13
Language Conflict and Violence: The Straw that Strengthens the Camels Back

David D.Laitin

A Tower of Babel in a single country—in which groups of people speak radically different languages—is all too often portrayed as incendiary. Selig Harrison wrote ominously about the “dangerous decades” that India would face because of its conflicts over language.1 Popular representations of language conflicts in Belgium, Quebec, and Catalonia suggest that cultural issues of this sort unleash irrational passions, leading otherwise sober people away from the realm of civic engagement. And the unjust underpinnings of language laws are often said, even by the combatants themselves, to induce violent rebellion. In late 1999, for example, the leader of the Kurdish rebels in Turkey, Abdullah Ocalan, referred to the restrictions on the Kurdish language as the principal motivating factor for the war against Turkish rule. He told the judges:

These kinds of laws give birth to rebellion and anarchy…. The most important of these is the language ban. It provokes this revolt. The way to resolve this problem is to develop Kurdish as a normal language for private conversation and broadcasting.2

That language conflict is one manifestation of a genre of uncivil politics is a principal theme in the iconic “The Integrative Revolution” by Clifford Geertz, where language was included with a set of other “primordial” attachments that were seen as threats to civil society. “When we speak of communalism in India,” Geertz wrote:

we refer to religious contrasts; when we speak of it in Malaya, we are mainly concerned with racial ones, and in the Congo with tribal ones. But the grouping under a common rubric is not simply adventitious; the



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 531
International Conflict Resolution After the Cold War 13 Language Conflict and Violence: The Straw that Strengthens the Camels Back David D.Laitin A Tower of Babel in a single country—in which groups of people speak radically different languages—is all too often portrayed as incendiary. Selig Harrison wrote ominously about the “dangerous decades” that India would face because of its conflicts over language.1 Popular representations of language conflicts in Belgium, Quebec, and Catalonia suggest that cultural issues of this sort unleash irrational passions, leading otherwise sober people away from the realm of civic engagement. And the unjust underpinnings of language laws are often said, even by the combatants themselves, to induce violent rebellion. In late 1999, for example, the leader of the Kurdish rebels in Turkey, Abdullah Ocalan, referred to the restrictions on the Kurdish language as the principal motivating factor for the war against Turkish rule. He told the judges: These kinds of laws give birth to rebellion and anarchy…. The most important of these is the language ban. It provokes this revolt. The way to resolve this problem is to develop Kurdish as a normal language for private conversation and broadcasting.2 That language conflict is one manifestation of a genre of uncivil politics is a principal theme in the iconic “The Integrative Revolution” by Clifford Geertz, where language was included with a set of other “primordial” attachments that were seen as threats to civil society. “When we speak of communalism in India,” Geertz wrote: we refer to religious contrasts; when we speak of it in Malaya, we are mainly concerned with racial ones, and in the Congo with tribal ones. But the grouping under a common rubric is not simply adventitious; the

OCR for page 531
International Conflict Resolution After the Cold War phenomena referred to are in some way similar. Regionalism has been the main theme in Indonesian disaffection, differences in custom in Moroccan. The Tamil minority in Ceylon is set off from the Sinhalese majority by religion, language, race, region, and social custom; the Shiite minority in Iraq is set off from the dominant Sunnis virtually by an intra-Islamic sectarian difference alone. Pan-national movements in Africa are largely based on race, in Kurdistan, on tribalism; in Laos, the Shan States, and Thailand, on language. Yet all these phenomena, too, are in some sense of a piece. They form a definable field of investigation.3 Language difference is perceived here as one of those symbolic cultural realms in which conflict can all too easily leave the realm of politics and become a threat to peace. In this paper I present powerful evidence to the contrary. Language conflict is not of a piece with religious or other forms of cultural conflict; it has its own particular dynamic. Furthermore, conflict over language is not a prescription for violence. In fact, under certain potentially incendiary conditions, language conflict can help to contain violence. The empirical source of my challenge to the conventional wisdom is the Minorities at Risk (MAR) database developed by Ted Gurr, which analyzes the status and conflicts of 268 politically active communal groups in 148 different countries. Among the 449 original variables included in the dataset, there are assessments of cultural, economic, and political differences between minority and dominant groups; group grievances and organizational strength; transnational support of minority goals; polity characteristics; and protest, communal violence, and rebellion.4 In the analysis that follows, rebellion of minority groups against the state is the dependent variable. Linguistic differences between minority and dominant groups as well as grievances of minority groups over state language policies are the independent variables. In the second section of this paper I review the standard theory linking modernization to language conflict, suggesting why conflicts over language issues become incendiary. In the third section I explore the MAR database. The findings are stunning: The greater the language difference between the language of the minority group and that of the dominant group, the lower is the probability of violence. Language grievances held by the minority regarding the official language of the state or the medium of instruction in state schools are not associated with group violence, but there is a weak negative relationship between language grievances and rebellion. Language grievances are strongly associated with increased levels

