4
A Theory of Irregular War I
Collaboration
Informers, they ought to be hanged. It is no sin to kill them.
Quoted in Ranajit Guha, Elementary Aspects of Peasant Insurgency in Colonial India
You can't tell who's who.
First Lieutenant Quinn Eddy, U.S. Army, Afghanistan, 2001
This chapter lays out the first part of a theory of irregular war as the foundation on which to build a theory of civil war violence. I begin by discussing the relation between irregular war and geographical space and I derive key implications for the nature of sovereignty in civil war. I then turn to the thorny issue of popular support, where I distinguish between attitudinal support (preferences) and behavioral support (actions). I argue in favor of a framework that makes no assumptions about the underlying preferences of the vast majority of the population and only minimal assumptions about behavioral support, in which complex, ambiguous, and shifting behavior by the majority is assumed, along with strong commitment by a small minority. I conclude with a discussion of the institutional context within which interactions between political actors and civilians take place.
4.I. SOVEREIGNTY IN CIVIL WAR
Analytically, the distinct character of irregular war is marked by the lack of front lines. A veteran of the campaigns against the American Indians remarked that "the front is all around, and the rear is nowhere" (in Paludan 1981140); this feature was captured by a rhyme sung by German soldiers stationed in the occupied Soviet Union:
Russians ahead Russians behind And in between Shooting
(Cooper 1979:92)
87
88
The Logic of Violence in Civil War
However, rather than being nonexistent, the boundaries separating two (or more) sides in an irregular war are blurred and fluid. Put otherwise, irregular war fragments space. This fragmentation can easily be seen on maps depicting countries that are undergoing civil wars: whereas conventional wars neatly divide space into two well defined and clearly demarcated spaces, irregular wars show up as messy patchworks; the more detailed the map, the messier it looks (e.g., Giustozzi 2000:291; Cooper 1979:62; Li 1975:154). Mark Danner (1994:17) describes the region of northern Morazan in El Salvador as a "crazy-quilted map," "where villages 'belonged' to the government or the guerrillas or to neither or to both, where the officers saw the towns and hamlets in varying shades of pink and red."
The fragmentation of space reflects the fact that irregular war alters the nature of sovereignty in a fundamental way. At its core lies the breakdown of the monopoly of violence by way of territorially based armed challenge. The simplest way to conceptualize the division of sovereignty in civil war is to distinguish between zones of incumbent control, zones of insurgent control, and zones in which control is contested. Where the government is able to exercise effective control and where its troops and administrators are able to move with safety day and night, we are in a zone of incumbent control. Where insurgents are able to effectively prevent the operation of government forces day and night, and the government is absent and unable to perform basic state functions, such as collect taxes and draft young men into its army, we are in a zone of insurgent control. In both zones sovereignty is undivided, though the sovereign in each is different.
In between these two zones lies an "intermediate" area, often referred to as a "contested" or "twilight" zone (Armstrong 1964:30). Deemed as the "most important arena of struggle" (McColl 1969:624), this is the zone of contested control. Unlike in the two other zones, the nature of sovereignty has been radically altered as conveyed by the following descriptions from the German-occupied Soviet Union and British-controlled Malaya:
A far greater number of people lived in what may be called the twilight zones where neither Germans nor partisans held permanent sway. In some instances, German garrisons would have nominal control, but partisans would be able effectively to raid and take reprisals by night; in others, neither side had sufficient forces to command constant popular obedience. Generally, the Germans would only occasionally send in troops and civilian officials to recruit forced labor, round up food, or simply conduct reconnaissance. (Dallin et al. 1964:330)
The terrorists were secure in their jungle. The Army and Police and Government Administration were secure in their towns. Between was the no-man's-land of village, road, railway, plantation, rubber, paddy. The terrorists at their strongest could paralyze the communications of all Malaya, but they could never hope to take the towns by storm. At their strongest, the security forces could confine the terrorists to deep jungle, but they could never hope to find them all in one massive offensive - the jungle was too thick. (Crawford 1958:82)
Political actors face three distinct population sets: populations under their full control; populations they must "share" with their rival; and populations completely outside their control. These three situations constitute two general types of sovereignty: segmented and fragmented. Sovereignty is segmented when two
Collaboration
89
political actors (or more) exercise full sovereignty over distinct parts of the territory of the state. It is fragmented when two political actors (or more) exercise limited sovereignty over the same part of the territory of the state.
4.2. THE IDENTIFICATION PROBLEM
Irregular combatants and the spies and agents of either side hide among the civilian population. This feature of irregular war, which can be termed the "identification problem," was concisely described by an American officer patrolling an Afghan village (in Zucchino 2004^8): "Two out of 10 people here hate you and want to kill you. You just have to figure out which two." An American soldier made the same point during a brutal house-to-house search in Iraq: "I feel bad for these people, I really do. It's so hard to separate the good from the bad" (in Filkins 2005:57). A few years earlier, Soviet soldiers had referred to their Afghan adversaries as dukhi, the Russian word for ghosts (Baker 2002 Ai), and summarized the problem they faced as follows: "You see me, but I don't see you" (in Wines 2001 :B7).
The inability to tell friend from enemy is a recurring element of irregular war, as evidenced by the following observations of U.S. military personnel in Vietnam in 1968, in Afghanistan in 2003, and in Iraq in 2003:
Wherever I went and young Vietnamese men would look at me I grew scared. There really was no way to tell who was who. You could be in a room with one and not know whether he was really a Charlie or not. It became easy to sense the distrust that must exist in the outlying areas. How could one really fight in the fields and know whether at any time the men beside you were not going to turn tail and train their guns on you? Whom did you begin to trust and where did you draw the line? Another ludicrous aspect of the war. (John Kerry in Brinkley 2003:50)
But even without a common language between them, the villagers seem to know what the Americans have come to do. Silently, turbaned men in long gray tunics open doors in compound walls for five- or six-man groups of helmeted men in desert camouflage.... All seems accepted: in bitter helplessness against what the Americans are doing or - as the Americans hope - in gratitude for the American defeat of the repressive Taliban. It is impossible for the soldiers to know. Gonzalez speaks of trying to guess the sentiments of the locals not by their smiles but by the firmness of their handshakes. (Bergner 2003:44)
You have enemies but they're ghosts. They hit us and they run. They don't come out and fight us. (in Zaretsky 2oo3:A4)
This is not a recent development. A French general stationed in Spain in 1810, remarked that "the great difficulty [was] not fighting [the guerrillas] but finding them" (in Tone 1994:105); American soldiers serving in the Philippines in the early 1900s spoke of "chasing a phantom" (May 1991:161); and a British soldier in Malaya recalls: "Somewhere in that gigantic morass were fifty veteran terrorists. How were they to be killed? How were they even to be found?" (Crawford 1958:87).1 A Pakistani officer in Bangladesh (Salik 1978:103) observed that
1 See also Linn (1989:58); Calder (1984:138; 158); Salik (1978:101); Henriksen (1976:397); Meyerson (1970:79); Trinquier (1964:26); Kitson (1960:192).
9°
The Logic of Violence in Civil War
"the main problem was to isolate the rebels from the innocent people____It was
difficult to distinguish one from the other as all of them looked alike. A rebel carrying a sten gun under his arm could, in emergency, throw his weapon in the field and start working like an innocent farmer." The following exchange between a journalist and an Indian officer in Kashmir captures the identification problem:
1 asked him how many terrorists he thought there were. "Very few, these days," he replied.
Why, then, did the government need to keep half a million men here? "Because," he replied quietly, "you don't know who they are." (Hilton 2002:73)
These are not mere anecdotes. The CIA estimated that less than 1 percent of nearly 2 million small-unit operations in Vietnam conducted in 1966-8 resulted in contact with the insurgents (Ellsberg 2003:240).2 Not surprisingly, irregular warfare has been called "war in the shadows" (Asprey 1994) or "phantom" war (Cooper 1979).
As the preceding examples suggest, the identification problem hurts primarily the incumbents: it is their opponents who, being weaker, hide. "It was an extremely one-sided type of warfare," a German officer pointed out about the partisan war in the Soviet Union, "because the German soldier was easily recognizable, and the partisan fighter, because he wore civilian clothes, was not" (in Cooper 1979:89). This explains the difficulty that incumbents have in defeating insurgents, despite their often tremendous advantages in resources. Vietnam is a classic example but far from the only one. In 1965 Peru spent more than $10 million to defeat about one hundred poorly armed guerrillas (M. F. Brown and Fernandez 1991:190), and in July 1993 the British domestic intelligence service (MI5) revealed that the greatest part of its budget, "forty-four percent of an undeclared total of many hundreds of millions of pounds, was targeted against [the IRA,] a small, impoverished, working class guerrilla organization of around six hundred fighters with an estimated budget of five million pounds" (Toolis 1997:285).
Yet, insurgents also face an acute identification problem. The populations in the midst of which they hide may turn them in; spies and agents hiding among these populations may also identify them. According to Lucian Pye (1964:177), "The advantages guerrillas and terrorists may possess in opposing the far greater resources of the government can largely be countered if the government has adequate intelligence. At later stages in their insurrection, whatever advantages of mobility, surprise, and esprit de corps the guerrillas possess can usually be more than offset if the government has the crucial intelligence at the right moment."
Insurgents are vulnerable if they are identified. Between 1942 and 1944 the French resistance suffered more losses as a result of betrayal within its own
2 Monthly figures for minor operations in Vietnam during 1964 are telling: 59,996 operations for 451 contacts with the Vietcong; 72,794 operations for 406 contacts; 73,726 operations for 491 contacts, and so on (R. Thompson 1966:88). An American officer said "that he had spent the entire year in Vietnam and never seen a single live Vietcong" (Herrington 1997:3»).
Collaboration
91
ranks than through German action (Laqueur 1998:230). By 1983 the Afghan Communist regime had deployed 1,300 agents in insurgent units, 1,226 along communication lines, 714 in underground political organizations, and 28 in Pakistan (Giustozzi 2000:98). As a result, betrayal becomes a pervasive obsession among insurgents.3 Insurgent memoirs are replete with an overarching concern about information leaks (e.g., BarnettandNjama 1966:61); conversely, counterin-surgent memoirs (e.g., Aussaresses 2001; Flower 1987) brim with descriptions of thorough penetration of their opponents' organizations - particularly in urban environments.
There are two dimensions to the identification problem: first is the categorical refusal of at least one side, the insurgents, to be reduced to a single identity, that of combatant (Andreopoulos 1994:195). This entails the transformation process that American soldiers fighting the Filipino insurgency in 1900 described as "chameleon act" (May 1991:142-3, 161).5 Second is the refusal of the surrounding population to identify them to their opponents. Either the people do not know who is really an insurgent, which is sometimes true about spies and clandestine agents; or, much more commonly, they refrain from identifying the insurgent combatants who hide among them - out of diverse motivations, including sympathy and fear. Herein lays the relevance of popular "support."
4.3. SUPPORT
"The battlefield today is no longer restricted," observed a French officer in Algeria (Trinquier 1964:29); "like it or not, the two camps are compelled to make [civilians] participate in combat." The fight is conducted through the people; as a Cypriot peasant told the writer Lawrence Durrell (1996:224), it is "like a man who has to hit an opponent through the body of the referee."
It is widely argued that the outcome of irregular war hinges on the behavior of civilians; put otherwise, "civilian" or "popular support" is "the sine qua
3 Bouaziz and Mahé (2004:253); J. L. Anderson (2004:176); Bizot (2003:112); Elliott (2003:961); Tucker (2001:87); Portelli (1997:138); Todorov (1996:90); Schroeder (1996:428); Saul and Leys (1995:53); Stubbs (1989:189); Paludan (1981:78). Sometimes, this obsession can lead to excesses. In the Philippines, the Communist New Popular Army launched a "terrifying" purge, killing hundreds of its own members and supporters in 1986-8 because of fears about informing (Jones 1989:265-75).
4 Identification is not a problem limited to "ideological" wars; it is part and parcel of many (though not all) ethnic civil wars. In many ethnic civil wars armies systematically recruit among their ethnic rivals, fighters switch sides, and civilians collaborate with the army of their ethnic rivals (Kalyvas 2003). At least one political actor (usually the incumbents) seeks to control the "underlying" population of the ethnic rival, rather than exterminate it or remove it. Despite claims positing the impossibility of defection in ethnic (Kaufmann 1996; Ranzato 1994) or even nonethnic conflicts (Zulaika 1988:32), such defection is possible when actively solicited. Defectors do not lose their original ethnic identity but alter it, through the addition of qualifiers such as "moderate," "loyal," "antiextremist," or through their migration to another identity dimension.
5 Civilian collaborators are equally hard to identify, as Leakey (1954:121) recalls about Kenya: "There is no outward sign by which one can tell whether a man is a Mau Mau supporter or not, for the original practice of making seven cuts on those who were 'initiated' into the movement was very quickly abandoned, because it made recognition by the police too easy."
92
The Logic of Violence in Civil War
non of victory" (Trinquier 1964:8). Almost all writers converge in asserting that no insurgent movement can survive without "civilian support," and neither can incumbent victory be achieved without it (Wickham-Crowley 1992:8; Bard O'Neill 1990:70-89). As an IRA man pointed out, "Without the community we were irrelevant. We carried the guns and planted the bombs, but the community fed us, hid us, opened their homes to us, turned a blind eye to our operations" (Collins 1999:225). Even the "Manual for Warfare against Bands" used by German troops during the Second World War made a central point of the fact that "the attitude of the population is of great importance in the fight against bands. Bands cannot continue in existence for any length of time in the midst of a population which entertains good relations with us" (in Heilbrunn 1967:150).
Yet below this unanimity lies extreme confusion, for there are two very different ways of thinking about popular support. One way is to think of it as an attitude, preference, or allegiance, and the other is to emphasize behavior or action. Obviously, there is a connection between the two, but in an irregular war the gap between attitudes and actions tends to swell; in many ways, this is the main consequence of the war.
Conceptualizing support in attitudinal terms is common. It has been argued that "guerrilla victory depends upon the loyalties of civilians in the area of operations," whereas in conventional wars "military operations go on without regard for the hapless civilian population. No one asks it to take sides in the struggle -at any rate not at first, while the battle rages____In revolutionary war... the allegiance of the population becomes one of the most vital objectives of the whole struggle." (C.Johnson 1962:649; Fall in Trinquier i964:ix). As a Missouri Unionist argued, "if counties known as disloyal would cease to sympathize with treason, and become earnest supporters of the Government, guerrilla warfare would soon cease to exist" (in Fellman 1989:91; emphasis added). As there is only a short step between attitudes and ideas, it often is asserted that ideology is central to civil war. In this view, ideology matters more than territory (Angstrom 2001:106); in the crude formulation of an American practitioner, "The only territory you want to hold [in a civil war] is the six inches between the ears of the campesino" (in Siegel and Hackel 1988:119).
The adoption of a primarily attitudinal stance is problematic. Attitudes are unobservable and must somehow be inferred, a hard task when it comes to civil war. A common, yet flawed, solution is to reason backward from the outbreak of the war to its causes, by positing the force of popular beliefs and grievances.6 Civil wars are evidence of a deep "crisis of legitimacy"; substantial segments of the population (often, just "the people"), this argument goes, are intensely opposed to the regime in place and, consequently, reallocate their support toward rebels; in this sense, civil wars are really "peoples'wars" (Tone 1994:4; C. Schmitt 1992:213; van Creveld 1991:143). The implication is that people actually "choose" what faction to support based on its political and social profile or ideology - as if they were
6 An extension consists in linking support to the outcome of a war: defeats are conveniently ascribed to the dearth of popular support and victories to its abundance.
Collaboration
93
voting in elections (C.Johnson 1962), and this choice has a tremendous impact on the military conflict: "Military and strategic factors are far less important than popular attitudes in a civil war. If an army is welcomed by the local population, its resources and strength are automatically increased. If, on the other hand, it is unwelcome, its strength is tied up in pacifying and policing the conquered territory" (Brovkin 1994:91).
However, there are good reasons to question the view that the outbreak of a civil war ought to be taken as unproblematic evidence of a deep "crisis of legitimacy" or of massive support in favor of the insurgents. Explanations of rebellion in terms of grievance parallel explanations of regime breakdown in terms of legitimacy; they are, as Adam Przeworski (1991:54-5) points out, either tautological or false; only when organized political forces challenge the sovereign does "political choice become available to isolated individuals." The absence of alternatives often produces collaboration irrespective of the level of popular satisfaction or lack thereof, which may be then wrongly interpreted as a reflection of legitimacy. Moreover, joining a rebellion can be the product of the ongoing war as much as it can be its cause. Stoll (1993:20) is right to observe that "once an armed conflict is underway, the violence exercised by both sides can easily become the most important factor in recruitment. People may join the revolutionary movement
less because they share its ideals than to save their lives____Hence, just because
an insurgency grows rapidly does not mean that it represents popular aspirations and has broad popular support."
An equally misguided solution to the problem of unobservable attitudes is to derive them uncritically from "revealed" or observed behavior (Sen 1986). Often, the observation that some people collaborate with a political actor is interpreted as evidence of loyalty toward this actor. However, observed behavior is consistent with contradictory attitudes; as a result, it is a poor mechanism of preference revelation (Lichbach 1995:287). In fact, civil wars produce strong incentives for preference falsification (e.g., Calder 1984:155). Clearly, observed behavior is not just an imperfect indicator of preferences, but most probably an inaccurate approximation as well (Kuran 1991).7 Vietnamese peasants proved particularly adept in this respect:
Both sides recruited military manpower relentlessly, and both sides equally demanded the loyally of the peasants. Thus, a rice farmer in Hiep Hoa could easily find himself sitting under a banner at midnight, participating in an antigovernment rally during which he might play the role of an outraged and exploited peasant, under the watchful eye of a Communist propaganda cadre. The following morning, the same farmer could send his children to the new, government-built school and then walk to the village office to vote in a local election - this time under the watchful eye of government hamlet chief. The village Vietcong would boast in their report that ". . . So far, 90 percent of the villagers
7 More systematic data can undermine widespread assumptions about preferences, but they are rarely fully reliable or available. The contents of private correspondence opened and read by the Vichy authorities contradicted the assumptions of observers about popular preferences in occupied France: out of 4,352 letters written in December 1943 and mentioning acts that Vichy listed as terrorism in the Montpellier region, 3,976 were hostile to the resistance and only 142 were sympathetic (Kedward 1993:113).
94
The Logic of Violence in Civil War
have actively thrown their support to the cause of the revolution." At the same time the Hiep Hoa village chief would inform his superiors that "more than 95 percent of the villagers voted in the recent election, with anti-Communist candidates receiving the near unanimous support of the people." (Herrington 1997:37)
No wonder that careful observers are baffled, like Kevin Toolis (1997:255-6) was in Northern Ireland: "Who can tell the truth in a world filled with double deceptions, handlers, confused loyalties, liars, self-loathing, professional deceivers, disinformation, black propaganda and betrayers? At the end of this journey I began to doubt the motivations of almost everyone I talked to."
This problem of observational equivalence has led to interminable debates, usually contaminated by partisan bias, on how to interpret observed support: on the one hand, those favorable to the rebel side claim that observed civilian collaboration with the rebels reflects genuine support derived from existing grievances and the belief that rebellion is the way to right existing wrongs, and the same observers imply that similar civilian collaboration with the incumbents is instead a result of coercion; on the other hand, those observers siding with the incumbents claim exactly the opposite. For example, arguing that Vietnamese peasants supported the Vietcong rebels because they believed in their program and ideas (e.g., FitzGerald 1989) rather than because they were coerced into supporting them (e.g., Klonis 1972:155) was a key rhetorical weapon in the polarized debate surrounding the U.S. intervention in Vietnam.
No matter how attitudes are inferred, an implication of the focus on attitu-dinal support is the twin claim that victory requires attitudinal shifts - "hearts and minds" - which can only be achieved through nonviolent persuasion: a "genuine and timely effort to satisfy the grievances of the people" is required since "popular support for the guerrilla is predicated upon the moral alienation of the masses from the existing government" (Ahmad 1970:15); in Robert Thompson's (1966:169) words, "Force of arms alone will not prevail." The main policy implication of this view is that incumbents need to persuade hostile populations to switch their sympathies through programs of political liberalization, economic development, and civic action.8 However, no matter how deep the beliefs and how powerful the allegiances, they are not the only determinants of behavior. People can be coerced, and violence is used to force people to alter their behavior and behave in ways that may not be consistent with their preferences. As pointed out, election makes a poor analogy for civil war. Even Communist revolutionary doctrine, at the root of the "hearts and minds" approach and best exemplified by Mao's "fish in the sea" dictum, stresses violence much more than is often assumed, when it points out that political power grows out of the barrel of a gun. Participants always recognize that violence plays a key role. In Trinquier's
8 This view has led to a false dichotomy between military and political responses to insurgencies. In fact, the two often go together, since political programs cannot be implemented in "insecure" environments. In Malaya, where the term "hearts and minds" originated, "it became clear that the strategy entailed the use of both the stick and the carrot" (Stubbs 1989:164-5). Conversely, a U.S. officer observed during the 1899-1902 Philippines War: "This business of fighting and civilizing and educating at the same time doesn't mix very well. Peace is needed first" (in Linn 1989:128).
Collaboration
95
(1964:8) words, civilian "support may be spontaneous, although that is quite rare and probably a temporary condition. If it doesn't exist, it must be secured by every possible means, the most effective of which is terrorism."
The difficulties associated with attitudinal support flow not just from the inference process but also from conceptual problems related to the very motivations that underlie support. For assuming popularity says nothing about how popularity translates into action on the ground. Both joining a rebel army and collaborating with it result from variable and complex sets of heterogeneous and interacting motivations,9 which are affected by preferences over outcomes, beliefs about outcomes,10 the behavior of others and the networks into which people are embedded,11 and security considerations12 in an environment where chance
9 Barton (1953:141) lists five different kinds of motivations, Maranto and Tuchman (1992:251) eight, and R. Berman (1974:58, 67) twenty-seven!
10 According to Robert Thompson (1966:170): "Much can be learnt merely from the faces of the population in villages that are visited at regular intervals [by incumbent forces]. Faces which at first are resigned and apathetic, or even sullen, six months or a year later are full of cheerful welcoming smiles. The people know who is winning."
11 Processes of joining are highly likely to be rooted in network dynamics (Petersen 2001). Stark (1997) shows how social network ties (especially friendship and kin ties) are the best predictors of religious conversion. Insurgents consistently point to the importance of local networks in producing recruitment, based on "the desire of persons to unite with friends, neighbors and kinsmen" (Barnett and Njama 1966:158), and their practice is consistent with this view (Perry 1984:445). Hart's (1999:209) analysis of IRA unit rolls (in 1916-23) shows that brothers constituted between 37 and 58 percent of the battalions he examined. He adds that "the question of personal motivation is oddly absent from most memoirs and memories of the period. Volunteers seem to have regarded their political commitment as completely natural and their motives as self-evident, requiring little reflection.... Veterans are rarely able to recall exactly when and how they joined but they remember vividly how it felt to belong: 'There was a spirit in the air alright.'... For most IRA men, joining the movement in its early days required little deliberate choice or effort. If you had the right connections or belonged to a certain family or circle of friends you became a Volunteer along with the rest of your crowd. If not, you probably stayed outside or on the fringes" (Hart 1999:203, 220). The list of Colombian guerrillas incarcerated after the Volencia in a Medellm prison was "rife with networks of uncles and nephews, sons and fathers, brothers and cousins" (Roldan 2002:243). Horton (1998:186) found that the former contra combatants she interviewed had an average of five other relatives in the contra army; likewise, the Sandinista rebels had used "multiclass networks of kinship, friendship, and patron-client ties to their own ends" (Horton 1998:69). Paul Berman (1996:66) adds that the country people of Nicaragua "were loyal to their own enormous clans," and this was reflected in patterns of recruitment: "Brothers follow brothers" (P. Berman 1996:78). Goltz (1998:150) found that the Azerbaijani militias of the early 1990s "seemed to be composed more of extended families than soldiers," and Aviout-skii and Mili (2003) stress the importance of clan solidarity in rebel recruitment in Chechnya. Examples abound: Nepal (Sengupta 20050:64), Chechnya (Tishkov 2004:94), Bosnia (Claverie 2002:48), Colombia (Sanchez and Meertens 2001:17; Rubio 1999:102; Pecaut 1996:257), Latin America in general (Wickham-Crowley 1992:152), the Balkans and the Baltics (Petersen 2001), the Congo (Bazenguissa-Ganga 19993:42), Algeria (Faivre 1994:145), Mozambique from the 1960s to the 1990s (Finnegan 1992:118; Henriksen 1983:96), the Philippines (Kerkvliet 1977:205), Malaya (Stubbs 1989:49), Kenya (Kitson 1960:126), China (Wou 1994:252), and even revolutionary France (Cobb 1972:26).
12 According to a Dominican guerrilla (in Calder 1984:126), after a Marine captain had threatened his life, he believed that "his only remaining option was to flee into the hills."
96
The Logic of Violence in Civil War
and contingency cannot be underestimated.13 Of course, many fighters are conscripted or abducted.14 Additional factors include curiosity and the prospect of excitement and adventure,15 the lure of danger,16 the acquisition of a new and more rewarding individual identity or moral worldview,17 the pleasure of acting as one's own agent,18 and purely criminal motives.19 Collaboration may provide access to public goods (such as dispute resolution, or protection against common crime, which explodes when state authority wanes),20 or individual material benefits (including land, lower taxes, higher prices for produce, or debt forgiveness),21
13 A Nicaraguan peasant did not want to become involved in the war, but after encountering by chance some contra rebels in the forest, he gave them food and gradually slipped into more sustained collaboration. As Horton (1998:183) puts it, "once this first step was taken, [he] found it difficult to retreat into neutrality.... Without having made a specific decision to collaborate with the contras, [he] found himself in the role of the contra correo. And two years later, in 1983, because of pressures from both State Security and the contras, [he] became a full-fledged combatant with the contras." See also Todorov (1996:94), Fenoglio (1973:60), and Clutterbuck (1966:94) for similar examples.
14 It is estimated that after the spring of 1942, at least 80 percent of Soviet partisans "joined either unwillingly or because they had no alternative.... The Soviets made no pretense about recruitment for the partisan movement. Compulsion was paramount" (Cooper 1979:71). Even in ethnic civil wars, where individuals are supposed to have extremely strong preferences, participation often results from conscription. Although Somali "warlords could use the language and sentiment of clan to rally allegiance along blood lines [they] built their authority on the power of the gun" (Besteman 1996:590-1). In Bosnia, "many people found themselves carrying a gun whether they liked it or not. If you were of combat age, meaning only that you possessed the strength to fight, kill and possibly survive, then you were conscripted into whichever army represented your denomination, Muslim, Serb or Croat" (Loyd 2001:85). See also Waldman (2003:Ai); Seidman (2002:40); Horton (1998:9); Nordstrom (1997:50); Senaratne (1997:99).
15 Sengupta (2005^64); Kitson (1960:126).
16 Tishkov (2004:98).
17 As a former IRA cadre put it: "At the very least such activity gave a strange edge to my life: I lived each day in a heightened state of alertness. Everything I did, however trivial, could seem meaningful. Life outside the IRA could often feel terribly mundane" (Collins 1999:362). See also Sanchez and Meertens (2001:22); Mahmood (2000:73); Mirzeler and Young (2000:419); McKenna (1998:184); Peters and Richards (1998); Armony (1997:207); Enzensberger (1994:42); Ash (1995:205); Wickham-Crowley (1992:20-1); Ortiz Sarmiento (1990:116); Henriksen (1983:160).
18 Wood (2003:18).
19 Criminal motives have long had their place in descriptions of civil war. Consider the following description of the American War of Independence in South Carolina (McCrady 1969:139): "There came with the true patriots a host of false friends and plunderers. And this was true of both sides in this terrible struggle. The outlaw Whig and the outlaw Tory, or rather the outlaws who were pretended Whigs and Tories as the occasion served, were laying waste the country almost as much as those who were fighting for the one side or the other." See also Reig Tapia (1996:583) for the Spanish Civil War. See also Mueller (2004); Silke (1998); Fisher (1997:87); Nordstrom (1997:56-7); Cribb (1991, 1990); Jones (1989); Paul and Demarest (1988); Ash (1988); Henderson (1985).
20 According to Degregori (1998:135), the most important targets of Sendero in Peru were "abusive merchants, cattle thieves, corrupt judges, and drunk husbands." See also Smyth and Fay (2000:123); Rubio (1999:129); Manrique (1998:204); Berlow (1998:95); Senaratne (1997:75); Gallagher (1995:50); Stoll (1993:80); Wickham-Crowley (1991:44); Jones (1989:127); Kheng (1983:148); Kerkvliet (1977:70, 164); Rudebeck (1975:445); Taber (1965:40); Lear (1961:92).
21 Kedward (1993:96); Stoll (1993:78); Popkin (1979); Race (1973:123-5).
Collaboration
97
protection against indiscriminate violence from the opposite side,22 escape from obligations that are seen as more onerous (such as military or labor drafts),23 acquisition of higher status24 including what status can be obtained via access to guns,25 the weight of personal or local disputes,26 or simply the response to emotions such as anger, moral outrage caused by public humiliation,27 and the desire to take revenge.28 Furthermore, support is partially endogenous to the war. This can take many forms, not least of which is the purposeful use of violence to generate support (Snyder 2003), even when and where sympathy for an organization already runs high (Collins 1999:128; 170; Harris 1989:89). Clearly, observed support corresponds to a complex mix of preferences and constraints.
Like the much better-studied motivations for joining the rebels, individual motivations for joining progovernment militias are also heterogeneous and mixed (Rubio 1999; Stoll 1999; Starn 1998; Mackenzie 1997; Cribb 1991).29 The men
22 Goodwin (2001); see Chapter 5 for a discussion and evidence.
23 Del Pino (1998:170); Berman (1996:69); Jankowski (1989:123-4); Cooper(i979:25); Race (1973:172).
24 Collins (1999:164) points out that "in the nationalist community [of Northern Ireland], in republican circles anyway, IRA men have considerable status, and for those Provos who look for sexual advantages from it, there is no shortage of women willing to give more than the time of day to IRA volunteers." When it appeared to be winning, the Chinese Communist revolution "created an unprecedented and unparalleled opportunity for status advancement and social mobility for tens if not hundreds of thousands of persons at both the local and higher levels" (Levine 1987:173). According to Sheehan (1989:177), "there is considerable evidence that many young Vietnamese of peasant origin join the Vietcong because the Communists who have been forced by the nature of their revolution to develop leadership from the countryside, offer them their best hope of avoiding a life on the rung of the ladder where they began - at the bottom." Many young people joined Shining Path in Peru motivated by the "concrete exercise of power in their own localities" (Degregori 1998:130). Todorov (1996:100) tells the story of a Frenchman who worked as an interpreter for the Gestapo and "then discovered that the shame and humiliation he had experienced as an adolescent were soothed by the power he enjoyed in his position with the Gestapo."
25 Johnson (2001:202); Mirzeler and Young (2000:419); Rubio (1999:115); Finnegan (1992:70); Zulaika (1988:25).
26 Kalyvas (2003).
27 Thaxton (1997:308-9) reports that a Chinese peasant cited as his main motivation for joining the Communists the fact that a government officer "kicked over his sweet melon basket and scolded him for daring to sell his 'dirty melons' leaving him to pick up the scattered melons one by one." Humiliation of traditional chiefs by "upstart" village administrators played an important role in Mozambique (Geffray 1990:32). See also T. Brown (2001:42); Horton (1998:106-9); R. Berman
097475)-
28 Adams (1994:7) recalls his relationship with a South Vietnamese officer: "I spent many hours talking to Lieutenant Lam. Gradually, he opened up. Late one evening over supper, Lam told me how much he hated the [Vietcong]. They had killed his brother, he said." West (1985:56) describes a South Vietnamese policeman who had tattooed on his chest the words "Kill Communists." The Vietcong had apparently slain his wife and all his children except one. The same policeman was described by a U.S. sergeant in the following terms (in West 1985:160): "Thanh's definitely mean. He hates. He lives only to kill VC."
29 An analysis of individual causes of joining the French Waffen-SS in Marseille is replete with mixed, even contradictory, motivations. Consider the following statement: "After I'd gambled away my weekly pay my mother berated me and told me that if I didn't recover the money she would send me to my father [from whom she was separated]... unable to find the money, I signed up" (in Jankowski 1989:12 8-9). Jankowski's is one of the very few studies of motives for joining an armed
98
The Logic of Violence in Civil War
who joined the militia in the Philippine island of Negros "cast their lot with the military neither out of ideology nor out of any sense of debt or obligation, but because they had been ambushed or extorted by the NPA, needed work, had been directed to join by their employers, or saw an opportunity to exact revenge against their personal enemies" (Berlow 1998:182). An American officer in Iraq described the motivations of the Sunni commandos fighting on the American side against the insurgents by pointing to their need for a paycheck, their desire for the social status that comes from being members of a professional military, and their yearning for a routine - rather than ideology: "For some, there's definitely a desire to make Iraq better, but for a lot of them, it's just life as they know. For most of them, the cause isn't really that important. They're more used to working in this role.... I think for a lot of them, they couldn't fathom doing something different with their lives" (in Maass 2005:82).
Coercion is often cited as a chief motivation - though this can also be a self-serving claim. "You really didn't have a choice," Mehmet Refiktas, a Kurdish "village guard" in Turkey said. "Asked what happened to the homes of men in his mountain village, Islamkoy, who declined the government's offer, explained, 'Oh, they were burned'" (Vick 2002A18). David Stoll (1993:128) tells a similar story about Guatemala: "Joining the patrol became a way to protect self, family, and community from the guerrilla contacts that triggered [government] massacres." Indeed, many Guatemalan militiamen were survivors of army massacres (Stoll 1993:162). Security matters in the opposite direction as well: the Portuguese trusted village headmen in Mozambique and gave them weapons because many had been murdered by the rebels and those remaining lived in fear of the same fate (Cann 1997:161).
Economic motivations also matter. Two men who explained why they had joined the pro-Israeli South Lebanese Army: "We were poor. The only work was their army. If you're in the army, you can live. If you are not, you can't" (in Sontag 2000A1). Turkey pays the 95,000-strong militia it maintains in its Kurdish areas a monthly salary of $115, which is a "valuable sum." Indeed, it was reported that "Turkish officials and foreign diplomats have begun to worry that if these salaries are suddenly cut off and nothing replaces them, some unhappy veterans might grab their weapons and take to the hills" (Kinzer 2000A8). Economic compensation may also take the form of loot. When the army promised the content of guerrilla storage pits in the outskirts of a Guatemalan town to militiamen who found them, "the race was on, to dig each one up before someone else did" (Stoll 1993:107). Ideology seems to play a minor role for militiamen, probably less than it does for rebels. In fact, militias are often composed of turncoats, either ex-rebels or former supporters of the rebels.30
organization during a civil war based on contemporary written records; he estimates that the percentage of those whose motives were mixed to be as high as 80 percent (Jankowski 1989:123-4). Almost every single "thick description" that I have read points to a complex and fluid mix of motives. See Tucker (2001:38); Ellis (1999:127); Hammond (1999:260); Horton (1998:6); Faivre (1994:121); Geffray (1990:105-13); Meyerson (1970:95); Barnett and Njama (1966:149). 30 Berlow (1998:233); Moyar (1997:68); Henriksen (1983:136).
Collaboration
99
Finally, fear and revenge are important motivations. The fear of insurgents that follows their violence often drives people into militias (e.g., Sengupta 2oc>5a:A3; D. Anderson 2005:73) and revenge is a central motivation, maybe more so for militiamen than for rebels, as suggested by the following examples from Algeria and Chechnya:
Each day when he comes here, Abdurahim (13) (whose family was killed by the Islamists in the village of Haouche Fanir, on May 14, 1997) dreams of joining the patriots. But for his two surviving brothers, who are both militiamen, that seed of revenge is already growing - pointing to a cycle of violence that will be difficult to break. "If I kill 1,000 terrorists, it won't be enough for my one brother," says a brother named Arabah, cradling a gun and wearing a clear-plastic waist pouch stuffed with colorful shotgun shells. "Do I look for revenge? Of course," he says. "Of course." (Peterson 1997a)
Mr. Tovzayev described himself as a fighter against the "bandits" - the rebels - on behalf of Mother Russia. He maintained close ties to the Russian military and was particularly proud of the armored jeep that was presented to him as a gift by Col. Gen. Gennadi N. Troshev, commander of Russian forces in the Northern Caucasus.... [He] spoke of taking up arms against the rebellion in 1995 after a rebel leader, Allaudin Khamzatov, entered his village and assassinated Mr. Tovzayev's father in front of his family. Within three months, he raised a small force of fighters and killed Mr. Khamzatov in an ambush. "This is how this kind of life started - fighting bandits," he said. (Tyler 2001: A8)
Revenge can be triggered by family feuds and local conflicts that become intertwined with the war (Abdi 1997; Leakey 1954:114). The French found that many of the Calabrians who volunteered to join the militia they formed in 1806 "saw the army as an excellent opportunity to settle a long-standing vendetta with some nearby family members" (Finley 1994:29). Often, revenge is exclusively related to the insurgents' past actions (Roldan 2002:258; Linn 1989:54). Because it is possible for rebel agents and sympathizers to hurt, alienate, or humiliate others, incumbents capitalize on discontent with insurgent rule in forming militias in recently "liberated" areas, and the militias' actions are keyed toward revenge. Stephen Ash (1988:155) describes the Unionist militiamen of middle Tennessee as "avenging angels." While the staunchest recruits for the village-based militias formed by the French in their war against the Vietminh were in provinces previously under Vietminh administration (R. Thompson 1966:168). The Algerian men who joined the French army in 1959 complained about the coercion exercised by the FLN rebels, especially the taxes, fines, and the tight control on everyday life that they imposed (Faivre 1994:143; Hamoumou 1993). An NPA commander in the southern Tagalog region of the Philippines "noted that the rise of vigilante groups in some rural barrios of Lopez (Quezon) was due in part to NPA heavy-handedness. The rebel commander recalled there was 'too much taxing' and that 'some of our comrades mistreated some of the people. The opportunity was taken by the military'" (G. Jones 1989:249). Even if such animosities are absent, they can be generated by the very act of militia formation. A U.S. officer reported from the Philippines that "having compromised themselves by collaboration and knowing that they risked guerrilla retaliation, [Filipino defectors] 'seemed most anxious to ferret out all insurrectos in this vicinity'"
IOO
The Logic of Violence in Civil War
(in Linn 1989:43-4). The Earl of Carlisle, an envoy of the British government to America in 1778, made the same point about Loyalist militiamen: "In our present condition the only friends we have, or are likely to have, are those who are absolutely ruined for us" (in Shy 1976:186).31
Whatever their initial motivations, with time many militiamen develop genuine loyalties, at least to one another. In Cotzal, Guatemala, "an unwanted civil patrol had, paradoxically, become an institution of solidarity" (Stoll 1993:144). Likewise, in Peru "no one imagined that these patrols, too, would turn into a massive movement with an important degree of popular participation and autonomy from the state" (Starn 1998:236).32
These complexities of preference formation suggest the need to shift the focus from attitudes to behavior. But understanding support in terms of observed action also entails several problems. First, it is not easy to map behavior during a civil war because data are lacking. Real-time ethnographic observation can partly address this problem (e.g., Sluka 1989), but the limited sample on which it is based and the practical constraints posed by armed combat can defeat the implementation of rigorous research designs (Wood 2003:42). This is why most work relying on retrospective reconstructions is an enterprise fraught with difficulties.
Second, observed support is not dichotomous but can be conceptualized as a continuum that stretches from full association with one political actor to full association with the opposite political actor, while encompassing various shades of association, including neutrality (Petersen 2001:8). This is reflected in distinctions such as those between hard and soft supporters (Sluka 1989:291-4), between passive and active supporters (Bard O'Neill 1990:71-2), between direct and indirect participants and those "caught in the middle" (Kerkvliet 1977:166-7), and between constituents, sympathizers, members, activists, and militants (Lichbach 1995:17). The ethnographer of a Belfast Catholic ghetto widely considered to be a hotbed of insurgent support (Sluka 1989:291) found that "not all the residents who support the guerrillas support them in all of their roles.... Many people in Divis support the IRA and INLA in one area or role, while simultaneously condemning them in another." Moreover, observed support is dynamic and relational, hence a person's actions are influenced by others (Petersen 2001).
Third, we must distinguish between reasons for joining an organization and reasons for remaining in it (Molnar 1965:77-82). Students of military history, in particular, have adopted John Lynn's (1984) distinction between initial motivations (why people join an army), sustaining motivation
31 In fact, violence may be used to generate commitment. In Guatemala, to overcome the militiamen's reluctance to kill the first victims, "an officer ordered them to pick executioners by lot. Subsequently, a victim would be tied to a tree and everyone in the patrol ordered to stab him with machetes. Before long, some patrollers were volunteering to kill" (Stoll 1993:107). Note that this explains the form of violence and the perpetrator; the victim could have been picked up as part of a scheme to deter defection.
