the widespread fear of violence than its actual occurrence (Gluckman 1956, chap. 1; Hoebel 1971; Moore 1972; Colson 1974:40-43).
More recent evidence indicates, however, that violence occurs at extremely high levels in at least some stateless societies. The evi- dence-indirect but comparatively hard-is of two kinds: violent mortality statistics and per capita rates of lethal conflict. Because anthropologists are likely to report informa- tion of this kind only when they are struck by how much lethal conflict there is, these sta- tistics may represent the most violent state- less societies. Even so, they are instructive.
Cross-cultural studies further suggest that the state reduces internal homicide gradually rather than suddenly. Otterbein and Otterbein (1965) and Otterbein (1968) show that feuding (i.e., blood revenge following a homicide within a po- litical community) and internal warfare (i.e., armed combat between political communities within the same culture) are compatible with the existence of a state in preindustrial societies. This makes sense: Historical evidence shows that states establish their authority slowly, particularly in preindustrial settings when states are relatively small and remote from many of the regions they purport to rule (Stone 1965, chap. 5; Tilly 1986: 17 1-72). Early states therefore experienced more citizen violence than do modern states (Giddens 1985; Mann 1993:lS).
Gebusi (New Guinea)
Achuara Jivaro (Ecuador-Peru)
Grand Valley Dani (New Guinea)
Mae Enga (New Guinea)
Huli (New Guinea)
Achuara Jivaro (Ecuador-Peru)
Period in Conflict Source
5 generations Yost (1981 :687)
1 940-1962 Knauft (1985:l 16)
Bennett Ross (1984:96)
Heider ([ 19791 1991 : 1 14)
Meggitt (1 977:l 10)
Bennett Ross (1984:96)
.' Not available. Based on "a mortality sample of over 250 relatives of Achuara informants" (Bennett Ross 1984:96). Not available. Based on genealogies of 350 deceased males and 201 deceased females. Not available. Based on "a sample of genealogies drawn in 1955-57 from [14 clans] with reference to the stated cause of death of men before 1950" (Meggitt 1977:109; also see p. 12).
* Not available. Based on genealogies of 409 male deaths and 360 female deaths.
Table 2 presents mortality statistics from war and homicide compiled by anthropolo- gists for several stateless societies. The soci- eties in Table 2 exhibit considerable diver- sity. They are based on both foraging and ag- ricultural economies, located in different re- gions of the globe, and have varying degrees of contact with state societies. Thus, at the time covered by the data, the Albanians (c. 1900) lived in a region dominated by states. The South Americans were more removed, but several were connected by trade to state societies (Ferguson 1995). The most remote were the New Guinea groups some of whom appear to have been almost entirely free of external influence (Knauft 1993). State con- tact is therefore unlikely to explain the vio- lence of stateless societies (cf. Ferguson and Whitehead 1992; Ferguson 1995).
The statistics in Table 2 may depict the ex- treme rather than the typical stateless society. However, a survey of the available archaeo- logical evidence reveals comparable levels of violence: In 14 societies widely separated in space and time, the percentage of violent deaths ranges from 2 to 41 percent (Keeley 1996: 197). Thus, even if the societies in Table 2 are unusual, they are by no means unique (also see Keeley 1996: 195-97).
By contrast, violent mortality statistics in modern democratic state societies are much lower. Even in the United States, where criminal homicide is considerably more fre- quent than in most industrialized countries (Reiss and Roth 1993:52-53), only 1 percent of deaths-are caused by homicide (calculated from U.S. Bureau of the Census 1995:94).
Comparisons of this kind are, however, somewhat misleading because the statistics for stateless societies embrace all violent deaths, including those incurred in warfare, while those for modern states include only people killed in criminal homicide (Lee 1979:397-99). But bringing war back in does not change the overall picture (Harris 1993: 299).1°
Table 3 compares the annualized rate of death from war and homicide per 100,000 people for several pre-state societies for
I0 The most lethal international conflict of the twentieth century was World War 11, but in that war just "9 percent of the USSR's and 5 percent of Germany's population were killed, while for England and France it was only 1 percent, and for the United States an infinitesimal 0.2 percent" (Livingstone 1968:5).
Selected Stateless Societies and Democratic State Societies
Gebusi (New Guinea)
Omu (New Guinea)
Goilala (New Guinea)
Auyana (New Guinea)
Detnocratic State Societies
Period per 100,000
1940- 1962 683 1896-1946 620 1896-1 946 533 1924- 1949 420 1906- 1926 330 1970-1974 166 1920-1955 42
1900-1 990 58 1900-1 990 26 1900- 1990 13
Hallpike (1977: 120)
Robbins (1982: 21 1)
Warner ([I9371 1958: 157-58)
Melancon (1982:33, 42)"
Lee (1979:382, 397)
Talculated by Knauft (1987:464). Because the Yanomamo data cover only four years, interpretive cau- tion should be exercised. However, a high incidence of violence among the Yanomamo has been docu- mented independently by Chagnon (1988) and Lizot (1994).
For the three modern democracies, criminal homicide rates are taken from Reiss and Roth (1993:52). War homicide figures for the United States come from the U.S. Bureau of the Census (1995:366) (U.S. total = 426,000). For Britain, war data are from Clodfelter (1992:558, 590, 625-26, 627, 628, 629, 631, 632, 637, 638, 641, 784, 957, 981, 1012, 1040, 1086, 1124, 1131-32, 1193) (Britain total = 1,118,950). Likewise, for France, war data are from Clodfelter (1992:629-30, 634-35, 640, 619, 783, 955, 997, 1012, 1013, 1122) (France total = 2,105,180). In all three cases, calculation of the rate is based on the 1950 population (to provide a few years to recover from the effects of the war) as provided by U.S. Bureau of the Census (151 million) (1995:8), and estimated from Mitchell (1992:4, 8) for France (41 million) and Britain (49 million).
which it can be calculated, with the annual- ized combined criminal homicide and war death rate of three twentieth-century demo- cratic states. The numbers for the three de- mocracies are approximations because infor- mation on the number of civilians killed in World War I is incomplete. Moreover, miss- ing data for the early part of the century means that an average criminal homicide rate cannot be calculated for the entire period. Although criminal homicide has increased in the United States (and probably in Britain and France as well) since the early 1960s (Zahn 1989:219), the contemporary rate is used so as not to bias the results in favor of the argument. Note that the democratic rate of lethal conflict does not include (1) "justi- fiable homicide- (i,e,, killing deemed to be noncriminal, such as many police-civilian homicides) or (2) killing occurring in the
course of Neither figure be likely to have much On the rate, however. For instance, in 1994, there were 816 justifiable homicides known to po-
lice in the United States (Federal Bureau of Investigation 1995:22). Round this figure up to 1,000, double it to correct for under- reporting, and the homicide rate still in- creases by only .8 per 100,000.
Although Table 3 contains information on only three modern democracies, states at or near the top of the world political order, such as the United States, Britain, and France dur- ing the twentieth century, are typically the ones most involved in war, and experience the greatest number of war casualties (Singer and Small 1972:287). If these three states are atypical, then it is probably because they suf- fered more rather than fewer war deaths than other democracies. "
I' Because Table 3 is based on the number of people a state loses in war (the measure most rel- evant to the issue addressed) it does not directly refute idea that states externalize violence,
However, the greater number of violent deaths generally experienced by leading powers calls his theory into question. Moreover, Lee (1979:398) cites Vietnam as an example of a country that suf-
The high level of lethal conflict for the stateless societies in Table 3 does not appear to be a product of temporary and exceptional circumstances (although longer observation periods would be helpful).12 Among the Gebusi, for example, most lethal conflict consists of the killing of people suspected to have killed others through sorcery. This is an ingrained cultural pattern that has endured for at least 40 years (it has survived the ar- rival of the state).
The absence of Western medical knowl- edge and facilities does not seem to explain the difference in lethal conflict between the stateless societies and state societies in Table 3 either, and for two reasons. First, the ac- companying ethnographic evidence reveals that most of these stateless societies experi- ence such elevated levels of aggression that their rates of lethal conflict would surely be high even if they did have scientific medical care. Second, the more advanced medical re- sources of modern state societies is offset to some degree by their more sophisticated (and lethal) weapons technology. To assess the relative impact of these contrary effects is difficult. For lethal conflict involving indi-
fered from American externalization. But includ- ing the North and South Vietnamese casualties does not generate a lethal conflict rate anywhere close to those of the stateless societies in Table 3 (except for Lee's example, the Jul'hoansi). Ap- proximately 1.55 million Vietnamese died in the war involving the United States, 1965-1975 (Clodfelter 1992: 1323). If, as Lee recommends, these casualties are added to those of the United States (58,000) and if the combined population of Vietnam (39 million; Europa Year Book 1971: 1664, 1670) and the United States in 1970 (203 million; U.S. Bureau of the Census 1995:8) is used as a base, the annual rate of war deaths is still only about 62 per 100,000 for the years
1965-1975. Widen the time frame and the rate drops even lower.