OCR for page 531
International Conflict Resolution After the Cold War of political protest, suggesting that the remedy for these grievances is more likely to be sought in the political realm rather than by guerrilla action. Language grievances when compounded by religious grievances (which are a reasonable predictor of rebellion) strongly and significantly reduce the magnitude of rebellion. In this sense when language and religious grievances are cumulative, language grievances lower the probability of large-scale violence. Language grievances are, therefore, straws that appear to strengthen the camel’s back. In the fourth section I formalize an “official language game” and then speculate on why the relationship between language grievance and group violence is not positive. Central mechanisms have to do with the ability of the state to commit to compromises and the inability of minority language entrepreneurs to solve collective-action problems. A theoretical sketch shows why language grievances (as opposed to, say, religious grievances) tend to redirect conflict from the military to the political/ bureaucratic realm. In the fifth section I discuss specific cases—India and Sri Lanka—to show that the statistical and theoretical analyses compel us to see oft-told national histories in new ways. Then I present new data on language policies and their associations with violent conflict and suggest the relevance of my findings for public policy. Policies that are equitable may not, the data show, have equally beneficial consequences in terms of reducing the probability of ethnic violence. To be sure, international intervention may be called for if the implementers of unfair language policies use minority protest as an invitation for all-out war against the minority group, but the unfair language policies themselves are not a threat to peace. I conclude that those interested in peace should encourage the open expression of language grievances and the subsequent political bargaining over the official language and the language of education. THE RELATIONSHIP OF LANGUAGE TO POLITICAL CONFLICT In the premodern era language was not politicized. As Ernest Gellner has masterfully demonstrated, in preindustrial times for most people the language of official state business was of no concern.5 Many states with considerable ethnic (especially linguistic) heterogeneity within their boundaries legislated official languages of state business without inducing the ire of their populations. This was no different from establishing a basic law of the state, of establishing uniform weights and measures, and other standardizing practices that Max Weber called “rationalization.”6 Furthermore, these states induced (over much longer periods) the vast majority of the population living within their territories to adopt the state language as their

OCR for page 531
International Conflict Resolution After the Cold War own, often with objections from the church but rarely with strong popular protest.7 This is part of what is today called “nation building.” It was so painless (compared to relations with other states or religious issues) that many political scientists writing in the 1960s erroneously coded early developers as having “natural” boundaries, linking nation and state. Within-state heterogeneity may have been substantial, but the language rationalization aspect of nation building was relatively benign. Take, for example, the infamous (to contemporary Catalans) Decree of the New Foundation, issued by King Philip V of Spain in 1716. Among other articles in a decree that sought to transform Spain from a decentralized kingdom to one based more on Bourbon principles, it required that all legal papers submitted to the king’s court be written in Spanish. Late nineteenth- and twentieth-century Catalan nationalists point to this decree as signaling the death of the Catalan nation. Yet historical reality reveals a quite different picture. A large database of royal court submissions in Spain from the midseventeenth through the mideighteenth centuries shows that Philip V was demanding a practice that had already become normal a quarter century earlier. In the 1660s, when most petitions brought to the king’s attention were requests for payment in recompense for quartering the king’s troops in the war in the Pyrenees against France that ended in 1659, Catalan petitioners hired notaries to translate their requests into Spanish. By the 1680s virtually all such documents were routinely produced in Spanish. It is no wonder that at the time of the New Foundation’s issuance there was hardly a murmur from Catalonia about the burdens that would be imposed on Catalans by having to communicate with the political center in Spanish.8 Although revival movements in Catalonia (as well as Basque Country and Galicia) politicized language in nineteenth- and twentieth-century Spain, it is historically remarkable how painless rationalization was; and even though nation building was never a full success in Spain,9 by the twentieth century virtually all Spanish citizens were fluent in Spanish. It is for states that consolidated rule in the modern era that language rationalization became a grave political problem. The source of the problem is in large part due to the fact that, as Gellner has highlighted, in the modern age social mobility and economic success have been dependent on literacy. Thus, clerks replaced peasants as the backbone of modern economies in the industrial age. Furthermore, as states got into the business of providing education, the language of state business became a much broader concern for far more people than when states were not providing such services to individual citizens. Under modern conditions, people have become quite sensitive to the language of state business, and if it is not their own they feel alienated from the state. They feel as well a