32 Two systematic differences in patterns of joining rebel organizations and incumbent militias may be the absence of high risk for first joiners and the absence of ideological indoctrination; both are consistent with an account that would point to the role of the state in assuming many of the costs of militia formation.
Collaboration
101
(why they stay in despite the cost), and combat motivation (why they fight on the battlefield)"
To summarize: inferring preferences from observed behavior is exceedingly difficult; preferences are open to manipulation and falsification; actual behavior is difficult to observe in civil war environments; and even when reliably observed, support is the outcome of a dynamic, shifting, fluid, and often inconsistent confluence of multiple and varying preferences and constraints. This turns the search for one overriding motivation across individuals, time, and space that dominates much of the literature on rebellion into a highly improbable and potentially misleading enterprise. Given the theoretical problems and the state of the empirical record, a sensible solution for a study of violence is to bracket the question of individual motivations and attitudes and to adopt minimal, yet sensible, assumptions about support.
First, it is enough to assume, following Tilly (1978:201), that launching an insurgency and eventually winning requires only "the commitment of a significant part of the population, regardless of motives, to exclusive alternative claims to the control over the government currently exerted by the members of the polity. "34
Second, it is not necessary to assume stable preferences. There is a dynamic dimension to support, not as an automatic and fixed translation of underlying preferences but as a malleable field of (often strategic) action that corresponds to both preferences and constraints. Indeed, civilians vary the level and the direction of their commitment throughout the war, as suggested by a report written in 1900 by U.S. Major General Elwell S. Otis about the insurgency in the Philippines: "A review of the telegraphic dispatches shows ... that our men were gladly received by the mass of people upon entering the provinces, then later, a portion of the people under insurgent impressment contributed in men and money to drive the Americans out, and finally, that the great majority, gaining confidence, united with our troops to destroy the Tagalo[g]s and the robber bands they directed" (in Linn 1989:29).
Third, this commitment may result from varying combinations of persuasion and coercion. In fact, consistent with J. Scott's (1990; 1985) analysis of peasant attitudes, many accounts of how people collaborate with armed actors point to
33 It is also possible to think of "compounds" that aggregate single motives. Margaret Levi (1997) identifies four models of compliance: habitual obedience, ideological consent, opportunistic obedience, and contingent consent. Contingent consent is a compound that includes the political actor's ability to credibly threaten sanctions, its trustworthiness, the presence of ethical reciprocity between citizens, and the availability of socially provided information. Moreover, even "clean" attitudes, such as willing consent, are compatible with a variety of second-order motivations, which have themselves left a trail of lingering academic debates: class interest (Wolf 1969), narrow individual self-interest (Popkin 1979), or a community's "moral" economy (J. Scott 1976).
34 A reflection of such an attitude in Vietnam is provided by Sheehan (1989:49-50): "While all of the peasantry in the northern Delta did not sympathize with the guerrillas, the majority either favored the Vet Cong cause or tacitly aided the Communists through the silence of a neutrality that worked against the Saigon government. Whether the neutrality was created by fear of guerrilla terrorism or by sympathy made no practical difference: the Saigon government lacked the cooperation of the peasantry, and cooperation was necessary to suppress the Communist-led intervention."
102
The Logic of Violence in Civil War
qualified, cautious, and ambivalent collaboration along the two poles of sympathy and fear. Finnegan (1992:102) recalls a conversation with a Mozambican peasant: "Wouldn't the bandits be recognized and turned over to the police [if they came in town]? The [man] to whom I put this question said, 'Not necessarily. The police are not popular here.' Were the bandits popular, then? 'Not necessarily.'" Consider the point made by a former Vietcong when asked if he joined voluntarily or not: "This is a subtle point. One cannot say that support is voluntary, and one cannot say it is not voluntary" (in Race 1973:129).35 The ambiguous coexistence of acceptance and fear is perhaps best expressed in Toolis's (1997:68) description of fighters in Northern Ireland: "They were the local thugs turned community warriors. Neither of their respective communities would wholly endorse their action but most ordinary citizens protected them by shutting their eyes and ears. No one, Catholic or Protestant, would have informed on their respective paramilitaries out of communal solidarity and for one other very good reason: if the paramilitaries found out they would have shot the informant in the head."
The coexistence of sympathy and sanctions reflects the mix of persuasion and coercion that political actors typically settle upon once they achieve an acceptable level of control. County Armagh has long been a hotbed of IRA support, yet this is also a place where, as a Catholic man put it, local IRA rule means that "nobody speaks out, because if they speak out, they go down a hole" (in Lavery 2005A5). Kenneth Matthews, a BBC correspondent who was kidnapped by Greek Communist rebels in 1948 and, as a result, visited rebel-held areas, reached a similar conclusion during his debriefing by British officials after he was released:
In this large stretch of country there are practically no Government forces of any kind nor any functioning officials of the central government.... Throughout the area the rebels exercise simple but effective administrative control.... As regards the feelings of the population in rebel territory, it is clear that Mr. Matthews has a most vivid impression of an almost universal feeling of what he could only describe as "horror" at the situation in which it finds itself. This does not mean that the rebel rule is a terroristic one. Mr. Matthews thinks that if the population thought that the rebel rule had come to stay, most of them would settle down under it more or less, although they would not like it. He did not think that more than 1 % of the population could be regarded as really in favour of the rebels.36
This last point suggests that deep and unflinching commitment is only required from a few people. These are the "pure, fervent idealists" who occupy a disproportionate position in many journalistic and historical accounts. Yet, an empirical regularity supported by considerable evidence is that only a small minority of people are actively involved in civil wars, either as fighters or active supporters. Lichbach
35 Recruitment of South Vietnamese peasants into the Vietcong army, Richard Berman (1974:198) found, was "neither the spontaneous volunteering often attributed to revolutionary movements nor the conscription of villagers into enforced servitude. Indeed it involved a mixture of coercion and persuasion."
36 "Notes on Conversation with Mr. Kenneth Matthews on the 1st November, 1948," PRO, FO 371/72217/R1237.
Collaboration
(1995:18) cites extensive evidence in favor of what he dubs the "five percent rule," according to which only about 5 percent of the population is made of active and militant supporters. A study of the percentage of combatants in seven insurgencies between 1940 and 1962 suggests an average of 7 percent can be classified as strong supporters, a combined total for both insurgents and incumbents (Greene 1990:75). This observation is common. The English Civil War "was not simply a struggle between gallant Cavaliers and psalm-singing Roundheads ... only a small minority of provincial gentry can be exactly classified in either of these conventional categories" (Everitt 1997:19). In Civil War Missouri, "some clearly identified with one side or the other, maintaining a notion of loyalty of belief and behavior. Many more sought to be disengaged, neutral" (Fellman i989:xviii). Brian Hall (1994:210) argues that the proportion of people that exhibited intense preferences and violent behavior in the former Yugoslavia was in the range of "one-to-five percent" of the population. In Colombia, only a "tiny minority" of civilians "actively collaborate as committed informers or partisans for the armed groups"; in contrast, the bulk of the civilian population seeks to remain neutral (Fichtl 2004:3).
Even in highly polarized environments and under less dangerous conditions, active participation remains low. Elisabeth Wood (2003) estimates that in the areas of El Salvador she studied, a larger minority (about one-third of the peasants who had not fled these areas) supported the insurgents. During the Spanish Civil War, "only a small minority was unconditionally political and identified with parties and unions.... Even the famous milicianos, the volunteer forces that helped to save the Republic when the military rebellion exploded, often had a shaky commitment to the cause" (Seidman 2002:6, 11-12). The total number of fighters in all the Lebanese militias at no time exceeded 30,000, and during fifteen years of war only 90,000 to 100,000 people (close to 3 percent of the population) were ever members of a militia; overall, less than 20 percent of the population was actively involved in supporting one faction or another (Nasr 1990:7). Similar conclusions have been reached about the Bosnian and Chechen wars (Mueller 2004; Claverie 2002:48; Tishkov 1997). Where electoral results are available, they sometimes suggest limited prewar support for very effective insurgents.37
Most "ordinary" people appear to display a combination of weak preferences and opportunism, both of which are subject to survival considerations (Chapter 5).38 Their association with risk-taking minorities tends to be loose and subject to the fortunes of the war and its impact on one's welfare (e.g., Serrano 2002:375; Lison-Tolosana 1983:48). This is the case in both ethnic and nonethnic conflicts,
37 Communist parties in German-occupied Europe are a case in point. There were only 830 Communists in the entire Bosnia Herzegovina at the time of the Axis invasion in April 1941, yet the Communist partisans proved extremely successful there (Hoare 2001:2). Of course, this was true of the Bolsheviks in Russia (Schmemann 1999:208). Insurgents usually start minuscule and grow very quickly. See Berkeley (2001:47); Horton (1998:74); Asprey (1994:537); Stubbs (1989:183); Paget (1967:35); Clutterbuck (1966:5); Barnett and Njama (1966:152); Kitson (1960:126).
38 Lubkemann (2005:504); Raleigh (2002:140); Schmemann (1999:208); Pyszczynski, Greenberg, and Solomon (1997); Malefakis (1996:26-7); Griffin (1976:137).
The Logic of Violence in Civil War
as suggested by the following vignettes from the American Revolution, German-occupied Ukraine, and Lebanon:
What emerges from the British record... is a picture of the great middle group of Americans. Almost certainly a majority of the population, these were the people who were dubious, afraid, uncertain, indecisive, many of whom felt that there was nothing at stake that could justify involving themselves and their families in extreme hazard and suffering. These are the people lost from sight in the Revolutionary record or dismissed as "the timid." With not even poverty to redeem them, they are also passed over by historians who believe that the inert mass of people in any epoch deserve nothing better than obscurity. These people, however, did count, because they made up a large proportion of a revolutionary republic whose very existence depended on counting. (Shy 1976:215-16)
Taking all of the evidence presented here into account, the following overall conclusion seems tenable: the attitude of the civil populace in this area is best described as docile and malleable. With the exception of the partisans themselves and small numbers of pro-Soviet and pro-German activists who were willing to risk death in order to serve their causes, most people seem to have been willing to obey whichever antagonist appeared most credible at a given time. (T. Anderson 1999:622-3)
Clearly, there were far more victims than perpetrators in the Lebanese civil wars, and most people simply wanted the killing to end. They might have subscribed to patriarchal and kinship values that encouraged ethnic or confessional identity, but unless they were suddenly swept up in both emotion and circumstance they were unlikely to become directly involved in the fighting. (M.Johnson 2001:230)
4.4. FORMS OF COLLABORATION AND DEFECTION
Political actors seek the exclusive and complete collaboration of all civilians. In practice, they are looking for active collaboration from a small number of dedicated supporters, and passive but exclusive collaboration from the population at large; they also seek to prevent civilians from collaborating with their rivals. They also prefer exclusive but incomplete collaboration to nonexclusive collaboration (such as neutrality and hedging); obviously, they prefer a low level of collaboration to no collaboration at all. The minimum core of collaboration is generally nonbetrayal to the enemy (Stubbs 1989:2; Leites and Wolf 1970:10). Insofar as civil war tends to be a polarizing process, collaboration and noncollaboration tends to be zero-sum.
The flip side of collaboration is defection, which can be disaggregated into at least three types: noncompliance, informing, and switching sides (Table 4.1); the last two are clearly acts of collaboration with the rival actor, although noncompliance is often construed as such. In this book, I understand defection as active collaboration with the rival actor.
Noncompliance can be public and private, collective and individual; informing is usually private and individual; and switching sides is both individual and collective but usually public. Noncompliance includes actions such as complaining and critiquing, tax evading, shirking, and fleeing. It may be individual or collective (an entire village shirks), private or public. Economic considerations and survival are usually the main motives. While it is the most benign form of noncollaboration, if
Collaboration
table 4.1. Types of Defection
Type
Scope
Noncompliance Informing Switching sides
Individual and collective Individual
Individual and collective
Public and private
Private
Public
left unpunished, noncompliance, may trigger cascades of more serious instances of noncollaboration.
Informing is the act of supplying information about one side to its rival; it is typically a private act that presupposes that information about one side and access to the other are simultaneously possible - something that entails an absence of front lines.39 While informing indicates some form of association with the political actor to whom information is being directed, it differs from switching sides in that it is usually a private act that requires secrecy. It is also individual rather than collective, and its effectiveness (or damage) tends to be unrelated to the number of informers. In Vietnam, the Vetcong were satisfied when they had one or two secret collaborators in hamlets controlled by the government (Race 1973:147).
Informing matters not just because it provides a direct military advantage (e.g., preventing or facilitating ambushes), but primarily because it solves the identification problem. A positive externality is that knowledge among the population that one side has crucial access to information undermines the population's willingness to collaborate with the other side.
The motivations behind informing, like those behind collaboration in general, are mixed. They may reflect genuine political preferences, expectations of personal gain, private grudges, coercion and blackmail, or survival considerations.40 And like defection on the whole, informing tends to be responsive to risk. The effective use of violence may successfully deter informing.
Amilcar Cabral, a nationalist leader from Guinea-Bissau, once said that a revolution is like a train journey. At every stop, some people get on, and other people get off (in Finnegan 1992:133). Switching sides is common in civil wars and involves both individuals and entire communities that openly start collaborating with a rival political actor. It is usually a public and visible act: individuals may defect from one army to its rival or entire villages may set up a militia and openly signal that they have defected. Switching is widespread in civil wars; rebel "turncoats" have been used extensively by incumbent forces, and they are usually
39 The failure to provide information one possesses is seen as a consequential act of defection. An officer of the Sandinista army in Nicaragua explained: "What kills directly is the tongue, because if no one tells me that someone is waiting there armed, and I don't notice anything, then I die [in an ambush]" (in Horton 1998:210).
40 As a Basque man told Zulaika (1988:83), "If you belong to the side of the losers, you have only one possibility of passing to the winners' side - to inform against your friends. In this way you gain power over them, and money."
io6
The Logic of Violence in Civil War
associated with considerable violence.41 During the Russian Civil War, entire local revolutionary committees "which had been appointed by the Bolsheviks from among the local people went over to the side of the insurgents"; in fact, "it was not uncommon in central Russia and especially Ukraine that the same individual served in several or all armies, Red, White, and Green" (Brovkin 1994:105, 418). In China, many Communists joined the Nationalist side, especially after losing out in factional conflicts; they were the rebels' worst enemy, "for they knew the guerrillas' ways and were thirsty for revenge" (Benton 1992:475). In Vietnam, "defections from one side to the other occurred frequently as did shifts in the loyalties of villagers" (Berman 1974:31), and the Vietcong "counted defection as one of their greatest problems" (Moyar 1997:250-1). Again, motives vary widely42
Those switching sides provide obvious services: as sources of information, as helpers in getting their former colleagues to defect, and, by the sheer fact of their existence, as propaganda. Ponciano Del Pino (1998:169) also notes that having experienced their former organization from the inside, they are able to overcome the fear that such organizations often invoke in outsiders. Because switching sides is a dramatic and consequential act, the harshest punishment appears to be reserved for those who switch at crucial junctures in the conflict - especially for village leaders or even entire villages. Robert Thompson (1966:25) reports that when the Vietcong regained control over a village that had defected to the government, they seized "the headman and his family, disemboweled his wife in front of him, hacked off his children's arms and legs and then emasculated him."
4.5. THE INSTITUTIONAL SETTING OF COLLABORATION
A rather unexplored aspect of irregular war concerns the institutional context within which interactions between political actors and civilians take place, what I describe as the "meso" level. Sometimes, this interaction is informal. More commonly, however, this process is institutionalized and takes two basic forms: militias and committees. Being armed, militias are typically empowered to use violence directly, whereas committees are not. Political actors rely on both, though incumbents appear to prefer militias and insurgents committees, a preference that may be only due to the availability of weapons.
41 Myers (2oo5:A4); Hedman (2000:132-3); Bearak (2000); Gossman (2000); Mahmood (2000:83); Clayton (1999:50); McKenna (1998:180-1); Del Pino (1998:169); Zur (1998:106-7); Starn (1998:244); Gacemi (1998); Berlow (1998:182); Moyar (1997:167); Cann (1997:101-2); Swedenburg (1995:156-64, 195); Le Bot (1994:176); Stoll (1993:140); Hamoumou (1993); Cribb (1991:143); Rosenberg (1991:46, 92); Blaufarb and Tanham (1989:63); Flower (1987:115); Crow (1985:170); Calder (1984:158); Henriksen (1983:136); Salik (1978:105); Heilbrunn (1967:69-70); Paget (1967:91-2); Clutterbuck (1966); Kitson (i960).
42 Among the reasons for defecting from the Vietcong, Moyar (1997:111) includes the following: "disagreements over promotions or demotions, mistreatment of one's family members, accusations about collaboration with the GVN, and animosity between subordinates and superiors."
Collaboration
4.5.1. Militias
Militias are primarily a political rather than a military institution. They are part of a strategy of local rule and state building.43 As an Algerian argued about the country's militias: "People can't eradicate the terrorists without the army, and the army can't exterminate the terrorists without the people" (in Peterson 1997b). In Guatemala, the main objective of the civil patrols, as militias were known, was "to inform on guerrilla sympathizers in the community" (Carmack 1988^63). The primary purpose of militias is "population control" (Jones and Molnar 1966:25). While the individual militia members may be focused on defending their villages or families, the fact that they are permanently present in their villages and are operating in places they know well allows incumbents to tap into private information.
Although insurgents rely on local militias (e.g., Geffray 1990; Stubbs 1989:87-8), the term is usually associated with incumbents, who use them as auxiliaries.44 The various irregular and semiregular groups of anti-rebels, referred to by such diverse names as paramilitaries, militias, death squads, and home, civil, or village guards, are the "opposite" face of the rebels (Zahar 2001; Rubio 1999:20) -"counter-gangs" in Frank Kitson's (i960) formulation. The formation of militias along with the creation of "fortified villages," often described as "local" or "self-defense" programs (Armstrong 1964:30), is an essential part of counterin-surgency efforts (Hedman 2000:133; Barton 1953). Typically, militias are formed at the local (usually village) level, comprise local men (and sometimes women), and their activities are closely tied to their locality.
Militias often reach massive size. It is estimated that by 1985, 1 million rural Guatemalans were involved in patrolling their communities (Warren 1998:89). Militias are also prevalent in ethnic conflicts where states are often able to provoke interethnic defection (Kalyvas 2004). For example, the Indian security services in Kashmir have been successful in getting Muslim militants to switch sides and become "countermilitants" (called "renegades" by the locals and "friendlies" by the government) (Gossman 2000:275). Militias are also a key tool for enforcing occupation. In fact, occupiers are surprised to discover how easy it is to recruit natives and often find that they get more recruits than they have places to fill (e.g., Finley 1994:29) 45
43 This is visible in their functions. In Guatemala, militias exercised judiciary powers. Whereas before the war, people would go before the local mayor, who served as a judge of first instance, or to a higher-level judge in the provincial capital to settle differences, during the civil war they appealed to local "civil patrol" leaders (S. Davis 1988:29-30).
44 In this respect, militias in irregular war differ from those in symmetric nonconventional ones, where they often take on an autonomous role.
45 More than 1 million Soviet citizens fought on the German side and the total number of collaborationists was about double the number of partisans (Klonis 1972:91). In Algeria, more Algerians fought on the French side than against it: "At no time from 1954 to 1962 did the numbers of Algerians fighting with the ALN for independence match the number of Algerians fighting on the French side" (Home 1987:255). Half of the Portuguese soldiers fighting against independentist rebels in Guinea and two-thirds in Mozambique were natives. By 1974 the independendist rebels had reached a peak of 22,000 men in Angola, as opposed to 61,816 locally recruited troops fighting
io8
The Logic of Violence in Civil War
Militias can be costly. Although they are formed to engage primarily in "protective violence," they often mete out "predatory" and abusive violence, including extortion.46 Their reputation for atrocity is well established.47 They may also cause an escalation in violence because they use their power to fight personal or local conflicts 48 "Missouri militiamen had a great need to exact revenge against their rebel sympathizing neighbors, and they knew which scores they wished to settle" (Fellman 1989:129). Many cases of abuse were reported from the Kurdish areas of Turkey, where the Turkish government formed progovernment village militias to fight a Kurdish insurgency. In the village of Ugrak, Vick (2002 Ai 8) reports, the state armed the Guclu family, which, by most accounts, "wielded no particular clout until the state made them the law.... The policy had the effect of emptying the village of everyone not named Guclu. The families who left describe being pushed off their land by neighbors who used police powers to commandeer better land and bigger houses." As one of their victims put it, "These people given weapons by the state use the weapons for their own benefit." The village guards, "whom many locals describe as mafias... do as they please under the color of law, enjoying virtual immunity from prosecution, according
to human rights activists and local residents____Reports of rape at the hands of
village guards are rising, and critics describe leaders of prominent clans using guard status to cement their already considerable power, in some cases running smuggling rings unchallenged by state authorities afraid to try to disarm them."
Ironically, the local character of militias that permits the gathering of information so necessary to political actors may also turn them into indiscriminate weapons with counterproductive effects. For example, a British journalist "had no doubt at all," in 1948 Greece, "that the activities of the Right Wing bands ... are responsible for the rebels' strength. He said that recruits are continually coming in and that he saw many arrive himself. He was sure from his conversations with
on the Portuguese side, and "there were always more African volunteers for the Portuguese troops than there were openings" (Clayton 1999:51-4; Cann 1997:103-4; Henriksen 1983:60-1).
46 Trinquier (1964:34) recognizes that in such settings "abuses are always possible." The Confederate rangers created by the Virginia legislature during the American Civil War "used their recognition by the state as a license to steal and murder. They took assumed Unionists from their homes, tried and convicted them on the spot, and meted out whatever punishment struck their fancy. They were not very discriminating in their victims, however, and Confederate sympathizers in the region soon began to ask for protection from their 'protectors'" (Paludan 1981:52). Informers working for the Japanese in the Philippines during World War II would often blackmail people into bribing them by promising nonbetrayal (Lear 1961:27). The same occurs in China (Seybolt 2001:218). In Malaya, "newly recruited officers and sergeants without proper training and a rapidly expanding and poorly supervised rank and file provided fertile ground of corruption. The Emergency made extortion and bribery much easier for those who wished to line their own pockets. If bribes were received there were no arrests, but an uncooperative 'donor' could always be shot as a communist sympathizer" (Stubbs 1989:72).
47 Roldan (2002:161-2); Zur (1994); Mason and Krane (1989:185); Calder (1984:130); Perry (1984:433); Kerkvliet (1977:196); Shy (1976:187).
48 The most vivid description in this respect is the one by Paul and Demarest (1988) of the events in the Guatemalan town of San Pedro la Laguna during the civil war. See also Dupuy (1997:158); Fellman (1989:185); Stubbs (1989:72); S. Davis (1988:28); Kerkvliet (1977:196).
Collaboration
them that Right Wing excesses and arbitrary and unjust acts of Government representatives are still rapidly swelling the rebel ranks in the Peloponnese."49 Because of this tendency, political actors tend to crack down on excesses (e.g., Paul and Demarest 1988).
Because militias threaten insurgents, they quickly become the insurgents' primary targets. Many massacres committed by insurgents take place in villages whose denizens defected by joining newly formed militias (Kalyvas 1999). Hence, even if villagers initially joined the militias under coercion, they may quickly learn to fear and hate the rebels. In Guatemala, Stoll (1993:100) shows, the army relied on this mechanism to solve the problem of trust and prevent militiamen from handing their weapons to the rebels: the militiamen were initially not armed until enough blood had been spilled - militiamen killing guerrillas and vice versa - to confirm that they were on the army's side.50 Overall, there is a consensus is that, in the end, militias are a rather effective weapon against rebels.51
4.5.2. Committees
Local, usually village-based committees handle and screen information for armed actors.52 Such committees can be found in most settings. In revolutionary France, "vigilance" committees of local patriots "were to be found in every city and town down to most small bourgs, as well as in significant numbers of villages in some areas"; they were "endowed with arrest power and became the lower rung on the ladder of revolutionary repression that led through prison to the revolutionary tribunal and up the steps of the guillotine" (Lucas 1997:33). In revolutionary Russia, Chekas drew the lists of persons to be arrested (Werth 1998:172). During the American Civil War, "vigilance committees" were set up in frontier districts (Ash 1995:123). In Kenya, the Mau Mau insurgents setup local committees that "gave the orders for killings, raids, money collection, and recruitment"
49 "Notes on Conversation with Mr. Kenneth Matthews on the 1st November, 1948," PRO, FO 371/72217/R1237.
50 Trust is the key issue: as Robert Thompson (1966:136) recommended about Vietnam, "where the people are reliable and can be trusted, then the number of persons armed should be sufficient to defend the hamlet. Where the people are not yet to be trusted, then no one should be armed... There can be no half-way measures where the people are not trusted. This will only lead to treachery and disaster." I was able to find few reports of militiamen collaborating with rebels. Henriksen (1983:159) reports evidence of direct contacts between residents of aldeamen-tos (strategic hamlets) and insurgents. He adds that "even members of the self-defense militia collaborated with guerrillas, feeding, informing, and occasionally turning firearms over to them." Likewise, it has also been reported that some Chechen pro-Russian militiamen collaborate with the rebels (Nougayrede 2002), while FARC guerrillas appear to have infiltrated progovernmental peasant militias (S. Wilson 2003). The most striking case was investigated by Lacoste-Dujardin (1997), who tells how an Algerian Berber village that was armed by the French joined the FLN on the basis of advice from an anthropologist. It is surprising, however, how unusual these cases seem to be. The problems caused by lack of trust run the other way as well (Race 1973:256).
51 Richards (1996:171); Blaufarb and Tanham (1989:79); Linn (1989:54); Jones (1989:273-5); Home (1987:255); Cooper (1979:115); Race (1973:270); Paget (1967:91-2).
52 Sometimes militias call themselves committees, as in Nepal (Sengupta 2005a). I use the term here to refer to small, information-processing groups rather than armed groups.
no
The Logic of Violence in Civil War
(Kitson 1960:45), while in Malaya the British formed similar local "committees of review" that screened civilians arrested during sweeps (Stubbs 1989:74). The Filipino rebels of the NPA relied on committees to determine assassination targets: these committees based their decisions "largely on complaints from sympathizers and rebel intelligence" (G. Jones 1989:249). The same was true in El Salvador (J. L. Anderson 2004:136). Likewise, the South Vietnamese set up "screening committees" consisting of officials at the hamlet, village, and district levels, which reviewed evidence pertaining to the activities of people suspected of collaborating with the Vietcong (Moyar 1997:204). At the same time, the Viet-cong set up a vast network of committees beginning at the village level (West 1985:21) in order to ensure "that critical decisions... were made by local people, with relatively more flexibility and with some sensitivity to the demands of the particular situation" (Race 1972:164); according to Berman (1974:50), "such a structure placed prime responsibility on low-level cadre." In exchange for their monitoring and information, local agents obtain a valued immanent good: the power to rule over their communities.
Although there is substantial evidence regarding the existence of such committees,53 we know little about how they actually operate. Perhaps their most important feature is that they often have a role in determining what violence is visited on the locality in which they operate, but how this power is wielded varies. In many cases, these committees have veto power over the use of violence in their community. I return to this issue in Chapters 7 and 8.
4.6. CONCLUSION
This chapter has specified the first part of a theory of irregular war by systematizing well-known but scattered insights about insurgency and introducing a novel conceptualization of sovereignty in irregular war. After raising various problems associated with the concept of "popular support," I discussed identification, the key issue facing political actors. In the following chapter, I keep my focus on collaboration and explore its relationship with control.
53 Fitzpatrick and Gellately (1997); Rosenau (1994:315); Rosenberg (1991:199); Geffray (1990); Gross (1988); Henriksen (1983:148); Clutterbuck (1966:6).
5
A Theory of Irregular War II
Control
Non, decidement, on ne tue pas les mouches a coups de marteau (We definitely don't kill flies with hammers).
Lieutenant Colonel Bigeard, French army, Algeria
This is a political war and it calls for discrimination in killing. The best weapon for killing would be a knife, but I'm afraid we can't do it that way. The worst is an airplane. The next worst is artillery. Barring a knife, the best is a rifle - you know who you're killing.
John Paul Vann, U.S. adviser in Vietnam
This chapter analyzes the relation between collaboration and control and argues that military resources generally trump the population's prewar political and social preferences in spawning control. In turn, control has a decisive impact on the population's collaboration with a political actor. However, the amount of military resources required for the imposition of full and permanent control in a country torn by civil war is enormous and, therefore, typically lacking. This places a premium on the effective use of violence as a key instrument for establishing and maintaining control - and thus for generating collaboration and deterring defection; in turn, effective violence requires discrimination.
5.I. THE ALLOCATION OF COLLABORATION
A robust empirical observation is that the allocation of collaboration among belligerents is closely related to the distribution of control, that is, the extent to which actors are able to establish exclusive rule on a territory. This relationship can be formulated as a hypothesis: the higher the level of control exercised by a political actor in an area, the higher the level of civilian collaboration with this political actor will be.
An immediate concern is the direction of causality. Does control spawn collaboration or is it the other way around? For example, Brovkin (1994:126) observes
in
112
The Logic of Violence in Civil War
about the Russian Civil War that "an army of 100,000 could not possibly have taken control of a territory with a population of 40 million people in three months if there had not been a universal resentment of the preceding administration." Likewise, it is claimed that "the most important locational factor" of insurgent base areas "is that political objectives clearly override purely geographic (terrain) advantages" (McColl 1967:156).
There is little doubt that collaboration and control are self-reinforcing. More objectionable, however, is the view that control emerges exclusively from collaboration and never shapes it; similarly objectionable is the "median voter" view of civil war, namely that patterns of control during the course of the war reflect majoritarian preferences, especially as reflected in the prewar period. Indeed, it is not necessarily the case that political majorities enjoy a military advantage over minorities; in fact, the opposite may be true (Massey et al. 1999). The prewar political preferences of the Spanish population, with the partial exception of Catalonia and the Basque country, turned out to be a poor predictor of the distribution of control among Nationalists and Republicans during the first months of the war (Derriennic 2001:168); in Bosnia, the Muslims had a clear numerical advantage but were unable to translate it into a military one. More important, this argument disregards the effects of the war and fails to account for the many instances of preferences that are endogenous to war, as discussed in the previous chapter.
A more encompassing and dynamic hypothesis is that initial patterns of control are predicted by some combination of prewar preferences and existing military resources,1 but as the war evolves, control is more likely to trump prewar preferences in determining collaboration. Even though collaboration and control are interlinked, it is possible to disaggregate their interaction into a simplified temporal sequence. An example is as follows: at tI? an insurgent group gains control of a locality through the successful use of military means at to, as a result of existing popular preferences, or through a combination of both. As a result, collaboration with that group at t2 increases. However, the government army may counterattack, chase the insurgents out, and impose its own control through purely military means at t3. Now, this will spawn collaboration with the army at t4; even though the population may have had a preference for the insurgents. If the army maintains its control for a long time, the preferences of the population may possibly shift "endogenously" toward the army at t5.
1 Geffray's (1990:53-4) analysis of the arrival of RENAMO in theMariri area of theNampula district in northern Mozambique points to this reinforcing process: the local chief Mahia had been alienated by the policies of the government, hence he welcomed the RENAMO insurgents before they were in a position to protect him; at the same time, however, RENAMO decided to come to this area because of the convergence of favorable geographical features: distance from cities and towns, thick forest, water resources, and proximity of a mountain with many caves. Ultimately, it was the presence of RENAMO that "gave people the military means to place themselves outside the state's scope." In Geffray's (1990:93) words: "Tens of thousands of people moved under the protection of Renamo's weapons in this area... outside the range of the Frelimo state. These populations could have never entered in a state of insurrection on their own, without the intervention of an armed force capable to maintain Frelimo's forces at a distance."
Control
This is a process reminiscent of the cuius regio eius religio principle, whereby entire populations became Protestant or Catholic following their ruler's choice. Michael Seidman (2002:40) calls the widespread tendency during the Spanish Civil War to side with the camp that dominated the city or region where one lived "geographical loyalty." Finnegan's (1999:50) observation about the dynamics of popular support in Sudan points to the same direction: "People's political views would be highly contingent on the power arrayed around them."
The point is, in short, that although control and collaboration interact, control may trump the political preferences of the population in generating collaboration during the war. This insight is consistent with arguments that stress state capacity (Fearon and Laitin 2003; Coleman 1990:479) and the related observation that insurgencies are likely to develop and acquire civilian support where state control has declined or collapsed (e.g., Del Pino 1998:170; Skocpol i979)-2 The implication is that prewar popular preferences may be an inaccurate predictor of the distribution of control during the war.3
Highlighting the importance of control in no way implies that coercion is the only factor or that popular grievances are irrelevant. Thousands joined the South African-financed RENAMO insurgents in Mozambique because it allowed them to destroy the deeply unpopular new villages created by the Mozambican government. However, they did so only after the insurgents were able to challenge the government and establish local military control, thus shutting out the army (Geffray 1990:39). Popular dissatisfaction with new villages was equally strong in Tanzania, but because no insurgent group challenged the state, this dissatisfaction was not expressed in the context of an insurgency.
Because emphasis on how control may trump prewar preferences and shape collaboration during the war may appear questionable at first sight, I provide a few examples to demonstrate the argument's plausibility before specifying a set of causal mechanisms that translate control into collaboration.
During the Spanish Civil War, many leftists joined right-wing militias (and vice versa) because they found themselves on the wrong part of the front line and wanted to survive (Cenarro 2002:75). Young French men wishing to avoid labor conscription during the Second World War were more likely to join the collaborationist Milice if they lived in the cities, and the Resistance if they lived in
2 James Coleman (1990:479) dubs this "power theory," contrasting it to the "frustration theory." This is particularly visible in the case of weak occupation or colonial administrations and explains the ease with which insurgencies are able to spread. For instance, Cann (1997:21) points out that the insurgency in Angola during the Portuguese colonial rule began in areas where the Portuguese presence was "so sparse that it was physically impossible for [administrators] to maintain anything but the most casual control over their districts." As an American analyst pointed out about the inability of Colombia's military to control the country, "The military can't substitute for the presence of the state" (in Forero 2ooi:A3). Similar points are made by Evans (1985:211) on Virginia during the American Revolution and Horton (1998:126) on the contras in Nicaragua. Obviously the lack of state presence indicates a key condition for the emergence of insurgencies rather than their timing - why the breakout rather than when. This suggests that joining an insurgency entails lower risks than usually assumed when incumbent forces are absent or very weak (e.g., Degregori 1998:130; Horton 1998:126; Herrington 1997:29).
3 Elliott (2003:408); Geffray (1990:39); Li (1975:188).
ii4
The Logic of Violence in Civil War
the countryside.4 Two men explained why they had joined the pro-Israeli South Lebanese Army (SLA) (in Sontag 2000A1): "We grew up on guns. Guns were muscles. And in this area, the guns were in the hands of the S.L.A."
Practitioners are well aware of this point. Mao Zedong (in Bruno Shaw 1975:209) argued that "the presence of anti-Japanese armed forces" was the first condition for the establishment of a base area. "If there is no armed force or if the armed force is weak," he pointed out, "nothing can be done." A 1968 CIA report from Vietnam (in Moyar 1997:321) observed that "most of the people respond to power and authority, whether that of the Vet Cong or the GVN." A French general described how, following a French military operation in Algeria, in 1959,
The strong rebel zones of Beni Merai - Babor and Arbaoun-Tamesguida have been seriously dismantled. The rebel elements have either retreated into surrounding sanctuaries or have been broken down into small groups that avoid contact. The rebel political organization, lacking the support of the military apparatus, is partly neutralized and its members are hiding. The logistical infrastructure is deeply disorganized. The population has been freed to a certain extent from the rebel constraint and has begun a clear move back toward our side.... The population's shift is, however, not irreversible; to maintain and accelerate it, we must pursue simultaneously and at every level the destruction of the bands, (in Faivre 1994:148-9)
Ronald Wintrobe (1998:45) identifies two instruments through which a dictator can accumulate power: repression and loyalty. Loyalty can be acquired in a variety of ways, including the provision of material benefits (especially monopoly rents) and ideological appeals. Due to its multilateral dimension, civil war is a context that turns the permanent and stable acquisition of loyalty into a very difficult enterprise. Once the war is underway, war-related resources such as violence tend to replace the provision of material and nonmaterial benefits, inducing individuals, for whom survival is important, to collaborate less with the political actor they prefer and more with the political actor they fear; in other words, the provision of benefits loses out gradually to the effective use of violence. After all, violence is a weapon that is easy to use, yet "promises returns far out of proportion to the amount of time, energy and materials" invested by political actors who rely on it (Thornton 1964:88); and political actors would rather be disliked but feared than liked but not feared when their rival is feared.5 An American
4 Using judicial data, Jankowski (1989:123-4) found that the largest contingent ofMiliciens, between a quarter and a third, had signed up to escape labor conscription to Germany, some even as an insurance against receiving a summons, and some the very day they received it. As one of these men put it, "So I joined the Milice to avoid going to Germany, because it was simpler for me, because it allowed me to be close to my [family] and my work. I didn't have all the uncertainty [that would arise from] leaving for an unknown maquis with possible reprisals against my family." The percentage of those who joined out of conviction is estimated by Jankowski to have been close to 5 percent; another 5 percent joined under pressure from family and friends, another 10 percent to take advantage of jobs and privileges, and the rest for multiple reasons.
5 The worse outcome, of course, is being both disliked and not feared. West (1985:157) describes such an instance. Following a failed Vietcong mortar attack against a village, the local police chief "said the story that the Viet Cong had tried to kill the villagers would be known in every hamlet
Control
journalist covering the counterinsurgency in Iraq asked the leader of a unit of Iraqi commandos, Colonel Adnan Thabit, about a local sheik, after the two had met in Samarra and the former had threatened the latter. Adnan replied that "it is not important whether he is with us or against us. We are the authority. We are the government, and everybody must cooperate with us. He is beginning to cooperate with us" (in Maass 2005:56). Richard Nixon's adviser Charles Colson conveyed this insight crudely but clearly: "When you have them by the balls, their hearts and minds will follow" (in Chang 1992:403).
Of course, fear alone does not suffice to sustain rule in the long term (Wintrobe 1998:37; C. Friedrich 1972:60); however, it operates as a first-order condition that makes the production of loyalty possible. Leaving aside the argument that most benefits a political actor is able to deliver will not be sufficient to offset a high risk of violent reprisals at the hands of the rival actor (Mason and Krane 1989:179), one can still note that material benefits become increasingly scarce during civil wars.6 These wars tend to deplete local economies and thus produce a rise in poverty and a reduction of goods available for distribution; this may turn people away from the political actor with whom they initially sided (Chingono 1996). In response to such shifts, even insurgents may become more coercive vis-a-vis their social base, "alienating the very people whose support they needed most," as Berlow (1998:179) noted in the Philippines: "Villagers would be asked to choose: pulo ukon polo, ten pesos or the barrel of the gun." "As the war dragged on," writes David Elliott (2003:348) about Vetnam, "even the poor peasants began to question the benefits of Party membership as compensation for sacrifices. The burdens increased and the rewards decreased." He shows that wealthier peasants fled their villages, in the process freeing land for the Vetcong to redistribute; "but the very insecurity that had made this land available also made it dangerous and unprofitable to cultivate____Thus the land question diminished in importance, and the daily struggle for survival replaced land as the most critical issue for the rural population" (2003:521). Although the Vietcong used mainly persuasion to recruit followers and fighters between 1957 and 1962, they became more coercive at later stages, when they often relied on an overt and compulsory draft (Berman 1974:50). In Peru, Sendero's sanctions "became increasingly cruel as the years went on" (Del Pino 1998:185). The Rhodesian intelligence chief (Flower 1987:122) describes his government's changing belief "that there had to be less 'carrot' and more 'stick'" as a response to the rebels' increasing brutality. Ideological benefits also lose much of their initial power. The fiercer and longer the conflict, the more likely that "limiting damage" will prevail for individuals over "positive" motivations, such as getting benefits or acting according to ideals (Leites and Wolf 1970:127).
the next day. That they had tried and failed was the worst possible combination for them. The villagers had been given the most powerful reason not to like them while not being made to fear them more."