I2 The observation period for the state societ- ies in Table 3 is 90 years (on the theory that the longer the period, the more reliable it is); for the stateless societies it averages (mean and median) approximately 30 years. However, this difference does not appear to bias the comparison. Consider the most bloody 30 (or so) years for any of the three states: France, 1914-1945. During that pe- riod, France had a rate per 100,000 of approxi- mately 155, a figure that does not change its rank- ing as the fourth least lethal society listed in Table 3.
viduals or small groups, such as most mod- ern criminal homicides and raids in preindustrial societies, the medical effect may be more important because of the small numbers involved. In addition, although there is some evidence that the introduction of more sophisticated weaponry into preindustrial settings results, at least under some conditions, in elevated levels of lethal conflict (Vayda 1976, chap. 4; Bennet Ross 1984:92-93), guns may not have much im- pact on modern homicide rates (Kleck 1991). As regards war between large armies, how- ever, weapons may be more important. Most of the British and French lethal conflict rates in Table 3 are made up by war killing, a good proportion of which is surely attributable to the extremely lethal weapons used in twenti- eth-century wars.
The data in Table 3, then, are influenced by the technology found in the two types of society. But whatever its precise effects, technology clearly cuts bvth ways, both in- flating and depressing the comparative le- thality of stateless societies. It is therefore highly probable that the great difference in rates of lethal conflict evident in Table 3 is due, at least in part, to underlying differences in the amount of aggressive social control found in these societies. Thus, the conclusion to be drawn from Table 3 seems clear: The more violent stateless societies have extremely high lethal conflict as compared to the more violent democracies, even when the destructiveness of modern war is taken into account.
Stateless Societies Generally
Are the stateless societies in Tables 2 and 3 exceptional? Unfortunately, incomplete data preclude a definitive answer. The available evidence, however, suggests that the typical stateless society experiences comparatively high levels of violent conflict. For example, although neither mortality percentages nor lethal conflict rates can be calculated, some other groups are known to have been ex- tremely warlike over sustained periods of time. Thus, violence looms large in certain stateless societies in New Guinea (Harrison 1993; Pospisil 1994), Amazonia (Murphy 1957; Harner 1972;), Australia (Warner [I9371 1958; Hart and Pilling 1979), and
North America (Nabokov 1967; Jorgensen 1980, chap. 9).
At the opposite end of the spectrum, highly peaceful stateless societies appear to be rela- tively rare. In a preliminary analysis, Fabbro (1978) identified seven societies with little or no violence, internal or external. But more intensive investigation has revealed that one of these-the Jut'hoansi-in fact has a non- negligible rate of homicide (Table 3). And a second case is questionable, as one observer suggested that homicide is relatively frequent among the Copper Eskimos (Fabbro 1978:75). In general, current scholarship views lethal conflict, offensive or defensive, as a feature of most stateless societies (e.g., Gregor 1990: 106-107). After examining 186 mostly preindustrial societies, Ember and Ember (1992), for instance, argue that ". . . war is a nearly universal fact of life in the ethnographic (anthropological) record" (p. 242). Other cross-cultural studies, based on samples of 50 or more societies, reach the same conclusion, typically showing that about 90 percent of preindustrial societies (most of which are stateless) experienced violent conflict, internal or external, when they were first described (Ross 1983:179, 182-83; Otterbein 1989: 143-49). Moreover, during war preindustrial societies tend to mobilize a higher percentage of their popu- lations and deploy as combatants a higher percentage of those mobilized than do mod- ern state societies (Keeley 1996:33-36). Fac- tors such as these lead the most comprehen- sive reviewer of the evidence to conclude that "a member of a typical tribal society, es- pecially a male, had a far higher probability of dying 'by the sword' than a citizen of an average modern state" (Keeley 1996:93).
The Advent of the State
The synchronic material just considered, al- though helpful, suffers from the limitation that the state itself represents only one of many differences between stateless and state societies that might affect the incidence of lethal conflict. For instance, the citizens of modern democracies are vastly more func- tionally interdependent than are their coun- terparts in stateless societies. Several theo- rists have argued that longer and more elabo- rate chains of interdependency reduce vio- lence (Durkheim [I 8931 1964, book l; Elias [I9391 1982; Black, 1990:47). Thus, crimi- nal homicide shows a long-term decline as societies develop a modern division of labor (Chesnais 1981). Although the English state, for example, was well-established by the fourteenth century, English homicide rates today are about 10 times lower than they were 600 years ago (Gurr 1981). Perhaps, then, modern rates of lethal conflict are com- paratively low solely because of factors like increased interdependence.
Diachronic anthropological evidence on lethal violence before and after the advent of the state suggests, however, that some por- tion of the reduced rates of violence exueri- enced by modern democracies shoulh be credited to the state (Tilly 1990:68-70). A theme in many ethnographies has been the systematic suppression and reduction of le- thal conflict following the arrival of a state (e.g., Middleton 1965:48; Harner 1972:210; Sorenson 1972; Hart and Pilling 1979:83, note 3; Rodman and Cooper 1979; Heider [I9791 1991:96, 160; Boehm 1984:6-7; Chagnon 1988). For example, after the ad- vent of the colonial Australian administra- tion, the Fore of Highland New Guinea in- creasingly brought their conflicts before pa- trol officers for settlement, and "an antifighting ethic quickly spread throughout the region" (Sorenson 1972:362).
Coming under the jurisdiction of a state is, however, only one of a series of changes that a stateless society typically undergoe-s when it establishes contact with a more technologi- cally developed society. So while the reduc- tion in lethal conflict cannot usually be at- tributed to developments internal to the soci- ety itself, it may be the result of external changes brought about by contact other than the arrival of the state (e.g., access to West- ern medicine). Here the case of the Gebusi of New Guinea is particularly valuable be- cause it presents an unusually pure case of state impisition. Except for a brief period of wage labor, virtually the only contact Gebusi have had with Western culture has been with government patrols. At the time Knauft (1985:12-16) conducted his research, their subsistence, settlement, and cultural patterns remained largely unchanged by contact- they had no specialization, stores, missions, or clinics, spoke only their own language, and did not migrate from their communities to seek employment. Even so:
The rate of homicide has decreased gradually since contact, from a high of 39.0 percent (971 249) of all adult deaths in the pre-contact era
(c. 1940-1962) to 23.3 percent (241103) during the period of Australian administration (1963- 1975), and to 19.0 percent (8142) under the na- tional administration of Papua New Guinea (1975-1982). The possibility of a five- to ten- year prison term for willful murder is clearly recognized by Gebusi as a cause of this decline, though the beliefs and motives that underlie homicide are unaffected by outside cultural in- fluences. (Knauft 1985: 116)
Note that the state's presence did not just re- duce the Gebusi homicide rates in the short term, but did so for at least 20 years (i.e., un- til Knauft finished his field work in 1982). Note further that despite the introduction of the state, Gebusi homicide rates remained ex- tremely high, thereby supporting the weaker version of Hobbesianism in which the state is but one factor influencing lethal conflict.
But even in its weaker form the Hobbesian argument is not devoid of difficulty. There are at least four reasons for thinking that the advent of the state does not reduce, and may even increase, rates of lethal conflict.
First, people sometimes resist the state, preferring to prosecute their conflicts vio- lently. In New Guinea, for instance, the co- lonial Australian government's initial success in pacifying many warring tribal groups proved to be ephemeral (Strathern 1992). Other illustrations of intractable violence in- clude fighting between rural Mexican fac- tions (Greenberg 1989), urban American gangs (Shakur 1993), and political violence in various parts of the world (van Creveld
1991; Tilly 1995).
Because the state is but one relevant fac- tor, the existence of violent conflict within state societies is, however, inconsistent with the weak version of Hobbesianism only if the scale of killing is greater than it would have been in the absence of the state, a possible but improbable scenario. States can usually limit violent conflict even when they cannot eliminate it.
Second, contact with Western societies can
AMERICAN SOCIOLOGICAL REVIEW
and often does have catastrophic effects on stateless people, effects that include outright slaughter. Throughout the globe, there are many examples of the devastation and even total elimination of indigenous groups fol- lowing contact with technologically more developed societies (Bodley 1990, chap. 3).13
It is indisputable that contact with Western culture has often destroyed stateless peoples, especially those occupying territory coveted by newcomers who engoy the protection of a state. However, once the issue of land own- ership is settled, the state typically exerts its authority by prohibiting violence and provid- ing alternative methods of dispute resolution, measures that usually result in less killing. In other words, the increased level of lethal conflict that stateless people experience in these circumstances appears to be attribut- able to the conflict over land or other mate- rial resources rather than the coming of the state. But should evidence emerge that the arrival of the state caused lethal conflict to increase, even when land or resources were not being sought, this argument would be se- riously challenged (provided the state was not highly centralized).
Third, contact with Western culture can sharply increase interpersonal homicide. For instance, Thomas (1994:78-79) recorded only 6 homicides among the Jul'hoansi group she studied during the hunting-gather- ing period. But in the 20 years since the group has attached itself to the fringes of sedentary communities, at least 20 of the 200 people she had known have become victims of lethal violence, usually inflicted in the course of drunken arguments.