OCR for page 531
International Conflict Resolution After the Cold War sense of unfair competition for jobs that are more easily garnered by those whose mother tongue is the state language. Indeed, the classic case of the unraveling of the Habsburg Empire, where peasants from non-German-speaking areas who became urban migrants were most receptive to pleas for the official recognition of their languages, fits this theory to a T.10 Postcolonial states that emerged after World War II, committed to the provision of public education and social welfare, were heavily constrained from following the path of Philip V and other earlier rationalizers. Newly elected political leaders were handed bureaucracies with a vested interest in continued reliance on colonial languages, as fluency in these languages differentiated the high-paid civil servants from their poorly paid brethren in the countryside. Furthermore, these same national leaders were held under suspicion by leaders from regions in which distinct languages were spoken. To impose one indigenous language on all groups would surely threaten the incumbency of any would-be rationalizer. Yet the goals of many postcolonial leaders included superseding the colonial language with an indigenous one. This difficult problem of choosing an official language (used for public administration and as a medium of instruction in schools), under conditions in which greater access to the official language translates into higher prospects for social mobility, has led many analysts to link language conflict with the potentiality of inducing ethnic violence. Nevertheless, their blunt theory is unable to make specific predictions about levels or types of conflict. THE ROUTE FROM ETHNIC CONFLICT TO ETHNIC VIOLENCE As I indicated in the Introduction, the standard literature on ethnic conflict often conflates all forms of ethnic contestation as a form of zero-sum intractable conflict, all with equally high potentialities for engendering violence. The leading theories provide a basis for understanding why language gets politicized in the modern era, but links to violent conflict are weakly theorized. In this section, relying on MAR data (supplemented with new variables), I show that language conflict does not translate inexorably into a higher probability of ethnic violence. The dependent variable for this section is REBELLION.11 The scale goes from 0 (no rebellion) through 4 (small-scale guerrilla activity) and up to 7 (protracted civil war). The question I ask is whether language difference, language grievance, or language grievance in association with other factors helps explain the values on REBELLION. To address this question I provide evidence from analysis of the MAR database that supports the four findings announced in the Introduction.

OCR for page 531
International Conflict Resolution After the Cold War Language Difference and Violence The independent variable describing the level of language difference is LANGSIM. Here I consider the hypothesis that linguistic distance between people living in the same country is a source of tension and that therefore people with different languages cannot easily live together in the same political unit. The MAR database lends some support to this thesis, but its coding on linguistic distance is invalid.12 Recognizing the failures of the MAR indicator to assess linguistic distance, James D.Fearon and I took the world classification of languages, produced by Ethnologue,13 a society of linguists interested in producing versions of the Bible in all languages of the world. Ethnologue linguists rely on linguistic trees, classifying languages by structure, with branch points for language family (e.g., Indo-European from Afro-Asiatic), language groups, and even sub-dialects. LANGSIM has three values depending on whether the language of the minority is the same as the language of the dominant group, the language is different but of the same family of languages (the initial branching point in the Ethnologue codings), or the language is of a different family.14 Linguistic difference alone between the dominant and minority groups in a country is not a predictor of intergroup violence. If we correlate LANGSIM with REBELLION, in fact, the trend is opposite what Gurr’s data show and what would be predicted from a theory that cultural difference promotes conflict. In a bivariate relationship between rebellion and linguistic similarity, the correlation is positive (.1359, significant at p=.03). Table 13.1 illustrates this relationship through the comparison of mean scores on REBELLION (here the maximum rebellion scores from 1945 through 1995). In the table we see that the mean score for rebellion is lowest when the groups are from a different language family and highest when the two groups share the same language. Thus, without introducing controls, the data show that greater linguistic similarity raises the probability of violence. Examination of the list of cases in each of LANGSIM’s categories helps show why an important cultural difference such as language does TABLE 13.1 Language Similarity and Rebellion (1945–1995): Comparison of Mean Scores on REBELLION LANGSIM Mean Value of Rebellion Number of Cases Entire population 2.49 244 1. Different family 2.06 111 2. Same family 2.62 93 3. Same language 3.40 40