6 A partial exception may be the local production of illegal goods (such as opium or coca) and primary commodities, especially minerals, which normally would accrue to the central government.
n6
The Logic of Violence in Civil War
5.2. SURVIVAL
Reporting from German-occupied Greece in 1944, a British officer wrote: "None is ever free from the struggle for existence: everything else is secondary to it."7 As violence becomes the "main game in town," survival becomes increasingly central for civilians (e.g., Kheng 1983:173). This is particularly true of peasants whose everyday attitude has been described with terms such as "pragmatism," "fatalism," or "resistant adaptation" (Del Pino 1998:178; Herrington 1997:29; Siu 1989:113).
Thucydides (3:83) describes war-torn Corcyra as a place where "everyone had come to the conclusion that it was hopeless to expect a permanent settlement and so, instead of being able to feel confident in others, they devoted their energies to providing against being injured themselves." Fatigue and suffering, the natural consequences of protracted war, effectively undermine preferences and sympathies. Civilians, as Fellman (i989:xviii) nicely puts it, become numb, "separating their consciences from their actions." By 1781 the typical settler of North Carolina, "whatever his initial loyalties, felt a profound need for order and regularity in his daily affairs. For some, the savagery of war begot more savagery, but for most it fueled contrary yearnings for peace and stability" (Ekirch 1985:110). All over the country, people "got angry when British or Hessian or Tory troops misbehaved, but they also grew weary of being bullied by local committees of safety, by corrupt deputy assistant commissaries of supply, and by bands of ragged strangers with guns in their hands calling themselves soldiers of the Revolution" (Shy 1976:13). Survival was similarly a key consideration in the areas affected by the guerrilla war during the American Civil War. As a Tennessee woman wrote in her diary, in 1865 (in Ash 1995=204): "I can see every day people are for them selves and no boddy else.... [M]ost [of the] people have turned out to steal and lied[,] not many that care for any one but themselves." Surveying extensive evidence from Civil War Missouri, Fellman (1989:46, 49) reaches a similar conclusion: "It is my clear impression that there were... more survivors than heroes - if maintaining loyalty under these circumstances would be the appropriate test of moral probity.... It made more sense to be a living liar than a dead hero, and the stakes were that great." Similar sentiments are expressed in a letter from the Russian region of Saratov that was intercepted by the Cheka in 1921: "The number of arrests is growing in Saratov. Several university professors were arrested. The average inhabitant sits quietly and only curses the Communists, but is so
7 "Woodhouse report on the situation in Greece, January to May 44 (5 July 1944)," PRO, FO 371/43689/R10469.
8 It has been argued that peasants do not use an economic "maximizing" way of reasoning and that their behavior is irrational as far as land, loans, "fair prices," and income are concerned (Shanin 1975:273). Although this may be true, it does not contradict the fact that when it comes to survival, most peasants, most of the time, prefer to live rather than die and act accordingly. In fact, this claim is consistent with survival-maximizing behavior, which is closer to risk aversion than income maximization. Indeed many studies have found that peasants generally maximize security and minimize risk (Kerkvliet 1977:255). For example, Siu (1989:113) points out that what mattered most for Chinese peasants during the Japanese occupation was "secure livelihood; they hoped at least for political stability and social order, regardless of the slogans the new leaders proposed."
Control
117
cowardly that when he reads the papers that are posted, his face takes on a loyal expression as if he might see a Bolshevik who might suspect him of disloyalty" (in Raleigh 2002:393).
Compare these examples with behavior in more recent wars. Truman Anderson's (1999:623) detailed account of partisan warfare in the Nazi-occupied Ukraine concludes that "the pragmatic, day-to-day calculus of personal survival played a much more important role than did either pro-German sentiment (rooted in Ukrainian regional hostility to the Soviet regime) or Soviet patriotism." Likewise, the Spanish anti-Franco guerrillas of the 1940s dealt with peasants who "preferred to eat than fight for their freedom, while being favorable to whatever side was in control" (Serrano 2002:374). Even when people have a strong preference for one side over the other, they may find that the circumstances make collaboration exceedingly difficult. In his largely autobiographical novel, the Italian writer Beppe Fenoglio (1973:380) includes the following retort of an Italian peasant to a partisan, following a successful Fascist raid: "We know that you are better than them, we know. But we are afraid, we live in perpetual fear." The internal documents of the Chinese Communist Party suggest that, in contrast to official rhetoric, "the number of willing heroes always decreased precipitously as the risks of martyrdom increased. They repeatedly assert the elusiveness of heroism" (Hartford 1989:112).
By the end of the Biafran War, the majority "of the Biafran people were no longer enthusiastic about the war. What they cared for most was their own 'survival'" (Essien 1987:151). As a Vietnamese journalist remarked: "After twenty years of this war, there is no right cause, no ideal. Neither side can speak in the name of anything in this endless agony. The only right cause that remains is the cause for settling the war urgently" (Chung i97o:xi). "Other than risking death or fleeing from their homes," Nordstrom (1997:52) reports from Mozambique, "the villagers had little option but to meet the demands of each passing group as best they could." "Possibly the most astute observation on the ideology of the violence-afflicted citizens," she adds (1992:266), "came from a young man living in Beira, Mozambique: 'the only ideology the people have is an anti-atrocity ideology.'" As a Chechen woman put it: "Anything but war. I would like to live the rest of my life in peace. I would agree to live on tea and bread. Just anything but war" - a feeling confirmed by a humanitarian worker: "The suffering is so intense and the suffering is not about politics. The vast majority of people we were meeting, they wanted to stay alive; they wanted a life in which their houses weren't bombed and there wasn't chaos and shooting on the street. Politics was not nearly as much on their mind as staying out of harm's way" (Wines 2003 A3; Gall 2001:25). "We need to have peace," said a man from Darfur. "We have suffered too much for this war" (in Polgreen 2005A3). In sum, the inescapable result is that, like in the Spanish Civil War, "the low reservoirs of popular commitment [are] quickly drained" (Seidman 2002:27).
The combined effect of the reduction in available benefits, the increasing role of violence, and the civilians' orientation toward survival is a situation in which effective threats translate into collaboration. In turn, the effectiveness of threats hinges on control.
n8
The Logic of Violence in Civil War
5.3. HOW CONTROL SHAPES COLLABORATION
The anecdotal empirical record provides substantial evidence that control spawns collaboration independently of prewar patterns of support. First, there is evidence showing that collaboration follows the spatial variation in control. This point was made explicitly by the writer of an economic report of a Greek village: "The whole of this area, being plainsland, was not suited to guerrilla warfare, which adapted itself mostly to mountainous terrain. Besides this, the village was relatively safe due to its nearness to Salonica, a large Army center. The leftist side was rather pronounced but no move was made by anyone so there was no visible result" (Tchobanoglou 1951:1). Consider Shy's analysis (1976:178) of the distribution of British loyalism (Toryism) during the American Revolution:
What appears as we look at places like Peterborough, where Tories are hardly visible, and at other places where Toryism was rampant, is a pattern - not so much an ethnic, religious, or ideological pattern, but a pattern of raw power. Wherever the British and their allies were strong enough to penetrate in force - along the seacoast, in the Hudson, Mohawk, and lower Delaware valleys, in Georgia, the Carolinas, and the transappalachian West -there Toryism flourished. But geographically less exposed areas, if population density made self-defense feasible - most of New England, the Pennsylvania hinterland, and Piedmont Virginia - where the enemy hardly appeared or not at all, there Tories either ran away, kept quiet, even serving in the rebel armies, or occasionally took a brave but hopeless stand against Revolutionary committees and their gunmen.
During the Civil War many Unionist sympathizers in the Appalachian counties of North Carolina ended up supporting the Confederacy: "Because of the influence of secessionists, it was wise to decide that one's sympathies lay with the South. Faced with the presence of Confederate power and the lack of armed organized Federal protection, reasonable men not seeking martyrdom might become southern patriots" (Paludan 1981:64). In the eastern regions of the Dominican Republic, the insurgents controlled the countryside, preventing those "who might have been willing to cooperate with the marines from doing so" (Calder 1984:159). In the occupied regions of the Soviet Union, the Germans initially thought that there was a direct connection between popular dissatisfaction with their rule and the rise of the partisan movement (Cooper 1979:24); however, they eventually realized that this connection was mediated by control. As a Soviet agent noted in 1942, "in those areas where the partisans are not active, the people are against them. In the imagination of the population the partisans are like bandits and robbers" (in Dallin et al. 1964:331). A 1941 Soviet report made the same point: "There are, however, many elements among the population who sympathize with the partisan movement and the Soviet regime. But, since they fear the consequences, they are using utmost caution in their activities" (in Cooper 1979:78). Likewise, Conservative villages in areas controlled by Liberal bands in Colombia during the 1940s were forced to "convert" to Liberalism (and vice versa): the most active Conservatives were killed or fled, and the rest of the population shifted its partisan allegiances (Ortiz Sarmiento 1990:176-7). Robert Thompson (1966:15) observed that support for the Vietcong "greatly expanded" in those areas of the countryside that came under their control. As one of them
Control
119
put it, "There are some, particularly the middle and rich peasants, who do not like the communists, because the communists hurt their interests ... but they don't dare oppose them because, if they oppose the communists, they must go live in a government area. But do they have enough money to go and live in Saigon? Probably not, and so they must be content and remain" (in Race 1973:130). Conversely, Moyar (1997:339) reports that Vietcong cadres serving in areas of heavy U.S. and South Vietnamese activity "defected in large numbers and brought with them great amounts of information," contrary to cadres working in areas of less intense military activity.9
Second, there is substantial evidence that collaboration follows the temporal variation in control. Gaining control over an area brings collaboration, and losing control of an area brings much of that collaboration to an end. In North Carolina, during the American Revolution, one Revolutionary veteran recalled that the loyalists "gained more confidence and they became more bold, more daring, and more numerous" following the American defeat in Camden in August 1780; in his own militia company, in contrast, only eight members remained "good and true" Whigs - "the rest had joined the Tories" (in Crow 1985:160). The tide shifted once more, until 1781, when Cornwallis's capture "disheartened" the Tories (Crow 1985:160-1). A Union general during the American Civil War found that in Virginia "the majority of people along our track to be reasonably neutral" and noted "the rapid development of loyal sentiments as we progressed with our raid" (in Wills 2001:204).10 Using a list drawn by the state authorities to assess the political allegiance of villages in the Ille-et-Vilaine in July 1795, Roger Dupuy (1997:194-7) found that their allegiance during the Vendee War is predicted less by their expressed preferences prior to the war (1789-92), and more by the local military balance of power and the village's geographical position - in other words, the extent of control exercised in the area by a political actor. When the counterrevolutionary rebels exploited the military weakness of the republicans and took over the countryside, all neutral villages as well as many staunchly republican villages turned counterrevolutionary. The archival material quoted by Dupuy is replete with remarks about "republican communes" whose "republicanism is diminishing" because of their "geographical position" and the "difficulty of communications," and of villages that will turn counterrevolutionary "if additional forces are not sent in."11 In fact, the French revolutionary Gracchus Babeuf
9 The spatial relationship between control and support holds beyond civil wars. Kamen (1997:180), for instance, found that one factor accounting for the inability of the Spanish Inquisition to strike fear among most Spanish people was its absence from the majority of localities - "the sheer impossibility of one inquisitor being able with any degree of frequency to visit the vast areas involved."
10 Wills (2001:204) comments: "For some this loyalty was certainly genuine; for others, any such demonstration need last only as long as the blue-coated soldiers remained in the vicinity."
11 A couple of examples from this report are worth quoting (Dupuy 1997:194-5): Amanlis: "its patriotism appears to have diminished, its municipality is not corresponding with us anymore; however it is presumably easy to recall [this commune] to the republican principles, but its geographical position and especially the difficulty of communications does not allow us to do so"; Bais: "a large commune which has shown all the energy of patriotism," but "its patriots will be forced to abandon
120
The Logic of Violence in Civil War
(1987:120) noted how many republicans began joining the counterrevolutionaries as early as 1797 because of the direction of the war.
At the beginning of the twentieth century, Macedonian villages would change "national camps" and become "Greek" or "Bulgarian" as many times as they were visited by Greek or Bulgarian fighters; a Greek participant referred to one village that kept welcoming both sides by saying that its inhabitants behaved "in a political manner" (Livanios 1999:205). In 1941 a high political cadre of the Chinese Communist army observed that with the removal of the Communist troops, resistance work stopped in the undefended areas, peasant morale plummeted, and local party organizations displayed open hostility toward the military leadership; local Communist soldiers often deserted, circulating stories to justify their flight that further damaged the army's reputation among the population (Hartford 1989:111). A remark by a Soviet partisan in a 1943 letter underscores this point: "When we came here from the Soviet hinterland, the Germans were everywhere and it was not very pleasant. There were many police and other riffraff who fought side by side with the Germans. The population was also against us. This lastyear [1943] has brought about perceptible changes. Our partisan area has become large. Now you see no Germans in the rayon centre. This work had to be carried out under difficult conditions; now, however, it has become easier; the population of the whole area stands behind us" (in Cooper 1979:64-5). Susan Freeman (1970:24-5) describes the behavior of a small village near the front lines of the Spanish Civil War:
Valdemora set about the business of survival.... Their approach was to remain aloof when possible and to yield when necessary. This they did as a community. When troops (of both sides), camped in the Sierra, demanded food, the alcalde [mayor] assessed all families for equal amounts and sent food into the hills. This is regarded as the only intelligent thing to have done... .In 1936, several individuals in the area who refused troops' demands were killed, and Valdemorans regard them as having wasted their lives for want of a little intelligent realism. As the siege of Madrid continued and the front lines moved away from the Sierra (before the war was one year old), acquiescence to the government in power was life's simple rule, as it had been under both the Monarchy and the Republic.
After the South Vietnamese and U.S. military came to the strongly pro-Vietcong village of My Thuy Phuong, "most of the revolutionary cells died, and the people became divided," a Vietcong rebel pointed out (Trullinger 1994:143). A similar trend was observed elsewhere in Vietnam: prior to January i960, in the South Vietnamese province of Long An, "a great many people were favorably inclined toward the [Communist] movement but elected not to cooperate overtly
because of the risks entailed by the continued government presence____Yet as the
presence of the central government was eliminated, the probability of achieving what the movement promised greatly increased, at the same time as the
their homes," and Bais will turn counterrevolutionary "if additional forces are not sent in." Moyar (1997:301) quotes a Vietcong district-level cadre making a similar point: "These observations of mine made me think that the Front is very active and harmful in quiet areas, while it is weakening there where the GVN [the South Vietnamese military] is active." See also Jon Anderson (2004:140) on El Salvador and Ekirch (1985:114) on revolutionary North Carolina.
Control
121
risks of involvement greatly decreased" (Race 1973:191). "The people began to draw away from us and to fear our presence, knowing that we would attract government forces and more fighting," recalled a former Vietcong (Herrington 1997:30). Elliott's (2003:1006) massive research on the Mekong Delta confirms this observation: there was "a clear decline in popular support for the revolution that resulted from the loss of physical control." After the United States threw its weight behind the Northern Alliance in Afghanistan, many Taliban switched sides. As one of them explained, "I joined the Taliban because they were stronger. I am joining the Northern Alliance because they are stronger now" (in Filkins 2001: Ai).
The obverse of every gain in collaboration that one side experiences with gains in control is the loss of collaboration experienced by the other side along with its loss of control. Del Pino (1998:178) describes how the population of the Ene Valley in Peru "joined" the Shining Path after it took control of the valley in 1988, placing forces at its entrance and exit points. When in 1991 the Peruvian army began to "liberate" the valley and "recuperate" the population, these same peasants joined the local militias in their fight against the rebels. In Kenya, "with the provision of security came a change of conviction, away from some 90 percent general Kikuyu support for Mau Mau's methods (though not necessarily its aims) to support for the government" (Clayton 1999:14). Kerkvliet (1977:237) shows how peasants had to withdraw their support from the Huk rebels in the Philippines because of a shift in control. A related example is the description of the evolution of support in the village of Punta Dumalag located in the Davao area of the Philippines, a reputed Communist stronghold. This village was first organized by Communist cadres in the late 1970s. A clandestine barrio revolutionary organization acted as a shadow government, and most villagers actively participated in the insurgency in a variety of functions. In early 1988, a few months before the writer Gregg Jones's visit, control shifted: the rebels "had been forced to abandon their once impregnable stronghold, and a fiercely anticommunist autocracy led by Alsa Masa [Risen Masses] vigilantes ruled the barrio." Jones found that although the most committed villagers had fled, most villagers remained and "now they too professed allegiance to Alsa Masa. Jones concludes that "if Punta Dumalag residents were secretly unhappy with Alsa Masa, as Davao NDF faithful suggested, it appeared that [they] at least had adjusted to life under the new order" (G.Jones 1989:270-5).
British reports on the Greek Civil War, during its last phase (1946-9), provide a good example of how control shapes collaboration. The following excerpts from these reports to the Foreign Office require no comment:
May 1947. A huge block of nationalist inhabitants in Laconia, although ready and willing to render assistance for the suppression of banditry, fails to find the means to do so. Meanwhile the bandits are intensifying their activity. The nationalists become discouraged and capitulate and in view of the Government's weakness, are giving no assistance to the local authorities.12
12 "Greece: Security Situation in the Peloponnese; Sir C. Norton to Mr. Bevin (26 June 1947), Attached Greek Gendarmerie report (16 May 1947)," PRO, FO 371/67006/R8651.
122
The Logic of Violence in Civil War
June 1947. As it is, the population of an essentially anti-Communist area are losing confidence in the Government, and is in the mood to capitulate to the Communists for lack of any better alternative.13
November 1948. Whilst the majority of the two million inhabitants of the Peloponnese are at heart patriotic and opposed to subjection to Russian communism ... they are being forced by fear and increasing misery to accept the role of the communist guerrillas in order to survive at all.14
February 1949. Civilian morale is steadily on the increase... .This is reflected in the greatly increased assistance which is being given by the civilian population to the Army and the Gendarmerie in the collection of intelligence, and the large increase in the number of bandits surrendered.... The reaction to the military operations has been what might have been anticipated. For the first few days after the occupation of bandit areas by the troops the attitude of the civilian population was surly and suspicious, but as soon as it was realised that the Government forces had come this time with the intention of remaining, their attitude completely changed - increased cooperation, particularly in the matter of intelligence.15
Political actors are obviously well aware that control spawns collaboration. The Vietcong security doctrine made a clear connection between the two: the imposition of control made possible the creation of "a sympathetic environment, that is, an environment (the population) composed of sympathetic and neutral elements, from which the hostile elements have been removed" (Race 1973:146). Likewise, counterinsurgency experts point out that the primary objective of governmental militias is population control so as to deny insurgents "the support of the civilian population" (Jones and Molnar 1966:25).
Forced population removal (often called "resettlement," "population control," or "villagization"), a method used by incumbents in some civil wars, further confirms the endogeneity of collaboration to control.16 This method, whose use in modern times was pioneered by the British and the United States around the turn of the twentieth century, seeks to deprive insurgents of their population basis; in counterinsurgency parlance, "the population problem... may be solved by physical or psychological separation of the two elements: guerrillas and population" (Condit 1961:24).17 Euphemisms, such as "emptying the tank"
13 "Greece: Security Situation in the Peloponnese; Sir C. Norton to Mr. Bevin (26 June 1947)," PRO, FO 371/67006/R8651.
14 "Report from Patras (2 November 1948)," PRO, FO 371/72328/R13201.
15 "Report from the Military Attache on the Military Situation in the Peloponnese (visit: 18-21 February 1949)," PRO, FO 371/78357/R2293.
16 Because this method requires extensive resources, it is used primarily by wealthy (or foreign-supported) incumbents, or in times of growth (e.g., Stubbs 1989:113).
17 The British moved a large part of the Settler South African population into concentration camps during the Boer War (Klonis 1972:53); the U.S. forced thousands of Filipinos into "protected zones," where as many as 11,000 died as a result of malnutrition, poor sanitary conditions, disease, and demoralization (Linn 1989:154-5); similar tactics were used, on a smaller scale, in the Dominican Republic (Calder i984:xxii). Union troops relied on resettlement in areas prone to guerilla war during the American Civil War (Fellman 1989:95). In the 1950s, the British reintroduced the method of resettlement in fortified villages - both in Malaya and Kenya; the French used it in Algeria during the war of independence, as did the Americans in Vietnam (using British advisers
Control
and "drowning the fish," abound. Despite generating considerable grievances, moving the population into areas where it can be controlled by the incumbents (either in fortified villages or in refugee camps around towns) appears to produce collaboration with the incumbents. Indeed, the forced relocation of peasants in slums surrounding major cities (e.g., in Greece, Vietnam, Turkey) did not seem to generate serious security problems for incumbents. The same peasants who supported the insurgency back in their villages turned quiescent, even though they now had more reasons to complain about the appalling conditions of their new life.18
One of the best-researched cases in this respect is the Vietnam War. Summarizing a wealth of surveys and research on the attitudes of refugees, Wiesner (1988) notes that forcible evacuation not only failed to alter peasants' preferences, but it turned many peasants who were resentful of the government for their removal into Vietcong sympathizers. Moreover, it brought Vietcong supporters and cadres into camps located in government-held territory, thus reducing security. As a result, forcible evacuation was criticized by many U.S. officials. However, despite being populated with Vietcong sympathizers (and some cadres), refugee camps never became a security problem for the incumbents. Sympathy did not translate into collaboration with the Vietcong. On the contrary, the Vietcong saw these camps as a threat, as evidenced by the fact that they sometimes bombed them. In fact, when switching from a description of attitudes to that of behavior, the U.S. reports found that most refugees tended to collaborate with the incumbent authorities.
Observers sometimes note this relationship between control and collaboration, but they tend to miss its significance or misinterpret its causal direction. Milton Finley (1994:28-9) notes that the Napoleonic French troops in Calabria were able to recruit local volunteers only from the towns but fails to connect this pattern to the fact that "even nominal French control stopped at the edge of the town; the countryside belonged to the brigands." Consider the following remark about Vietnam made in April 1964 by the American journalist Walter Lippman (in Taber 1965:17): "The truth, which is being obscured from the American people, is that the Saigon government has the allegiance of probably no more than 30 percent of the people and controls (even in the daylight) not much more than a quarter of the [national] territory." It is easy to see the connection between the
with experience in Malaya) under the designation "Strategic Hamlet Program," and various colonial, African, and Latin American regimes used it as well - Portugal, Ethiopia, and Guatemala, most notably. Even leftist incumbents, like the Sandinistas of Nicaragua, relied on this method (Horton 1998:229). A recent example of the use of this method is Turkey's evacuation of 1,779 villages and hamlets and 6,153 settlements in the eastern part of the country in the 1990s, in the war against the Kurdish PKK insurgency (Jongerden 2001:80).
18 John Cann (1997:155) summarizes some of the difficulties in his discussion of the application of the method by the Portuguese in Africa: "Moving people was invariably an emotional process because of their attachment to ancestral lands. Timing was also a large factor, and moving a population after it had been subverted was pointless and generally backfired. Often implemented in a rush, the program experienced unnecessary teething problems that required a sizable amount of time and money to correct."
19 Wiesner (1988:113, 136-8, 144, 243-4, 357)-
124
The Logic of Violence in Civil War
similar percentages of allegiance and territorial control that Lippman may have perceived as two unrelated matters.
5.4. CAUSAL PATHS FROM CONTROL TO COLLABORATION
A common causal mechanism that translates control into collaboration is coercion and survival maximization: the imposition of control allows the effective use of violence, thus deterring defection; opponents are identified and flee, are neutralized, or switch sides. The rest of the population complies, while some people may switch their preferences to side with the ruler. Although violence is an important channel though which control spawns collaboration, it is not the only one. I identify six additional mechanisms that translate control into collaboration: shielding, "mechanical ascription," credibility of rule, the provision of benefits, monitoring, and self-reinforcing by-products. By enabling these mechanisms, violence matters indirectly more than directly.
First, as already suggested, the force inherent in control solves collective action problems and deters opposition via coercion. In Tilly's (1992:70) words, "coercion works; those who apply substantial force to their fellows get compliance." Shy's (1976:179) study of individual motivations during the American Revolution confirms "the brutally direct effects on behavior, if not on opinions, of military power." As a Nicaraguan peasant put it, "Those with arms give the orders" (Horton 1998:207).20 In short, collaboration can be tacit, the product of "no alternative," or "the necessity of the moment." As a resident of the rebel-held "demilitarized zone" in Colombia put it (Forero 2000A.3), "People don't really have the chance to make their own decisions. They don't have a choice. They just go along with this thing."21 Clearly, combining credible threats of violence with the option to switch sides appears to be very effective. Survival can bend the people's posture (Henriksen i983:75).22
Threats are not the entire story, as indicated by the second mechanism. Control also lowers the cost of collaboration with the established authority by shielding the population from competing sovereignty claims. It does so by providing protection from threats and violence made by the rival actor. "By the early 1780s," writes Roger Ekirch (1985:121) about North Carolina, "a majority of settlers were inclined to support whichever side could ensure a modicum of stability." German reports from occupied Ukraine stressed the vulnerability of potential collaborators to partisan reprisals owing to the absence of any German garrison
20 Note that conscription requires control. Vietcong recruits were more likely to have been conscripted in rebel-controlled areas than in government-controlled ones (R. Berman 1974:69).
21 See also Horton (1998:136); Kedward (1993:60); Shy (1976:13); Barnett and Njama (1966:151).
22 Note that the same is true of many ethnic civil wars where incumbents offer the option of collaboration to ethnic minorities. In Punjab, "except for committed [Sikh] guerrillas, defiance was possible only when there was a measure of protection" (Pettigrew 2000:211); in Chechnya, the pro-Russian militia of Ramzan Kadyrov includes rebels who switched sides (Myers 2oo5:A4). Consequently, ethnic insurgents must also rely on terror (Collins 1999:128), even when they enjoy widespread sympathy among the population (e.g., Herrington 1997:22-3).
Control
in the vicinity". Infrequent German patrols were simply not credible protection, and at that point, some villages had yet to see any German troops at all (T. Anderson 1995). Once the Vietcong were able to control large parts of Long An province in South Vietnam in i960, "a vastly larger number of people moved into the Party and Party-controlled groups, because the threat of exposure and capture had been greatly reduced by the elimination of the government's 'eyes and ears'" (Race 1973:116)^ former Vetcong confirms this point by describing how collaboration in the Mekong village of Ban Long depended on who controlled the larger nearby village of Vnh Kim: "Once Vnh Kim fell under the control of the Front, Ban Long's security would be assured and the tasks of motivating the people (meetings, celebrations, labor recruiting) would be carried out freely and easily. On the other hand, if Vnh Kim was under GVN control, Ban Long would have to pay a lot of attention to safeguarding itself from traitors, keeping secrets, and defending itself (in Elliott 2003:268).
Third, overtime, control produces "mechanical ascription" (Zulaika 1988:32). Long-lasting control spawns robust informational monopolies that socialize populations accordingly. In such circumstances, joining an armed group appears as a natural course of action for many: "In theory, people in the age group to fight have an option of two parties. In practice, with the exception of a few politically sophisticated ones, in rural areas such as Itziar [Basque country, Spain] they simply obeyed the official army's orders" (Zulaika 1988:32). This single option becomes the dominant cultural message and "the undisputed model of heroic activity" for adolescents. "Joining the IRA was not difficult in Meenagh Park," points outToolis (1997:39); "it was the obvious career move for a young man with time on his hands." He adds: "At the kitchen table, I sat asking the same question over and over again - why had Tony joined the IRA? The logic of the question was unintelligible to the Doris family. In their minds the mere description of life in Coalisland was sufficient to explain why Tony had joined the IRA. My naive question shook this natural assumption. They searched for ways to explain something that was so obvious it was inexplicable" (1997:40).
In many areas of South Vetnam it was natural for young men to join the Vetcong, which had been the effective government for more than twenty years (Bilton and Sim 1992:57; Meyerson 1970:91). A Taliban defector explained his actions in the following terms: "When the Taliban conquered Afghanistan, all the fighting men in Badakhshan joined them. There was not a single good man in the province from the Northern Alliance" (Filkins and Gall 200i:B2).23
The exclusive access to a recruitment pool generates cascades of support because the families of fighters tend to support the armed factions where their
23 For Degregori (1998:131-2) this process results from a "demonstration" rather than a socialization effect - but, in effect, the same mechanism operates. Peruvian youth in the Andes, he points out, were inspired to join the Shining Path, "an organization that was on the rise, prestigious, with a demonstrated effectiveness. Such an organization would empower and transform them. Joining Shining Path had elements of a rite of passage or of initiation into a religious sect: an armed sect." A certain degree of coercion can help. In controlled territories, the fighters' families may act as hostages, to be punished if a fighter deserts (Cooper 1979:74).
126
The Logic of Violence in Civil War
younger members are fighting: "For the Chinese who had relatives or friends who had 'gone inside,'" Stubbs (1989:89) notes about Malaya, "there was no question where their allegiance lay." "Whatever her politics," Clutterbuck (1966:93) confirms, "it was no surprise if a mother smuggled food to her son, even though she faced years in jail if caught." As a Nicaraguan farmer told Hor-ton (i998:xiii) about the contras: "Those boys are our sons, our neighbors"; in "rural communities where the contras had already established a foothold." Horton (1998:175) adds, "young men were reluctant to join an army fighting against their neighbors, friends, and even family." Conversely, when the U.S. and the South Vietnamese governments improved their "pacification" efforts and brought more areas under their control, they relied increasingly on local militias, which produced a similar effect (Moyar 1997:313). This is well understood by political actors. The Chinese Communists produced a propaganda verse to that effect:
If a father gets his son to enlist
The revolution will take this to heart.
If a son gets his father to enlist
He'll be forever revolutionized.
If an elder brother gets a younger brother to enlist,
The roots of poverty will soon be excised.
If a younger sister gets an elder brother to enlist,
Only then may the roots of wealth reach deep.
If a younger brother gets an older brother to enlist,
The Nationalist Army will be smashed to bits.
If an elder sister gets a younger brother to enlist,
Victory will soon be ours.
If a wife persuades her husband to enlist,
There'll be no worries in the family.
(Levine 1987:155)
This process explains two oft-noted features of rebel movements: their bulk is usually made up of natives of the region where they operate rather than people who come from other areas to join them (Geffray 1990:39; Barton 1953:70); and, after crossing a certain threshold, tiny groups tend to grow exponentially: Degregori (1998:132) compares the rise of the Shining Path in the Andes during the early 1980s to wildfire. An Algerian woman described to Baya Gacemi (1998:109, 185) the remarkable speed with which her co-villagers first supported the Islamist GIA rebels and, three years later, when the rebels began to be defeated, how they joined the government militia and began informing on the rebels. This initial growth feature is often mistakenly interpreted as proof that these insurgencies articulate very real and widespread popular grievances. Of course, they may, but the point is that exponential growth is observationally equivalent with, and potentially endogenous to, the cascade mechanism.
Fourth, control signals credibility - both the short-term credibility of immediate sanctions, as well as the long-term credibility of benefits and sanctions based on expectations about the outcome of the war. Civilians would rather side with
Control
127
the (expected) winner than the loser.24 Russian peasants were more inclined to collaborate with the Germans opportunistically while German fortunes seemed favorable and German forces had the upper hand in the partisan war; in contrast, when the Germans were perceived as failing to subdue the partisans, their confidence in the strength of the German army declined and they were more inclined to support or even join the partisans (Hill 2002:43; Cooper 1979:27). U.S. advisers in Vietnam "noticed that rises and declines in the level of Allied military activity tended to produce rises and declines in the rates of Communist desertion in an area" (Moyar 1997:110). In his partly autobiographical novel, Fenoglio (1973:296-7) observes that peasants in the Italian mountains helped the anti-Fascist partisans,
uniquely in exchange of the guarantee that we were going to win, that they would find their harvest, their flocks, their peaceful trade between fairs and markets, once this dirty story of Germans and Fascists had ended once and for all. Now, after the rough [partisan defeat] in Alba, they still had to give, to help, to risk their heads and dwellings, but the victory and the liberation were hazily remote. For months, they helped us smiling and laughing, confidently asking many questions; now they started to help silently, then almost grudgingly, at least with mute complaints, increasingly less mute.
It is often noted that insurgents receive most new recruits after two types of events: indiscriminate incumbent violence against civilians, and successful insurgent engagements against the incumbents.25 Between 10 and 20 percent of the partisans in 1944 were former collaborators of the Germans; as it became clear that the Germans would lose the war, even more people flocked to the partisans: many were known as "hubbies," Soviet soldiers who had fallen behind, married local girls, and expected to sit out the war, but were eventually compelled to join the partisans (Cooper 1979:70-2). This is why groups trumpet their victories and attempt to hide their defeats (e.g., Tone 1994:109). Obviously, the best indicator that a group is winning the war is a decisive shift in control. The Vietcong popularity in the Hua Nghia province of South Vietnam soared in 1965-6, when many hamlets became virtually off limits to government officials: "At this time, it was not necessary to use threats or terror to obtain" support, a peasant told Herrington (1997:29); "it was given willingly because the people were nearly certain that the future lay with the Communists." Race (1973:39-40) describes how the Vetminh won the support of peasants in Long An province during the war against the French: "The peasantry had seen the landlords run, they had seen the village councils forced to sleep in outposts and to move in the countryside with armed escorts. As night fell in the countryside, the peasants saw where lay the power of the conflicting sides: the Vetminh slept with the people, the village councils slept with the soldiers in the outposts."
Conversely, when incumbents are able to signal credibly that they will win, many civilians will shift their support away from the rebels and toward them
24 More precisely: few would ever join a side they thought was losing (Hartford 1989:122). See also Manrique (1998:204); Herrington (1997:25); Lichbach (1995:68); Coleman (1990); Sansom (1970:226-7).
25 Tucker (2001:90); Laqueur (1998:317); Wickham-Crowley (1991:43); Debray (1967).
128
The Logic of Violence in Civil War
(Cann 1997:104). In the Philippines, denunciations of actual or suspected insurgents "rose sharply" after the antiguerrilla campaign began to show signs of success (Barton 1953:129). Many Peruvian peasants shifted their support toward the army because "by 1990, most villagers realized that the military was not about to 'collapse before the glorious advances of the people's war,' as the first cadre had promised in 1982" (Starn 1998:229-30). A Pakistani officer who fought in Bangladesh in 1971 recalls that "The Bengalis' behaviour followed the fluctuations in the fate of insurgency operations. They usually sided with the winning party. If our troops were around, the people were apparently with us, but when they were withdrawn, they welcomed their new masters (the [rebel] Mukti Bahini) with full warmth" (Salik 1978:101). In a world where expectations about the outcome matter and where information is mostly local, local control may signal dominance and eventual victory.20
Fifth, control makes possible the provision, when available, of all kinds of benefits intended to generate loyalty - "hearts and minds." Under conditions of incomplete or no control, such programs are guaranteed to fail (Harmon 1992; Clutterbuck 1966). As Machiavelli argues in The Prince, there can be no good laws where there are no good armies. Insurgents are able to lower or eliminate tenant rents to landlords only where they are able to exercise control (Wood 2003). The Vietcong were able to implement land reform programs where they exercised control rather than in those places where the peasants were more exploited (Elliott 2003:504). In fact, rents to landlords were directly related to the degree of control exercised by the two rivals: the higher the degree of Vietcong control, the lower the rents. The direction of causality ran from control to rents, as peasants living close to major roads and military posts paid much higher rents (Sansom 1970:60-1).
Sixth, control facilitates direct monitoring and population control. Direct monitoring requires better and more extensive administration, which is impossible in the absence of control; in turn, administration reinforces control. Once an area is placed under control, processes such as the registration of inhabitants and the compilation of detailed lists of the population of every locality become possible. The Japanese were able to "detect resistance organization early enough to nip it in the bud" only in areas that the Chinese Communists considered "enemy-occupied" or "weak guerrilla areas" (Hartford 1989:95). In Malaya, the government "extended its administrative net over the population" in the context of its counterinsurgency policy (Stubbs 1989:163); "in areas where the population were reasonably secure and where the methods used by police and military organizations for collecting intelligence were efficient, the greatest amount of information was collected" (Jones and Molnar 1966:29). The rise of the Vietcong reflected a similar process: "First the GVN posts were neutralized, then the intelligence stopped coming. With no intelligence, the larger GVN units were rendered ineffective, and the dangers for GVN agents increased even more" (Elliott 2003:424). A 1942 German report from the occupied Soviet Union made
26 What is true of civilians is also true of combatants. See Finley (1994:101) and R. Berman (1974: 178-9).
Control
129
this point: "The appointment of reliable mayors and indigenous policemen in communities recently cleared of partisans has proved to be an effective device for preventing the formation of new bands in such communities and in the adjacent woods. The mayors and police, in conjunction with German troops in the vicinity and with secret field police and military police detachments, watch closely over the pacified areas, paying particular attention to the registering and screening of all persons newly arrived in the area" (in Cooper 1979:46; emphasis added).
Seventh, control spawns a self-reinforcing dynamic. Because some areas are controlled early on by one political actor, they may develop a reputation of being loyal to this political actor: the Djacovica valley in Kosovo was reputedly pro-KLA, and the villages of the Shamali plain in Afghanistan were perceived as supporting the Northern Alliance. Irrespective of whether this reputation truly reflects a majority preference, it may lead to indiscriminate reprisals (or the expectation thereof), turning potentially accidental or misperceived strongholds into real ones. "Even if it were not true that everyone in the town 'adored' the guerrillas," writes Mary Roldan (2002:243-4) about a Colombian town, "the very fact that the authorities thought so and assumed that the town as a whole could not be trusted, encouraged and reinforced a sense of local identity and collective purpose. This sense of collective involvement enabled local inhabitants to justify having taken up arms against the government." On the Korean island of Chejudo, insurgents were strongest in the villages closest to the mountains; as a result the government forces labeled as enemy territory all areas lying five kilometers from the coast and treated them accordingly (Yoo 2001). In Kenya, Mau Mau instructions included the rule that "warriors were entitled to take by force any foodstuffs in the gardens and livestock concentrated at any Government centers irrespective of whether they belonged to friend or foe" (Barnett and Njama 1966:195). This was the case in the Filipino village of San Ricardo studied by Kerkvliet (1977:166), the Vietnamese village of My Thuy Phuong studied by Trullinger (1994), and the RENAMO-held areas in northern Mozambique studied by Geffray (1990:71).
Such behavior tends to reinforce the association between a political actor and the underlying population. As Dallin et al. (1964:329) point out about the German-occupied Soviet territories, "the survival of the partisans became a prerequisite of [the population's] own survival, since their fate was certain should the Germans reoccupy the area." "Residents in Japanese garrisoned districts [on the island of Leyte, Philippines], particularly the poblacion or town center, and the local officials," Lear (1961:27) points out, "came to be branded as ipso facto pro-Japanese." "Lots of guerrillas" a participant recalled, "told me: 'Tacloban people, pro-Jap. They do not fight Jap, they live in Jap town, therefore they pro-Jap. If I catch Tacloban man, I kill him.'" As a result, the inhabitants of Tacloban "did not like the guerrillas. They were afraid of guerrillas, and they had reason" (Lear 1961:28).
Thus, what may have been initially an accident of location may generate new and enduring political identities. Consider the following description of how a small town in northern Spain went from being deeply divided to supporting only one side once the civil war began (Lison-Tolosana 1983). The shift resulted from the aggregation of individual strategies of survival maximization after the
i3o
The Logic of Violence in Civil War
town found itself in the Nationalist zone; once the local Republican leaders were decimated in the first days of the war, all the young men, irrespective of their family's prewar political affiliation, were drafted into the Nationalist army, fought against the Republic, and "became" Nationalists. This endogenous preference shift was reflected in subsequent patterns of religious practice, where observance serves as a proxy for support for the Nationalist cause: after the Nationalist troops occupied the town, the number of those not fulfilling their Easter Duties dropped from 302 in 1936 to 58 in 1937. Initially the result of repression and fear, this shift eventually produced new, real, and enduring identities. Lison-Tolosana (1983:190, 196, 290) was able to establish that, whereas the generation that held power in the town before the civil war was divided between Republicans and Nationalists, the next generation was united and Nationalist. This pattern is confirmed by Freeman's (1970:24) study of a Castilian village that likewise fell under Nationalist control as soon as the war began. Young men were drafted into the Nationalist army, and "their former loyalties are hard to discern today." In contrast, Seidman (2002:38) reports that 80 to 85 percent of wage earners who found themselves in the Republican zone during the Spanish Civil War joined a party or union only after the civil war erupted - and did so for practical rather than ideological motivations. Note that whether villages or towns found themselves in the Nationalist or Republican zone was, for most of them, a matter of accident. These examples are consistent with an interpretation of the Spanish Civil War as shaping preferences rather than just reflecting them.