Although the increased incidence of Jul 'hoansi lethal conflict occurred after contact with a state society, several other factors ap- pear to be responsible for the increase: (a) Adopting a sedentary life can increase vio- lence among former hunter-gatherers because it undermines one of their principal peace- keeping mechanisms-the ability of dispu-
tants to move away from or to avoid one an-
'" related point is that the advent of a state may intensify the warfare of neighboring people not under its direct control by, for example, de- pleting resources or increasing inequality based on access to new goods (Bennett Ross 1984; Ferguson and Whitehead 1992; Ferguson 1995).
other (Black 1990:49-53). (b) Proletarianiza- tion-their transformation into "the poorest of the poor" (Thomas 1994:79)-means that the Jul'hoansi have become a low status mi- nority group within a larger state society, a structural position commonly associated with elevated rates of homicide (Reiss and Roth 1993:69-72; Silverman and Kennedy 1993, chap. 8; Strang 1993:25,3 1). (c) Cooperation and interdependence have given way to in- creased individualism (e.g., hoarding resources) and dependence of women (e.g., only men can find jobs), factors also known to raise levels of violent conflict (Black 1990:43-49).
Fourth, several writers have speculated that the advent of the state may weaken indi- genous forms of social control and thereby increase violence (Kluckhohn 1944:94-95; Stirling 1960:73-75; Black 1989:80). They argue that what was previously held in check by the menace of violent retaliation is subse- quently only weakly deterred by the threat of a prison sentence, usually of short duration. Consequently, where indigenous rates of ho- micide are low (i.e., lower than the Ju'lhoansi rate), the coming of the state might cause them to rise.
Although plausible, this argument is not supported by any firsthand evidence.I4 Were supporting evidence to emerge it would qualify the weak Hobbesian thesis by sug- gesting that the state has mixed effects: re- ducing homicide rates when they are high and raising them when they are low.
On balance, then, the arguments against the weak Hobbesian argument fail. Some fail for a lack of data. Others fail because factors apart from the state (e.g., proletarianization, resistance to the expropriation of land) seem to lie behind increased levels of lethal con- flict following contact with the state. Should new findings emerge to alter these conclu- sions, however, the weak Hobbesian thesis would have to be narrowed or abandoned.
A final set of supporting data is the distribu- tion of homicide within modern state societ-
l4 However, there is evidence that the incidence of rape increased after the introduction of colonial rule among the Gusii of Kenya (Le Vine 1959).
ies. Black (1983) points out that the amount of violence, including homicide, among dif- ferent groups tends to be inversely related to the amount of law found among those groups. Black's (1976) earlier work proposed a theory of the social distribution of law. Cit- ing a large body of cross-cultural informa- tion, Black showed that law tends to be found among strangers, the wealthy, the in- tegrated, the conventional, and the respect- able, as well as among members of ethnic and racial majorities. Conversely, the poor, intimates, marginals, the unconventional, the unrespectable, and members of ethnic and ra- cial minorities tend, to occupy what he termed "stateless locations" within modern state societies. (To distinguish these from ac- tual stateless societies they might be called "virtual stateless locations".) The character- istics of these virtual stateless locations are that people use law relatively rarely among themselves, have hostile relationships with legal authorities (e.g., are often reluctant to provide information or testimony to legal of- ficials), and receive relatively little assistance or satisfaction for their problems when they do invoke the legal system (Shakur 1993; Canada 1995). Alienated from law, virtually stateless people are more likely to resort to violence to settle their disputes.
Homicide data support this idea that mod- ern people kill when the state is weak or dis- tant. In all modern societies for which there is infdrmation, homicide is strongly concen- trated among the poor, unemployed, members of cultural minorities, intimates and ac- quaintances, young people in their late teens and early twenties, and those with criminal records (Cooney forthcoming a).I5
To summarize the argument thus far: Within modern state societies, homicide rates are typically highest among people occupy- ing virtual stateless locations. The cross-cul- tural evidence suggests that actual stateless societies generally have elevated levels of le- thal conflict. Democratic state societies, by
l5 Gender appears to be an exception. Because men are generally wealthier and more indepen- dent of informal social control than are women, law is more readily available to them (Black 1976, chaps. 2 and 6). Even so, men are more likely to kill and be killed (Daly and Wilson 1988, chaps. 6 and 7).
contrast, tend to generate relatively little le- thal conflict, even when warfare is taken into account. However, states in which political authority is strongly developed are in an al- together different category.
In centralized political systems, lethal con- flict is once again frequent but for reasons that are typically distinct from those prevail- ing in stateless societies. Whereas the ab- sence of government means that people have to solve disputes on their own, centralized government makes the state a party, actual or potential, to every conflict. If the former tends to produce "warre," the latter veers to- ward tyranny. Tyranny manifests itself, above all, in large-scale executions. But it also creates more war fatalities and about as much rebellious homicide as noncentralized states do.
Although the twentieth century may be the most lethal of recent centuries (Tilly 1990:70-76), it holds no monopoly on state killing. The historical record abounds with accounts of large-scale slaughter-from the Zulu (Walter 1969, chaps. 6-10) and Asante (Wilks 1975) kingdoms of Africa to the Az- tec empire of Central America (Davies 198 1, chap. 9), from the Mongol (Saunders 1971) and Chinese (Rummel 1994:51-54) empires in the East to the Spanish Inquisition in the West (O'Brien 1973). However, the recent information, for all its imperfections, is in- comparably better than that for earlier peri- ods. Hence, in comparing execution, war, and rebellion in centralized and noncentral- ized regimes, I concentrate here on the twen- tieth century.
Available data suggest that rates of interper- sonal homicide are relatively low in central- ized states (Archer and Gartner 1984, part 3; Reiss and Roth 1993:52). However, the ho- micide rates at least of socialist countries do not appear to be any lower than those of democratic countries at the same level of so- cioeconomic development. In other words, state centralization does not seem to pur- chase any additional reduction in criminal homicide.
AMERICAN SOCIOLOGICAL REVIEW
More certain is the fact that highly central- ized states commonly exhibit high rates of state-citizen killing (execution). Black (1993, chap. 8), building on earlier theoretical for- mulations found in Durkheim (1 899-1900), Wittfogel (1957), and his own writings (Black 1976:92-97, 1990:47-49), proposes that inequality-including state centraliza- tion-increases the tendency to create and punish enemies (also see Rummel 1994, 1995; cf. Tilly 1995: 167).
The evidence supports this proposition. Because centralized states are highly suspi- cious of their populations, everybody is vul- nerable to denunciation by his or her enemies (Gross 1984). The result is a high rate of ex- ecutions, often for political crimes for which there is little evidence of actual wrongdoing (Beck and Godin 1951; Bergesen 1977).16 But precisely how high the rates are is diffi- cult to establish. The political scientist, R. J. Rummel (1990, 199 1, 1992, 1994) estimates the number of people killed by governments in the twentieth century (actually from 1900 to 1987). He notes that even for the best- studied case-the Nazi genocide of the Jews-estimates of the numbers killed given in five of the most thorough studies vary by as much as 40 percent:
This is for a genocide carefully administered by a regime that was better than most about keeping records and statistics, whose surviving archives were completely available after the Nazi defeat, and about which there has been for nearly half a century many historians dedicated to uncovering the truth. If then the estimate of the Jewish Holocaust can vary so much, we should hardly expect to get the true figure on other genocides or mass murder. (Rummel 1992:5-6)
Rummel reviews all the available data be- fore "narrowing the range of estimates to what a hypothetical, reasonable analyst would arrive at from the available informa-
'"he cross-cultural literature also provides some supporting evidence for the link between centralization and execution, although typically it does not yield estimates of the number of people killed. Thus, in a study of 53 societies, Otterbein (1986:94) found that despotic states, in which the leader alone makes decisions to execute, have more capital offenses than do polities in which decisions are shared between the political leader and other groups (e.g, victim's kin).
tion, and then defining within this range a prudent figure that somewhat reflects the cen- tral thrust of the statistics and historical events" (1992:6). On this basis, Rummel cal- culates that the German government killed over 20 million European citizens during the years 1939 to 1945. This translates into an annual lethal violence rate of 1,008 per 100,000. For the Japanese occupation of China, Korea, Indonesia, Burma, and else- where in Asia in the years 1937 to 1945 he estimates the lethal violence rate as 999 per 100,000 (Rummel 1992:20-21). For Cambo- dia under the Khmer Rouge regime (1975- 1979) his estimate is 8,160 per 100,000 (Rummel 1994: 194).17 These rates are calcu- lated over short periods of time and hence are inflated. But his estimate for Communist China of 120 per 100,000 is based on 39 years (1949-1987) (Rummel 1991, chap. 8). And, most striking of all, his estimate for the So- viet Union of 450 per 100,000 covers 71 years (1917-1987) (Rummel 1990, chap. l).I8
l7 Unlike the comparison between stateless so- cieties and noncentralized state societies, that be- tween noncentralized states and centralized states is not strongly affected by differences in medical resources. In highly centralized regimes, medical care. along with adequate food, clothing, and shelter, are often deliberately withheld from en- emies of the state. For example, in Cambodia un- der the Khmer Rouge government, for a doctor to practice medicine among the general population or even to admit that he or she had done so under the previous regime was to invite almost certain death, most likely after prolonged torture (Ngor 1987). Indeed, in some cases, the best-known of which is the Nazi genocide of the Jews, central- ized states have employed medical knowledge and personnel to facilitate their mass killing (Lifton 1986). The absence of medical care for victims of homicide under modern centralized re- gimes should be seen, therefore, not so much a confounding factor as an indicator of the violent inclinations of the state.
l8 It must be emphasized that the statistics on which Rummel relies are inexact. Thus, some writers, drawing on newly available material, ar- gue that the scale of state repression and killing under the Soviet regime has been considerably exaggerated (cf. Conquest 1991; Wheatcroft 1992; Nove 1993, 1994). Even if this criticism is correct, however, it is clear that highly central- ized societies, of which the Soviet Union is but one example, are characterized by rates of state killing that greatly exceed those found in more democratic regimes.