OCR for page 531
International Conflict Resolution After the Cold War not provoke violent group conflicts. On the one hand, there are many cases where there are vast differences in language but where the conditions do not permit large-scale rebellion. One type includes postindustrial settlers who moved into urban areas and had no territorial base in which to mobilize for military action against the state. These groups differed greatly linguistically from the dominant groups that controlled the state. Another category contains groups living in states ruled by “settler” populations whose language is different from any of the autochthonous groups—their state-building activities achieved success in earlier eras, and they are less likely to face ethnic rebellions in the post-1945 period. Third, former slave groups, many of them classified as having Creole languages, are of a different language family from the dominant groups of their societies, yet they have not been in a position to radically oppose the state in the post-World War II period. Finally, some nomadic groups (the Romani) may well be subject to pogroms but do not have the resources to challenge the state through violent action. They too differ considerably linguistically from the dominant groups in the societies in which they live. On the other hand, there are many indigenous populations (defined by Gurr as “conquered descendants of original inhabitants of a region who typically live in peripheral regions, practice subsistence agriculture or herding, and have cultures sharply distinct from dominant groups”)15 who speak languages in close proximity to their conquerors yet harbor long-standing grievances and have a rural base to rebel. Here, despite language similarity, we see a breeding ground for violent confrontation. Once we control for factors such as urban versus rural base of the minority population, language similarity will be shown to have no explanatory power. But presenting the bivariate correlations helps to undermine the oft-expressed opinion that cultural differences in and of themselves are prescriptions for violent confrontations, especially in an age of identity politics. Language Grievances and Violence The Gurr dataset has two variables measuring language grievance, each measured for two-year time periods (1990–1991, 1992–1993, 1994– 1995). The first variable measures the level of demands by the minority to have its language given greater official status. The second variable measures the level of demands by the minority to have its language used as a medium of instruction in state schools. I constructed a composite variable, MAXLANG, which is the maximum value of grievance on either of the variables in any of the time periods. The bivariate relationship between MAXLANG and REBELLION is –.055. Not only is the relationship

OCR for page 531
International Conflict Resolution After the Cold War not significant in a positive direction, but the sign is the opposite of what the Gellnerian approach to modernization would have led us to expect.16 Perhaps language grievances alone are not a causal factor explaining group violence but in conjunction with other factors can raise its probability. To find out it is necessary to go beyond bivariate correlations and comparisons of means and examine these relationships through regression analysis. In the multiple regressions that follow, using a cross-sectional study of rebellion that I completed in collaboration with James Fearon,17 I enter three control variables that have the greatest predictive power: namely, the log of gross domestic product (GDP) taken from 1960, the rate of GDP growth from 1960 to 1980, and a dummy variable I call RURBASE that indicates whether the group has had a long-term rural settlement in a specific region of the country.18 With these controls the coefficient of MAXLANG is not always negative when regressed against REBELLION. But, as we will see, as I analyze Table 13.2, language grievances do not add to the straws on the camel’s back, fostering violence. The intuition behind the camel’s back approach to ethnic violence is that language grievances alone are not a sufficient cause for rebellion but in conjunction with other factors can add to an atmosphere that induces rebellion. The specification listed in Table 13.2 examines this intuition. Besides the control variables, it includes a variable MAXRELGR, which is the maximum score of religious grievances expressed by the minority group in the first half-decade of the 1990s. It also includes a value taken from the polity database on the degree to which the country was democratic in 1989, called here DEMOCRACY. Finally, it includes an interaction term called LGxRG, which is the product of MAXRELGR and MAXLANG.19 TABLE 13.2 Rebellion and Cultural Grievances—Dependent Variable: REBELLION; Linear Regression: OLS Independent Variable Coefficient (B) Standard Error MAXLANG .202 .167 MAXRELGR** .418 .152 LGxRG* –.200 .101 DEMOCRACY 1989 .062 .050 GDPCHANGE 1960–1980** –.874 .249 RURBASE** 1.277 .396 LANGSIM .125 .227 LOGGDP60** –.682 .221 Constant** 6.249 1.63 R2=.27105. *Significant at p<.05; **Significant at p<.01