These new identities may turn ascriptive. As Germaine de Stael (1818:33) remarked, "to kill is not to extirpate ... for the children and the friends of the victims are stronger by their resentments, than those who suffered were by their opinions." Many people branded as kulaks in the Soviet countryside between 1927 and 1935 were not wealthy peasants but rather White army veterans or their relatives (Viola 1993:78). The magnitude of such "genealogical witch-hunts" is suggested by Stalin's 1935 announcement that "sons were no longer responsible for the sins of their fathers" (Viola 1993:80). Hart (1999:294) recounts how, when he was in a pub in Cork County, Ireland, his companions pointed to a middle-aged man and announced, "Here comes the informer now." This man was far too young to have been alive in the 1920s during the Irish Civil War. "Oh yes, I was told afterwards," Hart recounts, "it was his father who had been an informer. They were not sure what he had done to warrant the charge, but 'the informer' was what he had been called behind his back even after, and 'the informer' his son remained." The Chinese Communists institutionalized the identities that emerged from the civil war by developing a nomenclature of "five red types," three of which derived directly from choices made during the civil war: revolutionary cadres, revolutionary soldiers, and dependents of revolutionary martyrs (the two remaining types were workers and poor and lower-middle-class peasants). These categories evolved into ascriptive groups according to the "blood pedigree theory," which was conveyed by couplets such as the following one:
If the father's a hero, the son's a good chap; If the father's a reactionary, the son's a bad egg. (Chang 1992:285; L. White 1989:222)
Control
The relationship between control and collaboration is theoretically significant because it undermines the widespread assumption that joining an insurgent organization is always a highly risky behavior (thus, automatically turning recruitment into a collective action problem). Consider William H. McNeill's (1947:80-1) description of the process of joining the ELAS insurgent army in Greece, for whom the puzzle to be explained is not recruitment but the absence of more recruits:
In actual fact, a soldier in ELAS lived a good deal better than did the ordinary peasant, and did not have to work with the same drudging toil. He further had the psychological exhilaration of believing himself a hero and the true descendant of the robber klefti who had fought in the War of Independence and were enshrined in the Greeknational tradition. Under the circumstances, many a peasant's son found himself irresistibly attracted to the guerrilla life; and an over abundant peasant population made recruitment easy. Fewer came from towns; life was relatively comfortable there, and EAM had other work for townsmen, organizing strikes or serving as propagandists among the more illiterate peasants. From the very beginning the chief factor that limited the number of the guerrillas was lack of weapons.
Given the impact of control on collaboration, it is not surprising that observations about popular support and individual commitment often point to accident, contingency, and chance.27 Alexander Dallin et al. (1964:336) argue that decisions to side with the Germans or the partisans in the occupied Soviet Union were largely dependent on the "accident of which regime was stronger and happened to control a given area." Lear (1961:237) reached a similar conclusion about the Japanese occupation of the Philippine island of Leyte: "What we are trying to point out by these and other possible examples is that to a large extent chance decreed what motives would be victorious in the inner struggle of competing motives determining whether an individual in Leyte was to be guerrilla or collaborationist." In Greece, "almost half as many young men from [the village of] Kerasia served in the national army as joined the guerrillas. Accidents of call-up and timing probably decided who served in which forces as much or more than ideological conviction. But once committed, one way or the other, a man found it difficultto change sides safely" (McNeill 1978:154). Chris Woodhouse (1948:58-9), the commander of the Allied Military Mission to the Greek Partisans during the country's occupation, describes the "choice" of a Greek peasant:
He was living in his mountain village in 1942.... [H]e joined the left-wing resistance movement, because it was the first in the neighborhood. (They were on top then, so he was right.)... It happened that he joined the movement which was dominated by the Communists, though he was no Communist: he might as easily have happened in other circumstances to join the Security Battalions formed to fight the Communists, though he would still not have been pro-German. He would not have been a recognisably different individual if things had happened otherwise; but recognisably different things would have befallen him. His fate did not rest in his own hands, but in the chances that brought him into contact with men from above the horizontal line [i.e., outsiders]; chances that were largely geographical. If he lived in one part of the mountains, he was more likely to be
27 Loyd (2001:48-50); Tucker (2001:61); Livanios (1999:197); Laqueur (1998:99); Mackenzie (1997); Todorov (1996:94); Chang (1992:449); Henderson (1985:41); McNeill (1947:134).
The Logic of Violence in Civil War
in contact with the Communist influence first; if in another, with the non-Communist resistance; if in the plains, with the Security Battalions and the collaborating authorities; and so on. (emphasis added)
As this passage suggests, "chance" is another name for a warring party's access to a segment of the population, which, in turn, is largely determined by control. From the preceding discussion, I derive the following proposition:
Proposition i The higher the level of control exercised by an actor, the higher the rate of collaboration with this actor - and, inversely, the lower the rate of defection.
5.5. THE DISTRIBUTION OF CONTROL
If collaboration is endogenous to control, then what determines the distribution of control? There are good reasons to think that control hinges largely on military effectiveness; in turn, this type of effectiveness is often (but not always) determined by geography. Because it is improbable that political preferences shape geography, the direction of causality seems obvious. This is not to say that political preferences are irrelevant. As pointed out previously, where political preferences and military resources overlap, as in the case of ethnic minorities that live clustered in isolated and rough terrain, prewar cleavages and geography will likely reinforce each other (Toft 2003). However, where no overlap exists, either because geography is unfavorable (e.g., an ethnic minority that is concentrated in cities) or because a rival actor can effectively muster superior military resources (e.g., a strong incumbent presence in a mountainous ethnic minority enclave), geography will tend to trump popular preferences in producing control.
Control has a clear territorial foundation: rule presupposes a constant and credible armed presence - a fact well understood by practitioners. Mao used to point out that geographical conditions "are an important, not to say the most important, condition for facilitating guerrilla war" (in Benton 200:714); he stressed that without "base areas" it was impossible to sustain guerrilla warfare (in Bruno Shaw 1975:208-9). As an American participant in the Vietnam War realized, control at the microlevel means establishing "suzerainty" over each village (West 1985:191). Armed actors can threaten credible sanctions only where they are able to sustain a military presence; their absence is an open invitation to their rivals. A striking and recurring feature of irregular war is how space shapes control. Towns, plains, key communication lines, and accessible terrain in general tend to be associated with incumbent control, whereas mountains and rugged terrain are generally insurgent strongholds; the location of insurgents is best predicted by variables such as terrain and distance from provincial military bases.28 The incumbents' presence in remote or inaccessible areas is, at least
28 Kocher (2004); Fearon and Laitin (2003); Hill (2002:44); Shaw (2001:154); Yoo (2001); Zur (1998:82); Tone (1994:13); Tong (1991; 1988); Brustein and Levi (1987); Schofield (1984:315); Crow (1985:129); O'Sullivan (1983); Wolf (1969:292-3); Salik (1978:101). Rough terrain is not synonymous with mountains; plains can sometimes offer an environment favorable to guerrilla
Control
!33
initially, limited to fortified villages and towns, while the insurgents' influence in towns is, at least initially, limited to clandestine organizations. Of course, geography should not simply be understood to mean "terrain." Geffray (1990:53) found that the location of RENAMO bases in Mozambique was a function not just of remoteness and distance from local administrative centers where incumbent forces were garrisoned, but also of proximity to administrative district boundaries. Apparently, RENAMO strategists found that these locations allowed them to benefit from the government's bureaucratic ineptitude, as local authorities tended to reject counterinsurgency jurisdiction when they could.
That military effectiveness as determined by geography generally trumps prewar political and social support in spawning control is best suggested by the following regularity: incumbents tend to control cities, even when these cities happen to be the social, religious, or ethnic strongholds of their opponents, whereas the insurgents' strongholds tend to be in inaccessible rural areas, even when rural populations are inimical to them.29
Insurgents tend to be uniformly weak in cities, although cities are often their prewar strongholds. Observers often note that many big cities in countries in the midst of civil wars look normal and peaceful (e.g., Butaud and Rialland 1998:124). Urban areas are inimical to rebels because it is easier for incumbents to police and monitor the population (Kocher 2004; Trinquier 1964:18; Kitson 1960:78); the collection of information through blackmail and bribes is facilitated because regular contacts between handlers and informers are possible. As a result, urban insurgents are particularly vulnerable to penetration and information leaks, as suggested by the cases of Northern Ireland and Palestine; and once identified, insurgents can be easily defeated by the superior force of incumbents. Urban guerrilla warfare is uncommon - and summarily dismissed by counterinsurgency experts (Blaufarb and Tanham 1989:15-16). As Trinquier (1964:71) puts it, "the most vulnerable part of the enemy organization is in the towns. It is always within the control of the army troops to occupy it, and a police operation ... can destroy it." Fidel Castro remarked that the city was the "grave of the guerrilla" (Laqueur i998:xix, 333).30
war, as suggested by the Vendean "bocage" in France (fields surrounded by tall hedges of bushes, narrow sunken roads, and dispersed villages, hamlets, and farmsteads). Other examples include the thick forests and swamps of Ukraine, Belorussia, and Russia; the rice fields of the Mekong Delta in Vietnam; the swamps and inundated plains of the Henan in China or Malaysia; and the thick orange groves of the Mitidja in Algeria.
29 In this respect, civil war stands opposite to crime: cities are much more difficult to police for states than rural areas (C. Friedrich 1972:26-7).
30 From this perspective, the inability of the United States to pacify several Iraqi cities in 2003-5 is a clear indicator of the numerical inadequacy of its military forces. This was eventually acknowledged publicly. A U.S. Marines commander said that "cities like Ramadi and Samarra had been allowed to slip into insurgents' hands largely by default, as the Americans began to concentrate their limited resources on other areas, like protecting the new government and critical pieces of infrastructure. Offensive operations based on intelligence," he added, "were a lower priority" (Filkins 2004:15). American forces were stretched so thin that soldiers even logged scheduled patrols that never took place - known as "ghost patrols" (Packer 2003:72).
!34
The Logic of Violence in Civil War
The Chinese case is instructive. Mao knew well the dynamics of cities and insurgency: "As for the big cities, the railway stops and the areas in the plains which are strongly garrisoned by the enemy, guerrilla warfare can only extend to the fringes and not right into these places which have relatively stable puppet regimes" (in Bruno Shaw 1975:209). The destruction of the Chinese Communist Party's urban infrastructure was close to total by 1927 (Schran 1976), and even though the Communists were able to return to the cities following the defeat of the Japanese, they had to abandon most of them once again, unable to withstand the pressure from the Kuomintang (Chang 1992:103). Benton (1999:729-30) shows how "armies, not classes, made the Chinese Revolution." He concludes that "class struggle ... did not well up from below, as a precondition for [the Chinese Communists'] triumph, but was whipped up from above, after they had achieved power by coopting and reorganizing the groups and networks that honeycombed rural China." As a Chinese Communist strategist summarized, "The enemies are the city gods, but we are the village deities" (in Wou 1994:222). The Algerian case offers a test of sorts. During the first half of the 1950s, two nationalist organizations were competing to lead the struggle against the French, the rural-based FLN and the urban based MTLD. Gilbert Meynier (2004:422-3) argues that the FLN prevailed over the MTLD precisely because of its rural connections.
The experience of occupied southern towns during the American Civil War supports this point as well. Garrisoned southern towns, whose citizens lived constantly in the presence and under the thumb of the occupying northern army, were places seething with hostility. On top of their ideological dislike of the Yankees, these towns suffered from unemployment, a severe housing shortage, skyrocketing inflation, and reduced supplies of food, fuel, clothing, medicine, and other basic commodities. As a Union general put it, "The people are suffering for want of almost all the necessaries of life." Naturally, all this further reinforced the existing feelings of bitter hostility vis-a-vis the occupiers. Yet, Ash (1995:82) concludes, "of all the citizens of the occupied South, those in the garrisoned towns posed the least threat to the Federal army, for armed resistance there was out of [the] question." The likelihood of internecine conflict between secessionist majorities and unionist minorities was also greatly reduced because "the secessionists were fundamentally impotent and the Unionists fundamentally invulnerable, thanks to the constant presence of Federal troops" (Ash 1995:122).
The anecdotal evidence is substantial. When the Greek Communist guerrillas attacked the town of Edessa, a known leftist stronghold, they found that the local population failed to assist them, and discovered that the government troops were never threatened from the rear (Vettas 2002:211). In Vietnam, "urban" became a synonym for government-controlled areas (Elliott 2003:1051; Meyer-son 1970:16). The Algerian FLN rebels were unable to control any cities and were eventually defeated in Algier's Casbah despite their initial strength there (Aussaresses 2001:41).31 Although cities in the Portuguese colonies of Africa were critical for the preparation of the insurgency, they did not experience any
31 Rejali (2004b) points out that the French won the battle of Algiers because they were able to destroy the insurgent infrastructure through population control and the recruitment of a large
Control
*35
significant action because the use of informers, curfews, dragnets, and censorship "impaired guerrilla mobilisation" (Henriksen 1976:384). Among the reasons why Biafrans refrained from engaging in guerrilla warfare was the "close concentration of towns" within Biafra and the concomitant absence of "hide-outs" - rather than the absence of a population supporting the independentist cause (Madiebo 1980:105). In El Salvador, state repression deterred opposition in urban areas while intensifying resistance in many rural areas (Stanley 1996:4). In Colombia, "state forces frequently control the centers of larger towns and cities, where municipal government buildings are located," but "the state's authority evaporates" in outlying neighborhoods (Fichtl 2004:3).
Further confirming the importance of military resources in generating control and hence collaboration is the oft-noted propensity of villages located near central roads to collaborate with incumbents. A British officer (Hammond 1993:137) noted that Greek villages that had "the misfortune" to lie "on or near the main roads" of Macedonia tended to collaborate with the German occupation army, the same situation as in Vietnam (Sansom 1970:60-1) and Rhodesia, where such villages were also more susceptible to being labeled "sell-outs" by Zimbabwean guerrillas because "their location close to the roads meant that they were more often visited by soldiers than guerrillas" (Kriger 1992:20s).32 Whereas "modernizing" villages near main roads had been among the first to respond to revolutionary appeals, they were also more likely to be controlled by the government, and "as the risks of political action escalated during the middle and late 1960s, the gap between political attitudes and behavior widened, and many revolutionary sympathizers became inactive when the dangers became too great or, in some cases, adopted a clandestine role so deeply hidden that it often amounted to a temporary cessation of revolutionary activities" (Elliott 2003:589). The availability of external support for insurgents turns the combination of terrain and proximity to borders into a strong predictor of insurgent control.
In pronounced contrast, rural areas tend to be inimical to incumbents, often regardless of their prewar political preferences. A high-ranking American officer serving in the Dominican Republic in 1921, argued that the construction of roads would stifle the insurgency: "A highway would bring the people more in contact with the Capital, thus giving the Central Government an opportunity to control political conditions" (in Calder 1984:164). An examination of both prewar and postwar electoral returns in the Peloponnese region of southern Greece suggests that the Right tended to be stronger in the mountains, and the Center-Left and Left stronger in the plains and towns. Yet, the war reversed this relation. A British agent in occupied Greece reported about the collaborationist Security Battalions that "they have lost their popularity in the mountains but in coastal areas and large towns they are looked upon as the lesser evils."33
number of informers, not through torture as suggested by Gilles Pontecorvo in his famous film The Battle of Algiers.
32 The same appears to have been the case in the occupied Soviet Union (Cooper 1979:45).
33 "Second Report of Colonel J. M. Stevens on Present Conditions in Peloponnese (24 June 1944)," PRO, HS 5/669/S6557.
i36
The Logic of Violence in Civil War
"Rurality" is a proxy for various causal mechanisms, including the ability of combatants to hide without being denounced because of rural norms of solidarity and honor; higher levels of tolerance among rural people to threats of violence; a tradition of rebellion reinforced by norms of reciprocity, which leads to mass participation in antistate activities ranging from contraband and banditry to full-fledged rebellion; and the fact that an economy based on subsistence farming tends to favor armed resistance more than one based on wage labor. Perhaps most important, the dispersion of population settlements in rural environments impedes policing (Kocher 2004); it is easier to enforce a curfew in a town than in a large rural area because taxing and monitoring hundreds or even thousands of hamlets exposes small army detachments to ambush (Tone 1994:13).34 Hofheinz (1969:76) attributes the realization of "the highest rates of mobilization and participation in Communist politics" in the "rear area base counties" because "of the security provided by terrain and distance." Communist insurgents on the Korean island of Chejudo and in Malaya were linked so closely to the mountains that they came to be known as "mountain," "hill," or "jungle people" (Yoo 2001; Kheng 1983:168). My informants in Greece often referred to the insurgent and incumbent camps using exclusively geographical identifiers: they talked about those "up" and those "down." Even within rural regions, insurgents are more likely to obtain collaboration in the roughest and most remote areas (Horton 1998:126; Nordstrom 1997:99; Escott and Crow 1986:376; Kitson 1960:124).
This insight allows the reinterpretation of some findings that take ideology or ethnicity as the main causal variable of violence. Timothy Gulden (2002) found that in Guatemala more than half of the army killings took place in municipalities in which the Mayas made up between 80 and 90 percent of the population (Mayas make up less than 8 percent of the total population in the country as a whole). Based in part on this finding, he claims that this violence constitutes an instance of genocide. However, these municipalities are mostly rural and located far from centers of government control. They could just as easily have been targeted because they were located in areas of guerrilla presence as because they were Mayan. This raises the issue of endogeneity of grievances: did the guerrillas pick their location based on the presence of Mayan grievances or did they educate the Mayas - who just happened to live in terrain that favored insurgent activity - about their plight? Empirical evidence supplied by Stoll (1993:87) allows a partial separation of the two: the army's repression did not focus on areas where
34 Gambetta (1993:109) finds that the Sicilian countryside is more difficult to police and, therefore, more agreeable to the Mafia than the cities. Not surprisingly, the Mafia is even able to exercise territorial control over some rural areas in Sicily. Tone (1994:162-6) compares the mountainous village of Echauri to the town of Corella, both in the Navarre; he attributes Echauri's insurgent outlook during the French presence to "solid community institutions [that] acted systematically to shelter individuals from the French regime" and Corella's collaborationist behavior to the character of its elites. However, the critical comparison for disentangling the effect of social structure from that of military resources would have been between a village similar to Echauri located just outside a French garrison. If my argument is correct, such a village would have collaborated with the French despite its social structure.
Control
!37
indigenous organizations (and presumably grievances) were strong and guerrillas had little presence, but rather in areas where the guerrillas were trying to organize the peasants despite weak indigenous organizations. In fact, the four areas of greatest government violence follow the insurgents' swath as it moved south to cut the Pan-American Highway.
The absence of overlap between prewar and war strongholds is visible where detailed studies are available. The Appalachians, the Cumberlands, and the Ozarks, Beckett (2001:11) notes, saw the rise of Confederate guerrillas during the American Civil War, even though these were the very areas within the Confederacy that most Union sympathizers inhabited. Following their defeat in the cities, the Chinese Communists staged a comeback from backward and isolated "border areas" where the prewar support was minimal if not nonexistent (Schran 1976). The urban populations of the German-occupied Soviet territories were more likely than rural ones to dislike the occupying authorities, partly because of their closer earlier identification with the Soviet regime and partly because of the more miserable conditions of life and work in the towns; yet, as Dallin et al. (1964:335) point out, "paradoxically, the partisan movement was largely a rural phenomenon. Research in a northern Greek region (Antóniou 2001) suggests that the electoral score of the Communist Party in 1936 was a bad predictor of the number of local men who joined the Communist-led resistance in 1942-4; instead, the distance from the town that served as the main base of the incumbent army proved to be an almost perfect predictor: the further away from the incumbent base a village was, the higher the proportion of local men who joined the rebels (prewar preferences appear to account for residual differences between equidistant villages). The French Communist guerrillas were very successful in the rural departement of Lot, where "communist candidates had stood in only two of the three constituencies in 1936 and had polled only 4,183 votes out of 30,293" (Kedward 1993:131). Elliott (2003:908) reports that the government's bombing and pacification campaign in the Mekong Delta of South Vietnam caused a disjuncture between the Vietcongs' class basis and control zones. The insurgency in El Salvador did not take place in the western departments of Ahuachapán and Sonsonate, homes of a mass peasant rebellion and subsequent massacre in 1932 (as well as of large coffee estates), but began in the isolated and underpopulated departments of Morazan and Chalatenango, which were peopled mostly by smallholders and which provided favorable terrain for organized groups to launch a rebellion (Grenier 1999:84). Likewise, the RENAMO insurgency against the FRELIMO government in Mozambique developed in the same areas where the FRELFMO anticolonial insurgency had been strong; in contrast, areas that supported the Portuguese incumbents during the anticolonial war tended to side with the FRELFMO incumbents during the RENAMO insurgency (Geffray 1990:41). Nordstrom's (1997:98-9) research in the isolated province of Niassa in the same country corroborates this point by demonstrating that RENAMO was able to generate very strong control and collaboration (with minimal violence) in an area that was both isolated and a prewar stronghold of FRELFMO. Geography was a clear proxy for military effectiveness: "As far as I could tell," Nordstrom notes (1997:99), little government "military interest
i38
The Logic of Violence in Civil War
extended to these regions. RENAMO-held zones were essentially left to their own devices."
The Nicaraguan case allows a type of natural experiment, because it is possible to compare the behavior of the Sandinistas in their successive roles as insurgents and, later, as incumbents. This comparison suggests that popular allegiances were largely endogenous to the exercise of territorial control. During the "contra" phase of the war, the (incumbent) Sandinistas firmly controlled the towns but were absent from the mountains: "The only Sandinista presence in the mountains would be a military one" (Horton 1998:137). As a result, people in those areas supported the contras. Many of these mountainous contra zones, however, had supported the Sandinista guerrillas in the 1970s (1998:21-2). The opposite was true of the towns, which were controlled by the (Somozista) incumbents in the first phase of the war and the Sandinista (incumbents) in the second one. In Horton's (1998:21) words, "Hundreds of Sandinista Army soldiers were stationed in the town of Quilali and as a result the town itself always remained firmly under FSMLN control." The population had no choice but to collaborate with them. In other words, whereas the Sandinistas qua insurgents based themselves in inaccessible rural terrain, they found themselves limited to cities when, qua incumbents, they faced the contra insurgency. In both cases, however, they obtained the collaboration of the population they ruled.
5.6. CONSTRAINTS ON MILITARY OPTIONS
If it is the case that full and permanent control over an area shapes civilian collaboration, then victory in civil war ought to be primarily a military task entailing the extension of control over the entire territory of a country. This much is conventional wisdom. In the words of a counterinsurgent, "There must be, above all, absolute determination to establish and retain a government police post intact and uncorrupt in every inhabited village. Authority must be re-established patiently, village by village, into the 'liberated' area, dealing with the easiest areas first" (Clutterbuck 1966:176).35
Civil wars, however, tend to take place in poor countries and they are protracted and inconclusive (Fearon and Laitin 2003; Fearon 2001). This stalemate reflects the rival actors' inability to establish full control over large areas of the country.36 Fenoglio's (1973:157) description of the Italian Civil War in 1943-5 is widely applicable: "The partisans were too strong to be attacked on their hills - at least this was the impression they gave; at the same time, they were too weak and
35 This view is, obviously, a reflection of a similar argument that links crime to police presence (C. Friedrich 1972:26).
36 Long duration in civil war may reflect two different processes. While some areas are "frozen" under the control of one or both actors, others shift back and forth. For example, the Ukrainian capital Kiev changed hands fourteen times in two years during the Russian Civil War (Werth 1998:111; Figes 1996:698). The only "wildly changing detail" in the thick ledgers maintained by UN teams in Sudan "was the column marked 'held by' which described whose fiefdom each village fell from month to month" (Peterson 2000:237).
Control
139
technically inept to be able to attack and dislodge the fascist garrisons in the towns of the plain."37
Once a civil war is on, the military requirements for the establishment and preservation of control over the entire territory of a country are staggering.38 This general problem, common to foreign occupiers and native sovereigns, is aptly summarized by Toolis (1997:70): "No army can patrol all of the roads all of the time." A military attache in Mozambique described it as the "big country, small army" problem (in T. Young 1997:150). A South Vietnamese official described it as a puzzle: "[W]e cannot stay with the people all the time. We come and go with operations by day, but we do not have enough strength to protect the people by night. I have yet to figure out how to protect a hamlet with thirty people. From a purely military viewpoint how can it be done? The Vietcong wait and wait, perhaps six months before they attack. We can build for two years, but they can destroy in one night. The person who finds the key to that puzzle has solved the problem" (in Race 1973:135).
Remarks about the numerical inadequacy of incumbent armies are commonplace. Of the counterrevolution in the country's western provinces, a French Republican general reported that the patriots "are so afraid, that we would need an entire garrison to guard every house" (in Dupuy 1997:133). A Union officer stationed in Missouri pointed out in 1863 that it would be impossible to exterminate the Confederate guerrillas "unless the government can afford to send ten soldiers for one guerrilla" (Fellman 1989:126). A British military report about the situation in Ireland in 1920 noted that "the police and military forces are too small to cope" (in Hart 1999:73). A Pakistani officer estimated that in order to successfully face the insurgency in Bangladesh, the Pakistani army would have needed 375,640 men as opposed to its actual force of 41,060 (Salik 1978:101). An American journalist reported that the Mozambican army needed more than a million men just to defend the country's infrastructure from the rebels but could only field thirty thousand (Finnegan 1992:95-6). The Rhodesian intelligence chief realized that demonstrations of force via army sweeps and air force bombings would be ineffective against Zimbabwean rebels because "the people in the rural areas would soon realise that such demonstrations could not be sustained and that a transient military presence could not enforce government policy" (Flower 1987:122). Finally, Forero (2001 A3) reports that in Colombia "the army is simply too small to cover the country.... And even when the army has carried out successful offensives, it has often been unable to set up a permanent presence."
A good illustration of this problem is supplied by the German occupation during World War II. The German forces assigned to fight against the various resistance movements were hopelessly inadequate for the task - nowhere more
37 Perpetual war in weak states lacking military resources is averted through the use of indirect rule (Kocher 2004).
38 Note that in times of peace, states are able to control their territory with much fewer resources that the same task requires from them during times of civil war. This suggests that the emergence of insurgencies cannot be simply accounted for by low levels of state control. Put otherwise, state capacity is a better argument for the dynamics of civil war than it appears to be for its onset.
140
The Logic of Violence in Civil War
so than in the Soviet Union. In an area of 43,000 square miles containing more than 1,500 villages and collective farms, the Germans had fewer than 1,700 men available for security duties, of whom only 300 were assigned to active measures against the partisans. In the central Soviet Union, the number of men of all types available for security duties was just 2 for every 3 square miles. "Although Hitler and the military and SS authorities came to understand the necessity of assigning considerable forces to secure the rear area, they were never able to make them available." Not surprisingly, the Germans were able to exercise only limited and very superficial control; vast areas that they simply abandoned quickly came under partisan control. Following a "mopping-up" operation in 1943, the German 221st Security Division reported that "the partisans had the opportunity... of reoccupying their former areas and thus making the success of these operations illusory.... Any removal of troops or a temporary withdrawal of troops from pacified areas resulted in reoccupation by partisans." As a German general put it, "With enough good troops, anything is possible." "The cause of the German failure," Cooper concludes "was both easy to analyse and impossible to rectify; it was simply, lack of troops."39
Even the U.S. military, with overwhelming resources in Vietnam, found it hard to overcome this problem: "We come here on an operation, and what does it prove?" remarked a U.S. soldier about a raid in a Vietcong-controlled village, in March 1969. "The VC will be back in control here tonight" (in M. Young 1991:240-1). The American military commander in Vietnam, General Westmoreland, was often ridiculed for his incessant demands for more troops, but his defense was not without merit: "I never had the luxury of enough troops to maintain an American, Allied or ARVN presence everywhere all the time. Had I at my disposal virtually unlimited manpower, I could have stationed troops permanently in every district or province and thus provided an alternative strategy. That would have enabled the troops to get to know people intimately, facilitating the task of identifying the subversives and protecting the others against intimidation. Yet to have done that would have required literally millions of men" (in Bilton and Sim 1992:34)4° The same problem, only more acute, can be seen in American-occupied Iraq. For instance, the 800-men strong Fourth Infantry Division's 1-8 Battalion, based in the area of Balad, in the restive Iraqi province of Anbar, was responsible for nearly 750 square kilometers (Filkins 2005:55). Likewise, only about 800 soldiers covered the area around the town of Rawa, which is the size of Vermont - and only 300 left their outpost on operations and never all
39 Cooper (1979:45, 143-4, !53-4)-
40 Sheehan (2000:179) confirms this point. The "lack of sufficient American troops to occupy and hold ground when it has been wrested from the Communists," he pointed out, "is one of the major reasons for the extent of damage to civilian life and property." An additional problem is posed by the large proportion of personnel devoted to support in modern armies. Luttwak (2003) estimated that out of the 133,000 American men and women in Iraq, only 56,000 are combat-trained troops available for security duties, while the number of troops on patrol at any one time is no more than 28,000. See as well, Shepherd (2002:351); Tucker (2001:90); Fall (2000:199); Vargas Llosa (1998:137); Fisher (1997:50); Finley (i994:xi, 29); Tone (1994:80, 143-4); Ortiz Sarmiento (1990:132); Ekirch (1985:114); Li (1975:187); Beaufre (1972:66).
Control
141
at the same time. As a result, "there is only a sporadic American military presence outside the few towns now occupied. Neither the Army nor the Marines maintain any permanent checkpoints" in the main regional road. The dearth of soldiers is critical because, as Colonel Stephen Davis, commander of Marine Regimental Combat Team 2, put it: "You can go through these towns again and again, but you can't get results unless you are there to stay" (in C. Smith 2005^6).
The difficulty of establishing full and permanent control through sheer numbers puts a premium on the shrewd use and allocation of existing military resources, on resilience, as well as on the ability to claim outside assistance, especially at crucial moments. Limited resources place a premium on the effective use of violence. But what makes violence effective?
5.7. VIOLENCE AND DISCRIMINATION
According to Michael Hechter (1987:162), a key determinant of collaboration is the perceived probability of sanction. Cesare Beccaria pointed out that "the political intent of punishments is to instill fear in other men," while Jeremy Bentham defined deterrence in terms of the "intimidation or terror of the law" (Zimring and Hawkins 19 7 3:7 5). In its simplest formulation, the theory of deterrence posits that threats can reduce the likelihood that certain actions will be undertaken. In a different formulation, deterrence by punishment is a method of retrospective inference via threats, so that whenever a wrong has been actually committed, the wrongdoer shall incur punishment (Kenny 1907). To Bentham we owe the main hypothesis: "The profit of the crime is the force which urges a man to delinquency: the pain of the punishment is the force employed to restrain him from it. If the first of these forces be the greater the crime will be committed; if the second, the crime will not be committed" (in Zimring and Hawkins 1973:75)-
Yet we know that many crimes are committed despite known and credible threats. Katz (1988:12-51) shows that a substantial number of homicides are carried out by people who are indifferent to sanctions; these homicides, which he calls "righteous slaughters," emerge quickly, are fiercely impassioned, and lack premeditation. Bentham's account of deterrence has also been criticized as "mechanical" and based "upon false psychology"; it is argued instead that threats may sometimes generate a desire of noncompliance and that criminal phenomena are completely independent of penal laws. At the same time, it is widely recognized that most people refrain from crime to avoid sanctions. Thus, a reasonable degree of deterrence can be achieved.41
When are threats effective? Beccaria (1986:81) argued that sanctions must be public, prompt, necessary, minimal under the given circumstances, proportionate to the crimes, and established by law. Zimring and Hawkins (1973) stress three conditions: threats must be publicly known, persuasive, and personalized. Hechter (1987:151) argues that compliance is more likely when people are
41 In Zimring and Hawkins's (1973:95) formulation, "It appears that the introduction of a threat as a barrier to committing a particular behavior is likely to cause members of a threatened audience to revise attitudes toward the desirability of the behavior."
I42
The Logic of Violence in Civil War
required to meet highly specific obligations rather than nonspecific ones. These features can be subsumed, to some degree, under the distinction between selective (or discriminate) violence and indiscriminate violence.42
Both selective and indiscriminate violence are, in principle, instrumental forms of violence aiming to generate collaboration via deterrence. The distinction is based on the level at which "guilt" (and hence targeting) is determined.43 Violence is selective when there is an intention to ascertain individual guilt. Because intentions are not always visible (though in many cases indiscriminate violence is publicized by political actors), one way to operationalize this distinction is by noting that selective violence entails personalized targeting, whereas indiscriminate violence implies collective targeting.44
In indiscriminate violence, also described by its legal designation, "reprisals," the concept of individual guilt is replaced by the concept of guilt by association: "If such people as are guilty cannot be found," proclaimed the German command in occupied Greece, "those persons must be resorted to, who, without being connected with the actual deed, nevertheless are to be regarded as co-responsible" (in Condit 1961:265-6). The specific rule of association varies and ranges from family to village, region, and nation. The most extreme form of indiscriminate violence is probably the one that selects its victims on the basis of membership in a nation or an ethnic or a religious group; it is often described as "random" violence and its archetypal example is a strain of Nazi terror in parts of occupied Europe. "On more than one occasion in the town of Athens," writes McNeill (1947:57), "a German patrol was sent out to the scene of the death of a German soldier, and there they arrested the first fifty persons who happened to walk down the street, lined them up against a wall and shot them out of hand." German terror in Warsaw during the same period is starkly described by Czeslaw Milosz (1990:90):
Once, in the first year of the War, we were returning from a visit to a mutual friend who lived in the country. As I remember, we were arguing about the choice of a train. We decided against the advice of our host to take a train leaving half an hour later. We arrived in Warsaw and walked along the streets feeling very satisfied with life. It was a beautiful summer morning. We did not know that this day was to be remembered as one of the blackest in the history of our city. Scarcely had I closed the door behind me when I heard shrieks in the street. Looking out the window, I saw that a general man-hunt was on. This was the first man-hunt for Auschwitz. Later millions of Europeans were to be killed there, but at the time this concentration camp was just starting to operate. From the first huge transport of people caught on the streets that day no one, it appears, escaped alive. Alpha
42 Selective violence is personalized, but needs not be public, prompt, or necessary - though it often is; it is certainly minimal and proportionate when compared with indiscriminate violence.
43 Because what matters is the level at which the targeting takes place, one can talk of violence that discriminates at the individual, local, or national level. However, I use the distinction between selective and indiscriminate violence because it captures the essential differences of targeting at the individual level versus any supra-individual level.
44 Note that, contrary to widespread perception, selective violence can end up being, and often is, massive in scale. For example, the Vietcong are estimated to have selectively assassinated as many as 50,000 people in a decade and a half (Wickham-Crowley 1990:215).
Control
H3
and I had strolled those streets five minutes before the beginning of the hunt; perhaps his umbrella and his insouciance brought us luck.
Because such threats are completely unpredictable, they produce, at least initially, a paralyzing, turbulent, and irrational fear, scarcely permitting any thought, leading to the atomization of society (E. Walter 1969:25-6; Thornton 1964:81). In a book published in 1947, a group of Greek psychiatrists reported the results of a remarkable study on the effects of German terror on the population of Athens; they found that most people were paralyzed by the daily expectation of an "unpredictable and unknown misfortune" and the "incredible anxiety in front of the unknown which afflicted every individual fate" (Skouras et al. 1947:124-36). As long as the victims have no way to react against such violence, its effect is "to increase compliance with authority among those who feel they may be threatened" (L. White 1989:328). In other words, the population may be pushed into total passivity and political abdication.
Although random violence may work for a dictator (McAuley 1992:50; B. Moore 1954:169-70), it is much less likely to achieve its aims in the midst of a civil war, where the presence of a rival makes defection possible. First, random violence defeats deterrence because it destroys the possibility of anticipation of a forthcoming evil and, hence, the ability to avoid it: it erases the relationship between crime and punishment, thus abolishing the concept of transgression. Its sheer unpredictability makes everyone fear lethal sanctions regardless of their behavior: innocence is irrelevant and compliance is utterly impossible. A German report (in Cooper 1979:27) described the attitude of the average citizen in the occupied areas of the Soviet Union: "If I stay with the Germans, I shall be shot when the Bolsheviks come; if the Bolsheviks don't come, I shall be shot sooner or later by the Germans. Thus, if I stay with the Germans, it means certain death; if I join the partisans, I shall probably save myself." Under such conditions, "abstention ceases to seem a protection. Recruitment of insurgents goes up as risks of passivity and insurgency begin to equalise" (Aron 1966:170). Indeed, Nazi terror in Poland "left the Poles no other alternative but to ignore the occupier -either actively, by opposing him, or passively, by behaving as if he did not exist" (Jan Gross 1979:238):
One would expect that noncompliance with German demands carried such drastic penalties that scarcely anyone would dare to defy them. But full compliance was impossible; terror continued and even intensified with time. The population quickly recognized the new logic of the situation: whether one tried to meet German demands or not, one was equally exposed to violence.... It makes no sense, in the context of random punishment, to style one's life according to the possibility of being victimized, any more than it makes sense to orient all of one's everyday acts to the possibility of an accident. (Jan Gross 1979:212)
Second, whereas compliance guarantees no security under conditions of indiscriminate violence, collaboration with the rival faction may both increase one's chances of survival and allow for a sense of normative integrity (Jan Gross 1979:202). In Poland, membership in the resistance made people more prudent
i44
The Logic of Violence in Civil War
and erased the false sense of security that was often fatal to those not involved in it; "conspirators" actively avoided capture by the Germans, while nonconspir-ators were much less careful in avoiding accidental contacts with the occupiers because they often felt that should they be arrested they would spend a few days in detention and later, once their innocence was established, they would be released. However, this assumption often proved fatal, given that there was little relationship between crime and punishment. "Conspirators" very often had much better identification papers than nonconspirators and, if apprehended, had already prepared satisfactory answers to the typical questions police would ask. When they were caught in a roundup, someone in the network would try to get them out of prison in time, or their families would be given money to bribe the appropriate officials. When threatened with arrest, blackmail, or denunciation, conspirators had vast organizational resources at their disposal: the organization would help them to disappear, find them a new place to live, and give them new employment or new documents (Jan Gross 1979:234-5).
Hence it is possible to argue that indiscriminate violence is of limited value since it decreases the opportunity costs of collaboration with the rival actor. A British counterinsurgent compared indiscriminate violence to "trying to catch fish in a weedy pond by splashing about with a rather widemeshed net as opposed to adopting the tactics of the pike, and lurking quietly in the weeds ready to snatch unsuspecting fish as they swim by" (Paget 1967:110). It is, therefore, possible to formulate the following proposition:
Proposition 2   Indiscriminate violence is counterproductive in civil war.
This proposition is a conjecture. Theoretical work on the related nexus between repression and dissent remains inconclusive (Lichbach 1987:297). Empirically, we lack controlled comparisons of outcomes in the presence and absence of such violence. Little attention has been paid to counterfactuals. For instance, we do not know how many insurgent armed actions would have taken place and how many people would have joined the rebels in the absence of indiscriminate violence. A few detailed studies show that indiscriminate violence was sometimes more successful than generally thought (Hill 2002; Hartford 1989).45 At the same time, the anecdotal evidence weighs heavily in favor of this proposition. In the following chapter I return to this issue.
In contrast to indiscriminate violence, selective violence personalizes threats; if people are targeted on the basis of their actions, then refraining from such actions guarantees safety. Practitioners and observers agree that selective violence is the most efficient way to deter defection. In Robert Thompson's formulation (1966:25), "Terror is more effective when selective." As an American colonel in
45 A Vietcong cadre argued that indiscriminate shelling by government forces in South Vietnam weakened the insurgency: "From experience, I have realized that the Front is strongest in villages which haven't been shelled and that, on the contrary, it weakens where the shellings happen frequently. To wage Front propaganda and to sow hatred against the GVN, Front cadres need quiet" (in Elliott 2003:767). Elliott concludes that little shelling favors the Vietcong but too much does not, presumably because no protection could be offered.
Control
*45
Vietnam put it, "You really have to use a surgeon's scalpel" (in Race 1973:238); Che Guevara (1998:91) recommended that "assaults and terrorism in indiscriminate form should not be employed." The Vetcong produced many official documents explaining the advantages of selective violence (e.g., Elliott 2003:266).
In practice, the distinction between selective and indiscriminate violence hinges on public perceptions since it is possible to pretend to be selective by indiscriminately targeting isolated individuals. As long as people perceive such violence to be selective, it will have the same effects as selective violence. If people do not perceive it as selective, the results will be the opposite, much like when they perceive selective violence to be indiscriminate. I discuss these issues in detail in Chapter 7.
The choice of whether to use selective or indiscriminate violence is heavily dependent on the quality of information available - one cannot discriminate without the information to discriminate - which itself is heavily dependent on the nature of the sovereignty exercised. Information requires collaboration, which requires a level of control sufficient to reassure those who can supply that collaboration. Although actors are less bound in their ability to perpetrate it, indiscriminate violence is less likely to work under circumstances of fragmented sovereignty. I address these questions in the following chapters.