Table 4. Average Annual Rate per 100,000 Population of People Killed by the State by Type of Regime, 1900 to 1987
Type of Regime Rate per 100,000
Note: These figures (and those in Table 5) are based on the number of years a regime fell into a particular regime category. The rate for "Commu- nist," therefore, includes the enormous amount of slaughter perpetrated by short-lived Communist re- gimes, such as that in Cambodia from 1975 to 1979.
Source: Rummel (1994:15).
Rummel (1994) argues that the central fac- tor in explaining state killings is the degree of unrestrained power, or centralization, of the state. Drawing on classifications used in political science, Rummel divides regimes into four categories of increasing "regime power": democratic, authoritarian, totalitar- ian, and Communist. Table 4 displays his es- timates of the rate of killing engaged in by these four types of regime in the first 88 years of the twentieth century.
In a subsequent multivariate analysis, Rummel (1995) reports that even after con- trolling for various measures of a society's social diversity, culture, and socioeconomic development, the best single predictor of do- mestic governmental mass killing (which he calls "democide") in 214 state regimes from 1900 to 1987 was the centralization of po- litical authority (as measured by the concen- tration of political decision-making and the extent of state control over social institu- tions). Democide is rare in democracies, fre- quent in semi-centralized regimes, and ex- tremely common in centralized states.19 In
l9 Rummel (1995) also found that the nature of the regime does not explain external democide- the amount of governmental killing of foreign citizens. He argues, however, that this too sup- ports the hypothesis because external democide is strongly correlated with war, and during war- time, the military tends to be given strong author- ity. Even in otherwise democratic societies, then, the military will often form an island of totali- tarianism during wartime.
short, the more centralized the regime, the more democide it commits (Harff 1988, as cited in Fein [I9901 1993:40).
Centralized states are more violent toward their own citizens than are democratic states, but it is possible that democracies lose more people in war so that in the end, their citi- zens may be just as likely to die violently as those in centralized states.
One reason for believing that, for democ- racies, deaths in warfare do not outweigh ex- ecutions in centralized states is that, at least in the twentieth century, the killing of citi- zens by states has claimed vastly more lives than the killing of soldiers in combat. Rummel's estimates include the number of people killed from 1900 to 1987 in both war- fare and political violence. His results are ap- proximations, but they yield a very clear pat- tern: Even excluding the estimated 39 mil- lion civilians of other countries that govern- ments have killed, almost four times as many people have died at the hands of their own governments than have been killed in battle combat (129 million versus 34 million) (Rummel 1994: 15). Noncentralized states would therefore have to be substantially more warlike than centralized states to com- pensate for the greater internal violence of centralized states.20 In fact, as Rummel's data also show (Table 5), centralized regimes tend to suffer more war casualties than do their less centralized counterparts.
Data on the number of casualties arising from rebellions against governments are not
*" The research literature typically frames the issue in terms of how warlike democratic and non- democratic nations are. One of the strongest find- ings is that democratic nations rarely fight one another (Small and Singer 1976; Ember, Ember, and Russett 1992:573). But it is unclear whether democratic nations are less warlike in general. Most writers seem to think they are not-that democracies are just as warlike overall as nondemocracies (Small and Singer 1976; Bueno de Mesquita and Lalman 1992:152-53). Others find, however, that democracies are less likely to go to war (Bremer 1992; Schweller 1992).
AMERICAN SOCIOLOGICAL REVIEW
Table 5. Approximate Average Percentage of
Population Killed in International
Wars, by Type of Regime, 1900 to 1987
Average Type of Regime Percentage Killed
.33 Totalitarian .64 Communist .53
Note: See note to Table 4.
Source: Rummel (1 994:15).
as plentiful as those on executions by gov- ernments. What information is available points toward two unsurprising conclusions: In general, (1) killings due to rebellions are considerably less frequent than killings due to executions, and (2) killings due to rebel- lions are more frequent in nondemocratic than in democratic states.
The most direct empirical evidence on the comparative frequency of executions and re- bellions comes from a large cross-national study of lethal political conflict (e.g., gue- rilla warfare, coups d'Ctat) in 50 countries during the 1960s. The study found that 71 percent of casualties were rebels, while 29 percent were soldiers and police. Thus, for every 100 people killed by soldiers and po- lice, only 41 soldiers and police were killed by rebels (Gurr and Lichbach 1979:155, in- cluding note 5).2'
" Using The New York Tirnes Index and other sources, Taylor and Jodice (1983) code the num- ber of victims of executions and assassinations worldwide from 1948 through 1977. For their purposes, a "political execution is an event in which a person or group is put to death under or- ders of the national authorities while in their cus- tody" (p. 63), while a "political assassination is a politically motivated murder of a national leader, a high government official, or a politician" (p. 43). Neither concept fully describes all relevant political killing and hence must be interpreted with caution. Nevertheless, it is of some interest that Taylor and Jodice (1983:75, 46) report a to- tal of 1.69 million executions compared to 2,866 assassinations. Most of the executions come from two extreme cases, China 1948-1952 (1.63 mil- lion executions) and Poland 1953-1957 (3 1,000 executions). But even excluding these outliers, executions (27,973) are still almost 10 times more numerous than are assassinations.
In the United States today, most lethal con- flict involving the state and its citizens does not arise from collective uolitical conflict but from individual crime. Even so, the same gen- eral pattern is evident. Although there are no reliable national statistics on the number of citizens U.S. police kill annually, researchers have used a variety of data sources to estab- lish that uolice kill at least three times more citizens every year than citizens kill police (Milton et al. 1977:32-34; Fyfe 1988: 174- 80; Peterson and Bailey 1988:209).
On the issue of which states experience most rebellions. research reveals that levels of lethal rebellion are generally low in demo- cratic states. Killing in the course of rebel- lions seems to be most common in states with intermediate degrees of centralization. Thus, there appears to be a curvilinear relationship between state authority and lethal rebellions: Rates of lethal rebellion are low in demo- cratic states (in which other forms of politi- cal opposition are available), and low again in the most centralized states (in which oppo- sition is severely repressed); between these extremes, lethal rebellion is more pronounced (Muller 1985; Boswell and Dixon 1990).
The evidence on lethal rebellions is con- sistent, then, with the view that democratic states experience less lethal conflict than those with a high degree of centralization.
Stateless and Centralized Societies
My argument here is that high rates of lethal conflict are found when state authority is ab- sent and also when it is extremely strong. This does not imply that the rates of the most homicidal centralized states will be the same as those observed in the most homicide- prone stateless societies. Nevertheless, a comparison of the two kinds of societies and their rates of lethal conflict is instructive.
First, though, I must reiterate a point made earlier. Rates of lethal conflict per capita are an extremely crude indicator of rates of con- flict across different types of societies be- cause, even when accurate, they also measure the effectiveness of medical care available to victims of violence and the sophistication of weaponry available to perpetrators. Thus, numerical data should be interpreted cau- tiously as indicating broad trends rather than exact patterns.
With that qualification in mind, consider the Soviet Union, one of the most homicidal states in the twentieth century. Approxi- mately 20 million Soviet citizens are be- lieved to have perished during World War I1 (Rummel 1990: 18 1). Combining this figure with the previously cited number of citizens executed by the Soviet state outside of war yields a total estimate for the Soviet Union of 81.5 million people killed in the period from 1917 to 1987. This represents a rate of approximately 620 per 100,000, comparable to the highest rate recorded in stateless soci- eties. Moreover, the Soviet rate emerges from a society with an average population of 184 million spread across many millions of square miles.