5.8. CONCLUSION
This chapter has specified a theory of irregular war stressing the role of control in shaping civilian collaboration. A key point is that control - regardless of the "true" preferences of the population - precludes options other than collaboration by creating credible benefits for collaborators and, more importantly, sanctions for defectors. The distribution of control can be shaped by military means, because sufficient military presence raises the credibility of sanctions for defection; at the same time, however, military resources for the establishment of total control are typically lacking. Political actors thus turn to violence, but to be effective, violence must be selective.
The role of sovereignty in shaping the use of selective violence combines with the counterproductive effects of indiscriminate violence to set the remainder of the book's theoretical agenda: first, an account must be provided for the occurrence of indiscriminate violence; second, an analytical treatment of selective violence must be specified. The former is the object of the following chapter, while the latter is addressed in Chapter 7.
6
A Logic of Indiscriminate Violence
Je vois des malheureux, mais, en verite, je ne puis trouver des coupables.
Stendhal, L'abbesse de Castro
We have by our own imprudencies & irregular proceedings made more Enemies than have become so from mere inclination.
General Stephen Drayton, North Carolina, 1781
Look at me -1 hadn't wanted to fight, they made me!
A Chechen fighter, after a Russian atrocity
This chapter specifies the logic driving indiscriminate violence. Proposition 2 posits that indiscriminate violence is counterproductive in civil war contexts. If this is so, then why is it observed so often? Addressing this puzzle calls for a theory of indiscriminate violence.1 I begin by examining how and when indiscriminate violence is observed. Next, I discuss its logic and specify the conditions under which it is counterproductive. I then review four arguments that account for why indiscriminate violence is observed, despite its apparent counterproduc-tivity, including the specious observation of indiscriminate violence because of truncated or misinterpreted data, and its commission as a result of ignorance,
1 This argument applies within the book's scope conditions: it presupposes that at least one actor intends to control the population against which violence is used. Indiscriminate violence may also be used to deport or exterminate particular groups. For example, secessionist insurgents may use indiscriminate violence against ethnic rivals to drive them off the territory they seek to control (e.g., Senaratne 1997:88). The same is the case in instances of "reciprocal extermination." In Lotnik's (1999) account of the Polish-Ukrainian clash of 1943-4, massacres of villagers targeted primarily the rival group. A former Polish partisan, Lotnik (1999:65), recalls his officer's talk on the eve of one of the first massacres: "Don't burn, don't loot. Just shoot young, able-bodied men. If anyone resists, make sure you shoot him before he shoots you. We have to teach them that they cannot take out selected Polish citizens and kill and torture them. We must teach them that they can't get away with that." Such cases are outside the book's scope conditions, which posit civilian compliance as a central goal.
146
Indiscriminate Violence
147
cost, and institutional constraints. I argue that indiscriminate violence emerges, when it does, because it is much cheaper than its selective counterpart. Yet, any "gain" must be counterbalanced by its consequences. Thus, indiscriminate violence is more likely either under a steep imbalance of power between the two actors or where and when resources and information are low. In the absence of a resolution of the conflict, even indiscriminate actors are likely to switch to more selective violence.
6.1. THE INCIDENCE OF INDISCRIMINATE VIOLENCE
Like other forms of violence, indiscriminate violence may be used to achieve a variety of goals, such as exterminating particular groups, displacing people, plundering goods, or demonstrating a group's power and ability to hurt another group. Consistent with this book's scope conditions, my focus in this chapter is on the use of indiscriminate violence to control a population rather than simply to loot, displace, or eliminate it.2
Seen from this perspective, indiscriminate violence is, initially at least, a way to come to grips with the identification problem. "A major problem for the Philippine military," writes Berlow (1998:180), "was the one the Americans encountered in Vietnam: They couldn't figure out who the 'fish' were until they started shooting. To be on the safe side, Filipinos, like the Americans in Vetnam, erred on the side of overkill and assumed that anyone was an enemy until proven otherwise."3
Distinguishing between indiscriminate and selective violence at the aggregate level is difficult. It is, therefore, close to impossible to estimate the contribution made by each type of violence to the overall fatality count. Indiscriminate violence is much more visible than its selective counterpart and, as such, is thought to be more prevalent (Valentino 2004; Downes 2004). The emphasis on indiscriminate violence often reflects the tendency of many observers to designate as indiscriminate all kinds of extrajudicial killing, including instances of selective violence (e.g., Carlton 1994:1). For example, the killing by Iraqi insurgents of "Iraqi officers, civilians, Iraqi, American and coalition soldiers" is described
2 Recall that this is an ideal-typical distinction. There are several examples of indiscriminate violence that begins as counterinsurgency on the cheap only to evolve into a process of haphazard quasi extermination, such as Darfur (Prunier 2005).
3 Henderson's (1985:179-80) description of the attitude of the Colombian army during the Violencia could apply to almost any case: "The underlying assumption was that every farmer was a 'bandit,' or potentially one, and should be treated as such." As a man from Guatemala told Stoll (1993:97): "All the Ixils of Nebaj, Cotzal, and Chajul they considered guerrillas. They were afraid of their own shadow." According to Gardner (1962:152-3), "the average German soldier [in Greece] became something less than particular about whom he shot or captured. His reasoning was that any man found in the area was either an active guerrilla or in league with the local band. For this reason, German figures for guerrilla casualties were usually much higher than those announced by the andartes [Greek partisans]." Young men, in particular, caught in an operation zone in "enemy" territory during a mopping-up operation are particularly likely to be killed. In a letter to their son, the parents of a peasant from the Sarthe described to him in detail how three unarmed friends of his were killed by French Republican soldiers because they began to run when they saw the soldiers coming (Dupuy 1997:182-3).
i48
The Logic of Violence in Civil War
as indiscriminate (Lins de Albuquerque and Cheng 2005:11). Zulaika (1988:85) writes of the "indiscriminate killings of chivatos (informers) and civil guards carried out by ETA."
The tendency to code all violence as indiscriminate is assisted by the scarcity of information: "The confused, unstable, and dangerous situation," writes Jagath Senaratne (1997:146) about Sri Lanka, "led many to believe that the violence was random and meaningless. The imputations of randomness by some observers (mainly journalists) was a result of the inability to see the many different strands of the violence... [and] to disaggregate 'the violence' into its components." Indeed, it is safe to say that rarely is noneliminationist violence totally random. Generally, the victims of indiscriminate violence are selected on the basis of a criterion, usually location. For example, the mass violence perpetrated by the Germans in Athens during the summer of 1944 targeted specific neighborhoods suspected of harboring Communist activity. Furthermore, an important part of this violence targeted specific individuals; neighborhoods were cordoned off, and their inhabitants taken to the central square where local hooded informers would finger individual suspects (I return to this point in section 6.5.1).
We simply do not know what the universe of civil war violence looks like. Nevertheless, descriptions of indiscriminate violence in civil war are numerous enough to suggest that no matter how bad our data, genuinely indiscriminate violence takes place often enough to warrant attention.
6.2. INFORMATION AND INDISCRIMINATE VIOLENCE
The preceding examples suggest that violence is indiscriminate when selection criteria are rough. This is the case when precise information is unavailable. An observable implication is the oft-noted association of indiscriminate violence with incumbents rather than insurgents.4 Insurgents are almost always the first movers; having eliminated the state's presence in the areas they control, they set up village-based administrations that are able to collect the kind of information that allows them to address the identification problem effectively (Wickham-Crowley 1990:216-17). "While the party had a thousand eyes and a thousand ears," Carlos Ivan Degregori (1998:143-4) observes about the Peruvian Shining Path, "the Armed Forces were blind or, rather, color-blind. They saw only black
4 Surveys conducted in Vietnam found that refugees who moved away from their homes because of (indiscriminate) bombardment and ground operations tended to associate these actions with the incumbent regime, while refugees who moved because of (selective) terror and coercion tended to associate them with the insurgents (Wiesner 1988:111). See also Spencer (2000:131); Benton (1999:102-3); Horton (1998:127); Cribb (1991:151); Carmack (1988^60); Calder (1984:159); Hen-riksen (1983:118); Armstrong (1964:41); Dallin et al. (1964:328). Linking indiscriminate violence to lack of information is consistent with empirical evidence from the former Yugoslavia and Israel showing that there is much more indiscriminate violence between the same groups when the victimizing group operates outside rather than inside state borders (Ron 2003). Ron provides a different explanation for this pattern, namely that borders have a significant effect on the conduct of war from the perspective of the regime, but the availability of information could be the causal mechanism accounting for the difference.
Indiscriminate Violence
149
and white____They did not perceive nuances; when they saw dark skin, they
fired." Likewise, an observer noted that in Indochina "the French destroy at random because they don't have the necessary information" (in Leites and Wolf 1970:109), and a U.S. report (Barton 1953:138) pointed out that "the guerrillas have a more effective intelligence system than their opponents."
Incumbent indiscriminate violence usually takes place in the context of military operations known as "mopping up," "comb," "cordon and search," "search and destroy" or "scorched earth" campaigns that seek to encircle and liquidate insurgents and undercut an insurgency's civilian basis. These campaigns are often dubbed "pacification" campaigns.5 The result is almost always uniform: indiscriminate violence. A U.S. officer, stationed in the Philippines in the beginning of the twentieth century, pointed out that "we do not know insurrectos and bad men from good ones, so we are often compelled to arrest all alike" (quoted in Linn 1989:139); a Filipino captured this problem when he described the U.S. Army as a "blind giant," powerful enough to destroy the enemy, but unable to find him (quoted in Linn 1989:160). When the U.S. Marines arrived in the province of Segovia in Nicaragua in 1927, they "had no practical way to distinguish between rebel sympathizers, supporters, and soldiers and 'peaceful civilians.' Facing these uncertainties, they opted to wage a brutally violent offensive against Segovian campesinos generally" (Schroeder 2000:39). Even pacification campaigns that claim a higher moral ground have resulted in significant indiscriminate violence, as suggested by more recent U.S. counterinsurgencies in Vietnam, Afghanistan, and Iraq (Kalyvas and Kocher 2005).
That indiscriminate violence is related to lack of information (rather than, say, ideology) is confirmed by the fact that insurgents do not shy away from this practice.6 Insurgents use it when they lack information: against villages that openly support the incumbents by setting up local militias, in areas where their presence is limited (such as urban centers), and after their administrative apparatus has been destroyed, as in Algeria in 1997 (Kalyvas 1999) or Malaya (Clutterbuck 1966:63).
6.3. DETERRENCE AND INDISCRIMINATE VIOLENCE
In 1981, after the Atlacatl Battalion massacred hundreds of villagers in the Sal-vadoran village of El Mozote, its soldiers carried a green cloth with white letters
5 Unaware of its own irony, a U.S. report in Vietnam pointed out that "areas cannot be pacified if there are no people living in them" (quoted in Wiesner 1988:113). The Japanese used terms such as "operation clean-up" and "operation purification by elimination." Their "three clears" policy (for clearing all grain, draft animals, and people) was termed "three all policy" by their opponents (for "take all, burn all, and kill all"). The Indonesian army coined the term "operation extinction" in East Timor, and the Guatemalan army referred to "operation cinders."
6 Peterson (2000:220); Horton (1998:167); Manrique (1998:218); Del Pino (1998:163-4, 172); Berlow (1998:197); Richards (1996:181); Swedenburg (1995:153); Shalita (1994:142); De Waal (1991:48); Geffray (1990:214-5); Fellman (1989:25); Home (1987:221-2); Wiesner (1988:58, 123); West (1985:272); Kheng (1983:65); Rodriguez (1982:33-4); Lewy (1978:276); Paget (1967:93-4); Mallin (1966:60); R. Thompson (1966:25-7); Pye (1956:104); Leakey (1954:101).
15°
The Logic of Violence in Civil War
that said "If the guerrilla returns to Morazan, the Atlacatl will return to Morazan" (Binford 1996:23). "Even when the Renamo adopted a strategy of mass terror in the mid-1980s," Finnegan (1992:58) points out about Mozambique, "most of its brutalities had discernible motives. Someone was suspected of withholding information, or a village was suspected of withholding food, and the bandidos wanted to make sure the neighbors got the message." As these examples suggest, and contrary to much conventional wisdom (e.g., Gurr 1986:51), indiscriminate violence is not necessarily gratuitous, wanton, or solely vengeful; rather, it often aims to deter people from collaborating with the rival actor by collectively sanctioning suspected collaborators and those related to them.
The central aim of indiscriminate violence is to shape civilian behavior indirectly through association. "Burn some farms and some big villages in the Morbi-han and begin to make some examples," wrote Bonaparte to General Guillaume Brune who, as commander of the army of the West, was getting ready to quash the monarchist rebellion; "it is only by making war terrible," he added, "that the inhabitants themselves will rally against the brigands and will finally feel that their apathy is extremely costly to them" (quoted in Dupuy 1997:158-9). The use of indiscriminate violence against Indian tribes by U.S. troops "raised the hope that severe enough punishment of the group, even though innocent suffered along with the guilty, might produce true group responsibility and end the menace to the frontiers" (Paludan 1981:43). A similar point was made in Missouri during the Civil War: "There will be trouble in Missouri until the Secesh [Secessionists] are subjugated and made to know that they are not only powerless, but that any desperate attempts to make trouble here will only bring upon them certain destruction and this [certainty] of their condition must not be confined to Soldiers and fighting men, but must extend to non-combatant men and women''' (in Fellman 1989:201).
A March 1944 public announcement of the Germans in occupied Greece stated that sabotage would be punished with the execution by hanging of three residents of the closest village unless the perpetrators were arrested within forty-eight hours or it was proved that the villagers had actively discouraged sabotage actions. This kind of violence provides a basic incentive for collaboration, namely the prevention of the threatened harm. The Germans' announcement concluded: "Hence the duty of self-preservation of every Greek when learning about sabotage intentions is to warn immediately the closest military authority" (in Zervis 1998:179).
Here is, then, the logic of indiscriminate violence in a nutshell: if the "guilty" cannot be identified and arrested, then violence ought to target innocent people that are somehow associated with them. The underlying assumption is that the "innocent" will either force the "guilty" to alter their behavior or the "guilty" will change their course of action when they realize its impact upon "innocent" people they care about - or both. In addition to spreading responsibility, indiscriminate violence also introduces an explicit calculus of comparative sanctions: the targeted population will collaborate with the incumbents because it fears their sanctions more than the rebels'. As a German army order pointed out, "the population must be more frightened of our reprisals than of the partisans" (Heilbrunn 1967:150).
Indiscriminate Violence
6.4. COUNTERPRODUCTIVE EFFECTS OF INDISCRIMINATE VIOLENCE
Though appalling as a practice, indiscriminate violence is not lacking in logic. Yet few observations seem to enjoy wider currency than the perception - shared by perpetrators, individuals targeted by indiscriminate violence, and outside observers alike - that indiscriminate violence is at best ineffective and at worst counterproductive.
Writing about the Vendee War in 1797, Gracchus Babeuf (1987:119) observed that the violent measures of the Republicans against the Vendean insurgents "were used without discrimination and produced an effect that was completely opposite to what was expected." A Greek guerrilla leader in Ottoman Macedonia at the start of the twentieth century asserted that a judicious balance had to be used in the administration of violence "for indiscriminate killing does harm rather than good and makes more enemies"; another one remarked that "the art is to find who should be punished" (in Livanios 1999:206). "No measure is more self-defeating than collective punishments" argues a classic text of irregular war (Heilbrunn 1967:152). Henriksen (1983:129) affirms that in "revolutionary warfare," "reprisals serve the rebels' cause." He notes (1983:128) that in colonial Mozambique, "again and again, FRELIMO converts pointed to Portuguese acts as the prime factor for their decision. Non-Portuguese observers substantiated this assertion."James S. Coleman (1990:501) includes the precept "Do not engage in indiscriminate terror" among the four basic recommendations for action that ought to guide both incumbents and insurgents.
Insurgents are well aware of the features of indiscriminate violence: "The party was correct in its judgment that government doctrine ... would drive additional segments of the population into opposition," a Vetcong document pointed out, "where they would have no alternative but to follow the Party's leadership to obtain protection" (Race 1973:172). Che Guevara went so far as to locate a key mechanism driving peasant support for the rebels precisely in the indiscriminate behavior of incumbents (Wickham-Crowley 1992:139), a point echoed by arguments positing that "along with the organizational catalyst, what is required to convert normally risk-averse peasants into revolutionary soldiers is a high level of indiscriminately targeted repressive violence" (Mason and Krane 1989:176). As Truman Anderson (1995 1143) concludes, "the primary contribution" of indiscriminate violence to the prosecution of modern wars has actually been to aggravate insurgencies and leave lasting, bitter memories which time does not erase." Arendt (1970:56) must have had indiscriminate violence on her mind when she remarked that "violence can destroy power; it is utterly incapable of creating it."
Perhaps the most striking case for the counterproductive effects of indiscriminate violence is the oft-noted tendency of insurgents to actually welcome incumbent reprisals - or even provoke them by ambushing isolated enemy soldiers close to a village - because such reprisals bring in recruits.7
7 Aussaresses (2001:62); Hayden (1999:39, 57); Bennett (1999:143); Keen (1998:21); Senaratne (1997:95); Schofield (1996:246); C. Schmitt (1992:280). International sympathy caused by atrocities represents an additional benefit for insurgents.
The Logic of Violence in Civil War
The most infamous example of the futility of indiscriminate violence is possibly the Nazi reprisal policy in occupied Europe, aimed at deterring resistance against occupation. Reprisals appear to have been an utter and complete failure: they simply did not stiffle resistance activity and, more importantly, they appear to have actually induced people to join the resistance. "Whatever the purpose of the German policy of reprisals," Condit (1961:268) points out, "it did little to pacify Greece, fight communism, or control the population. In general, the result was just the opposite. Burning villages left many male inhabitants with little place to turn except guerrilla bands. Killing women, children, and old men fed the growing hatred of the Germans and the desire for vengeance."8 German observers in neighboring Yugoslavia "frankly concluded that rather than deterring resistance, reprisal policy was driving hitherto peaceful and politically indifferent Serbs into the arms of the partisans" (Browning 1990:68). Nazi reprisals produced a similar effect all over occupied Europe (Mazower 1998:179).9 Japanese reprisals had similar effects in occupied Asia.10
The counterproductive effect of indiscriminate violence holds beyond the excessive levels of Nazi and Japanese violence. Consider the following examples from Sudan (Darfur), Guatemala, Vietnam, and Venezuela:
During a week spent traveling on a pickup truck piled high with roughly 15 fighters in the Sudan Liberation Army, or S.L.A., one of Darfur's two rebel groups, one thing stood out as starkly as a full moon over the Sahara: much of the responsibility for the growth of this insurgency lies with the Arab-led government in Khartoum.... Among the foot soldiers of the insurrection, the tactics of the janjaweed [pro-government militia] and government forces have stirred a deep well of anger and distrust and fueled an impulse for redress. To acquaint oneself with the rebels for even a few days is to discover the formula for an insurrection: kill a boy's kin, take a man's cattle, and a rebel is born. "They killed my father, so I joined the S.L.A," is how young Khalid Saleh Banat, [a] 13-year-old, put it. (Sengupta 2004^1, A8)
Immediately after the Guatemalan army killed about 50 people, including women and children, in the village of La Estancia, forty young men and women left the village to join the guerrillas. (Carmack 1988^54-5)
"Every time the Army came they made more friends for the VC." a Vietnamese peasant said about South Vietnamese army raids in his village. (Trullinger 1994:85)
A Venezuelan guerrilla suggested that there was probably a new recruit for every woman raped by government soldiers. (Wickham-Crowley 1990:234)"
8 According to historians, reprisals in Greece produced only local and limited aftereffects of intimidation (Hondros 1993:155-6; McNeill 1947:57-8).
9 Soviet Union (Shepherd 2002; Cooper 1979; Armstrong 1964:30; Dallin et al. 1964:328), Poland (Lotnik 1999:87), Bosnia (Gumz 2001:1037), Italy (Minardi 2002:8; Klinkhammer 1997:83; Col-lotti 1996:27; Pavone 1994:478), and France (Kedward 1993:190).
10 China (Lary 2001:109-10; Li 1975:209-10, 231), the Philippines (McCoy 1980:215; Kerkvliet 1977:68), Malaya (Kheng 1983), Burma (Tucker 2001), and Vietnam (Herrington 1997:21).
11 For general statements, see Rich and Stubbs (1997:7); Andreopoulos (1994:196); Bard O'Neill (1990:80); Molnar (1965:117). Similar observations have been made about the Vendee War (Laqueur 1998:24), the American Revolution in New Jersey (Shy 1976:205-6), South Carolina (Weir 1985:74), and North Carolina (Escott and Crow 1986:393; Crow 1985:145,173);
Indiscriminate Violence
153
Yet exactly why and how indiscriminate violence fails remains unspecified. I identify and examine five possible mechanisms: the emotional reactions it provokes, its ambiguous structure of incentives, reverse discrimination, selective incentives for the rivals, and the overestimation by those who use it of the strength of ties between political actors and civilians.12
6.4.1. Emotional Reactions and Norms of Fairness
Machiavelli (The Prince, III: 19) argued that punishment "should be used with moderation, so as to avoid cause for hatred; for no ruler benefits by making himself odious." Because indiscriminate violence targets people independently of what they did or could have done, it is perceived as deeply unfair. Unfair and immoderate punishment always creates a "bad impression," in the words of a Vietcong cadre (in Elliott 2003:91). Worse, it may trigger an intense emotional
French-occupied Spain (Tone 1994:103); the American Civil War in Missouri, "where the Confederacy gained in popular appeal when Missouri was 'invaded' and occupied by often brutal military forces" (Fellman 1989:11), and in North Carolina during the same period, where "terror did not paralyze guerrillas; it gave them power" (Paludan 1981:101); the Irish Civil War of 1922-3 (Laqueur 1998:180); the U.S. counterinsurgency in the Philippines in 1899-1902 (Linn 1989:85); the Dominican Republic in 1917-22 (Calder i984:xiv, 123); and Nicaragua in the 1920s, where "the extreme violence of the invading and occupying forces spurred the rapid growth of Sandino's Defending Army (Schroeder 2000:38); the Russian Civil War (Werth 1998:115; Figes 1996:565, 583; Brovkin 1994:201); the Chinese Civil War (Thaxton 1997:308-9; Hua and Thireau 1996:302; Griffin 1976:146); the Soviet reprisals in the Baltic after 1944 (Petersen 2001); anti-Japanese and anticolonial insurrections in Malaya (Stubbs 1989:256; Kheng 1983:24, 65; Kheng 1980:97; R. Thompson 1966:25; Clutterbuck 1966:161; Barton 1953:136); Kenya (D. Anderson 2005:69; 192-3; Paget 1967:29; Barnett and Njama 1966:197), Mozambique (Lubkemann 2005:496; Hen-riksen 1983:128), and Angola (Cann 1997:28); the Algerian War of Independence (Butaud and Rialland 1998:103); the Colombian Violencia in the 1940s (Roldan 2002:209; Ortiz Sarmiento 1990:174; Henderson 1985:143, 180); the Vietnam War (Wiesner 1988:32; Race 1973:197; Klonis 1972:182; Taber 1965:95); Laos in the 1960s (M. Brown 2001:26); the Philippines in the 1950s (Kerkvliet 1977:143; Crozier 1960:217) and more recently (McKenna 1998:156, 191-2; Jones 1989:125); Burma in the 1960s and 1970s (Tucker 2001:43, 9°); Cyprus (Paget 1967:29); Cuba (Jones and Molnar 1966:71); Bangladesh in 1971 (Salik 1978:104); El Salvador (J. L. Anderson 2004; Wood 2003; Goodwin 2001; Stanley 1996; Siegel and Hackel 1988:115; Mason and Krane 1989); Cuba and Peru in 1965; Venezuela, Colombia, Guatemala in the 1960s; Nicaragua in the 1970s (Wickham-Crowley 1991:43) and the 1980s (T. Brown 2001:26; Horton 1998:13, 179); Afghanistan in the 1980s (Cordesman and Wagner 1990:185; Barry O'Neill 1990:83); Guatemala in the 1980s (Stoll 1993:15, 119); Peru in the 1980s (Manrique 1998:197; Starn 1998:230; Vargas Llosa 1994:221; Shave 1994:115); Colombia in the 2000s (Semana 2003); Sudan in the 1980s (Keen 1998:22); Liberia (Duyvesteyn 2000:100-1); Algeria in the 1990s (Martinez 1994:104); Sierra Leone in the 1990s (Richards 1996:3-5); Sri Lanka (Senaratne 1997:67; Daniel 1996:170; Barry O'Neill 1990:81); Northern Ireland (Collins 1999:5, 153); Kashmir (Mahmood 2000:78; Mishra 2000); Punjab in the 1980s (Pettigrew 2000:206); the UN intervention in Somalia in the 1990s (Peterson 2000:111); Kosovo (Hayden 1999:37); Chechnya (Gordon 1999a); and the U.S. occupation of Afghanistan (Achakzai 2003) and Iraq (Mahdi and Carroll 2005; Maass 2005:41; Georgy 2003).
12 Indiscriminate violence also kills people who otherwise may be valuable sources of information. In Kitson's (1960:95) crude formulation: "Although most people felt that Mau Mau were better dead, we preferred them alive. You can't get much information out of a corpse."
*54
The Logic of Violence in Civil War
reaction (from "ill will" to "moral outrage," "alienation," and "visceral anger"), making people more willing to undertake risky actions.
That indiscriminate violence causes resentment and anger is well documented (e.g., Tishkov 2004:142; Wiesner 1988:366). A Guatemalan peasant told Warren (1998:109) how indiscriminate violence could transform fear into anger: "This was so heavy, so heavy. You were disturbed, you wanted to have some way of defending yourself. The feeling emerged - it wasn't fear but anger. Why do they come persecuting if one is free of faults, if one works honorably? You felt bad, well we all did. Grief but also anger." In turn, anger triggers the desire for action, as one of the earliest theorists of irregular war, J. F. A. Le Miere de Corvey, noted in 1823: civilians normally would not take up arms against regular troops; it was difficult to imagine, for instance, the merchants of Paris constituting themselves into a fighting force. But this situation might suddenly change if the house of a civilian was destroyed and his wife or children killed (Laqueur 1998:113). The critical mechanism is often the desire for revenge. "As the NPFL came in," the Liberian insurgent leader Charles Taylor told Bill Berkeley (2001:49), "we didn't even have to act. People came to us and said, 'Give me a gun. How can I kill the man who killed my mother?'" A man who was captured by a loyalist band in North Carolina noted in 1781 that the band "consisted of persons who complained of the greatest cruelties, either to their persons or property. Some had been unlawfully Drafted, Others had been whipped and ill-treated, without tryal; Others had their houses burned, and all their property plundered and Barbarous and cruel Murders had been committed in their Neighborhoods" (in Crow 1985:145).
Anger and the desire for revenge produce armed reaction only in the presence of an organization thatmakes such action possible (Wickham-Crowley 1990:235; R. Thompson 1966:35; Gardner 1962:44). The absence or weakness of organizations leads to passivity or sloppy actions doomed to failure; no matter how outraged, civilians will have no choice but to collaborate with the indiscriminate actor. For example, armed leftist groups in Argentina consciously planned a terror campaign in order to create chaos and unleash indiscriminate violence by the army so as to create massive dissatisfaction and launch a revolutionary process. They were right about the army's ability to terrorize but were also eliminated in the process, and the population had no credible alternative; the Guatemalan rebels, as well as many other insurgents, made a similar miscalculation.
6.4.2. Ambiguous Structure of Incentives
Indiscriminate violence by the incumbent side often fails to generate a clear structure of incentives for noncollaboration with the rebels and may even produce strong incentives for collaboration with them - thus generating defection instead of deterring it. Compliance is almost as unsafe as noncompliance, because the "innocent" can do little to nothing to escape punishment and the "guilty" are no more (and sometimes less) threatened. "The wanton nature of the retaliation -the picking of victims at random," Condit (1961:268) argues, "meant that pro-German Greeks or their relatives suffered as much as anti-German Greeks.
Indiscriminate Violence
155
Under these circumstances there was little advantage in being a collaborator [of the Germans].... As the numbers of the homeless and dead grew, the Greek population became simultaneously more terror stricken and more anti-German." In Kenya, it had become dangerous not to admit having taken the Mau Mau oath because "a denial of having taken the oath was often replied [to, by U.K. troops] by a bullet or a club on the head" (Barnett and Njama 1966:130).
Furthermore, indiscriminate violence lacks almost every feature generally considered to be necessary for the effectiveness of sanctions: it is usually late (e.g., Contini 1997), often arbitrary, inconsistent, erratic, and totally disproportionate.13 Unintelligible and unpredictable violence is likely to arouse unfavorable reaction (Leites and Wolf 1970:109). Inconsistency is shocking, confusing, and may signal weakness (Lichbach 1987:287); it makes one suspect a campaign aimed at mere annihilation, in the face of which chances of survival may seem enhanced through resistance.
These problems are, in large part, a consequence of the fact that usually control fails to follow indiscriminate violence. Indeed, the logic of indiscriminate violence requires that its potential targets be able to prevent its recurrence by denouncing hostile acts planned by the insurgents about which they are supposedly privy. Besides the assumption of information, as discussed in the preceding chapter, this can only work if civilians obtain credible protection from the incumbents; otherwise they will be exposed to insurgent counter-violence. Credible protection requires the establishment of incumbent control. Often, however, incumbents raid an area, kill civilians in reprisal actions, and then depart. Insurgents usually escape unhurt and are quick to return (Binford 1996:25; Geffray 1990:94; Wiesner 1988:128; Dallin et al. 1964:328); they either capitalize on the people's discontent or force collaboration by threatening their own, more credible violence (Sheehan 1989:115). In 1971 Bangladesh, "a Razakar [pro-Pakistani volunteer] from Galimpur in Nawabganj Police Station had gone as a guide with an army column to sweep a rebel hideout. When he returned, he found his three sons killed and a daughter kidnapped" (Salik 1978:105). In 1941 a German officer serving in the Ukraine reasoned: "Were the troops simply to shoot a number of
13 Kedward (1993:181) points out that in occupied France "there was no consistency in the German response to acts of armed Resistance which allows a meaningful correlation between different kinds of maquis action and the incidence of reprisals." In occupied Serbia, the Germans adopted a particularly harsh reprisal policy to quell the resistance: they set the ratio of reprisals to 100 Serbs for each German killed. However, many German commanders fulfilled their quota by drawing from the prisons mainly male Jews, as the "most convenient pool for drawing victims" (Browning 1992:134). Browning (1992:135) adds that in one instance the reprisals resulted in such grotesque absurdities as the predominantly Austrian troops of the 718th division shooting refugee Austrian Jews in Sabac in reprisal for Serbian partisan attacks on the German army. Of all the German officials in Serbia, only one, Turner, seemed to perceive the anomaly, but he consoled himself that "the Jews we had in the camps, after all, they too are Serb nationals, and besides they have to disappear." Todorov (1996) reports a similar case in German-occupied France, with the decision for reprisals taken by French collaborationists. Lomasky (1991:86) describes in a riddle form a more recent but quite parallel absurdity, the May 1972 Lod airport assault: "Q: Why do Japanese commandos fire Czech submachine guns at Puerto Rican passengers departing an Air France flight in an Israeli airport? A: To strike at American Imperialism."
i56
The Logic of Violence in Civil War
uninvolved residents by way of a reprisal and then simply withdraw, the residents' interest in finding the bandits would be reduced if not completely extinguished, and the danger of further support for the bandits increased" (quoted in T. Anderson 1999:610). In a report sent to his headquarters in April-May 1944, a German field commander in occupied Greece pointed out that the policy of reprisals had no noticeable effect because it did not entail the establishment of permanent control in the areas affected (Zervis 1998:221). This is why counterinsurgency experts (Thompson 1966:114-17) strongly recommend "clear-and-hold" instead of "search-and-clear" operations and warn that when there is no prospect of holding an area that may be cleared, no effort should be made to involve the inhabitants on the side of the government because "it is merely asking them to commit suicide."
6.4.3. Reverse Discrimination
Incumbent indiscriminate violence often produces a reverse discrimination against "non-rebels" and "anti-rebels," who, believing that their "innocence" will shield them, fail to protect themselves effectively. Consider the following example from occupied Italy in 1944: a man from Neviano Arduini, a province of Parma, was waiting for the Germans at his front door. "He was a Fascist, so he welcomed them, when he saw them. They ordered him to show his documents, he got in and came out with his identity card in one hand. He was hardly out, when he was shot in the head and killed. Just so, in front of his children. Then they ordered his wife to cook some eggs and ate them, right there, with the corpse lying on the ground" (Minardi 2002:6).
German reprisals during antiguerrilla campaigns in the Soviet Union frequently victimized pro-German starostas (elders) (Armstrong 1964:40). Counterinsurgency sweeps by the British in Kenya tended to grab moderate nationalists, as the much more careful and fearful radical militants fled to the forests (D. Anderson 2005:63). In his detailed investigation of the El Mozote massacre in El Salvador, Binford (1996:115) concludes that the people who were killed by the army "were the least decidido ('persuaded,' 'convinced,' but meaning, in this
context, 'politically committed')____Prior to the massacre, about 70 percent of
the prewar inhabitants of El Mozote left; several dozen of these had enlisted in the ranks of the [insurgent] ERP or supported the government. Those who did none of these things were murdered." A Greek man (Papakonstantinou 1999:313) recalls in his memoirs how he learned, one day, that the Germans would arrest a number of people in his hometown in northern Greece. Having seen the names on the blacklist, he set out to warn these people that their life was in danger and they better run. One of them, a disillusioned former Communist, refused: "I have severed my links to the party, I am not involved in anything right now, why should I flee?" He was arrested and executed, whereas the real communists ran away. Likewise, a Greek villager (Svolos 1990:22) recounts: "One evening the Germans raided our village and caught all the men they found at home. In fact, they found and caught precisely those men who were not associated with [the partisans] and had, thus, no reason to fear. They found and caught
Indiscriminate Violence
*57
them because those who had made up their minds [and were associated with the partisans] used to leave the village at night and sleep outside." An American commander who served in the Dominican Republic summed up this problem in his report (quoted in Calder 1984:154) on concentration camps for the internment of civilians: "As a military measure the concentration was productive of no good results. The good males came in and the bad ones remained out, but were not found."
The result of such actions ought to be obvious. As Stoll puts it in discussing Guatemala (1993:120), "The army was so indiscriminate that I heard of cases where even close family members of EGP [rebels] targets fled to the [EGP] guerrillas for protection, because they were far more selective in defining their enemy."
6.4.4. Selective Incentives for Rivals
Indiscriminate violence allows insurgents to solve collective action problems by turning the protection of the civilian population into a selective incentive. Protection emerges as a good only because of indiscriminate violence. As it escalates, so does the value of protection against it. Survival-maximizing civilians will be likely to collaborate with a political actor who credibly offers them protection, when its rival produces only indiscriminate violence. In El Salvador, Cabarrus (1983:195) argued, the power of the revolutionary organization was its ability to provide security for its members. When asked why he joined, a Salvadoran insurgent answered that he "had no choice____It was a matter of survival. Those were
the days when not to go meant getting killed" (J. L. Anderson 2004:222). A former Muslim rebel in the southern Philippines remarked that he "joined because of the violence created by the Ilaga [Christian fighters]; because there was no place safe during the trouble at that time" (in McKenna 1998:183). In occupied France, "when the acts of reprisals are added to the indiscriminate round-ups and the residue of Vichy collaborationism, the pressure on the population in a multitude of localities to look to the maquis as a place of refuge, or as a receptive and mobilizing organization, was high" (Kedward 1993:190).
Under such circumstances, participation in rebellion entails no collective action problem, but nonparticipation does.14 What is more, the actor providing protection can decide whether to turn it into a public good available to all or use it as a sanction against particular individuals or communities.15 The latter option makes indiscriminate violence extremely counterproductive: the decision by insurgents not to protect a village that is unfriendly to them amounts to
14 On this point, see Tone (1994:78), Stoll (1993:20), and Davis (1988:23).
15 In Japanese-occupied China, the Communists were able to teach peasants how to face Japanese raids following the "run for shelter under enemy attack" paofan method. By inducing collective discipline and eliminating free-riding, they were able to turn peasants into a disciplined group; in turn, the peasants won safety, which they could not have achieved on their own (Wou 1994:231). Similar tactics have been used in many places, including such methods as in-site hiding through the building of underground community tunnels (Vietnam), bunkers (Lithuania), or foxholes and caves (Latin America) (Wickham-Crowley 1991:43; Lansdale 1964:85).
158 The Logic of Violence in Civil War
exposing it to incumbent violence: in other words, using one's enemies as one's own enforcers.16
6.4.5. Overestimating the Strength of Ties between Political Actors and Civilians
Beyond inducing civilians to provide information about hostile activities, the logic of indiscriminate violence assumes civilians to be able to lobby armed actors to decrease the level of their activities. This requires that civilians have access and influence on armed actors and, conversely, armed actors care about civilians. This assumption is reasonable because armed actors depend on their civilian collaborators and wish not to alienate them.
Indeed, there are instances whereby insurgents have reduced or even suspended their activities because of the damage imposed by massive indiscriminate violence on the civilian population. The Norwegian resistance rejected aggressive tactics in 1943 as a result of German indiscriminate violence and justified its decision as follows: "We are convinced that [active assault on the enemy] will bring disasters to the people and the country which will be out of proportion to the military gains, and that it will disrupt and destroy the longer-term work of civil and military preparations which promise to be of the greatest importance to the nation" (quoted in Riste and Nokleby 1973:68-9).17 Likewise, there is evidence that insurgents sometimes suspend some of their activities locally because of the negative impact of indiscriminate violence - especially when they are weak (e.g., Fenoglio 1973:166-7). In occupied Greece, British agents reported that reprisals had a negative impact on the popularity of guerrillas, and they were right: an internal Communist document reported that "the people of the village were supporting us, but after its destruction [by the Germans] they began to turn against us."18 When pressed to extend the struggle in the cities by initiating a total war, Greek partisans objected on the grounds that the expected reprisals would turn the population against them (Mathiopoulos i98o:ix). Furthermore, I was able to uncover, during my research in Greece, a few cases where civilians were successful in lobbying the rebels to suspend their activity and spare their villages from reprisals (e.g., Frangoulis 1988:52).19
However, insurgents may also disregard civilian demands, most likely when they come from villages with weak ties to them. The villagers of Malandreni, in the Argolid region of Greece, were told in April 1944 that a German officer would visit them on a set date. Upon learning of this visit, the Communist-led partisans decided to set up an ambush. Fearing German reprisals, the villagers
16 An interesting twist: as a sanction for tax evasion, the Vietcong sent offenders for "reeducation" to hamlets that were shelled by the government army (Elliott 2003:873).
17 The same logic appears to have led the Cetnik guerrillas in occupied Yugoslavia to tone down their activity.
18 "Report by Lt. Col. R. P. McMullen on Present Conditions in the Peloponnese," PRO, HS 5/699; "Report about Markopoulo, 13 October 1944," ASKI, KKE 418 24/2/106.
19 Petersen (2001:196-7) recounts similar incidents in the Lithuanian village of Samogitia during the guerrilla war against the Soviet regime right after the end of World War II.
Indiscriminate Violence
159
demanded that the local Communist Party branch intervene with the partisans and have them cancel the ambush. The village party secretary describes the reaction of his regional boss: "Who do you think you are, comrade?" he was told; "A representative of the Germans?" To which he replied: "No, comrade, I just came to compare the benefit [of ambushing the Germans] with its cost, this is why I came." "The Germans burned many other villages," the boss replied, "but these villages joined the partisans" (Nassis n.d.:i i). Likewise, when asked to release the hostages he was holding in order to save the town of Saint-Amand from German reprisals in the summer of 1944, the maquisard commander Francois replied: "I couldn't care less about Saint-Amand, the men needed only to go off to the maquis, as we did ourselves" (Todorov 1996:72). Not surprisingly then, civilians often blame the insurgents for incumbent massacres. As one of the inhabitants of the Saint-Amand put it after the maquisards fled the town: "On June 7, the maquis ordered the rounds of drinks and, on June 8, it left us the job of paying the check" (Todorov 1996:42-3). He is echoed, sixty years later, by the sheik of the town of Labado in Darfur (Sudan): "We are angry at the SLA because they cause us this bad situation. All of our wealth and our homes are taken, but they run away and don't defend us" (in Polgreen 2005A3).20
Insurgents are usually aware of the risks they force on the civilian population from the outset and are generally unwilling to stop fighting because of them. Yet the absence of information leads incumbents (initially, at least) to overestimate the strength of ties between civilians and insurgents, as suggested by these examples from Civil War Missouri, Malaya, and Ethiopia:
Assuming all Missourians to be enemies, Kansas regiments believed it was their task to suppress them, to strip them of the means of resistance to Union authority as systematically as possible.... For them all Missourians were by nature traitors. (Fellman 1989:35-6)
Every Chinese was a bandit or a potential bandit and there was only one treatment for them, they were to be 'bashed around.' If they would not take a sock in the jaw, a kick in the gut might have the desired result. (Stubbs 1989:73)
We definitively know civilians will get hurt. But, knowing that the people sympathize with the rebels, the order is to bomb everything that moves. (De Waal 1991:123)
In his participant-observation study of a Catholic ghetto in Belfast, Sluka (1989:288-9, 300) described how the use of indiscriminate violence by the incumbents helped form pro-insurgent identities among its civilian targets:
Because of the stereotype that "all" people in Divis either belong to or strongly support the IRA and the INLA, the Security Forces treat them all as guerrilla sympathizers, and the Loyalist paramilitaries consider them all to be legitimate targets for political assassination. This has resulted in turning many who did not support the IRA or INLA before into supporters, sympathizers, and in some cases even members today. One of the best ways to turn politically moderate or apathetic Divis residents into IRA and INLA supporters or
20 Cases of civilians blaming the rebels for having provoked incumbent reprisal violence are provided about German-occupied Italy (Contini 1997; Pavone 1994:482-3) and Greece (Liapis 1994:202-5), Vietnam (Elliott 2003:1135; Wiesner 1988:64), Nicaragua (Horton 1998:168), Guatemala (Debray 1975:331), Peru (M. F. Brown and Fernandez 1991:168), civil war Russia (Figes 1996:1098), and German-occupied Soviet Union (T. Anderson 1999:609).
i6o
The Logic of Violence in Civil War
members is for policemen and British soldiers to unjustly harass, intimidate, and brutalize them, and for Loyalist extremists to assassinate members of the community.... Repression of the Catholic population by the Security Forces is enough to generate enough support for the guerrillas to ensure their survival.
6.5. WHY DOES INDISCRIMINATE VIOLENCE OCCUR?
Despite the absence of systematic empirical evidence, it is plausible to claim that the deterrent aim of indiscriminate violence often fails. Confronted with high levels of indiscriminate violence, many people prefer to join the rival actor rather than die a defenseless death. As in the Vendee, where desperate peasants were forced by Republican indiscriminate violence to join the counterrevolutionaries, they prefer to "sell their life at the highest price by defending themselves with vehemence" (Babeuf 1987:120). How are we, then, to account for the frequency with which indiscriminate violence is used?
Most "explanations" of indiscriminate violence focus on the individual level. The combination of weak discipline and strong emotions generates frustration and stress, eventually leading to indiscriminate violence. According to Grossman (1995:179), "The recent loss of friends and beloved leaders in combat can also enable violence on the battlefield.... in many circumstances soldiers react with anger (which is one of the well-known response stages to death and dying), and
then the loss of comrades can enable killing____Revenge killing during a burst of
rage has been a recurring theme throughout history, and it needs to be considered in the overall equation of factors that enable killing on the battlefield."
A Guatemalan peasant justified the violence of the army in similar terms (quoted in Warren 1998:100): "When they killed people, it was because they were filled with anger because their fellow soldiers had been cut down in battle." This is particularly true where insurgents avoid open combat, and it is practically impossible to distinguish civilians from rebels (Paludan 1981:94; Li 1975:232); soldiers, this argument goes, will tend to vent their anger by using violence indiscriminately against civilians, especially when they reach the conclusion, as one American loyalist did in 1780, that "every man is a soldier" (Weir 1985:74). Fear is another emotion associated with indiscriminate violence (Fellman 1989:128), as is the pursuit of pleasure (Katz 1988; Leites and Wolf 1970:92-4). Many fighters in Missouri saw the war as a form of hunting (Fellman 1989:176-84); the Rhodesian elite Selous Scouts units attracted "vainglorious extroverts and a few psychopathic killers" (Flower 1987:124). Racist attitudes cannot be discounted either. As Sheehan (1989:110) notes about the South Vetnamese army, most "Saigon officers did not feel any guilt over this butchery and sadism... [T]hey regarded the peasantry as some sort of subspecies. They were not taking human life and destroying human homes. They were exterminating treacherous animals and stamping out their dens."21 As has been discussed, civil wars offer plenty of
21 John Kerry's remark in his war notes is telling: "The popular view was that somehow 'gooks' just didn't have very much personality - they were ignorant 'slopeheads,' just peasants with no feelings and no hopes." The military command's messages praising his team's killings ended with the words "Good Hunting." Kerry's comment: "Good Hunting? Good Christ - you'd think we were going after deer or something" (in Brinkley 2003:57-8).
Indiscriminate Violence
161
opportunities for extortion and blackmail, while exposure to danger and death causes brutalization. These attitudes are compounded by the lack of resources: soldiers forced to live off the land will not shy from indiscriminate violence (De Waal 1991:43). However, although these are plausible individual determinants of indiscriminate violence, they remain unsatisfactory, being silent as they are about collective-level incentives or constraints; it is also unclear whether emotions and attitudes, such as fear, anger, or racism, are the causes, the correlates, or the results of using indiscriminate violence.
The persistence of indiscriminate violence has also prompted speculation that it is an irrational reflection of particular ideologies (Klinkhammer 1997:101) or the result of the "adrenaline of war zones" (Loizos 1988:650); any logic of deterrence is just a "fig leaf" for outright genocide or pure unmitigated acts of revenge on a defenseless population (Paggi 1996). Before resorting to ideological irrationality, however, it makes sense to examine and reject alternative explanations. I review four possible explanations for why indiscriminate violence is being observed: it may be an artifact of truncated data, or reflect ignorance, cost, or institutional constraints.
6.5.1. Artifact
The low visibility of selective violence may lead to a gross overestimation of indiscriminate violence. For one, selective violence is much more widespread than assumed. For instance, the killings by the Germans of persons "denounced as partisans by their fellow villagers" in the area of the Ukraine studied by Truman Anderson (1999:621) cumulatively rivaled two major massacres in that area. More people were killed by Colombian rightist paramilitaries around the town of Dabeiba in an individualized way than were killed in visible massacres (S. Wilson 2002). In my own study in the Argolid region of Greece, I found that civilian fatalities were equally likely to be produced by selective and indiscriminate violence (49.86 percent killed selectively and 50.14 percent indiscriminately), hardly the impression conveyed by the best historical treatment of the events in the region (Meyer 2002), which is replete with instances of indiscriminate violence.
Moreover, many instances of violence maybe miscoded as indiscriminate. The 1997 massacres in Algeria targeted specific families and neighborhoods (Kalyvas 1999). The Mau Mau attack against the Kenyan village of Lari in 1953 caused the death of seventy-four men, while another fifty were wounded; this massacre was widely described as indiscriminate. However, David Anderson (2005:127-8) found that it was "far from random in its violence," targeting the families of local chiefs, ex-chiefs, headmen, councilors, and leaders of the local militia; he adds that "the victims had been selected with care, their homesteads identified and
singled out____Neighbours were left unmolested as the gangs went about their
business, each attacking group moving systematically between the two or three homesteads for which it had been assigned responsibility." The same is true of several massacres in rural Colombia during the 1950s (Henderson 1985:150). A leader of the men who attacked the Colombian hamlet of El Topacio in May 1952 "knew the place and its people" and strolled from house to house playing
l62
The Logic of Violence in Civil War
a musical instrument, the tiple. "On that day, the musician was both judge and jury, for, wherever he paused, the bandits dragged out and shot every man and boy. Ninety-one died in that incident alone." Also in Colombia, the massacre of 140 men and boys from the village of San Pablo in early 1953 seems indiscriminate, until one learns that the victims were all Liberals whose credentials had been "carefully checked to verify affiliation" (Henderson 1985:152).
When the Vietcong attacked a district of Binh Son in 1967, they burned one section of six houses but not the adjacent houses (West 1985:273).22 Likewise, the homes of about thirty people in the Afghan village of Shakar Daria were burned by the Taliban, but the rest of the village was left untouched (Waldman 2oo2b:A9). The violence unleashed by the Guatemalan regime in the early 1980s discriminated on the basis of location; one of its notable features was "that neighboring villages fared quite differently: one might be destroyed while another was left untouched, depending on the army's perceived understanding of guerrilla support" (Green 1995:114). Guatemalan villages that were located in areas of high guerrilla activity but "did not have a reputation of being held by guerrillas" were not attacked by the army (Davis 1988:25). When the Serb forces attacked the village of Bukos in Kosovo and "caused the Albanian villagers to flee," they did not touch a similar neighboring village, Novo Selo, probably "because there were no Kosovo Liberation Army guerrillas in the village, residents said" (Gall 1999A6). The seemingly indiscriminate violence of the Russian Army in Chechnya was not blind: some villages stood "untouched, a reward, Russian officials say, to those who refused to aid the rebels and cooperated with the Russian army" (Gordon i999b:Ai).
Finally, armed actors often refrain from resorting to indiscriminate violence even when they have the ability to exercise it (e.g., McGrath 1997:112), something that usually goes unreported. For example, the Germans often refrained from collective reprisals (Lotnik 1999:61; Pavone 1994:481; Fleischer 1979). As Rana Mitter (2000:180) points out about Manchuria, "The impression given, in other words, is that the Japanese exercised random violence in Manchuria, whereas the evidence suggests that violence was part of a whole repertoire of techniques of coercion, and that co-optation remained their preferred option when available."
In sum, it is likely that the significance of indiscriminate violence is overstated because of a specious reading of available data. This type of violence may be less prevalent than generally thought. Even if this is true, however, one still needs to explain its occurrence.
6.5.2. Ignorance
Robert Thompson (1966:84) reports a joke: "There are only two types of generals in counter-insurgency - those who haven't yet learnt and those who never will!" Most accounts of indiscriminate violence explain it by reference to ignorance and
22 West (1985:273) also noted the surprising absence of reaction to this pattern: "No one asked why the VC had singled them out."
Indiscriminate Violence
organizational incompetence. The Vietnam War provides a prime example. For years, the U.S. military leadership failed to grasp the nature of the war (West 1985:2 56). As a general recalled, "Soon after I arrived in Vietnam it became obvious to me that I had neither a real understanding of the nature of the war nor any clear idea as to how to win it" (quoted in Thayer 1985:3). "Let's go out and kill some Viet Cong, then we can worry about intelligence," quipped a newly arrived general (R. Thompson 1966:84). The absence of front lines proved to be a major cognitive obstacle for officers trained in conventional war. As a result, much of the data generated by the conflict was not properly processed (Thayer 1985:4). Hence, "a theoretical basis for the violence program, consistent both internally and with objective conditions, was never articulated, despite the number of lives it consumed daily. The basis for using violence was a residue of military doctrines developed to deal with friendly military units operating on hostile foreign territory" (Race 1973:227). Several metaphors describe the difficulty that conventional armies have fighting irregular wars, from T. E. Lawrence's quip that irregular war is "messy and slow like eating soup with a knife" to Lieutenant Colonel Bigeard's aphorism ("We don't kill flies with hammers") and to the more recent remark by Lieutenant Colonel Todd McCaffrey in Iraq that fighting a war driven by intelligence with a conventional army is akin to "teaching an elephant to ballet dance" (in Negus 2004:5).
Proximate determinants of this ignorance include undue optimism and lack of preparation, along with the perception that the threat posed by a rebellion is low;23 fundamental misunderstandings about the nature of irregular war;24 inadequate organization and training25 or just sheer professional incompetence and corruption;26 the military's oft-noted weak institutional memory and lag in learning and updating of war doctrine - a tendency epitomized in the saying that the military fights not the present war but the last one;27 the prevalence of authoritarian structures within the military;28 its politicization and/or corruption;29 and, finally, straight racism.30 A problem with such explanations is that they seem unable to account for the bewildering variation in levels of indiscriminate violence. For example, in the occupied Soviet Union, the Germans varied the kind and intensity of violence they used considerably.
23 Fall (2000:115); Cann (1997:63); Paget (1967:33).
24 Harmon (1992:44); Sarkesian (1989:44-5).
25 Paget (1967:31).
26 Ellsberg (2003:115); Downie (1998:133); Stubbs (1989:70); Siegel and Hackel (1988:116-17); Leites and Wolf (1970:92-4); Paget (1967:78); Kitson (1960:192).
27 Downie (1998); Garvin (1991:9); Blaufarb and Tanham (1989:23); Trinquier (1964:61). It appears that the lessons learned in Korea went unheeded in Vietnam, and so did the Vietnam lessons in Central America (Downie 1998:158, 251; Katz 1975:589).
28 Mason and Krane (1989).
29 Blaufarb and Tanham (1989:19).
30 Heer and Naumann (2000); Li (1975:231); Welch (1974:237). An American soldier in Iraq explained an instance of American brutal behavior as follows: "I kind of looked at it as high schoolers picking on freshmen. Us being the seniors; the Iraqis being the freshmen" (in Filkins 2005:92).
164
The Logic of Violence in Civil War
Ultimately, ignorance must be qualified as a cause of indiscriminate violence because political actors often seem aware of its deleterious effects from the outset. During the Spanish Civil War, Catalan Republicans warned that indiscriminate violence against opponents in the Republican zone was bringing about a "counterrevolutionary climate in the rearguard" (de la Cueva 1998:360) - yet they did not refrain from using it. After a particularly bloody wave of reprisals in Greece, the German minister plenipotentiary for Southeast Europe, Hermann Neubacher, complained to the military commander of the relevant area: "It is utter insanity to murder babies... because heavily armed Red bandits billeted themselves, overnight, by force, in their houses, and because they killed two German soldiers near the village. The political effect of this senseless blood bath doubtlessly by far exceeds the effect of all propaganda efforts in our fight against Communism" (quoted in Condit 1961:26s).31 Yet the Germans kept resorting to mass reprisals. The sprawling American counterinsurgency literature of the 1950s and 1960s is replete with warnings about the negative effects of indiscriminate violence -including tens of studies by official or semiofficial outfits such as the Operations Research Office, the Special Operations Research Office, the Counterinsurgency Information Analysis Center, and the Center for Research in Social Systems.32 The widely distributed "Social Science Research Studies" conducted on various aspects of the Vietnam War argued "that more was being lost in terms of loyalty and respect for the GVN and the Americans than was gained in hurting the VC by bombing and shelling of villages, even where they were VC strongholds and fighting bases" (Wiesner 1988:122-3). These arguments were disseminated and popularized in journals like Foreign Affairs (e.g., Lansdale 1964). Indeed the military was well aware in the late 1960s that "the injury or killing of hapless civilians inevitably contributes to the communist cause" (in Bilton and Sim 1992:40). Yet, the U.S. forces indiscriminately shelled and bombed countless South Vet-namese villages for many years. More recently, the Russian army appears to have been aware of the effects of indiscriminate violence in Chechnya, and yet it has largely ignored this information.33 Hence the question must be restated: why use
31 Klinkhammer (1997:84) and T. Anderson (1995:342) document similar doubts among German officials in Italy and Ukraine.
32 "Guerrillas may initiate acts of violence in communities that are earnestly cooperating in order to provoke unjust retaliation against these communities. Unjust or misplaced punishment at the hands of the occupying force is vigorously exploited by the guerrillas to gain sympathizers and strengthen their own cause" ("Operations against Guerrilla Forces" quoted in Barton 1953:3). Major studies such as Project Camelot and Project Agile reached similar conclusions (M. F. Brown and Fernandez 1991:111, 204). See also Ferguson(i975);JonesandMolnar(i966);Molnar(i965); Gardner (1962); Condit (1961); Barton (1953).
33 A Russian general pointed out in August 1999 that "there should be no losses among the civilians. To destroy a bandit if he covers himself by a civilian, we must first separate the civilians from the bandit, and then take care of the bandits" (Bohlen 1999^1). In fact, Russian analysts argued that Chechen rebels might actually seek to provoke indiscriminate violence from the Russian military (Gordon 1999a). Yet, the Russians resorted to indiscriminate violence, which initially united the divided Chechens. Shamil Basayev, the best-known Chechen warlord, sarcastically said that he is "very grateful" to Russia for creating a new sense of unity among his people (Economist, 9-15 October 1999). Eventually, the war evolved into a stalemate, where, "paradoxically," the Russian
Indiscriminate Violence
165
indiscriminate violence in the presence of knowledge about its counterproductive effects? I point to two factors: cost and institutional distortions.
6.5.3. Cost
An overriding consideration in the use of indiscriminate violence is the cost of selective violence.34 Identifying, locating, and "neutralizing" enemies and their civilian collaborators one by one requires a complex and costly infrastructure. Most incumbents quickly realize that they lack the necessary resources. In a directive sent to the units occupying the Soviet Union, the German Central Command pointed out that "the Commanders must find the means of keeping order within the regions where security is their responsibility, not by demanding more forces, but by applying suitable draconian measures" (quoted in Cooper 1979:143). In short, indiscriminate violence initially appears as a handy substitute for individualized deterrence. Still, the low cost can explain the emergence of indiscriminate violence but not its persistence.
6.5.4. Institutional Distortions
Some cases of indiscriminate violence can be explained as resulting from internal institutional distortions. The Vietnam War provides an excellent illustration. Sheehan (1989) describes how the South Vetnamese military and the U.S. high command in Vetnam administered indiscriminate air and artillery bombardment on peasant hamlets at an estimated cost of about 2 5,000 civilians killed and 50,000 civilians wounded a year. An American provincial adviser talking about the area under his supervision remarked that "we shot a half-million dollar's worth of howitzer ammunition last month on unobserved targets. Yet the whole provincial budget for information- and intelligence-gathering is $300" (Fall 2000:110). This violence was premised on the theory that it would "terrorize the peasants out of supporting the Vet Cong" (Sheehan 1989:109). Of course, this alienated the population by killing and wounding large numbers of noncombatants and destroying farms and livestock (Taber 1965:95). Sheehan recounts how the U.S. military adviser John Paul Vann denounced the indiscriminate bombing and shelling of the countryside as both cruel and self-defeating. Initially, Vann had found it difficult to believe the utter lack of discrimination with which fighter-bombers and artillery were turned loose; apparently, a single shot from a sniper was enough to call for an air strike or an artillery barrage on the hamlet from which the sniper had fired. A province or district chief could start firing artillery shells in any direction at any hour of the day or night, not even needing an
tactics "hardened the resistance," and Chechen fighters were "no longer able to confront Russian troops head-on, but they remain determined to inflict as much pain as possible in the name of Chechen independence" (Myers 2002 :A6). 34 Militaries often quantify this cost. For example, the estimated cost of killing a single rebel in Kenya was £10,000, in Malaya it exceeded $200,000, while in Vietnam it reached $373,000 (Laqueur 1998:379; Paget 1967:101).
i66
The Logic of Violence in Civil War
unverified report stating that some guerrillas had gathered in a neighboring hamlet. Vann wondered how any American could think that Vietnamese peasants who lost family members and friends and homes would not be mad; in fact, most Vietnamese farmers had an alternative army and government asking for their allegiance and offering them revenge. Vann alerted his superiors to this fact by arguing that the bombing and shelling killed many more civilians than it ever did Vietcong and as a result made new Vietcong. However, he was usually overruled and the hamlets were bombed. As an American Air Force general put it: "The solution in Vietnam is more bombs, more shells, more napalm... till the other side cracks and gives up" (Sheehan 1989:619).
Why was such a policy allowed to go on? Sheehan argues that the underlying cause was the failure to curb "institutional proclivities." On the one hand, there was competition between different branches within the U.S. military, and the Air Force was quite successful in promoting bombings: it was in the personal interest of the Air Force chief and of his institution to believe that the bombing furthered the war effort, and so he believed it (Sheehan 1989:650). Moreover, processes of learning were undermined by the brief one-year or six-month rotation period for military personnel: as soon as a military adviser began to understand the situation, he had to leave (Meyerson 1970:37).35 Thus, the American military system does not seem to have encouraged learning. Likewise, South Vietnamese officers saw artillery shelling as an easy way to show that they were aggressive without running the risks of actual "search and destroy" operations. Commanders at all levels who only engaged in shelling could still retain their command and even be promoted, while those who took risks might be relieved if they suffered a setback or sustained heavy losses.36
Institutional distortions can be observed in other cases as well. The French revolutionary Bertrand Barere explained the initial failure to put down the rebellion in the Vendee by the "desire for a long war among a large part of the chiefs and administrators" (quoted in Tilly 1964:338). A Pakistani officer (Salik 1978:117) described the situation among his army in Bangladesh: "All the divisional commanders and the brigade commanders, except one major-general and one brigadier, invariably assured General Niazi that, despite their meager resources and heavy odds, they would be able to fulfill the task assigned to them.
35 According to a lieutenant colonel, "The day I got there, that man [his predecessor] was leaving. He had his hat and coat on, threw me the key and said, 'There's the shack. Good Luck. Every day is different around here.'That's all the training I had" (quoted inKatz 1975:591). Snow (1997:106-9) provides an extensive summary of the U.S. mediocre performance in counterinsurgency, central to which is the disdain of a military, high-tech organization for political, low-tech warfare involving civilians. Ellsberg's (2003:185-6) analysis confirms these points.
36 The South Vietnamese regime encouraged this misallocation of military resources because it was unwilling to commit its military to a full-fledged war; it was primarily concerned instead with preserving its elite troops to protect itself from a coup - as opposed to wasting it in fighting the war. In turn, this calculation could only be sustained because of the perverse effect of the U.S. involvement in Vietnam: the South Vietnamese government assumed that the United States, as the preeminent power in the world, could not afford to let its anti-Communist government fall to Vietcong.
Indiscriminate Violence
167
Protect
Outcome 1: Stalemate or Indeterminate
Fail to Protect
Outcome 2: Civilians Support Incumbents
Protect
Outcome 3: Civilians Support Insurgents
Fail to Protect
Outcome 4: Civilians Support Incumbents
figure 6.1. Civilian Behavior as a Function of Indiscriminate Violence and Protection
'Sir, don't worry about my sector, we will knock the hell out of the enemy when the time comes,' was the refrain at all these briefings. Any comment different from this was taken to imply lack of confidence and professional competence. Nobody wanted to jeopardize his prospects for future promotion."
Again, however, institutional distortions may explain the emergence of indiscriminate violence but not its continuation for a long time in light of inevitable evidence that it is counterproductive.
6.6. ACCOUNTING FOR THE PUZZLE
As I pointed out, the conjecture that indiscriminate violence is counterproductive is not based on systematic empirical research. Because of the inadequacy of data, it makes sense to turn to theory. Assume a setting where incumbents choose whether to use indiscriminate or selective violence, insurgents have the option of protecting civilians from incumbent indiscriminate violence, and civilians collaborate with the political actor who best guarantees their security. In such a setting, civilians will be likely to collaborate with the incumbents if the insurgents fail to protect them, whether incumbents are indiscriminate or selective; they will be likely to side with the insurgents when they are protected by them against indiscriminate incumbents; and the outcome is indeterminate when insurgents protect civilians and incumbents are selective (Figure 6.1).
This analysis yields the following prediction: incumbents can afford to be indifferent about the type of violence they use when insurgents are unable to offer any protection to civilians. Put otherwise, costly discrimination can be dispensed with when insurgents are weak. When this is the case, indiscriminate violence does succeed in paralyzing an unprotected population. When American indiscriminate violence made the Filipino civilians "thoroughly sick of the war," they "were forced to commit themselves to one side"; soon garrison commanders "received civilian delegations who disclosed the location of guerrilla hideouts or denounced members of the infrastructure" (Linn 1989:56-8).
Incumbents
i68
The Logic of Violence in Civil War
Likewise, most Missourians turned to the Union in their despair, Fellman (1989:78) notes, "not out of a change of faith but as the only possible source of protection." Guatemala provides the paradigmatic case in this respect (Stoll 1999; Le Bot 1994). After the Guatemalan army used massive indiscriminate violence against the population, civilians who had initially collaborated with the rebels were left with no choice but to defect, because the rebels utterly failed to protect the population from massacres (Watanabe 1992:181). As Stoll (1993:6) points out, "while the guerrillas could not be defeated militarily, they were unable to protect their supporters."37
Contrasting occupied Greece and Yugoslavia allows for a controlled comparison over both space and time. In October 1941 German troops burned two northern Greek villages, Ano Kerdilia and Kato Kerdilia, and shot all 207 male inhabitants in reprisal for the killing of German soldiers by a fledgling partisan group. The effect was stunning. Immediately after the reprisal, Greek men from surrounding villages were allowed to form militias and set up watch posts around their villages to prevent the partisans from entering and obtaining supplies. In some cases, they even caught a few partisans and handed them over to the Germans, who reported the population's "feelings of hatred toward the rebels" (quoted in Dordanas 1996:91-6). In one case, a partisan was denounced to the Germans (who caught and hanged him) by his cousin's husband (I-1; for list of interviews, see Table A.i in Appendix A). As a result, "the deterrent effect in northern Greece was swift and resistance faded away in the winter months" (Mazower 1993:87-8). The behavior of the villagers is best explained by their desire to avoid further dreadful German violence. Indeed, a woman I interviewed from a neighboring village (whose uncle was the leader of the partisan group and whose father was among those shot by the Germans) told me that "if the Kerdilia villagers had found my uncle, they would have skinned him alive. 'He was responsible for the massacre,' they said; 'We will find him, we will skin him alive, we will kill him!'" (I-i). This is an instance of effective use of indiscriminate violence in the face of an insurgent group that was extremely weak. A comparison to neighboring Yugoslavia in 1941 and Greece in 1943-4 °ffers a useful counterpoint: reprisals during this period clearly failed to produce similar effects. The reason is that in both Yugoslavia in 1941 and Greece in 1943-4, an important insurgent infrastructure faced German indiscriminate terror with a strategy that mixed its own selective violence with civilian protection.
In short, indiscriminate violence is likely to be effective when there is a steep imbalance of power between the two actors. Given reasonably strong insurgents, it should be unsustainable, as its counterproductive nature becomes clear. We would, therefore, expect rational incumbents who may initiate indiscriminate violence to muster additional resources and subject whatever institutional distortions they suffer from to the imperatives of their long-term interest. As a conflict waxes on, we should observe a shift toward selective violence, especially among
37 Similar processes are reported about Calabria in 1806-7 (Finley 1994:99), Peru in the 1960s (M. F. Brown and Fernandez 1991:140), and Angola in 1961 (Clayton 1999:35-9). The case of Darfur in 2004-5 may nt here, save for the huge international outcry.
Indiscriminate Violence
169
incumbents, the ones likely to initiate indiscriminate violence. This insight can be formulated as follows:
Hypothesis 1 Political actors are likely to gradually move from indiscriminate to selective violence.
Anecdotal evidence suggests the plausibility of this hypothesis (for an empirical test see Chapter 9). The transition to more selective forms of violence is particularly striking in the wars fought between German occupiers and European resistance movements during the Second World War. If there is one political actor whose extreme ideological outlook should have clouded its sense of instrumental rationality, clearly it would be the Nazis. They were also fighting a total war and their prospects of victory by 1944 were dim, to say the least. The German military was overstretched and many European countries were occupied by very thin forces. Hence, a policy of indiscriminate violence was overdetermined by both ideological and strategic factors. Yet a closer examination of their practices shows a remarkable and unexpected, if partial, evolution from indiscriminate violence to a mix of selective and indiscriminate violence, wherein the former came to play an increasingly important role.
In Greece, for instance, after a particularly bloody wave of indiscriminate violence in December 1943 that left more than 1,300 Greek villagers dead, German commanders were ordered "to seize the perpetrator himself and take reprisal measures only as a second course, if through reprisal measures the prevention of future attacks is to be expected"; in addition, the authority to order reprisals was removed from lower ranks and moved up to division commanders, who also had to get clearance from the competent administrative territorial commander (Con-dit 1961:265-6). Although these measures were never fully implemented (and final responsibility for reprisals usually rested with commanders on the spot), this indicated a willingness to change course in the face of the obvious effects of indiscriminate violence. The formation of a Greek auxiliary corps, the Security Battalions, and their explosive growth in the spring and summer of 1944 led to higher levels of discrimination in violence through those troops' access to local information.38
Asimilar process tookplace elsewhere in occupied Europe (Laqueur 1998:209; Heilbrunn 1967:147, 151; Dallin et al. 1964:327-33), as also occurred among the Japanese in Asia (Hartford 1989; Li 1975:204-9). In Vietnam duringthe late 1960s and early 1970s, the United States switched from indiscriminate violence to one of the most sophisticated programs of selective violence. In the Phoenix program, the goal was to kill, jail, or intimidate into defection the members of the Vietcong apparatus in the South "person by person" (Adams 1994:178). By 1971 the war was transformed into "one in which whom we killed was far more important than how many we killed" (Herrington 1997:69); a CIA operative claimed that "we had 75 percent of the key [Vietcong] cadres named" (Moyar 1997:146). A similar trend has been documented in El Salvador (Binford 1996:140) and Mozambique
38 Greek memoirs note that, unlike the Germans, the Greek auxiliaries targeted the homes of families whose men were guerrillas or sympathizers (Papandreou 1992:110; Svolos 1990:25).
The Logic of Violence in Civil War
(Geffray 1990), where many people had been reluctant to leave RENAMO-held places for government-controlled areas because they feared being indiscriminately targeted by government soldiers. More recently, the Russian army became more selective in Chechnya, switching away from zachistki or mopping-up raids ("its previously preferred method of hunting down rebels") to targeted disappearances and kidnappings of rebel suspects and the use of a Chechen militia (Gordon 1999c; Economist 2003:46). There is considerable evidence documenting similar shifts to higher levels of discrimination in violence in a variety of civil wars.39
If this argument is correct, we can explain the relative dearth of indiscriminate violence among insurgents by pointing to their better access to local information. Accordingly, we should expect to see insurgents relying on indiscriminate violence when and where they lack the ability to assess local information. Indeed, there is evidence that they use such violence against groups and places that are opaque to them, such as incumbent-controlled cities; yet even in such cases, they also tend to eventually switch to selective violence.40 Finally, the argument yields three implications about the incidence of suicide missions. First, insofar as it is a method of indiscriminate violence used to deter civilians (which is not always the case),
39 Such evidence is reported from various conflicts in Macedonia (Livanios 1999:205), the Philippines (McKenna 1998:158-9; Jones 1989:273; Linn 1989:77-8; Kerkvliet 1977:208, 240), China (Wou 1994:127-58), Malaya (Stubbs 1989:252; Pye 1964:177), Zimbabwe (Flower 1987:106-7), Guatemala (Stoll 1993:111, 139-40; Paul and Demarest 1988; Carol Smith 1988; Peralta and Beverly 1980), Peru (Starn 1998:230-8; Rosenberg 1991:207), and Iraq (Maass 2005). Note that the shift toward selective violence was not necessarily one toward correct targeting (Kalyvas and Kocher 2005). I discuss this issue in Chapter 7.
40 In the course of the Chinese Civil War, the Communist rebels discovered that assassinations of gentry members based solely on their identity forced otherwise rival gentry into a temporary coalition against them; this led to reprisals that were highly effective because gentry members could easily obtain reliable information about whom to target, thus "greatly eroding peasant morale and eventually putting a halt to the Communist peasant movement"; the same applied to grain seizures, which, "although highly appealing to peasants, invariably produced unintended adverse results. They often involved much killing and pillaging. Grain seizure might appeal to poor peasants in one locality, but random violence and killing destroyed villages in other localities and drove
settled peasants to the gentry side____Random violence in fact promoted community cohesion
by rallying peasants to the gentry. It also polarized local communities and made it impossible for the communists to expand their movement." As a result, the Communist Party explicitly forbade indiscriminate killing and criticized peasant cadres' perceptions that in conflicts with rival local militias it was normal to kill hundreds of peasants (Wou 1994:123, 142). In fact, the Communists recognized that the "red terror" resulting from "harsh indiscriminate action" was counterproductive and redefined their policy of violence; they were more selective during the Yenan period (1935-41) compared with the earlier Kiangsi Soviet (1924-33): "Rather than sticking stubbornly to past methods, the Communists appeared to learn and experiment" (Griffm 1976:93-4, 146). Likewise, in Malaya the Communist leaders decided that "blind and heated foolhardiness" was to be avoided in the future, while the emphasis was to be on "regulated and moderate methods" (Laqueur 1998:290; R. Thompson 1966:25). In Vietnam, the Communist Party exercised "much tighter control over the procedures for approving executions after 1954, because of the unfavorable consequences of the many careless executions that occurred during the Resistance" (Race 1973:189) and later abandoned the random bombing of urban centers (Fall 2000:111). The Algerian independentist rebels began to more rigorously check denunciations several years after the beginning of the insurgency, in 1957-8 (Hamoumou 1993:203-4).
Indiscriminate Violence
we may account for its relative scarcity by reference to its counterproductive effects (Kalyvas and Sanchez Cuenca 2005). Second, suicide missions should be observed in places and times where selective violence is extremely difficult or impossible, including areas where control is limited or nonexistent. This is consistent with evidence from Israel/Palestine. Finally, as long as an insurgency is in the ascendant, expanding its territorial control, we should expect it to replace suicide missions with more selective methods.
An important implication of this argument is the following: a major reason why wars of occupation turn into civil wars is that indiscriminate violence is counterproductive. The need for selective violence forces occupiers to rely on local agents, thus driving a wedge within the native population. In contrast, the persistent use of indiscriminate violence points to political actors who are fundamentally weak: this is the case with civil wars in failed states ("symmetric noncon-ventional wars"), where high levels of indiscriminate violence emerge because no actor has the capacity to set up the sort of administrative infrastructure required by selective violence. In this perspective, the subset of ethnic civil conflicts that display high levels of eliminationist violence could be endogenous to state failure.
The relatively recent spread of international norms against human rights violations has made indiscriminate violence even less desirable for those who use it. It is argued (J. L. Anderson 2004; Greenhill 2003) that weak challengers now have an overwhelming incentive in provoking incumbents into using indiscriminate violence. As recent conflicts in Kosovo and Darfur demonstrate, indiscriminate violence is now likely to attract considerable international negative publicity and even cause external intervention. If this trend continues, we are likely to observe either a decline in the use of indiscriminate violence in irregular wars as incumbents become acutely aware of its costs, or new sophisticated ways for hiding such violence from international scrutiny and ensuring "plausible deniability" (Ron 2000b). Conversely, as incumbent indiscriminate violence ebbs, indiscriminate violence by rebel groups may become more visible.
6.7. CONCLUSION
This chapter has sought to examine the workings of indiscriminate violence when used to generate civilian compliance. A key goal of indiscriminate violence in this context is to shape civilian behavior indirectly through association, and to shift responsibility for hostile actions to a wider group of people. It is likely to emerge when the information necessary for selective violence cannot be obtained with the allocated resources. However, indiscriminate violence seems to be counterproductive, with the exception of situations where there is a high imbalance of power. When violence is indiscriminate, compliance is almost as unsafe as noncompliance, because the "innocent" can do little or nothing to escape punishment and the "guilty" are no more (and sometimes less) threatened. If the rival political actor can provide credible protection against the violence, people will transfer their support. While at first such dynamics may not be clear, or institutional distortions may affect how political actors choose to pursue action on the ground, if
I72
The Logic of Violence in Civil War
a conflict waxes on, we ought to observe a transition toward selective violence as long-term interests begin to win out. Thus, instances of indiscriminate violence may be the product of a lag: political actors appear to engage in it because initially it seems much cheaper than its alternatives; however, they should eventually discern that it is counterproductive and switch to selective violence - the subject of the next chapter.
7
A Theory of Selective Violence
You've got to be on the ground to get the truth.
Lieutenant Colonel Greg Reilly, U.S. Army, Iraq
People talk and people die.
Eamon Collins, former IRA cadre
What kills directly is the tongue.
A Nicaraguan officer
This chapter develops a theory of selective violence as a joint process. Political actors operating in a regime of fragmented sovereignty must rely on selective violence to deter defection (i.e., active collaboration with the rival actor), despite lacking the resources for monitoring the population. Selective violence entails the personalization of violence and requires information that is asymmetrically distributed between political actors and individual civilians. Within the institutional context defined by irregular war, violence results from the convergence of two distinct but related processes: the political actors' attempts to deter individual defection, and individuals' decision to provide information to political actors. I supply a political economy of the joint production of violence, formulate a model that captures the key aspects of the theory, and specify a set of empirical predictions.
The argument is as follows. Selective violence presupposes the ability to collect fine-grained information. The most effective way to collect it is to solicit it from individuals, which explains the ubiquity of the practice of denunciation in civil war. Denunciation is central to all civil wars, with the probable exception of a subset of civil wars where no actor attempts to obtain the collaboration of members of groups that allegedly support its rival and where all relevant information is in the public domain, conveyed by visible individual identities. There are two distinct types of denunciation: political and malicious, both of which are accommodated by the theory of selective violence presented here. False denunciations are quite common, as individuals are tempted to settle private and local conflicts. However,
*73
i74
The Logic of Violence in Civil War
false information undermines the very premise of selective violence. Political actors cannot screen all the information they get, but they can mitigate this problem if they generate a credible perception that they are selective in their violence. This perception is conveyed by the presence of local agents, which signals the existence of a network of informants, the related ability of local agents to avoid blatant "mistakes" in targeting, and the clandestine and secret nature of the process of informing for the rival actor. A mix of accurate and erroneous hits is, thus, compatible with a perception of credible selection under these three conditions.
I then turn from political actors to individuals. Although motivations to denounce vary, the constraints faced by denouncers provide a good way to model the process. The key constraint is the likelihood of retaliation against the denouncer via the process of counterdenunciation to the rival actor by the family of the victim. Thus denunciation is a function of the control a political actor has over an area: control affects the likelihood of retaliation against the denouncer because counterdenouncers need access to the rival political actor. The theory predicts that denunciation leading to selective violence will be most likely where one actor exercises dominant but incomplete control. Where actors have total control, they can detect defection directly; this ability is public knowledge, which depresses the levels of defection. Where a political actor's control equals its rival's, no information will be forthcoming. Therefore, selective violence is unlikely where control levels enjoyed by one actor are high and, surprisingly, where the two actors share sovereignty. Put otherwise, the front line in irregular war is likely to be nonviolent. The theory also predicts the location of indiscriminate violence.
7.I. INFORMATION
Information is a key resource in irregular war (Eckstein 1965:158; Pye 1964:177); it is the link connecting one side's strength with the other side's weakness (Crawford 1958:179). It is widely accepted that no insurgency can be defeated unless the incumbents give top priority to and are successful in building an intelligence organization (R. Thompson 1966:84). Intelligence refers not only to "high-level military intelligence on maps, but [to] basic police intelligence at the [opponents'] own grass roots" (Clutterbuck 1966:4). The collection of such intelligence requires an enormous infrastructure: "We have to be everywhere informed," asserted a French officer in Algeria (Trinquier 1964:35); "therefore, we must have a vast intelligence network."
Monitoring is a fundamental problem of rule. As Tocqueville (1988:206) remarked, "the Sovereign can punish immediately any fault he discovers, but he cannot flatter himself into supposing that he sees all the faults he should punish." Indeed, information is as hard to come by as it is essential. As a British officer in Malaya eloquently observed, "We could not bring our military machine to bear without information, and we could not get information without the support of the population, and we could not get the support of the population unless they were free from terrorism, and we could not free them from terrorism until we had sent men to kill the terrorists. So it went round and round - a most
Selective Violence
*75
complicated combination of vicious circles. The key to breaking these vicious circles remained one thing: information" (Crawford 1958:180-1).
Where does information come from? There are many sources, as suggested by the same officer: "Information came from captured terrorists, who bought their lives with it; from spies; from informers; from every kind of civilian contact and grapevine; from photographs of the jungle; from single footprints in the jungle; from captured documents, weapons, camps, clothes, supplies; from the reports of the jungle-patrols quartering backwards and forwards over the same huge areas" (Crawford 1958:180).
It is possible, nevertheless, to distinguish between three major sources of information: material indices, violent extraction, and consensual provision. Material indices (photographs, captured documents, etc.) require high levels of technical sophistication to obtain, are difficult to interpret, and tend to be of limited value in contested zones. Violent extraction comes in many forms. Intimidation, blackmail, and bribes work better in urban environments, where regular and sustained contacts between handlers and informers are possible, than in rural environments, where such contacts are either impossible or easier to detect. Long detentions, even when feasible, tend to produce false confessions (Rose 2004:134). The "massive screening" of suspects is often counterproductive (Leakey 1954:122). Then, there is torture, for some "a methodological problem, not a moral dilemma" (West 1985:61).