In contrast, high rates for stateless societ- ies occur among small populations located in diminutive territories- The Gebusi homicide rate, for example, is based on a population of just 450 people occupying 65 square miles (Knauft 1985:6). In a group of this size, a single homicide represents a rate of over 220 per 100,000. With populations varying so radically in size, standardized homicide rates per 100,000 are necessarily somewhat mis- leading. No doubt there are small towns and villages within the Soviet Union that experi- enced rates well in excess of 620 per 100,000. A better comparison, then, would be to com- pute the lethal conflict rate of the whole of stateless New Guinea and compare it to a modern centralized state, but these figures are not available.22 In these circumstances, the safest conclusion is that, at the extremes, cen- tralized state societies generate quantities of lethal social control that are at least as high as those found in the most violent stateless societies.
Lethal conflict in stateless and centralized state societies must also be seen in light of the overall amount of conflict management or social control that they produce. (Recall that for Black [I9901 these terms are interchange- able.) In general, stateless societies have relatively little social control. Foraging people handle their conflicts primarily through a mixture of negotiation, criticism,
22 Similarly, same-scale data might reduce the contrast in the number of violent deaths experi- enced by stateless societies and noncentralized states evident in Tables 2 and 3.
and avoidance (Balikci 1970, chap. 9; Lee 1979; Woodburn 1979). Stateless agricultural people behave similarly, except that their more sedentary lifestyle breeds less reliance on avoidance and more on gossip and third- party settlement (e.g., mediation) (Colson 1953; Gulliver 1963; Jones 1974). In both types of society, people may kill one another, perhaps even at high rates, but typically it is because of the absence of third parties who can interrupt episodes of violence once in motion. Stateless people do not arrest, im- prison, torture, and kill for political crimes that nobody has committed. In centralized states, by contrast, lethal social control is just the peak of the mountain, the most extreme manifestation of a whole social machinery of repression that includes secret police, in- formers, detention without legal representa- tion, tortured confessions, rigged trials, and labor camps (Solzhenitsyn 1973; Conquest 1990; Feng 199 1 ;Kiernan 1996). To summa- rize: In stateless societies people kill one an- other in the absence of other ways of han- dling conflict; centralized states, by compari- son, kill their citizens despite a vast array of alternatives.
Lethal social control, then, can exhibit two wholly different patterns. The first is discon- tinuous: Social control is either mild or le- thal with little found in between the extremes. The second is continuous: Social control exists at all gradations of severity such that the high volume of lethal violence is matched by a high volume of social con- trol of every kind (see also Knauft 1987).
In sum, then, while the most homicidal centralized states generate about the same amount of lethal social control as the most homicidal stateless societies, they generate vastly more social control in general.
Social theorists from Hobbes onward have often argued that in the absence of the state, life would be unendurably violent. Yet until now the empirical basis of that claim has not been subjected to systematic analysis. Viewed in the light of Black's (1976, 1993) theoreti- cal work on conflict management, the evi- dence currently available suggests several conclusions:
(1) Statelessness is compatible with low
AMERICAN SOCIOLOGICAL REVIEW
rates of lethal conflict. Contrary to a strong interpretation of Hobbes, in the absence of the state, life is not automatically nasty, brut- ish, and short because peaceful nonlegal forms of handling conflict may be highly de- veloped.
The existence of the state is compat- ible with high rates of lethal conflict. Life in state societies is not invariably peaceful, civi- lized, and long-other factors can foster the violent resolution of conflict.
The advent of the state, however, tends to reduce rates of lethal violence by provid- ing alternative mechanisms for resolving conflict. As a weaker version of Hobbesian theory predicts, citizens of states, at least of less centralized states, generally appear to have a lower risk of being killed in violent conflict, whether it takes the form of war, execution, rebellion, or homicide, than do their counterparts in stateless societies.
States with strong or centralized sys- tems of authority tend to have a high inci- dence of lethal conflict, especially killings of citizens by the state itself. Centralized states create enemies, internal and external, and are quick to resort to lethal violence, sometimes on a massive scale.
Lethal conflict is most frequent, then, when state authority is either absent or very strong. In stateless societies, conflicts be- come lethal because other forms of conflict management are unavailable; centralized states kill as part of a general regime of re- pression and aggression.23
Lethal conflict is least frequent in noncentralized states. Noncentralized states tend to be less violent than stateless societ- ies because they provide their citizens with additional peaceful means of handling con- flict; they are less violent than centralized states because they are not dominated by a
23 My argument should not be taken to imply any evolutionary trend in social control. As with other aspects of the state, centralization does not follow a single trajectory, but waxes and wanes over time, exhibiting cyclical rather than linear regularities (Goldstone 1991:40-44). Early mod- ern states were often more highly centralized than many twentieth-century states. Authority in the absolutist states of early modern Europe, for in- stance, was both highly concentrated and extremely punitive (Spierenberg 1984). The same is true of many agrarian empires (Wittfogel 1957).
central power that creates and punishes en- emies at home and abroad.
This entire argument-that the relationship between state authoritv and lethal conflict is U-shaped-is based on a limited number of societies, and must be taken as tentative and provisional. It is possible, for example, that the stateless societies studied by anthropolo- gists and archaeologists to date are not rep- resentative of stateless societies in general. Moreover, the noncentralized regimes ana- lyzed earlier are all twentieth-century de- mocracies, and hence the argument, even if valid, may be limited to that historically un- usual type of state. In addition, plausible theoretical arguments suggest that the non- centralized state may, at least on occasion, actually increase lethal conflict. For all these reasons, the side of the curve that is consis- tent with weak Hobbesian theory must be re- garded as especially tentative. But though it rests on stronger evidence, questions remain about the centralized side of the curve as well. For instance, some extremely violent states are not all that highly centralized. The Rwandan genocidal massacre of 1994, for example, occurred in a state that, while cer- tainly not democratic, was not a dictatorial, single-ideology regime, like, say, the Soviet Union in the 1930s (Prunier 1995). Thus, there are several ways in which the argument could be fal~ified.~"he claim that rates of lethal conflict are lowest in the least central- ized states, especially, provides a hypothesis for further investigation, not just of the types of polity considered here, but of the broader universe of regimes not addressed (e.g., an- cient empires, city-states, feudal monar~hies).~~
24 Testing would require that other variables relevant to explaining violent conflict be held constant (Black 1990).
2s One implication of my argument is that the most common criticism of anarchism-that it would inevitably increase violence-appears to be true only under some conditions (Taylor 1982). In highly centralized nations that are ex- periencing a wave of political executions, lethal conflict would surely decrease in the absence of the state. But in democratic polities, the sudden disappearance of the state among people long ac- customed to legal means of resolving disputes would create a void likely to be filled, at least in part, by violence (Taylor 1976:141). However, a gradual reduction in state authority might allow
From a conflict management perspective, the state is important primarily because it provides an elaborate system of third-party dispute settlement (law). But even when the state settles conflicts peacefully, it is never the only institution to do so. In all societies, individuals and groups (e.g., elders, neigh- bors, teachers, employers) can, and often do, settle the disputes of others without recourse or reference to the state (Black 1990:56-58; also see Ellickson 1991: 138-47). Thus, in re- ducing lethal conflict the critical factor may not be the state but the presence of some sys- tem of third-party settlement, legal or other wise (Cooney forthcoming b, chap. 2). Diverse though the state's many forms are, then, a fuller understanding of violence will necessitate viewing the state in a more ab- stract and generalized light.
Mark Cooney is Assistant Professor of Sociology and Adjunct Assistant Professor of Law at the University of Georgia. His prirnary irzterests are law, violence, and the relationship between them. He is currently completing a book on the role of third parties in interpersonal violence, provision- ally entitled Warriors and Peacemakers: How Third Parties Shape Violence (New York Univer- sity Press, forthcoming).
Archer, Dane and Rosemary Gartner. 1984. Violence and Crime in Cross-National Perspec- tive. New Haven, CT: Yale University Press.
Balikci, Asen. 1970. The Netsilik Eskirnos. New York: Natural History Press. Baumgartner, M. P. 1988. The Moral Order of a Suburb. New York: Oxford University Press.
, 1992. "War and Peace in Early Child- hood." Pp. 1-38 in Virginia Review of Sociol- ogy, vol. 1, Law and Conflict Management, edited by J. Tucker. Greenwich, CT: JAI.
, 1993. "Violent Networks: The Origins and Management of Domestic Conflict." Pp. 209-31 in Aggression arzd Violence: Social lnteractionist Perspectives, edited by R. B. Felson and 9. T. Tedeschi. Washington, DC: American Psychological Association.
Beck, F, and W. Godin. 195 1. Russiarz Purge and the Extraction of Confessiorz. London, En- gland: Hurst and Blackett.
Bennett Ross, Jane. 1984. "Effect of Contact on
people to develop alternative nonviolent ways of handling conflict, such as negotiation, mediation, arbitration, avoidance, and toleration (Black 1989, chap. 5; Baumgartner 1992:29-31).
Revenge Hostilities among the Achuara Jivaro." Pp. 83-109 in Warfare, Culture, and Environment, edited by R. B. Ferguson. Or- lando, FL: Academic.
Bergesen, Albert. 1977. "Political Witch-Hunts: The Sacred and the Subversive in Cross-Na- tional Perspective." American Sociological Re- view 42:220-33.