Opinions on torture vary1 but tend to converge: from Pietro Verri's Osser-vazioni sulla tortura to Hannah Arendt's (1970:50) remark that torture is not a substitute for a "secret police and its net of informers," many authors believe that, besides being immoral, torture is an inefficient way to collect information; they argue that it produces false confessions extorted from victims desperate to save themselves further agony; that it discourages those disaffected with the enemy from turning themselves in and drives into the enemy camp those wrongly submitted to torture; that it signals inability to recruit informers and, hence, an institutional decay that causes sources of human intelligence to dry up; and, that it destroys the long-term use of a source for a doubtful short-term benefit.2 The Spanish Inquisition rejected as invalid confessions gained under torture, and it turns out that "in statistical terms, it would be correct to say that torture was
1 Aussaresses (2001) recognizes that the continued use of torture during the Algerian insurgency implies that it worked, yet his book includes more instances of information being gleaned from denunciation than torture. Moyar interviewed a number of U.S. and Vietnamese officials involved in intelligence operations. Some of them testified to the effectiveness of torture, especially during military operations when information was needed for immediate use, while others told him that torture only decreased the quality of intelligence obtained. Most U.S. advisers were not sure if prisoners revealed useful information when tortured. Of those who did think that they knew enough to verify the accuracy of prisoner testimony, a considerable number echoed the claim of American policy makers that torture did not provide any worthwhile intelligence and often yielded false information: "If you put people under physical duress, they'll tell you anything, just to get you to stop hurting them" (Moyar 1997:101-2).
2 Rejali (2004a); K. Brown (2003:167); G. Thompson (2003); Cann (1997:118); Blaufarb and Tannham (1989:27); Home (1987:205); Comision Nationalsobre la Desaparition dePersonas (1986:61); Clutterbuck (1966:97); Molnar (1965:247).
176
The Logic of Violence in Civil War
used infrequently" (Kamen 1998:188). The French reached negative conclusions about its efficiency in Algeria (Rejali 2004b; Crozier i960:19), as did some American interrogators in Afghanistan (Mackey and Miller 2004). Darius Rejali (2004a) sums up the existing evidence by concluding that "torture during interrogations rarely yields better information than traditional human intelligence." However, hard evidence is not available and it is also true that if torture always failed it would never be used. Nevertheless, it is plausible to argue that the regular use of torture requires a significant infrastructure, that it is difficult to implement in contested rural areas, and that it works in tandem with human intelligence.
The most common, and probably effective, way to access private information is consensual provision. A large body of criminological research shows that the likelihood of solving a crime decreases if the public does not identify suspects to the police (Rejali 2004a). The same is true for civil wars, as in Northern Ireland: "The recruitment of informers has long been the primary British method of gaining intelligence on their republican enemies. Over the centuries informers have been used, with devastating effect, to disrupt and destroy republican rebellions, and despite the electronic hardware of the twentieth century, the Crown's most powerful weapon in the present-day Troubles remained the human informer" (Toolis 1997:194). Indeed, "it is said that informers [within the IRA] supply over two-thirds of all intelligence" to the authorities, who make hundreds of informer recruitment bids per year; during the 1980s, the IRA executed close to forty of its own members that it suspected of being informers (Toolis 1997:212, 193; M. Dillon 1990:283).
Information can be provided by paid informers;3 however, these are hard to recruit (especially in contested rural areas), expensive to maintain, and easier to spot.4 A more common practice is denunciation, the casual and undirected provision of information from noncombatants.5 Denunciation turns the production of selective violence into an outcome jointly produced by political actors and civilians. In this sense, selective violence is a joint process.
7.2. DENUNCIATION
Denunciation is a complex social phenomenon that has thus far been little studied (Fitzpatrick and Gellately 1997:1). It is simultaneously surreptitious and shameful. It should not be confused with the practice of "public denunciation" in which
3 The term "informer" generally implies a regular, often paid, relationship to an authority (Fitzpatrick and Gellately 1997:1) - as opposed to the term "informant" or "denouncer." The (British) School of Service Intelligence (Army Intelligence Wing), defines an informant as "any individual who gives information. The term is generally used to describe a casual or undirected source as distinct from an informer, who is normally connected with criminal activities, can be directed and receives payment for his services" (inM. Dillon 1990:283-4).
4 Eric Schmitt (2003 :A2o) provides evidence about the difficulties and cost of relying on paid informers in Iraq.
5 The inverse of denunciation is "recommendation" or "certification." Following the Spanish Civil War, people suspected of having been Republicans could only find jobs if local families of proven loyalty to the regime "certified" them (Aguilar 1996:85). Both denunciation and recommendation are instances of a transfer of private information to political actors.
Selective Violence
177
people assembled in a public meeting accuse a peer (e.g., Madsen 1984:80). Community norms across cultures stigmatize the provision of information to outsiders, but also to insiders, as in Northern Ireland:
The waters cleave and the life of the informer, and their kith and kin, diverges from the tribe. (Toolis 1997:194-5)
To be an informer is to be the "Judas within, the betrayer," and the "lowest of the low." (Smyth and Fay 2000:27)
He ran to the house of a woman who spies on her neighbours for the republican movement, keeping a close eye on happenings offensive to her republican morality. She thinks her association with the IRA gives her clout in the community. In a sense, of course, she is right. But she does not know that behind her back she is loathed and despised. (Collins 1999:3)
These norms are reflected in the variety of pejorative terms invented to describe denouncers: rats, snitches, touts, soplones, chivitos, sapos (toads), orejas (ears), ruffians, mouchards, and the like. In the West Bank, individuals accused of informing the Israelis have a hard time finding lawyers when they are arrested; often, they are simply murdered, and hospitals have turned away their corpses; in one case, the mother of one informer refused to claim his body. The stigma attached to denunciation makes it almost impossible to find people willing to acknowledge having denounced. "Among my friends some had been with the Reds, such as Luzio and Isasi," said a right-wing Basque villager, "but I was friends with them as before. I never denounced anybody. I have a clean conscience about that at least" (in Zulaika 1988:25). This stigma drives even people who have denounced for "legitimate" political and ideological reasons to keep it concealed. During my fieldwork in Greece, I was unable to find anyone who openly acknowledged having denounced, even though I found several people willing to acknowledge participation in all kinds of unsavory activities and acts of violence.
Thus, denunciation is not easily observable, even ex post. The only exceptions are the salvaged archives of highly bureaucratized organizations that rely on the practice, such as the Catholic Church, the Gestapo, or the Stasi.6 A good indirect indicator of the presence of denunciation is the generalized suspicion in civil war contexts (e.g., Collins 1999:200; de Stael 1818:125). Consider the following statements:
We live in the middle of spies, the spies are among us like the devil among the Christians, (an Italian peasant in 1944-45, m Fenoglio 1973:386)
Those were the days... of the Whispering Terror. Whispers could bring about death, (a Malayan peasant in the 1940s, in Kheng 1983:141)
The villagers were fearful day and night, and wondered if they had done anything of which they could be accused, (a Vietnamese peasant on his village under Vietcong control, in Elliott 2003:259)
6 It should come as no surprise that almost all studies of denunciation have focused on Nazi Germany and the Soviet Union (Nerard 2004; Joshi 2003; Gellately 1991).
i78
The Logic of Violence in Civil War
It is often overlooked that the sort of fear that is so pervasive in civil wars is not just generic fear of armed actors but often fear of being denounced by one's own neighbors. During the recent civil war in Algeria, "everyone feared everyone, it was the law of silence. People suspected their neighbors, they were distrustful even within their own families" (Leclere 1997). In Guatemala, "spying and informing became endemic" (Zur 1998:73); fear drove "a wedge of distrust between members of families, between neighbors, among friends. Fear divide [d] communities through suspicion and apprehension, not only of strangers, but of each other" (Green 1995:105). As a provincial Colombian judge put it: "The people stay silent out of fear, because here you can't open your mouth much - if you open your mouth here it will fill with flies" (in Fichtl 2003).
It is possible to distinguish two broad categories of denunciation: those provoked by "political" motives and those by ulterior or personal motives. Denunciation is "political" ("disinterested," "selfless," "loyalty-driven," "pure") when a person denounces someone primarily out of loyalty to a cause or authority.7 A student who worked for the Vietcong said that he denounced his "friends" because he "looked on these people as the enemy and only wanted to kill them in order to have peace" (Elliott 2003:1099). Of course, private gain (including survival) may flow from the success of the actor to whom one denounces, but the primary motivation is to contribute to this success. Denunciation is "malicious" ("private," "manipulative," "interested") when it is primarily motivated by personal motives unconnected to broader political causes, for example, as payback for personal slights - though such motives can be disguised to appear as political. Sheila Fitzpatrick (1997:117) points out about the Soviet Union that "the motivation is to provoke a state response from which the denouncer would derive some specific benefit or satisfaction. The benefit might be the disgrace of a professional rival or competitor in village politics, the eviction of a neighbor from a crowded communal apartment, the punishment of a former spouse, or the settling of scores with a personal enemy."
The theory of selective violence stresses constraints and is agnostic about the motivations behind denunciation. As such, it accommodates both political and malicious motives. Nevertheless, accounts of malicious denunciation are as pervasive in the descriptive microliterature as they are absent in the theoretical and descriptive macroliteratures; as a result, it is worth examining malicious denunciation closely. I make a few initial observations here and return to this issue in Chapter 10.
Malicious denunciation may originate from individuals, families, village factions, or even entire villages;8 it may reflect a larger cleavage: for example,
7 A third type may be called "social" denunciation when a person denounces someone because she has transgressed a social norm. For example, Nerard (2004:374) sees denunciation, in a broader sense, as an act of social protest under conditions of authoritarianism, when other channels are closed. Obviously, the boundaries between these types can be fluid.
8 In some cases, private grievances can be locally collective rather than individual: in revolutionary France, for example, certain villages used the practice of denunciation as a way of removing from the village a perpetual troublemaker or petty thief (Lucas 1997). Similarly, during the Vietnam War, some U.S. advisers opposed the bombing of villages. "They contend that there is something
Selective Violence
179
personal envy can be the individual reflection of a class cleavage, but as long as one particular landowner (as opposed to any random landowner) is denounced, malice might also be involved. The same is true when a cleavage-based enmity is mixed with the expectation of private gain, as when one denounces a particular landowner or member of a "rival" political or ethnic group in the hope of acquiring property. Malicious denunciation can spring from preexisting conflicts (a recurring family feud) or it can be triggered by the war itself (revenge or retaliation for a previous action during the war, "counterdenunciation," an attempt to clear one's name by denouncing a neighbor, etc.). The Germans obtained a good deal of information about the resistance from Greeks who had been ill-treated by the partisans (Condit 1961:247) as have the Americans in Iraq (Finer 2005). Although some private denunciations are venal, entailing material benefits (e.g., the property of the denounced person), often the benefit for the denouncer is purely emotional (e.g., pleasure for punishing a despised rival).
Malicious denunciations need not be false, though they often are. One may falsely denounce an adulterous spouse as a spy, but a betrayed spouse may have known all along that her adulterous partner was a spy and only denounced him after she found out about his affair. The Norwegian village of Telavaag was burned to the ground and all its male inhabitants were deported by the Germans after they received correct information that it was used as a transit area for British agents; this information turned out to have come from a woman who found out that she had not received her share of the coffee and other goods brought in by the British agents (Riste and Nokleby 1973:51-2).
Denunciation usually implies a degree of intimacy, since it requires enough familiarity with the denounced person to have information about them. Indeed, a striking aspect of accessing the Securitate files in Romania and the Stasi files in East Germany was the discovery that informers were often associates, friends, even family members (Bran 2002; Garton Ash 1997).
Obviously, false denunciations generate moral hazard issues as CIA officials looking for information about the Vietcong knew:
The people who identified members of the [Vietcong] shadow government often had many types of non-Communist enemies in their area of operation, particularly if they worked in their native areas. Like most people, they had personal enemies: the men who had insulted their sisters, the men who had stolen their sweethearts, the farmers who had borrowed money from their families and failed to repay it, and even the GVN officials who had beaten their cousins. Family members of these enemies also could be fair game, especially when previous offenses had involved relatives. (Moyar 1997:114)
Denunciation is central to civil war: itis a common (e.g., Franzinelli 2002:197) rather than "a particular twist," as is sometimes thought (e.g., Wickham-Crowley 1990:209). This is why the absence of violence from a community is often explained by reference to the absence of denunciations. John Watanabe
basically wrong with a system by which paid Vietnamese informers can trigger air raids on villages and in which the United States acts primarily as a mechanical arm for the Vietnamese authorities. 'No agent ever calls an airstrike on his own village - it's always somebody else's,' said one American caustically" (Mohr 1966:3).
i8o
The Logic of Violence in Civil War
(1992:182), an anthropologist who worked in Guatemala, observed that "despite the army occupation, almost no one died in Chimbal, in contrast to all the towns around them. When I asked why, many responded that because they were 'good people' - or, more precisely, good 'Christian believers' of whatever persuasion -who had not denounced one another to the army as people had elsewhere."
Because most observers tend to focus on the actual perpetrators of violence and their motivations, they completely miss the fact that the information used to make violence happen may come from civilians, usually closely linked to the victims. Villagers killed in parts of German-occupied Ukraine were executed by German soldiers but "denounced as partisans by their fellow villagers" (T. Anderson 1999:621). The Guatemalan army killed thousands of Maya Indians during the 1980s: however, outside areas targeted for indiscriminate massacres, these were often "people fingered as subversives by local army informers" (Watanabe 1992:181). A Spanish woman explained the assassination of a woman in the town of Zamora during the civil war: "Viloria was a beast. He was paid to kill and he killed my father. He was the one of a group of men who were paid to murder. They shot without knowing who their victims were. 'Kill this one,' the Falangists would say. They would grab them and shoot twenty, thirty, forty, whomever they wanted. But he was not the one to denounce her. He shot her, but the person responsible for her death was he who denounced her. If there had not been denouncers, there would not have been assassins" (in Sender Barayon 1989:145).
It is necessary to distinguish between collaboration/defection and denunciation. Although both actions entail the provision of information to political actors, denunciation only refers to the provision of information about specific individuals, whereas collaboration/defection entails a much broader set of activities, from tax payment to providing information about the military activities of a rival organization. As will become clear, it is possible to collaborate/defect without denouncing, though not vice versa: the act of denunciation is, by its very nature, an act of collaboration/defection. Denunciation is riskier and more consequential than defection, both because of the social stigma attached to it and because it targets specific individuals who are members of the community.
Like any social practice, denunciation can take a variety of forms, ranging from very informal to highly institutionalized. When Iraqi soldiers raided the town of Aleze, north of Baghdad, knocking on doors and searching houses, the American reporter covering the raid noticed that "a corpulent woman whispered to the Iraqi soldiers that her neighbor disliked Americans and spoke of having grenades" (Glanz 2005A14).9 Iraqis are also reported to sometimes hand Americans lists of names at checkpoints before driving away (Negus 2004:5). Jon Lee Anderson (2004:140) describes the meeting between a Salvadoran rebel and an informer: "The peasant asks Diego to step aside for a moment, out of earshot. They stand together for a few minutes, the peasant whispering intently, Diego listening and
9 Glanz adds that "a search there turned up nothing, but the Iraqi soldiers were careful to come back and make a show of searching the woman's house as well, so that her neighbor would not suspect that she had spoken up."
Selective Violence
181
nodding. The peasant is a civilian collaborator, giving information about what is happening in the villages that lie ahead."
Denunciation can also be institutionalized. My research in Greece revealed the following procedure: a person denounced someone to a local committee member (or to someone with access to that committee), who brought the case to the committee that discussed the various cases. Committees had three options: they could send all cases to the relevant authority, select among the cases and sent out some, or send none and even veto the use of violence against anyone. Rules and procedures varied but sometimes included a formal vote. For example, when the Germans came to Ermioni in the Argolid, a small town of 2,212, in June 1944, they arrested several men who were accused of membership in EAM, the insurgent organization that had been ruling the town since 1943. An eighty-member assembly was formed to discuss whether fifteen local individuals should be handed over to the Germans or remain in town. The assembly met and decided by secret vote that more than half of them should be deported. This result, however, surprised everyone; a smaller committee was then formed to deliberate further and went to seek advice at the local capital, Kranidi. The Kranidi committee had just discovered that the Germans had, on their own initiative, shot six local men who had been handed to them in a similar fashion and advised the Ermioni men to free all prisoners, as a practical way of preventing any violence that would circumvent their consent. They did this, and everyone was freed (Frangoulis 1988:52-4).10 This example indicates the complexity of denunciation procedures and the willingness of locals to keep a degree of control on who is handed to outsiders and who is not.
7.3. DENUNCIATION IN ETHNIC CIVIL WARS
In most civil wars, ethnic and nonethnic alike, initial information about actual or potential defectors tends to be public. In ethnic civil wars, individual identities are often (though not always)11 signaled in a variety of publicly visible ways; in turn, these identities may convey (or be perceived as conveying) information about the likelihood of one's future behavior. The same is the case in some nonethnic civil wars, when polarization is pronounced ex ante and where political loyalties are in the public domain.12 In such environments, no private information is generally needed for violence to be selective. The first burst of violence will often target publicly known local leaders, as in Spain: "In Fuenmayor, when the Civil War broke out in July of 1936, order broke down completely, with predictably tragic results. After securing the town, the paramilitary Guardia Civil, acting in concert with local Rightists, dragged thirty labor leaders down to the municipal cemetery
10 This turned out to be a wise move in retrospect; when the rebels came back to Ermioni, the former prisoners were asked to intercede with the rebels in order to prevent reprisals, which they did.
11 During the war in Croatia, rival armed groups had to wear ribbons whose colors were changed every day to distinguish friend from enemy (Pervanic 1999:23).
12 There are many ways to identify "ideological" identities in nonethnic environments (e.g., Figes 1996:665; Rosenberg 1991:41).
l82
The Logic of Violence in Civil War
and shot them without trial in front of their families. The neighboring town of La Campana, held by the Left at this time, saw a brutal retaliation. The enraged Leftists there herded fifteen members of the landowning class along with the parish priest into the town jail and burned them alive" (Gilmore 1987:44).
Note that, in such environments, the expectation of being targeted based on publicly visible markers of identity will automatically deepen polarization, as individuals coalesce around their respective groups because of security considerations. Following this first round of violence, rival elites will be eliminated, and their "underlying populations" may or may not be given the choice to "surrender" to the rival army. If such an option is not given, these populations are either exterminated or deported, or they may flee.13 This process produces a front line, and collaboration resembles support during interstate wars. The logic of denunciation has limited application in these settings and applies only to marginal cases of spies and "fifth columnists." If, however, the rival population is given the option to comply, and if some people begin to collaborate with the actor in control, existing identity categories cease to convey information about future behavior (Kalyvas 2004). Such collaboration occurred during the Mau Mau rebellion in Kenya, as described by a British counterinsurgent:
One step of great long-term significance which was made at this time was the decision to form a Kikuyu Home Guard. It was a brave and imaginative move on the part of the Administration to set up, and later to arm, members of the tribe which had given birth to the Mau Mau, and 90 percent of which had taken some form of Mau Mau oath. But it was proved to be a right decision. Within a few months, the Home Guard numbered 10,000, and then rose to 20,000; they fought resolutely against their own tribe, first with spears and pangas, and only later with shotguns and rifles; but together with the Kikuyu Tribal Police, they had by the end of the Emergency killed no less than 4,686 Mau Mau, which amounted to 42 per cent of the total bag.... As a result of their resolute defiance from the start, the rebellion became a civil war within one tribe instead of being a nationalist movement. (Paget 1967:91-2)
13 Mass deportation is different from a decentralized process of segregation, though the two are often difficult to distinguish. Both ethnic and nonethnic civil wars tend to produce segregation. Chamoun (1992:23) recalls how the first months of the civil war in Lebanon led to confessional segregation, "everyone seeking refuge in neighborhoods where his religion was majoritarian." Darby (1990:98) writes that in the Belfast area he studied "more families left their homes, not because they had actually experienced violence, but from anticipation of trouble in the future." This is not restricted to ethnic civil wars. As a man from heavily secessionist Independence, Missouri, wrote to his brother that "All the people are leaving here that are for the Union" (in Fellman 1989:74). In Colombia, villages became politically homogeneous as a result of the civil war because known opponents fled (Sanchez and Meertens 2001:17). When the British occupied Boston, during the American War of Independence, thousands of Patriot supporters and their families fled the city; when the British left in 1776, thousands of Loyalist supporters followed the British troops (Carr 2004). Lear (1961:120) reports that the anti-Japanese guerrillas in the Philippines "encouraged the migration of loyal Filipinos from the enemy-controlled areas to the unoccupied districts." A pro-Japanese administrator reported, "At present there are only 30 families in the poblacion and our efforts to increase the number of returning families meet with little success because guerrilla elements controlling the barrios outside the poblacion are prohibiting or preventing the people to come in, or have contact with the authorities. They threaten to kill, kidnap, punish, or inflict injuries to those who are attached to, and cooperate with, the present regime" (in Lear 1961:208).
Selective Violence
183
In this case, the process of denunciation in ethnic civil wars follows the general lines described in this book.
7.4. IS SELECTIVE VIOLENCE POSSIBLE?
A key paradox of civil war is that it increases the need for monitoring the population while simultaneously undermining the actors' capacity to do so. To address this problem, political actors try to implement decentralization and indirect rule, delegating a measure of power to local committees or militias (Chapter 5). Decentralization produces more local information, but it simultaneously generates problems of moral hazard because inaccurate information leads to indiscriminate violence, causing counterproductive effects. An Afghan villager told U.S. troops that rival tribesmen were falsely claiming that the Taliban were active in the village and added: "Don't make the mistake the Russians made. They had informers and they arrested the wrong people and it turned everyone against them" (in Zucchino 2004A.9). A Guatemalan guerrilla said she joined the guerrillas "to avoid being killed by envidia (envy)" - denunciation by a personal enemy to the army (Stoll 1993:136). A village chief in Vietnam pointed out that false denunciations and extortion attempts by corrupt government officials had the same effect: "They would pick you up and then torture you until you had to confess. So a lot of people went over to the Vietcong, even though they didn't like them, because they had no choice. If they had stayed, they would have been arrested" (Race 1973:71-2). Indeed, a key complaint about abuse of power in Vietcong areas was "the killing of people whom the villagers knew to be innocent" (Elliott 2003:944).
Local delegation is a double-edged sword. On the one hand, it makes monitoring possible and, by creating agents who are constantly present on the ground, it facilitates denunciation by providing denouncers with plausible deniability along with an entity to shield them. Because they have access to local information, agents are able to evaluate the accuracy of the denunciations they receive. Colin Lucas (1997:35) suggests that in France, "revolutionary committees in smaller communities did usually seek to ignore or downplay denunciations that were overtly motivated by personal interests and emotions." As a rebel committee member in a Greek village told me, "A young man fell in love with a girl but her brother interfered, so he denounced her brother. He kept telling me that [her brother] was speaking up against the organization. I did not listen to him. I was objective and I was able to impose the law here in the village" (I-58). Moreover, local agents facilitate denunciation: by assuming a great part of the responsibility for the violence that follows, they partly shield the denouncer from her act and dilute individual responsibility.
On the other hand, delegation is not a silver bullet against the problem of informational inaccuracy. A particularly brazen case is that of local army agents in a Guatemalan town who intentionally provided misinformation (including the staging of fake battles and the writing of guerrilla graffiti on the walls) in order to manipulate the army into believing that the town was infiltrated by rebels so they could run a criminal protection and extortion racket (Paul and Demarest
i84
The Logic of Violence in Civil War
1988). Using local authority to settle private feuds takes place even among highly disciplined insurgents, as in Vietnam (Elliott 2003:259). It is simply very hard to monitor local agents, particularly during civil wars, when capacities are stretched thin and there is a lot of pressure to take action.
Political actors are generally aware that many denunciations are false. They know, like the Phoenix operatives did in Vietnam, that often "the distinction between the VC and private enemies became nebulous" (Moyar 1997:115); they dislike the "Hooded Men" system because "of the danger that the men in the hoods might pay off old scores against their enemies" (Kitson 1960:100). In a 1919 letter, Lenin castigated the Cheka of Ekaterinoslav for being a criminal organization that "executed every person they did not like, confiscated, looted, raped, imprisoned, forged money, demanded bribes and then blackmailed those who had been forced to pay bribes, freed those who could pay twenty times more" (in Werth 1998:120). Later on, regional Soviet officials "were well aware that peasants were using mutual denunciation as a tool to pursue village feuds" (Fitzpatrick 1997:107). When the Germans invaded the Soviet Union they were confronted with the same phenomenon. For example, when the 25th Motorized Division entered the territory of Bryansk oblast, it "complained that denunciation was simply rife among the population" (Terry 2005:8).
Sooner or later, political actors discover, as did U.S. officers who fought against Filipino rebels in 1899-1902, that some collaborating mayors "dragged them into local feuds" (Linn 1989:146). As their modern counterparts realized in Iraq, "These people dime each other out like there's no tomorrow" and "out of a hundred tips we've gotten from Iraqi intelligence, one has worked out" (in Packer 2003:71). An American soldier criticized the performance of local Iraqi informers during a military operation in the city of Tal Afar: "We almost never get anything good from them," he said. "I think they just pick people from another tribe or people who owe them money or something" (in Finer 2005A1). Again in Iraq, Captain John Prior of the U.S. Army realized that "he'd been pulled into a family feud" (Packer 2003:71), a point made colorfully by another officer, Captain Todd Brown:
Yes, that was a Jerry Springer action.... Sometimes that's what we call it when the informant just sends us on a wild-goose chase after guys that have done something to him, kind of a personal vendetta-type deal. So, when it's a personal vendetta, we call it the "Jerry Springer Show" reminiscent of all the - just the funny stuff that goes on in American society. Same thing going on here where it's a personal vendetta, and they just want to -a guy stole his cow or married a girl that he wanted to marry or stole some of his land or properly. He's just trying to get back by saying he's a leader of al Qaeda or something like that, and you go on a wild-goose chase with the informant. (CNN, 26 December 2003)
As a result, political actors mistrust their local agents and try to weed out unreliable information. They warn their subordinates, as a Colombian officer did, that to safeguard their independence, they should not allow themselves to be counseled by civilians (Roldan 2002:252). A CIA adviser in Vietnam recalls (in Moyar 1997:122): "There were times when I questioned a name on the blacklist of VCI. 'Is this guy actually VC infrastructure, or is he a political enemy or
Selective Violence
185
a business enemy of the province chief or district chief of somebody else?'"14 During a military operation, a Guatemalan officer warned assembled villagers as follows: "Everyone who did not present themselves today, those who really are guerrillas, bring them to me here. Tied up. But don't bring me innocent people. Don't bring me honorable people. And don't bring me people with whom you have some problem, over a piece of land, over a cow, over a woman, over money, none of that" (in Stoll 1993:102).
However, this is no easy problem to solve, because individuals are generally more practiced at deceiving than at detecting deception (deTurck and Miller 1990). One solution entails devising appeal procedures. The Chinese Communists introduced such procedures but they proved inadequate even during periods of relative stability and had to be suspended during periods of crisis; eventually much of their violence was arbitrary.15 Moreover, local agents can terrorize individuals who appeal.10
Another means of increasing accuracy is to introduce accountability by making the local agents' or the denouncers' identities public. Without the protection provided by anonymity, however, the pool of denouncers and of candidates for positions of local authority would dry up quickly. For example, U.S. commanders had planned to circulate a list of 1,400 people thought to have potential insurgent connections in the town of Tal Afar in Iraq, seeking verification or denials from local sheiks. "But they decided against it," Oppel (2oo5a:A8) points out, "because few sheiks would openly affirm or deny the status of insurgent suspects in front of other Iraqis." It is also possible to rotate agents on the assumption that they are less likely to have grudges outside their own turf.17 Besides the logistical issues entailed in such a method, this solution defeats the logic of delegation, which is
14 The guerrilla leaders in the Japanese-occupied Philippine island of Leyte "sought to bring some approximation to a rule of law in the territory they controlled," meaning that "espionage and collaboration must be discouraged through the judicious application of swift punishment to the guilty. But malice must not be permitted to level false accusations against the innocent, in order that neighbors might conveniently dispose of their personal enemies or improperly acquire their property" (Lear 1961:91).
15 "Since the responsibility for arrest and investigation, as well as the determination of guilt, were largely vested with one organ, the police, there were no effective checks to avoid unwarranted arrests and punishments" (Griffin 1976:139).
16 Helen Siu (1989:132) tells the story of a cadre in postrevolutionary China who supported a malicious accusation against a villager and forced him to sell his property in order to pay his accuser. The case was reviewed by a committee that reversed the verdict; it ruled that the villager had been victimized by the "arbitrary, opportunistic accusations of bad elements." This appears to be rather exceptional, however, for civil war contexts. Benjamin Paul and William Demarest (1988) list many cases in which individuals were unable to convince the Guatemalan army that their local agents were running a private racket.
17 After the inspector general of the Union army visited Missouri in 1864 and realized that "many of the soldiers and their families have suffered from the depredations committed upon them by rebels, and they have their enemies whom they desire to punish, and they are very prone to use their power, which their military positions give them to accomplish unwise purposes," he advocated the use of out-of-state troops who did not have three years' worth of grudges to avenge (Fellman 1989:87). According to a U.S. adviser in Vietnam (in Moyar 1997:222-3), "The police chief and the Special Police chief weren't living up in Phu Yen to pursue a vendetta. If they'd had a vendetta
i86
The Logic of Violence in Civil War
to save on monitoring costs: nonlocal agents are much less able to collect and assess information compared with local ones.
The most effective solution probably consists of cross-checking denunciations and applying sanctions when they are false. A Gestapo memorandum sent to all headquarters from Berlin in 1941 concerning denunciations among relatives, particularly husband and wives, suggested that denunciation was being used for private ends completely unanticipated by the regime. The memo introduced more thorough rules requiring married denouncers of their wives to answer under oath whether divorce proceedings had already commenced or were contemplated; moreover, the Minister of Justice added that even denunciations eventually resulting in the discovery of a serious crime did not automatically provide grounds for winning a case in a divorce court (Gellately 1991:143, 148-9).18 More forcefully, the Chinese Communists condemned as traitors those "who falsely accuse others as traitors" (Griffin 1976:173), and the Italian partisans executed people who joined them in order to conduct personal vendettas (Pavone 1994:451). The German occupation army in the Soviet Union arrested and imprisoned false accusers, and in some cases it had them publicly whipped in front of entire villages as a deterrent to future false denunciations (Terry 2005:8). An additional illustration comes from the Japanese-occupied Philippines: "There was a time when a soldier was made drunk by a fellow who ... wanted to eliminate [the soldier's] barrio lieutenant. The soldier was told that the said barrio official used to receive letters from town. Without investigating further the matter, the soldier looked for the 'Teniente' and shot him cold-bloodedly. Later, when the soldier became sober and perhaps realizing that he had committed a most heinous crime, he fetched in turn the informant and shot him also" (Lear 1961:94).
The empirical record is mixed as to how often and how effectively crosschecking is used. In Peru, a human rights worker remarked that "Sendero always investigates those it kills" (in Rosenau 1994:317). A Phoenix operative in Vietnam claims that "the overwhelming majority of those captured on Phoenix operations were picked up based upon tangible and credible evidence, rather than on the mere say-so of one person motivated by some sort of personal grudge" (Herrington 1997:196), though a U.S. military adviser observed that "falsification of data and targeting of personal enemies did occur, and when discovered usually resulted in some form of disciplinary action" (Moyar 1997:120). The British used a system of multiple hooded informers in Kenya, the idea being that "any genuine terrorist or committee member would be recognized by two or three of them" (Kitson 1960:10i).19 Berlow (1998:247) reports a conversation with an NPA cadre in the Philippines: "'We don't accept demands for retribution from families,' he said, explaining that people often make false accusations to the
against anybody, it would have been back home where they came from, not up in Phu Yen. They didn't know anybody up there - that's why the Government had put them there."
18 However, Nérard (2004:361) found that the cost of false denunciations in Stalin's Soviet Union was low to nonexistent.
19 Kitson (1960:102) adds that this was "a laborious business and very tiring for the hooded men, especially as the sun came out after a time, making them hot and thirsty."
Selective Violence
187
NPA - just as they do to the military - to try to settle purely personal grudges. 'We have our own policies for meting out penalties, including the death penalty'."
However, effective cross-checking requires a high level of control and an efficient bureaucracy; it is, therefore, very hard to achieve in civil war, when resources are stretched thin, especially in contested zones. The available evidence suggests that authoritarian regimes tend to expand more resources in screening denunciations compared with civil war actors. The Spanish Inquisition often knew when to discriminate between false and true denunciations (Kamen 1998:181), although a villager asserted in the 1480s that "in Castille fifteen hundred people have been burnt through false witness" (Kamen 1998:175). Likewise, James Given (1997:141) reports that one indication of the extent to which individuals attempted to manipulate the inquisition in medieval Languedoc was the frequency with which inquisitors imposed penances on individuals whose chief fault was bearing false witness against the innocent. "Manipulating the inquisition may have given some Languedocians access to a new and unusually effective political resource," he concludes (1997:142); "yet there was always the danger that the inquisitors might discover what was afoot. The price paid by an unlucky schemer for access to this particular resource could be very high."
In contrast, I found very few instances of effective cross-checking in civil wars, particularly outside of zones of full control. Republican militiamen in Spain made no systematic effort to find out whether the denunciations that had led to executions of villagers in a Spanish village were false or malicious (it turns out that they were both) (Harding 1984:75-6). An American officer serving in Iraq remarked that hooded local informants "are the first important step in the process of weeding [the insurgents] out." He added that "You obviously can't just go by what they say because they make plenty of mistakes, but since we don't know these places as well as they do, it helps to have them around" (in Finer 2005^1). Another officer admitted that he would never get to the bottom of the many contradictory stories told to him by various informers and their victims: "I am not freaking Sherlock Holmes," he exclaimed (in Packer 2003:72).
Instead of cross-checking, political actors turn to "secondary" profiling -secondary in the sense that it takes place once a list of names has been handed in: they look for visible features that may signal loyalty or disloyalty and separate true from false denunciations. The Indian security forces in the Punjab looked for "young Sikh men between the ages of 18 and 40, who have long beards and wear turbans" (Gossman 2000:267). The Nicaraguan contras considered as likely Sandinista supporters rural schoolteachers and health workers (Horton 1998:128) - two professions also targeted by the RENAMO insurgents in Mozambique (Nordstrom 1997:83). The Guatemalan army aimed at teachers, bilingual instructors, catechists, and officers of the cooperatives (Warren 1998:95; Paul and Demarest 1988:125-6). Wealthy Colombians are seen as fair game for insurgents, whereas paramilitaries target labor organizers and human rights workers (Fichtl 2004:5). The Vietcong were particularly suspicious of people who traveled to market towns, where the government was present (Elliott 2003:949-50). Obviously, such profiling is often ineffective. For example, the Languedocian Inquisition failed to screen out false denunciations when the
i88
The Logic of Violence in Civil War
victim was a person who "had offended important members of the local political establishment and thus made himself vulnerable to attack" (Given 1997:147). Furthermore, too much reliance on profiling defeats the basic premise of selective violence. Indeed, it appears that a compromise is made between the demands of selectivity and the limitations on available information. My fieldwork in Greece revealed several cases where profiling was mixed with local information. I found that the mayor of a village was killed by the rebels, probably after being maliciously denounced by the brother of a woman he had falsely promised to marry; as a mayor, he also came into frequent contact with the occupation authorities, a fact that may have tipped the balance against him when the moment of decision came (1-6,1-7).20
Ultimately, it is impossible to estimate the proportion of false positives and negatives. Nevertheless, there is evidence that political actors often choose to err in the direction of false positives rather than false negatives. U.S. commanders in the Philippines at the turn of the century drew the consequences of this situation clearly and officially: "To arrest anyone believed to be guilty of giving aid or assistance to the insurrection in any way or of giving food or comfort to the enemies of the government, it is not necessary to wait for sufficient evidence to lead to conviction by a court, but those strongly suspected of complicity with the insurrection may be arrested and confined as a military necessity and may be held as prisoners of war in the discretion of the station commanders until receipt of other orders from higher authority."21 An Italian partisan formulated the problem in stark terms: "The situation forces us to deal seriously with the problem of spies and denouncers: suspects must be arrested and killed on a minimal evidentiary basis. On the other hand, there is a risk of condemning innocent people: but how is it possible to wait for the proof of the betrayal? From the death or arrest of someone on our side?" (in Franzinelli 2002:204). In Kenya accusations made by others, including hooded informants, needed no corroboration (D. Anderson 2005:203). In Colombia, armed groups "prefer the simple 'justice' of summary executions of suspected collaborators over the convoluted machinations of trials or the awkwardness of taking accused collaborators captive" (Fichtl 2004:5). In 2003 the Colombian government proceeded to arrest hundreds of people in several localities on the basis of just a few denunciations; seventy-four people were arrested in the small town of Cartagena del Chaira on the strength of a single denunciation by a man who many locals accused of malice (Semana 2003).
20 In a sense, profiling reflects the joint production of violence onto the target: those most likely to be turned in are people who happen to both have personal enemies and fit a public profile of disloyalty. This is also an instance of how the master cleavage may shape violence, conditional on local dynamics.
21 In his description of the Athenian expedition in Sicily, Thucydides (6.53) tells the following story: "After the expedition had set sail, the Athenians had been just as anxious as before to investigate the facts about the mysteries and about the Hermae. Instead of checking up on the characters of their informers, they had regarded everything they were told as grounds for suspicion and on the evidence of complete rogues had arrested and imprisoned some of the best citizens, thinking it better to get to the bottom of things in this way rather than to let any accused person, however good his reputation might be, escape interrogation because of the bad character of the informer."
Selective Violence
189
"Better to kill mistakenly than release mistakenly" went a Vietnamese slogan, popular among some insurgents; for them, "justice was not an abstract ideal, but a tool in the political struggle"; "if it came down to a conflict between the revolution's prestige and abstract notions of justice, it was clear which would prevail" (Elliott 2003:91, 947). A U.S. commander in Iraq remarked about Iraqi counterinsurgents that "if they shoot somebody, I don't think they would have remorse, even if they killed someone who was innocent" (in Maass 2005:47).
Thus, selective violence targets many innocent people. Recounting the violence that took place in his village during the Greek Civil War, the writer of a local history concludes that the killings were caused "somewhat" by the political affiliation of the victims but "more" by the vengeful obsession of their enemies (Kanellopoulos 1981:609). The Phoenix program in Vetnam was often "rooting out the wrong people" (Adams 1994:179; FitzGerald 1989:516), the Huk guerrillas in the Philippines "killed people whom they thought were spies or enemies but were later shown not to be" (Kerkvliet 1977:177), and both the UNITA and the MPLA in Angola executed many innocent people as traitors based on false accusations resulting from personal enmities (Brinkman 2000:15). A local FLN commander in Algeria is said to have caused the execution of as many as 3,000 mainly innocent men and women in his campaign of terror, launched in 1958 and 1959 after the French were able to adroitly foster suspicions amongst the Algerians (Home 1987:323). In El Salvador, many false denunciations were "enough to seal one's fate, since government forces seldom investigated the charges and 'innocent until proven guilty' was not a principle recognized by the military, security forces, or ORDEN civilian irregulars" (Binford 1996:107; also Wood 2003:96-7). A report about Sri Lanka states that "by taking informers at their word, [security] forces allowed old grudges, land disputes and business rivalries to be bloodily settled" (University of Teachers for Human Rights 1993:38). Joseba Zulaika (1988:99) "found that the solidly established 'facts' about [a presumed informer in the Basque village], such as his traitor role in the events of i960, were plainly false."22 Quotas and rewards for "neutralizations" only make this problem more acute (Courtois 1998:21; Moyar 1997:116; Chang 1992:218).
There is some systematic evidence beyond the anecdotal record. Peter Hart (1999:17, 303) extensively researched the archives of the British police and found that among the victims of the IRA in 1916-23 "very few were actual informers. Most were innocent victims." By comparing data on IRA executions and British intelligence, he concludes that the great majority of true informers were never suspected or punished; most of those shot (or denounced, expelled, or burned out of their homes) never informed, and those blacklisted were also usually innocent. In Peru, it was reported that the special antiterrorism courts set up to combat the Sendero insurgency convicted hundreds of people who were later proved to be innocent of aiding rebel groups. By the summer of 2000, 1,089 of these "innocents" were released either by pardon or reversal of their sentences (Krauss 2000:3).
22 Additional instances of false denunciation are reported in Japanese-occupied Malaya (Kheng 1983:144, 180, 181-2), Guatemala (Warren 1998:99), and Sri Lanka (Senaratne 1997:147).
The Logic of Violence in Civil War
Overall, it is fair to surmise that political actors frequently fail to discriminate between the guilty and the innocent. Shall we then conclude that selective violence is an illusion and that all violence is, in fact, indiscriminate and, ultimately, counterproductive?