Black, Donald. 1976. The Behavior of Law. New York: Academic. , 1983,"Crime as Social Control." American Sociological Review 48:34-45.
, ed. 1984. Toward a General Theory of Social Control. 2 vols. Orlando, FL: Academic. , 1989. Sociological Justice. New York:
Oxford University Press.
, 1990. "The Elementary Forms of Con- flict Management." Pp. 43-69 in New Direc- tions in the Study of Justice, Law, and Social Control, prepared by the School of Justice Studies, Arizona State University. New York: Plenum.
, 1993. The Social Structure of Right and Wrong. San Diego, CA: Academic. , 1995. "The Epistemology of Pure Sociol- ogy." Law and Social Inquiry 20:829-70.
Block, Carolyn Rebecca. 1986. Homicide in Chi- cago: Aggregate and Titne Series Perspectives on Victitn, Offender, and Circumstances (1965-1981). Chicago, IL: Center for Urban Policy, Loyola University of Chicago.
Bodley, John H. 1990. Victitns of Progress. 3d ed. Mountain View, CA: Mayfield.
Boehm. Christopher. 1984. Blood Revenge: The Enactment and Matzagernent of Conflict in Montenegro and Other Tribal Societies. Philadelphia, PA: University of Pennsylvania Press.
Borg, Marian. 1992. "Conflict Management in the Modern World-System." Sociological Forurn 7:261-82.
Boswell, Terry and William J. Dixon. 1990. "De- pendency and Rebellion: a Cross-National Analysis." American Sociological Review 55540-59.
Bremer, Stuart A. 1992. "Dangerous Dyads: Con- ditions Affecting the Likelihood of Interstate Wars, 18 16-1 965." Journal of Conflict Reso- l~ttiotl36:309-41.
Bueno de Mesquita, Bruce and David Lalman. 1992. Wclr atzcl Reason: Dotnestic and Interna- tiotzal Imperatil~es. New Haven, CT: Yale Uni- versity Press.
Cain, Maureen and Alan Hunt, eds. 1979. Marx and Engels otz Law. London, England: Aca- demic. Canada, Geoffrey. 1995. Fist Stick Knife Gun: A Persorial Histoty of Violence in Atnerica. Boston, MA: Beacon. Carneiro, Robert L. 1970. "A Theory of the Ori- gin of the State." Science 169:733-38. Chagnon, Napoleon A. 1988. "Life Histories,
Blood Revenge, and Warfare in a Tribal Popu- lation." Science 239:985-92.
Chesnais, Jean-Claude. 1981. Histoire de la Vio- lence en Occident cle 1800 ri Nos Jours (History of Violence in the West from 1800 to the Present). Paris, France: Robert Laffont.
Clodfelter, Michael. 1992. Warfare and Arrned Conflict: A Statistical Reference to Casualty and Other Figures, 1681-1991. Vol. 2, 1900- 1991. Jefferson, NC: McFarland.
Colson, Elizabeth. 1953. "Social Control and Vengeance in Plateau Tonga Society." Africa 23:199-212. . 1974. Tradition and Contract: Tlze Prob- lern of Order. Chicago, IL: Aldine.
Conquest, Robert. 1990. The Great Terror: A Re- assessment. New York: Oxford University Press. . 1991. "Excess Deaths and Camp Num- bers: Some Comments." Soviet Studies 43:949
Coon, Carleton. 1950. The Mourztairz of Giants: A Racial and Cultural St~ldy of the North Alba- nian Mountain Ghegs. Cambridge, MA: Pea- body Museum of American Archaeology and Ethnology, Harvard University.
Cooney, Mark. 1995. "Legal Sociology and the New Institutionalism." Studies it1 Law, Politics, and Society 15:85-101. .Forthcoming a. "The Decline of Elite Ho- micide." Critninology 35. . Forthcoming b. Warrlors aticl Peacemak- ers: How Third Parties Shape Violence. New York: New York University Press.
Daly, Martin and Margo Wilson. 1988. Hornicicle. New York: Aldine De Gruyter. Davies, Nigel. 1981. Hurnarl Sacrifice it1 History and Today. New York: William Morrow.
Dentan, Robert K. 1978. "Notes on Childhood in a Nonviolent Context: The Semai Case." Pp. 94-143 in Learning Notz-Aggressiotl: The Ex- perience of Non-Literate Societies, edited by
A. Montagu. New York: Oxford University Press.
, 1979. The Setnai: A NOTI-Violetzt People of Malaya. Fieldwork ed. New York: Holt, Rinehart, and Winston.
. 1988. Comment on "Reconsidering Vio- lence in Simple Human Societies" by Bruce M. Knauft. Cllrret~tAtztl~ropology29:625-29.
Durkheim, Emile. [I8931 1964. The Division of Labor it1 Society. Reprint, New York: Free Press.
, 1899-1900. "Two Laws of Penal Evolu- tion." Utliversity of Cirlcir~tlnti L~HJ
Elias, Norbert. [I 9391 1982. The Civiliiing Pro- cess: Power and Civility. Vol. 2. Reprint, New York: Pantheon.
Ellickson, Robert C. 1991. Order w~ithout Law: How Ne~ghbors Settle Displttes. Cambridge,
MA: Harvard University Press.
Ember, Carol R. and Melvin Ember. 1992. "Re- source Unpredictability, Mistrust, and War." Journal of Conflict Resolution 36:242-62.
Ember, Carol R., Melvin Ember, and Bruce Rus- sett. 1992. "Peace between Participatory Poli- ties: a Cross-Cultural Test of the 'Democracies Rarely Fight Each Other' Hypothesis." World Politics 44:573-99.
Europa Year Book. 1971. A World Survey. Vol. 2, Africa, the Americas, Asia, Australasia. Eon-don, England: Europa Publications.
Fabbro, David. 1978. "Peaceful Societies: An In- troduction." Journal of Peace Research 15:67
Federal Bureau of Investigation. 1995. Crime in the United States, 1994. Washington, DC: U.S. Department of Justice.
Fein, Helen. [I9901 1993. Genocide: A Sociologi- cal Perspective. Reprint, London, England: Sage.
Feng, Jicai. 1991. Voicesfrom the Whirlwind: An Oral History of the Chinese Cultural Revolu- tion. New York: Pantheon.
Ferguson, R. Brian. 1995. Yarzomami Warfare: A Political History. Santa Fe, NM: School of American Research Press.
Ferguson, R. Brian and Neil L. Whitehead, eds. 1992. War in the Tribal Zone: Expanding States and Indigenous Warfare. Santa Fe, NM: School of American Research Press.
Fyfe, James J. 1988. "Police Use of Deadly Force: Research and Reform." Justice Quarterly 5: 165-205.
Giddens, Anthony. 1985. A Cor~temporary Cri- tique of Historical Materialism. Vol. 2, The Nation-State and Violence. Berkeley, CA: Uni- versity of California Press.
Glasse, Robert M. 1968. Huli of Papua: A Cog- natic Descent System. Paris, France: Mouton. Gluckman, Max. 1956. Custom and Conflict in
Africa. Oxford, England: Basil Blackwell.
Goldstone, Jack A. 1991. Revolution and Rebel- lion in the Early Modern World. Berkeley, CA: University of California Press.
Greenberg, James B. 1989. Blood Ties: Life and Violence in Rural Mexico. Tucson, AZ: Uni- versity of Arizona Press.
Gregor, Thomas. 1990. "Uneasy Peace: Inter- tribal Relations in Brazil's Upper Xingu." Pp. 105-24 in The Anthropology of War, edited by
J. Haas. Cambridge, England: Cambridge Uni- versity Press.
Gross, Jan. 1984. "Social Control Under Totali- tarianism." Pp. 59-77 in Toward a General Theory of Social Control. Vol. 2, Selected Problems, edited by D. Black. Orlando, FL: Academic.
Gulliver, P. H. 1963. Social Control in an Afri- can Society: A Study of the Agricultural Masai of Northern Tanganyika. Boston, MA: Boston
Gurr, Ted Robert. 1981. "Historical Trends in Violent Crime: A Critical Review of the Evi- dence." Pp. 295-353 in Crime and Justice: An Annual Review of Research, vol. 3, edited by
M. Tonry and N. Morris. Chicago, IL: Univer- sity of Chicago Press.
. 1988. "War, Revolution, and the Growth of the Coercive State." Comparative Political Studies 21 :45-65.
Gurr, Ted Robert and Mark Irving Lichbach. 1979. "Forecasting Domestic Political Con- flict." Pp. 153-93 in To Augur Well: Early Warning Indicators in World Politics, edited by J. D. Singer and M. Wallace. Beverly Hills, CA: Sage.
Hage, Per and Frank Harary. 1983. Structural Models in Anthropology. Cambridge, England: Cambridge University Press.
Hallpike, C. R. 1977. Bloodshed and Vengeance in the Papuan Mountains: The Generation of Conflict in Tauade Society. Oxford, England: Clarendon Press.
Harff, Barbara. 1988. "State Perpetrators of Mass Political Murder Since 1945." Presented at the Conference on State Organized Terror: The Case of Violent Internal Repression, Novem- ber, East Lansing, MI.
Harner, Michael. 1972. The Jivaro: People of the Sacred Waterfalls. Garden City, NY: Anchor.
Harris, Marvin. 1993. Culture, People, Nature: An Introduction to General Anthropology. 6th ed. New York: Harper and Row.
Harrison, Simon. 1993. The Mask of War: Vio- lence, Ritual and the Self in Melanesia. Manchester, England: Manchester University Press.
Hart, C. W. M. and Arnold R. Pilling. 1979. The Tiwi of North Australiu. Fieldwork ed. New Ydrk: Holt, Rinehart, and Winston.
Hasluck, Margaret. 1954. The Unwritten Law in Albania. Cambridge, England: Cambridge Uni- versity Press.
Heider, Karl. [I9791 1991. Grand Valley Dani: Peaceful Warr~ors.2d ed. New York: Holt, Rinehart, and Winston.
Hobbes, Thomas. [I65 11 1909. Leviatharz. Reprint, Oxford, England: Clarendon Press.
Hoebel, E. Adamson. 1971. "Feud: Concept, Re- ality, and Method in the Study of Primitive Law." Pp. 500-13 in Essays on Moderrlizatiorl of Underdeveloperl Societies, edited by A. R. Desai. Bombay, India: Thacker. Home Office Criminal Statistics. 1993. Statistics Relating to Crirne and Crirninal Proceedings for the Year 1993. London, England: Her Majesty's Stationary Office.
Horwitz, Allan V. 1982. The Social Control of Mental Illness. New York: Academic Press. . 1990. The Logic of Social Corztrol. New York: Plenum. Howell, Signe and Roy Willis. 1989. Societies at
Peace: A/tthropological Perspectives. London, England: Routledge.
Jones, Schuyler. 1974. Men of Influence in Nuri- stan: A Study of Social Control and Dispute Settlement in Waigal Valley, Afghanistan. New York: Academic.
Jorgensen, Joseph G. 1980. Western Indians: Comparative Environrnerzts, Languages, and Cultures of 172 Western Atnerican Indian Tribes. San Francisco, CA: W. H. Freeman.
Keeley, Lawrence H. 1996. War before Civiliza- tion. New York: Oxford University Press.
Kiernan, Ben. 1996. The Pol Pot Regime: Race, Power, and Genocide in Cambodia under the Khmer Rouge, 1975-1979. New Haven, CT: Yale University Press.
Kleck, Gary. 199 1. Point Blank: Guns and Vio- lence in America. New York: Aldine De Gruyter.
Kluckhohn, Clyde. 1944. Navaho Witchcraft. Boston, MA: Beacon.
Knauft, Bruce M. 1985. Good Company and Vio- lence: Sorcery and Social Action in a Lowland New Guinea Society. Berkeley, CA: University of California Press. . 1987. "Reconsidering Violence in Simple Human Societies." Current Anthropology 28:457-82. , 1990. "Melanesian Warfare: A Theoreti- cal History." Ocearzia 60:250-311. . 1993. "Inducements to Conflict." Review of War in the Tribal Zone: Expanding States ancl Incligertous Warfare, edited by R. B. Fer- guson and N. L. Whitehead. Science 260:1 181
Koch, Klaus-Friedrich. 1974. War and Peace in Jale'ntd: Tlte Management of Conflict in High- land New Guinea. Cambridge, MA: Harvard University Press.
Koch, Klaus-Friedrich and John A. Sodergen. 1976. "Political and Psychological Correlates of Conflict Management: A Cross-Cultural Study." Law ancl Society Review 10:443-66.
Kropotkin, Peter. [I8861 1975. "Law and Author- ity." Pp. 27-43 in Tlte Essential Kropotkin, edited by E. Capouya and K. Tompkins. New York: Liveright.
Lee, Richard Borshay. 1979. The !Kurzg San: Men, Wornen, arzd Work in a Foraging Society. Cambridge, England: Cambridge University Press. . 1993. The Dobe /u/'hoansi. 2d ed. Fort Worth, TX: Harcourt Brace College Publishers.
Le Vine, Robert A. 1959. "Gusii Sex Offenses: A Study in Social Control." Atnerican Anthro- pologist 61 :965-90. Lifton, Robert. Jay. 1986. The Nazi Doctors: Medical Killirzg and the Psychology of Geno- cide. New York: Basic Books.
Livingstone, Frank B. 1968. "The Effects of War- fare on the Biology of the Human Species." Pp.
3-15 in War: The Anthropology of Armed Con- flict and Aggression, edited by M. Fried, M. Harris, and R. Murphy. Garden City, NY: Natural History Press.
Lizot, Jacques. 1994. "On Warfare: An Answer to N. A. Chagnon." American Ethnologist 21 :845-62.
McArthur, A. M. 1961. "The Kunimaipa: The So- cial Structure of a Papuan People." Ph.D. dis- sertation, Department of Anthropology, Aus- tralian National University, Canberra, Austra- lia.
Maine, Henry Sumner. [I8611 1963. Ancient Law: Its Connection with the Early History of Society and Its Relation to Modern Ideas. Reprint, Boston, MA: Beacon.
Mann, Michael. 1993. Tile Sources of Social Power. Vol. 2, The Rise of Classes and Nation- States, 1760-1914. Cambridge, England: Cam- bridge University Press.
Masamura, Wilfred T. 1977. "Law and Violence: A Cross-Cultural Study." Journal of Anthropo- logical Researclt 33 :388-99.
Maxfield, Michael G. 1989. "Circumstances in Supplementary Homicide Reports: Variety and Validity." Criminology 27:67 1-95.
Meggitt, Mervyn. 1977. Blood Is Their Argu- ment: Warfare arnong the Mae Enga Tribesmen of the New Guinea Highlands. Palo Alto, CA: Mayfield.
Melancon, Thomas F. 1982. Marriage and Repro- duction arnong the Yanornamo Irzdiarls of Verz- ezuela. Ann Arbor, MI: University Microfilms.
Middleton, John. 1965. The Lugbara of Ugarzda. New York: Holt, Rinehart, and Winston.
Milton, Catherine, Jeanne Wahl Halleck, James Lardner, and Gary L. Abrecht. 1977. Police Use of Deadly Force. Washington DC: Police Foundation.
Mitchell, B. R. 1992. lr~ternatiotzal Historical Statistics: Europe 1750-1988. 3d ed. New York: Stockton.
Moore, Sally Falk. 1972. "Legal Liability and Evolutionary Interpretation: Some Aspects of Strict Liability, Self-Help, and Collective Re- sponsibility." Pp. 51-107 in The Allocation of Responsibility, edited by M. Gluckman. Man- chester, England: Manchester University Press.
Morrill, Calvin. 1995. The Executive Way: Corz- flict Matzagenzent in Corporations. Chicago, IL: University of Chicago Press.
Muller, Edward N. 1985. "Income Inequality, Re- gime Repressiveness, and Political Violence." Anlericarz Sociological Review 50:47-6 1.
Murphy, Robert F. 1957. "Intergroup Hostility and Social Cohesion." American A~ttkropolo- gist 59: 101 8-35.
Nabokov, Peter. 1967. Two Leggings: The Mak- ing of a Crow Warrior. Lincoln, NE: Univer- sity of Nebraska Press.
Ngor, Haing. 1987. A Ccitnboclian Odyssey. New
Nove, Alec. 1993. "Victims of Stalinism: How Many?" Pp. 261-74 in Stalinist Terror: New Perspectives, edited by J. A. Getty and R. T. Manning. Cambridge, England: Cambridge University Press.
. 1994. "Terror Victims-Is the Evidence Complete?" Europe-Asia Studies 46535-37. O'Brien, John A. 1973. The Inquisition. New York: Macmillan.
Otterbein, Keith F. 1968. "Internal War: A Cross- Cultural Study." American Anthropologist 70:277-89. , 1986. The Ultimate Coercive Sanction: A Cross-Cultural Study of Capital Punishment. New Haven, CT: HRAF. , 1989. The Evolution of War: A Cross- Cultural Study. 3d ed. New Haven: HRAF.
Otterbein, Keith F. and Charlotte Swanson Otter- bein. 1965. "An Eye for an Eye, a Tooth for a Tooth: A Cross-Cultural Study of Feuding." American Anthropologist 67: 1470-82.
Peterson, Ruth D. and William C. Bailey. 1988. "Structural Influences on the Killing of Police: A Comparison with General Homicides." Justice Quarterly 5:207-33.
Porter, Bruce. 1994. War, and the Rise of the State: The Military Foundations of Modern Politics. New York: Free Press.
Pospisil, Leopold. 1994. "'1 Am Very Sorry I Cannot Kill You Any More': War and Peace among the Kaupuku." Pp. 113-26 in Studying War: A~zthropological Perspectives, edited by
S. P. Reyna and R. E. Downs. Langhorne, PA: Gordon and Breach.
Prunier, GCrard. 1995. The Rwanda Crisis, 1959- 1994: History of a Genocide. London, En- gland: Hurst.