Such a conclusion would be erroneous. There is substantial evidence that political actors are successful in generating deterrence via selective violence in spite ofkilling many innocent people. This was clearly the case with the Phoenix program in Vietnam, which is described simultaneously as relatively inaccurate and very effective (Sheehan 1989:116; West 1985:95). The same was true about the IRA. The British authorities concluded in a 1921 intelligence report that the IRA was notoriously inaccurate: "In every case but one the person murdered [by the IRA] had given [them] no information"; at the same time, however, they recognized that, in spite of this, the IRA's war on informers was highly effective (Hart 1999:300). My own research in Greece corroborates this insight: many people were killed selectively but erroneously; yet their deaths were a deterrent, as intended by the perpetrators. In sum, though imperfect, selective violence is effective. But how?
To achieve deterrence, political actors must convince the targeted population that they are able to monitor and sanction their behavior with reasonable accuracy. In other words, they need to cultivate a perception of credible selection. They can achieve this goal without being perfectly accurate in their targeting. A mix of accurate and erroneous hits is compatible with a perception of credible selection under three conditions.23
First, the very presence of local agents signals the organization's willingness and potential capacity to be selective. Only if the moral hazard problems become excessive do political actors need to intervene; otherwise, the system remains in effect. Political actors advertise their selective capability as being a function of local agency: one of the principal slogans of Shining Path was "the party has a thousand eyes and a thousand ears" (Degregori 1998:143). In this sense, the importance of local agents is based less on what they actually do and more on their very existence. If the public believes that a network of informers is active, they will tend to infer that a victim was guilty (Herrington 1997:39)- or at least they will be too uncertain not to take such a possibility into account, an insight consistent with many observations about the paralyzing effect of the perception that an effective network of informers is active: "No one can be sure who is who" was the expression that described how people felt in Guatemala (Green 1995:105). A man described the situation in Ireland in 1922-3: "Perhaps the most reprehensible things one meets here is what is known as 'Intelligence.' One never knows to whom he is speaking. One never knows who is or who is not an 'Intelligence Officer.'... all eyes seem to gaze and all tongues to whisper in suspicion and doubt wherever
23 Obviously, the optimal strategy for political actors is random selection that appears to be selective (I thank Diego Gambetta for pointing this out to me). In practice, this is difficult to achieve; the administrative machinery required to create a credible perception of selection is extremely costly to set up and, once in place, it leads to a mix of correct and erroneous hits rather than purely random targeting.
Selective Violence
one happens to go" (in Hart 1999:124). Although violence is often public,24 there is a certain ambiguity about its true causes. Descriptions of terror consistently include rumors and blacklists compiled in obscurity, as well as "whisperings, innuendos, rumors" about who is on these lists (Green 1995:109; Faivre 1994:
J45)-
Second, local agents help political actors avoid blatant mistakes, which are easy to spot by the public and create a general perception of consistent mistargeting.25 As I discussed in the previous chapter, indiscriminate violence produces the highly visible mistargeting of potential or actual sympathizers of the indiscriminate actor, something that can be avoided through local delegation. Political actors do not want people to ask, as they did a local Palestinian leader about the assassination of a suspected collaborator of the Israelis, "Why did you kill this guy? He is innocent" (in Swedenburg 1995:199).26
Third, when defection takes place under constraints (i.e., where the rival actor has the upper hand), it is generally a secret activity; people cannot tell whether a particular victim was really a defector or not.27 In the areas of El Salvador studied by Leigh Binford (1996:112), "the majority of the politically 'undecided' population had no way of knowing whether or not the accusations that the victims [of the army] had collaborated with the guerrillas were merited, since the ERP maintained a low profile there." When uncertain about the victims' innocence or guilt but somewhat persuaded about the organization's credibility, most people tend to infer guilt and alter their behavior accordingly. Consider the following examples from Algeria, Vietnam, Sri Lanka, and Colombia:
When we were hearing that persons X or Y had been found murdered, we said to ourselves: "Who would have believed that they were traitors? But they must have been, since the FLN executed them." (in Hamoumou 1993:157)
24 They may leave messages on the victims' bodies, occasionally organizing (usually rigged) public trials and public confessions (Kheng 1983:180; Cobb 1972:1921). For example, the Vietcong would pin a "death notice" to the body, which listed the alleged "crimes" of the victim and stated that in the course of committing these crimes, the victim had "amassed many blood debts to the people" and therefore had to be condemned.
25 Note that the victim's family will not be convinced; however, they will not speak up if it is too dangerous to do so (Hamoumou 1993:157, 174). Plus, claims by a victim's family are not usually seen as being credible.
26 This is why the option of random selection under the pretense of actual selection cannot work. Note as well that the assassination of individuals who are known to have been victims of malicious denunciation is not necessarily interpreted as an instance of mistargeting since malicious denunciations can also be true.
27 Even members of the organization targeted by the violence may be unsure of the innocence of some of their colleagues. The degree of secrecy of defection varies depending on additional factors, such as the patterns of past organization. For example, the Japanese noticed that, following their mopping-up campaign, Communist organizations were stronger in Yongqing than in Hejian (both in Central Hebei, China), and this despite the fact that Communist tradition had been much stronger in Hejian. Because organizing in Hejian had started under fairly secure military conditions, Communist activists were known by all villagers; when the Communists were forced out, most local activists could be denounced; in contrast, because the Communist organization in Yongqing was built later and under much less favorable conditions, it included secret organizations that were better able to withstand the Japanese onslaught (Hartford 1989:117).
I92
The Logic of Violence in Civil War
The VC executed four persons in my village. They explained that these people were paid agents of the Government authorities. Nobody could figure out whether this was true or not. Everybody was afraid. No one dared say anything, (in Mallin 1966:72)
The insurgent JVP succeeded in conveying a "general presumption that, if someone were killed by [them], then s/he had done something which deserved punishment." (M. Moore 1993:628)
There is a widespread belief among the population that victims of the violence had "asked for trouble." Typical comments about people who had been murdered included "Algo debia" (He had something on his slate) or "Es que se habia polarisado" (He had himself polarized); killings would often be "explained" by designating the victim as a desintegrado, a descompuesto, a ladron (thief), a falton (someone who hadn't kept his word), a hablon (someone who talks too much) or a desechable (a disposable person). (G. Martin 2000:181)
In short, the effectiveness of selective violence hinges less on pinpoint accuracy and more on a perception among the population that a process of selection is taking place. The use of local agents is essential in generating this perception and helps explain the apparent paradox of campaigns of selective violence that are highly effective despite failures of accuracy.
7.5. A POLITICAL ECONOMY OF DENUNCIATION
The incidence of denunciation depends on both motivations and constraints. Motivations are plentiful and diverse (Chapter 10); even low levels of social conflict and high levels of solidarity may not prevent denunciations from taking place given the small number of people required to set this process in motion. Constraints are much more effective regulators of denunciation.
The supply of denunciations is subject to a fundamental constraint, namely the likelihood of retaliation faced either by the denouncer or by the local committee that vets denunciations. While the norm-driven loathing often triggered by this act is a potential source of risk for the denouncer, the real risk comes from credible threats of reprisals rather than diffuse feelings of dislike. This dimension, of course, is standard in organized crime: credible threats of retaliation discourage witnesses from testifying (e.g., Butterfield 2005). A similar mechanism can be found in civil war. A Greek villager explained why he refrained from denouncing to the right-wing authorities the leftist villagers who caused the death of his uncle: "There were partisans roaming around the village," he told me; "you did not know what could happen to you" (I-10). A Greek Communist villager (Nikolaidis 1977:55) recalls how he reacted after a Communist guerrilla had a local villager tried by a "popular court" and beat him up: "Do you have any idea how we will suffer because of your kangaroo court? You are leaving but we have to stay here." This process is apparent in the following examples from Kenya, Algeria, Vietnam, and Northern Ireland:
Even when such people are known to "loyalist" Kikuyu living in the towns to be Mau Mau followers, it is not easy for these people to give evidence against them or point them out. If they did so, swift retaliation would follow. (Leakey 1954:121)
The inhabitants will know them, since they suffer terribly from their activities, but will not denounce these agents unless they can do so without risk. Fear of reprisal will always
Selective Violence
193
prevent them from communicating to us information they possess.... To succeed, we must never lose sight of the fact that we will receive information only if people can give us information without risk to themselves. (Trinquier 1964:35, 78)
Rule number one was "Never inform the government of Communist activities." In Hiep Hoa, most of the villagers were well aware which families were revolutionary families and who constituted the village's party committee. But no one could be certain of the loyalties of every one of his neighbors.... Virtually every hamlet in Vietnam had at least one clandestine informant who would not hesitate to report to the Vietcong the name of a farmer who warned the Americans about a booby trap. The Vietcong's organization was thus the major device by which the revolution insured the silence of the people - and this silence was sufficient to frustrate our efforts. (Herrington 1997:39)
Thus the major factor in the initial decline of GVN intelligence on the situation in the countryside was a change in the security of the intelligence agents: the GVN lost the ability to protect them. Obviously the risk calculation changed. Given the seriousness of likely reprisals for such activity, those who were in it for the money must have found that it wasn't worth risking their lives. Those who bore grudges against the revolution found that the costs of exacting revenge had escalated dramatically. (Elliott 2003:424)
She lived in the area. She knew who killed her husband, but she couldn't say who killed him, because my brothers all lived there and my father lived in the area, so they would have had to leave the country. They wouldn't have been able to stay. So she couldn't really say anything about who killed him. She saw his killers every day and they used to scare her to make sure she kept her mouth closed, (in Smyth and Fay 2000:23)
The significance of the risk of retaliation as a determinant of denunciation is consistent with psychological studies according to which the relative strength of signs of retaliation inhibit revenge (Bandura 1983); with experimental evidence suggesting that the anticipation of retaliation is, under some circumstances, an effective regulator of aggression (Walters 1966); with sociological studies of rural contexts showing that peasants take seriously into consideration the power of their competitors in deciding whether to challenge them or refrain from doing so (e.g., Hua and Linshan 1996:180-2); and with studies of criminal or quasi-criminal environments demonstrating that credible threats of retaliation by criminals inhibit victims and witnesses from reporting the crime or providing evidence (e.g., Crisp 2000:620) - not to mention numerous casual yet insightful observations in literary texts (e.g., Stendhal 1996:38).
This risk explains why denouncers (as well as informers) seek anonymity. A Russian dictionary defines denunciation as "a secret revelation to government representatives of some kind of illegal activity" (Kozlov 1996). Political actors are often willing to provide anonymity to mitigate the risks of denunciation (Kamen 1998:182; Moyar 1997:74).28 The figure of the hooded informer fingering the people to be arrested (the infamous encapuchado in Latin America) is common across most civil wars.29 Political actors dislike total anonymity because it is "an open invitation to perjury and malicious testimony" (Kamen 1998:182).
28 An advertisement I saw in the New York subway included the following message: "You don't have to reveal your identity to help solve a violent crime. Call 1-800-577-tips. Rewards up to $2,000."
29 E.g., D. Anderson (2005:202); Wood (2003:114); Mahmood (2000:83); Zur (1998:80); Stoll (1993:62); Stubbs (1989:44); Kheng (1980:96); Kerkvliet (1977:66); Kitson (1960:100).
i94
The Logic of Violence in Civil War
However, anonymity is not easy to achieve, especially in small communities. "There were no secrets in a rural Vietnamese hamlet," recalls Stuart Herrington (1997:23). Of a stupid man, Cypriot peasants say: "He thought he could beat his wife, without his neighbors hearing" (Durrell 1996:224). It is often possible to guess the origin of a denunciation, particularly when personal feuds and small communities are concerned (e.g., Butterfield 2005:22; Argenti-Pillen 2003:61-2; Berlow 1998:44).30 Kevin Andrews (1984:122), who traveled across Greece in 1949, reproduces the following conversation he had in a village:
"Tell me one thing. The people who burned the houses, the men who killed your sister and her child - what has become of them now?"
"Become of them now?" He looked at me with Papastavros's same childlike gaze. "Nothing."
"What do you mean?"
"They're all there."
"Still in the village!"
"Where else would they go?"
"But does he know who did it?"
"Of course he does. In a village everything is known."
Petitions to higher authorities written by some victims in Civil War Missouri suggest that victims had guessed the probable identities of their victimizers (Fellman 1989:60); the masked members of a death squad in Guatemala were identified by the relatives of one of their victims (Paul and Demarest 1988:123); rumors about who betrayed Saddam Hussein's sons emerged immediately after they were killed;31 in one Greek village of my study, the hooded informer who came along with the Germans to finger resistance members was recognized by so many of the villagers who were gathered in the village's central square that he had to take his hood off.
Relatives and friends of a victim of denunciation naturally desire revenge against either the denouncer or the local agents who endorsed the denunciation. Hence, potential denouncers and local committees must take into account the risk of retaliation they face. Unlike revenge in blood feuds, which is direct, retaliation in the context of civil war tends to be mediated. As discussed in Chapter 3, blood feuds are generally ritualistic occurrences regulated by a concrete set of norms about which offenses are subject to retaliation; these norms explain why people are willing to retaliate given the potentially large costs they face (Gould 2003). The unwillingness, in most societies, of the great majority of people to commit violent acts and the absence of revenge can be explained by the absence of blood feud norms.32 Civil war increases the opportunities for revenge and lowers its costs significantly: a person need not directly bloody her hands.
30 This is the point where feuds and purely political activity diverge: secret informing is much more difficult to detect than malicious denunciations based on personal and local conflicts.
31 "Host betrayed Saddam's sons," BBC News, 24 July 2003, http://news.bbc.co.uk/i/hi/world/ middle_east/3O02 783.stm.
32 Consistent with the observation that mass killings are the work of relatively few people (Valentino 2004).
Selective Violence
*95
Retaliation takes the form of "counterdenunciation," that is, the denunciation of the initial denouncer to the rival political actor. Just as denouncers may use political actors to carry out their own ends, the family of a denunciation victim may "counterdenounce" the initial denouncer. In short, most people's fundamental dislike for committing violence with their own hands and the political actors' aspiration for monopolizing violence33 turn counterdenunciation into the main tool for retaliation.
Two conditions must be fulfilled for a counterdenunciation to take place. First, the counterdenouncer must have access to the rival actor (and this actor must have the ability to carry out the reprisal). Access to political actors is asymmetrical, and they are naturally unwilling to endogenize revenge cycles. Second, just as denouncers take into account the risks of retaliation, counterdenouncers must think about counterretaliation.34 When individuals feel that, even if recognized or revealed as denouncers, the political actor to whom they are denouncing their fellows has the ability to shield them from retaliation, they are more likely to denounce (or counterdenounce);35 if they worry that in denouncing they are likely to be unprotected and hence face retaliation through counterdenunciation (or if they believe that their denunciation or counterdenunciation is unlikely to be implemented), they are unlikely to denounce.36
Although what counts as an acceptable level of protection, given the keenness of the motivating impulse, will vary with individuals' risk tolerance, the baseline answer is that denunciations will be a function of control exercised by political actors. Control also affects the ability of an actor to carry out a reprisal. I integrate this insight into the formal illustration of the theory of selective violence that follows.
7.6. A MODEL OF SELECTIVE VIOLENCE IN CIVIL WAR
I formally illustrate the theory of selective violence and generate predictions about the likelihood of selective and indiscriminate violence across space and about the identity of the perpetrators (whether they are incumbents or insurgents);
33 Kathleen Hartford (1989:114) describes how the Chinese Communist Party placed "traitor-elimination" programs under direct control of the district party committee, not under the village party branch. Assassinations were vetted by the district, and independent revenge killings were not permitted.
34 Of course, the identity of the counterdenouncer must be equally visible to that of the original denouncer.
35 I assume that a low probability of being killed does not translate into very large expected costs. In fact, very low probabilities of death could actually make individuals behave as if the expected costs were zero. Substantial experimental evidence suggests that individuals tend to overestimate their chances of success for relatively beneficial actions and to underestimate their chances of success for potentially costly actions (e.g., Mirels 1980; Weinstein 1980; Larwood and Whitaker 1977; Miller and Ross 1975). When I asked several Greek informants why people denounced given a low probability of retaliation, they pointed to the driving habits of many locals who tend to drive very dangerously despite being aware of a low probability that they may suffer an accident.
36 Denunciation to an actor who is unable to carry out the reprisal is functionally equivalent to be left unprotected by this political actor.
196
The Logic of Violence in Civil War
the theory is agnostic about the intensity and timing of the violence. To keep the model simple, I disaggregate it into three distinct but related processes: the individual calculus of defection, the individual calculus of denunciation, and the organizational calculus of violence.
Preferences are straightforward. Political actors maximize territorial control; they seek to "conquer" territory and increase the level of control over the territory they rule. I assume no anarchy; when one actor abandons a territory, the rival actor moves in. Increasing control means obtaining the exclusive collaboration of civilians and eliminating defection, that is, collaboration with the rival actor; this is the main function of selective violence.
The production costs of selective violence are assumed to be inversely related to control; I take the distribution of control at to to be exogenous; once the process has begun, subsequent shifts of control are a function of two factors: first, exogenous military resources that allow an actor to "conquer" territory hitherto controlled by its rival and, second, the use of selective violence in territory that is already "conquered," which increases the degree of collaboration and hence control in the subsequent period tT - provided, of course, that the existing balance of power is not exogenously altered by one actor withdrawing forces or the rival actor bringing in additional forces.
Civilians are boundedly rational; they are reward-sensitive and seek to maximize a personal or political utility subject to their likelihood of survival; they also tend to conflate opportunities with their beliefs about opportunities. The model is agnostic about the motives of defection and denunciation: they can be political or personal, expressing ideology, revenge, or spite. However, I assume that denunciations take place locally between people who know each other. I also assume that individuals believe that the level of control exercised where they live is stable. Civilians must make two separate strategic decisions: whether to defect and whether to denounce. Political actors must decide whether to use violence and what kind to employ.
7.6.1. Defection
Consider a distribution of the geographical space into five discrete zones of control, ranging from 1 to 5. Zone 1 is an area of total incumbent control, and zone 5 is an area of total insurgent control. In between lie zones 2,3, and 4, which are contested areas where control varies as follows: zone 2 is primarily controlled by the incumbents (dominant incumbent control), zone 4 is primarily controlled by the insurgents (dominant insurgent control), and zone 3 is controlled equally by both sides (parity).
Following Proposition 1,1 assume that defection (i.e., collaboration with the rival actor) is shaped by the level of control exercised by the competing political actors. If there are k defectors in a village and c is the level of control an organization enjoys in the village, k(c) decreases as c increases. The benefits of defection include the material and/or nonmaterial advantages derived from helping an organization with which one is associated, while the costs of defecting and being caught - prison, torture, death - are steep. If i is the payoff for defecting
Selective Violence
197
and u is the cost of the defector being caught, then for the vast majority of people who prize survival u > i. The likelihood of being caught will, therefore, condition their willingness to defect given their preferences.
Political actors are willing to pay a premium for collaboration (in the form of more promises, promotion, or material goods) where their capacity to control decreases, even while their ability to deliver this premium decreases with control, as one moves away from zone 3 toward areas of weaker control. In contrast, their capacity to arrest defectors increases with control, as one moves away from zone 3 toward areas of stronger control. A defector is caught either by direct detection or by denunciation. If p is the probability that a defector will either be detected directly or denounced and caught, then the cost of defection is prohibitive where control for the rival actor is total: pu > (1 - p)i; p reaches its maximum value under total control and decreases until it reaches zero under the rival actor's total control. Figures 7.1 and 7.2 illustrate the relation between expected cost and benefits for collaboration with incumbents and insurgents respectively (or defection toward incumbents and insurgents) across the five zones of control.
It follows that only martyrs defect under total control (zones 1 and 5), though highly committed individuals defect under dominant control (zones 2, and 4). Defection picks up in zone 3 for both actors and explodes in zones 4 and 5 (toward the insurgents) and 2 and 1 (toward the incumbents) (Figure 7.3). Defection is a problem for incumbents in all zones except zone 1 and for insurgents in all zones except zone 5. Put otherwise, zones 1 and 5 are homogeneous, while zones 2, 3, and 4 are heterogeneous, consistent with their characterization as contested areas.
7.6.2. Denunciation
Consider the following formal illustration of the argument. There are two villagers, A and B; A chooses whether to denounce B or not, and B chooses whether to denounce A or not. Each villager has an exclusive political association with one political organization (villager A with organization A and villager B with organization B); in turn, each organization enjoys a certain level of control; rA is the degree to which organization A is able to control the village and exclude organization B, and rB is the degree to which organization B is able to control the village and exclude organization A. Consistent with the preceding discussion in this chapter, the values of and rB across the five zones of control are as follows: rA goes up in zones 1 and 2 and down in zones 4 and 5, while rB goes up in zones 4 and 5 and down in zones 1 and 2; zone 3 is a zone of parity where rA = rB. The spatial location of each villager (and hence rA and rB) is chosen by Nature.
Each villager supplies information to the organization, which carries out assassinations accordingly. I assume that the villagers can only inform the organization with which they are associated and that, once denounced, an individual will be targeted and assassinated by the actor to whom she is denounced with a probability p. Let pA be the probability that organization A targets and succeeds in assassinating B following a denunciation, and pB the probability that organization B targets
figure 7.1. Payoffs and Expected Cost of Collaboration with (or Defection to) Incumbents
200
The Logic of Violence in Civil War
k(c)ins k(c)inc
Control 1 2 3 4 5
_ Defection toward the insurgents
--------- Defection toward the incumbents
figure 7.3. Defection as a Function of Control
and succeeds in assassinating A following a denunciation. I assume visibility between the denouncer and the denouncee's family, thus allowing for potential retaliation: the family of a person who is denounced and assassinated has the option of retaliating by counterdenouncing the initial denouncer to the rival organization. Villagers choose between two strategies, denounce (D) and not denounce (N).
Let x be the value to individual A of A's organization's assassinating B and the value to B of B's organization's assassinating A; this is the satisfaction derived by the elimination of a local rival. Let y be the immediate cost to individual A of denouncing B and the immediate cost to individual B of denouncing A, which may consist of detection and sanctioning by the rival organization, which I assume to be death; y > x, because one's own death generally far outweighs whatever benefit is derived from a rival's denunciation and death. I assume x and y to be the constant across individuals. In addition to y, suppose qA is the probability of retaliation via counterdenunciation by individual A's family against B and qB is the probability of retaliation via counterdenunciation by individual B's family against A; qA is a decreasing function of rB, the degree to which organization B is able to control the village and qB is a decreasing function of rA, such as:
qA = qV) qB = qV)
Note that these functions are symmetric and that q is convex before zero and concave after zero. The probability of retaliation via counterdenunciation depends on whether the rival organization exercises a monopoly or quasi monopoly of force.
Selective Violence
201
It is, thus, possible to think of rA and rB as the extent to which an organization can shield an individual from retaliation by the other side.
There are four possible outcomes: {Denounce, Denounce} or (D,D), {Denounce, Not Denounce} or (D,N), {Not Denounce, Not Denounce} or (N,N), and (Not Denounce, Denounce} or (N,D). The payoffs of each outcome for each player are as follows:
Player A
PA(D,D) = pA(x - qB(rA)y) + pB(qA(rB)x - y)
PA(D,N) = pA(x-qB(rA)y) PA(N,N) = o
PA(N,D) = pA(qV)x-y) Player B
PB(D,D) = pB(x - qA(rB)y) + (qB(rA)x - y) PB(D,N) = pB(qB(rA)x - y) PB(N,N) = o
PB(N,D) = pB(x-qA(rB)y)
The equilibria are the following:
1. (D,D) is an equilibrium when x > qB(rA)y and x > qA(rB)y or   x/y > Max[qB(rA), qA(rB)]
2. (N,N) is an equilibrium when x < qB(rA)y and x < qA(rB)y or x/y < MinfqV), qA(rB)]
3. (D,N) is an equilibrium when x > qB(rA)y and x < qA(rB)y or qB(rA) < x/y < qA(rB)y
4. (N,D) is an equilibrium when x < qB(rA)y and x > qA(rB)y or qA(rB) < x/y < qVty
Individual A will denounce individual B without B denouncing A (D, N) when rA is large and rB is small, that is, when organization A has a monopoly or quasi monopoly of power and organization B cannot protect its supporters. Conversely (N,D) emerges when is small and rB is large. Given that x/y < 1, the mutual nondenunciation equilibrium (N,N) obtains when both organizations are unable to protect their collaborators (qA(rB) and qB(rA) are high and both rA and rB are small); in other words, individuals will refrain from denunciation given a high probability of a very steep cost, in a logic akin to the "mutually assured destruction" of nuclear competition.37 In contrast, the mutual denunciation equilibrium (D,D) requires both organizations to simultaneously have a capacity to protect
37 The assumption for (N,N) is that that the relationship between y, x, and q is such that at r — .5, x < y * q.
202
The Logic of Violence in Civil War
Organization ATs level of control 1 .75 .5 .25 0
J_,_I_,_L
	1 1		
CO _			
CD _ ^ .		I if \	
	A: Denounce	A: Not	A: Not
IN -^-^	B: Not	B: Not !\ 1      ' \ /i     i v	B: Denounce
i—. -	i i		
0 ,25 .5 .75 1
Organization B's level of control
Probability of counterdenunciation: - q_a,byA
q_b, by B
figure 7.4. Individuals' Equilibrium Strategies
their collaborators and prevent retaliation (qA(rB) and qB(rA) are low, and both rA and rB are high); the presence of two strong quasi states at the same time and place is highly unlikely in civil war, where achieving a monopoly of power is the central goal of the rival sides.38
A simple simulation using reasonable numerical values for y and x (100 and 33, respectively) shows the distribution of equilibria across the values of r for the two villagers given the values of q (Figure 7.4). Individual A will denounce and individual B will not denounce if organization A enjoys more control compared with organization B, and vice versa. The mutual nondenunciation equilibrium emerges when the two organizations approach parity in control. Given the values of and rB in the five discrete zones of control, the equilibrium (D,N) should emerge in zones 1 and 2, the equilibrium (N,D) should emerge in zones 4 and 5, and the equilibrium (N,N) should emerge in zone 3 (Figure 7.5). Note that the absence of denunciation from the zone of parity is consistent with the high rates of defection in that zone, as discussed previously.
7.6.3. Violence
I now turn to political actors. Let the benefit of using violence for an actor be b and the cost of violence v. Actors will use violence when b > v and will refrain from using violence when b < v; b includes the consolidation of their control which is
38 Note that (D,D) may also result in nonviolence, as the two sides effectively shield their collaborators from counterdenunciation, but it is unlikely to be the dynamic at work when we see nonviolence occur, as argued previously. In Chapters 8 and 9,1 search for direct evidence on the mechanism of nonviolence in zones of parity.
Selective Violence
203
{D,N}
{D,N}
{N,N}
{N,D}
{N,D}
Control 1
figure 7.5. Denunciation
Equilibria (Individual Calculus
5
achieved by the elimination of actual defectors and (especially) the deterrence of potential defectors; v captures the potential backfire effect of violence, as those affected by it may, under some conditions, defect toward the rival actor, even though they did not intend to defect prior to the violence; it also includes the alienation effect of violence perceived as being gratuitous, even when few opportunities of defection exist; v is a function of their ability to defect (which depends on access to the competing actor) and the perception that compliance is futile and does not guarantee survival, which depends on the selectivity of violence.
Information about defectors comes either from direct monitoring, when the level of control is high, or from denunciations when control is lower; this is the case because direct monitoring entails a large administrative apparatus that is unavailable when control is challenged, that is, in contested areas. If there are no denunciations, or if denunciations are known to be systematically false, then the cost of violence will exceed its benefit (b < v), hence there will be no violence. An indicator of the overall bias of denunciations is the actors' estimate of the likelihood of defection, k(c). Where the rival actor is absent, defection is unlikely: k(c) = o; hence most denunciations are likely false.39 From the discussion of defection it follows that k(c) = o for incumbents in zone 1 and for insurgents in zone 5. Therefore, selective violence should not be observed in these zones; selective violence should neither be observed in zone 3 where the theory predicts an absence of denunciations (and, hence, of information) or a local veto to violence due to the fear of counterdenunciation. Figure 7.6 illustrates the predicted relationship between control and violence.
In short, where levels of control are high, there is no defection, no denunciation, and no violence.40 If violence is observed in zones 1 and 5, it is likely to be indiscriminate violence exercised by the rival actor. Where one actor exercises hegemonic but incomplete control (zones 2 and 4), there will be defections and denunciations; hence political actors have both an incentive and the ability to use selective violence. Finally, in areas of parity (zone 3) there will be much defection but no denunciation. Although the incentive to use violence is high, its cost will be even higher. In the absence of information, using indiscriminate violence in zone 3 could result in mass defection toward the rival actor, hence its low likelihood. Indiscriminate violence should be observed in zones 2 (by insurgents) and 4 (by incumbents), though with lower probability compared to zones 1 and 5, following the conjecture that it is inversely related to the availability of information (Chapter 6). Figure 7.7 provides a depiction of how defection, denunciation, and selective violence are predicted to vary across the five zones of control.
39 After a few iterations where denunciations are not acted upon, denunciations should cease altogether.
40 To be more precise, there will be little homicidal violence; violence is likely to take nonhomicidal forms (imprisonment) and be used to achieve goals other than the deterrence of defection (e.g., the punishment of criminals).
204 The Logic of Violence in Civil War
Control 1 2 3 4 5
figure 7.6. Selective Violence and Control
The predictions can be restated as testable hypotheses.
Hypothesis 2 The higher the level of an actor's control, the less likely it is that this actor will resort to violence, selective or indiscriminate. Therefore, no incumbent violence is likely in zone 1 and no insurgent violence is likely in zone 5.
Hypothesis 3 The lower the level of an actor's control, the less likely that this actor will resort to selective violence and the more likely that its violence, if any, will be indiscriminate. Therefore, insurgent violence in zones 1 and 2, if any, is likely be indiscriminate and incumbent violence in zones 4 and 5, if any, is likely be indiscriminate.
Hypothesis 4 Under fragmented control, violence will be exercised primarily by the political actor enjoying an advantage in terms of control: incumbents in zone 2 and insurgents in zone 4.
Hypothesis 5 Parity of control between the actors (zone 3) is likely to produce no selective violence by any of the actors.
These predictions are counterintuitive insofar as neither political actors nor individuals resort to violence where they would like it most. In contrast to Arendt's (1970:56) implication that the highest level of contestation should breed the most violence because this is precisely where "power is in jeopardy," the most contested areas are predicted to be oases of peace in the midst of violence. An additional surprising prediction is that in zone 3, high levels of simultaneous defection toward both sides coexist with low levels of denunciations.41 In other words, individuals collaborate with both sides but their collaboration excludes denunciation. The prediction about the absence of violence at the very center of the war is interesting in two ways. First, it suggests a complete contrast between symmetric and asymmetric war when it comes to violence. In the ideal type of conventional war, all violence takes place on the front line; in the ideal type of irregular war, the functional equivalent of the front line turns out to be peaceful for civilians. Second, this prediction reflects the theoretical insight about the joint production of violence: selective violence takes place only where and when the
41 Note that even more intuitive predictions such as Hypothesis 2 are far from accepted wisdom; there is an extensive literature that links authoritarian state strength (a functional equivalent of full control) with high levels of violence (e.g., Rummel 1994; Duvall and Stohl 1983:175-6).
o
figure 7.7. Predicted Pattern of Selective Violence, Defection, and Denunciation
206
The Logic of Violence in Civil War
incentives of local and supralocal actors converge. No violence takes place where political actors alone want it most or where local actors alone are most willing to provide the information necessary for its production.42 Indeed, individuals will fail to get rid of their enemies where it is safest to denounce.
A key theoretical implication is that the logic of state terror (where power is directly translated to violence) is fundamentally at odds with the logic of civil war violence. The theory also points to the endogeneity of strategies of insurgent violence to the logic of control. For example, when insurgents know that they are unlikely to have an advantage in control, they may adopt a strategy of indiscriminate terrorism - as suggested by the cases of Northern Ireland, the Basque country, and Palestine. Furthermore, the predictions are at odds with the logic of the security dilemma, which posits the emergence of violence through preemption precisely in the most contested areas (zone 3), reasoning that where vulnerability is high every defensive move is likely to be interpreted by the opposite side as an offensive one, hence prompting violence. These predictions also contradict the security version of the technology of warfare thesis, which also sees violence peaking in the most contested areas (zone 3), where the actors are most vulnerable.43 Likewise, if revenge is Hobbesian behavior, it ought to be observed primarily where authority is more decentralized, that is, zone 3. Finally, if control reflects polarization, the most polarized area should be zone 3, where the population is divided and collaborates with the two rivals. However, exactly the opposite is predicted by the theory.
This theory aims to predict variation in violence within civil wars and should be able to tell us something about cross-national variation in violence? If the theory is correct, the deadliest civil wars will be those where one or more of the following conditions obtain: indiscriminate violence is high; control shifts frequently (zones 2 and 4 dominate); areas of equal distribution of control (zone 3) are limited; and areas of complete control (zones 1 and 5) are limited. Obviously, these patterns are consistent with many types of military interaction. Where indiscriminate violence is high, Chapter 6 suggests that this may be a function of insurgents being threatening but weak, itself a function of the particular geopolitical situation (e.g., an insurgency that seeks to control the state may be seen as more threatening compared with a secessionist one, ceteris paribus). Where control shifts frequently, it may be caused by external intervention at key junctures of the conflict, which allows the side that loses to reclaim territory that was lost. The American and North Vietnamese interventions during the South Vetnamese insurgency are a case in point. An interesting implication is that not all long wars are the same: some long civil wars are stalemates with few control
42 This point resonates with Roldan's (2002:90) observation that in Colombia "violence could not succeed when it was embraced by either a handful of local leaders or the regional government alone."
43 Recall that the security version of the technology of warfare thesis served as my basis for theoretical development. This is an example of how theoretical predictions can go beyond initial assumptions.
Selective Violence
shifts, and hence low violence,44 whereas other long civil wars may be cases entailing continuous control shifts and, thus, high levels of violence. Obviously, the latter can only be sustained in the presence of high levels of foreign assistance to the rival parties. The multiplicity of competing mechanisms underlying observationally equivalent aggregate outcomes suggests the enormous pitfalls of inductive cross-national studies.
7.7. CAVEATS
Theories simplify, and this one is no exception. Its simplicity constitutes its great strength. Nevertheless, it is worth highlighting these simplifications.
To begin with, the theory exogenizes military decisions concerning the allocation of resources across space and time. The theory also assumes that individuals are good at assessing risk - in this case, they are able to assess the risk of being caught when defecting and the risk of being counterdenounced if denouncing. Yet, there is evidence from psychological experiments that people are not good at assessing risk in general (Kahneman and Tversky 1974). Nonrational concerns cloud or even distort thinking, shorten time horizons, or undermine the instrumental efficiency of behavior. Keep in mind, however, that the urge to survive can be a powerful corrective. Second, the relationship between denunciation and control depicted here is static and assumes a stable strategic environment. Individuals look around, evaluate the level of present control exercised by the two rival actors, and if the risk of retaliation is low enough, they denounce. Likewise, political actors care primarily about deterring defection, not about acquiring the peoples' goodwill for future governance. This assumes that individuals ignore the future (the likelihood that control may shift and that they may face retribution) or the past (emotions such as the desire for revenge for the violence that just took place may prove to be so overpowering as to produce a drastic discount in risk assessments).
Without discounting the role of individual expectations and emotions (which I indirectly test for in Chapter 9), it is important to note that individuals tend to underestimate the duration and fluidity of civil wars and hence overestimate their own security, especially in the war's initial stages 45 During the same period, people are inexperienced and, hence, much more likely to believe the claims of political actors about the stability of their rule. For example, a Mozambican
44 This is the case with the regions of El Salvador studied by Wood: late in the war, a stalemate emerged, and there was low violence. At the microlevel, stalemate signals the possibility for individuals to remain neutral (Wood 2003:153).
45 For example, Kerkvliet (1977:165) reports that according to former Philippino Huk rebels "peasants thought the revolt would last only a short time." A former Mau Mau recalls that: "Few realized that the struggle might last two or three years. Most were thinking in terms of a few months" (Barnett and Njama 1966:151). In El Salvador, Binford (1996:112) reports, many peasants "were waiting the situation out and hoping that things would improve. No one - army, ERP combatants, civilians - could have predicted at that time that the Salvadoran civil war would endure for another ten years." Similar observations are made by Joes (2000:73), Upton (1980:275), Escott (1978:171), and Hunt (1974:45).
208
The Logic of Violence in Civil War
man recalls that when RENAMO insurgents came to his village in 1984, they organized a meeting and claimed that "Frelimo would never come again and cause trouble" (in Nordstrom 1997:90). The mental transition to a state of civil war takes time: civilians are generally facing a completely novel situation, unlike anything they have known. Moreover, given limitations in the flow of information, many people tend to form expectations about the future based exclusively on local reality. Even when a war has lasted for a long time, people tend to overemphasize the immediate over the long-term future. In his report, a British official who visited northern Greece in 1948, remarked: "I think it would be fair to say that the Western Macedonian peasant, like most people who have to exist in a situation of uncertainty, danger and disaster, is at present living very much on a day-to-day basis, and does not look beyond the immediate future." A BBC correspondent who was kidnapped by the Greek rebels in a different region around the same time stated that the peasants "live in a state of complete uncertainty, able to look only a few weeks ahead."46 Finally, it is important to stress the geographical fragmentation caused by civil war, whose main product is the fragmentation of information. A Greek observer pointed out in 1944 that the main consequence of the breakdown of communications between the provinces was "the isolation of the inhabitants who have no idea what is going on even in the districts next to them."47 Therefore, individual decisions are often made on the basis of highly localized information and local developments. The case is easier to make for political actors, whose victory (or survival) is a precondition for the application of any political program.
Finally, I assume only two actors, yet many civil wars give rise to a multiactor context. However, the theory can speak to such contexts as well. War entails a reductionist logic, and very often local environments reduce competition to between only two actors even when the national context is multiactor. Rarely are all actors in a multiactor conflict simultaneously active in every locality of a country and where they are, alliances tend to produce bipolar conflict.
Once again, note that the theory of selective violence is not intended as a complete representation of reality, but as a sensible simplification, a theorical baseline, and a useful tool for the derivation of theorically based empirical predictions. Comparing the actual empirical variation against this baseline allows the specification of the empirical fit of the theory; furthermore, the identification of its empirical failures is particularly productive (Chapter 9).
Nevertheless, the theory can be further refined and expanded. Modeling the very complex military dynamics of the war will help endogenize them into the theory and clarify how different types of war affect violence. In turn, this will allow the derivation of robust hypotheses about the variation of violence across wars, as well as across several types of violence - from organized crime to terrorism and genocide. It is also possible to specify more complex theories that incorporate
46 "Report by Mr. D. S. L. Dodson on a tour of Western Macedonia (26-29 November 1948)," PRO, FO 371/72328/R14275; "Notes on Conversation with Mr. Kenneth Matthews on the 1st November, 1948," PRO, FO 371/72217/R1237.
47 "General Report on Conditions in Athens," PRO, FO 371/43690.
Selective Violence
heterogeneous individual preferences, community structures, and fragmented organizations; additional dimensions of violence (e.g., displacement, hostage taking, imprisonment); a more realistic specification of expectations about the future and learning from the past; and additional parameters (e.g., multiple armed actors, the role of propaganda, modern mass media, diasporas, and transnational networks). Hopefully, this book will help spark a research agenda in these directions.
7.8. CONCLUSION
This chapter has specified a theory of selective violence in civil war as a joint process, created by the actions of both political actors and civilians. The key resources around which the process is arrayed are information and violence. Political actors need information in order to be able to target selectively, to distinguish from among the sea of civilians those who are abetting the enemy. Civilians have information, which they provide through denunciation, which can be either political, or, more likely, malicious, in hopes that the violence of the political actors will be directed against those denounced. There is, significantly, a great potential for abuse in such a system, but violence need only be perceived as selective in order to avoid the pitfalls of indiscriminate violence. Denunciation will only occur in such situations in which its benefits, be they psychological or material, outweigh the predicted costs; the most significant cost would be retaliation, quite possibly in the form of a counterdenunciation by the victim or the victim's family to the other political actor. Hence, denunciation will only occur when potential denouncers perceive the political actor as able to protect them from retaliation. This process is modeled in terms of control, with the number of defectors decreasing as control increases and the number of denouncers increasing as control increases. Selective violence can only take place in those areas where control is complete enough for denouncers to denounce, but not so complete that defectors have either fled or simply ceased to be of concern to the political actor. The theory thus predicts that political actors will not use violence where they need it most: it is there that denouncers are most exposed to retaliation, and in the absence of information necessary to make violence selective, no violence is likely to take place.