Rasler, Karen. 1986. "War, Accommodation, and Violence in the United States, 1890-1970." American Political Science Review 80:921-45. Reiss, Albert J., Jr. and Jeffrey A. Roth, eds. 1993. U~zderstandiizg and Preventing Violence. Washington, DC: National Academy Press.
Reyna, S. P. 1994. "A Mode of Domination Ap- proach to Organized Violence." Pp. 29-65 in
Studying War: Anthropological Perspectives,
edited by S. P. Reyna and R. E. Downs. Lang- horne, PA: Gordon and Breach.
Robarchek, Clayton A. 1977. "Frustration, Ag- gression, and the Nonviolent Semai." American Ethnologist 4:762-69.
Robarchek, Clayton A, and Robert K. Dentan. 1987. "Blood Drunkenness and the Blood- thirsty Semai." American Anthropologist 89:356-65.
Robbins, Stirling. 1982. Auyana: Those Who Held onto Home. Seattle, WA: University of Washington Press.
Rodman, Margaret and Matthew Cooper, eds. 1979. The Pacification of Melanesia. Ann Ar-
bor, MI: University of Michigan Press.
Rosenfeld, Richard and Steven F. Messner. 1991. "The Social Sources of Homicide in Different Types of Societies." Sociological Forum 6:51
Ross, Marc Howard. 1983. "Polilical Decision Making and Conflict: Some Additional Codes." Ethnology 22: 169-92.
Rummel, R. J. 1990. Lethal Politics: Soviet Genocide and Mass Murder Since 1917. New Brunswick, NJ: Transaction Books. . 1991. China's Bloody Century: Genocide and Mass Murder Since 1900. New Brunswick, NJ: Transaction Books. , 1992. Democide: Nazi Genocide and Mass Murder. New Brunswick, NJ: Transac- tion Books. . 1994. Death by Governmerzt.New Brunswick, NJ: Transaction Books. 1995. "Democracy, Power, Genocide, and Mass Murder." Journal of Conflict Resolu- tioiz 39:3-26.
Sahlins, Marshall. 1968. Tribesmen. Englewood Cliffs, NJ: Prentice-Hall.
Saunders, J. J. 1971. The History of the Mongol Coizquests. London, England: Routledge and Kegan Paul.
Schweller, Randall L. 1992. "Domestic Structure and Preventive War: Are Democracies More Pacific?" World Politics 44:235-69.
Senechal de la Roche, Roberta. 1996. "Collective Violence as Social Control." Sociological Fo- rum 11:97-128. Shakur, Sanyika. 1993. Monster: The Autobiog- raphy of an L.A. Gang Member. New York: Penguin. Silverman, Robert and Leslie Kennedy. 1993. Deadly Deeds: Murder in Carzacla. Scarboiough, Canada: Nelson Canada. Singer, J. David and Melvin Small. 1972. The Wages of War 1816-1965: A Statistical Hand- book. New York: John Wiley and Sons.
Small, Melvin and J. David Singer. 1976. "The War-Proneness of Democratic Regimes." Jerusaletn Journal of Interrzatior~al Relations 1 :SO
Solzhenitsyn, Aleksandr I. 1973. The G~tlng Ar- chipelago, 1918-1956: An Experirnerzt in Lit- eraT Investigation. Vols. 1 and 2. New York: Harper and Row.
Sorenson, E. Richard. 1972. "Socio-Ecological Change among the Fore of New Guinea." Current Anthropology 13:349-72.
Spierenberg, Peter. 1984. The Spectacle of Suj- fering: Executions and the E~~olutiorl of Repres- sion, From a Preindustrial Melropolis to the European Experience. Cambridge, England: Cambridge University Press.
Starr, Harvey. 1994. "Revolution and War: Re- thinking the Linkage between Internal and External Conflict." Political Resenrch Quarterly
Stirling, A. P. 1960. "A Death and a Youth Club: Feuding in a Turkish Village." Anthropological Quarterly 33:5 1-75.
Stone, Lawrence. 1965. The Crisis of the Aristoc- racy, 1558-1641. Oxford, England: Clarendon Press.
Strang, Heather. 1993. Homicides in Australia, 1991-92. Canberra, Australia: Australian Insti- tute of Criminology. Strathern, Andrew. 1992. "Let the Bow Go Down." Pp. 229-50 in War in the Tribal Zone: Expanding States and Indigenous Warfare, edited by R. B. Ferguson and N. L. Whitehead. Santa Fe, NM: School of American Research Press.
Taylor, Charles Lewis and David Jodice. 1983. World Handbook of Political and Social Indi- cators. Vol. 2, Political Protest and Govern- ntent Change. New Haven, CT: Yale Univer- sity Press.
Taylor, Michael. 1976. Anarchy and Coopera- tion. London, England: John Wiley and Sons. , 1982. Community, Anarchy, and Liberty. Cambridge, England: Cambridge University Press.
Thoden van Velzen, H. U. E and W. van Weter- ing. 1960. "'Residence, Power Groups, and In- tra-Societal Aggression: An Enquiry into the Conditions Leading to Peacefulness within Non-Stratified Societies." International Ar- chives of Ethnography 49: 169-200.
Thomas, Elizabeth Marshall. 1994. "Management of Violence among the JuIWasi of Nyae Nyae: The Old Way and the New Way." Pp. 69-84 in Studying War: Anthropological Perspectives, edited by S. P. Reyna and R. E. Downs. Lang- horne, PA: Gordon and Breach.
Tilly, Charles. 1975. "Reflections on the History of European State-Making." Pp. 3-83 in The Formation of National States in Western Eu- rope, edited by C. Tilly. Princeton, NJ: Prince- ton University Press. . 1986. "War Making and State Making as Organized Crime." Pp. 169-91 in Bringing the State Back In, edited by P. Evans, D. Rues- chemeyer, and T. Skocpol. New York: Cam- bridge University Press. , 1990. Coercion, Capital, and European States, AD 990-1992. Cambridge, England: Blackwell. , 1995. "State-Incited Violence, 19001999." Political Power and Social Theory 9:161-79.
Tucker, James. Forthcoming. The Therapeutic Corporation. New York: Oxford University Press.
Turnbull, Colin M. 1961. The Forest People. New York: Simon and Schuster. . 1965. Wayward Servants: The Two Worlds of the African Pygmies. Garden City,
NJ: Natural History Press.
. 1978. "The Politics of Non-Aggression." Pp. 16 1-221 in Learning Non-Aggression: The Experience of Non-Literate Societies, edited by
A. Montagu. New York: Oxford University Press.
U.S. Bureau of the Census. 1995. Statistical Ab- stract of the United States: 1995. Washington, DC: U.S. Department of Commerce.
van Creveld, Martin. 199 1. The Transformation of War. New York: Free Press.
Vayda, Andrew P. 1976. War in Ecological Per- spective: Persistence, Change, and Adaptive Processes in Three Oceanian Societies. New York: Plenum.
Walter, Eugene Victor. 1969. Terror and Resis- tance: A Study of Political Violence. New York: Oxford University Press.
Warner, W. Lloyd. [I9371 1958. A Black Civili- zation: A Social Study of an Australian Tribe. Rev. ed. New York: Harper and Brothers.
Weber, Max. [I9221 1968. Economy and Society: An Outline of Interpretive Sociology. Vol. 1. Reprint, New York: Bedminster. . 1954. Max Weber on Law in Economy and Society. 2d ed. Edited by M. Rheinstein. Reprint, Cambridge, MA: Harvard University Press.
Wheatcroft, Stephen G. 1992. "More Light on the Scale of Repression and Excess Mortality in the Soviet Union in the 1930s." Soviet Studies 42:355-67.
Wilks, Ivor. 1975. Asante in the Nineteenth Cen- tury: The Structure and Evolution of a Politi- cal Order. Cambridge, England: Cambridge University Press.
Wittfogel, Karl A. 1957. Oriental Despotism: A Comparative Study of Total Power. New Ha- ven, CT: Yale University Press.
Wolfgang, Marvin E. 1958. Patterns in Criminal Homicide. Philadelphia, PA: University of Pennsylvania Press.
Woodburn, James. 1979. "Minimal Politics: The Political Organization of the Hadza of North Tanzania." Pp. 244-66 in Politics in Leader- ship, edited by W. A. Schak and P. S. Cohen. Oxford, England: Clarendon.
Wrong, Dennis. 1994. The Problem of Order: What Unites and Divides Societies. New York: Macmillan.
Yost, James A. 1981. "Twenty Years of Contact: The Mechanisms of Change in Wao ("Auca") Culture." Pp. 677-704 in Cultural Transforma- tions and Ethnicity in Modern Ecuador, edited by N. E. Whitten, Jr. Urbana, IL: University of illinois Press.
Zahn, Margaret A. 1989. "Homicide in the Twen- tieth Century: Trends, Types, and Causes." Pp. 216-34 in Violence in America, vol. 1, The History of Crime, edited by T. R. Gurr. Newbury Park, CA: Sage.