OCR for page 531
International Conflict Resolution After the Cold War Several important relationships need to be elaborated from the analysis of Table 13.2. First, although the positive relationship of LANGSIM and REBELLION remains as in the bivariate analysis, the relationship is not significant at all once controls are added. Second, the relationship of MAXLANG and REBELLION is no longer a negative one, but that relationship as well is not statistically significant. Third, the relationship between MAXRELGR and REBELLION is significantly positive, suggesting that a one-point jump in MAXRELGR (on a four-point scale) raises the likely rebellion score by nearly half a point. The statistical effects of language grievance on rebellion are thus quite distinct from those of religious grievance on rebellion, a point I return to later. Fourth, the level of democracy has no significant relationship to the scale of REBELLION. I elaborate on this point later as well. Fifth, and most significant for challenging the straw and camel’s back intuition, is the strong negative relationship between the LGxRG interaction term and REBELLION. What this captures is a slope for MAXLANG that is marginally positive (.21) when there are no religious grievances at all (MAXRELGR=0). Yet the slope of MAXLANG changes sign (to –.4) when there are high levels of religious grievance (MAXRELGR=3).20 One way to see this relationship is to compare the REBELLION mean score of 2.58 when MAXRELGR=3 and MAXLANG=0 to the mean score of 1.05 for REBELLION when MAXRELGR stays at 3 but MAXLANG=3. Adding a powerful language grievance to a powerful religious grievance thereby reduces the REBELLION score by 1.53, a very powerful effect indeed. Suppose (in a stylized portrait of Sudan) that a Muslim-dominated country where Arabic is the official language dominates over a Christian region whose people speak a variety of languages, but none of them have Arabic as their mother tongue. Further suppose that the majority imposes Shari’a (i.e., Muslim law) on the minority, activating regional entrepreneurs to use the churches as recruiting grounds for a rebellion, overcoming the logic of collective inaction. Finally, suppose that the majority adds fuel to the fire by imposing Arabic as the sole official language for schools throughout the country. (Because the dominant region already was relying on Arabic, there is no problem of implementation.) Now not only the priests but the schoolteachers are mobilized. But this additional mobilized group need not add fuel to the revolutionary fire, as Table 13.2 counterintuitively demonstrated. First, there will be an incentive for some southerners to learn Arabic and get prized jobs, without a school hierarchy policing their linguistic defection. Second, the aggrieved schoolteachers face a difficult choice: whether to fight in the guerrilla camps (with the anti-Shari’a forces) on the religious front or in the state bureaucracies on the linguistic front. To the extent that they can win delays and concessions

OCR for page 531
International Conflict Resolution After the Cold War on the latter front, the oppressive language laws may take some potential rebels out of the rebellion. Not all cases of high religious and high language grievances are peaceful (Catholics in Northern Ireland, Sri Lankan Tamils in Sri Lanka, and Serbs in Croatia are members of this category), but most are quite peaceful (Russians in Uzbekistan, Indian Tamils in Sri Lanka, Germans in Kazakhstan, Malays in Singapore). If a score on REBELLION that is greater than three is taken as an indicator of large-scale ethnic war, under conditions of high language and religious grievances in only three out of 22 cases (14 percent) were there large-scale wars. For the entire sample of 267 cases, 65 (24 percent) were experiencing a high level of rebellion in the 1990s. The conjunction of both language and religious grievances yields a far lower probability of large-scale ethnic war than language alone. To sum up, although MAXLANG in isolation has no significant relationship to REBELLION, as an interaction term with MAXRELGR, language grievances reduce the potential for violence. Language Grievances and Political Protest Language grievances alone do not increase the probability of violence (but they appear to have some ameliorative effects), but this does not mean that grievances over language policy lead to political quiescence. In fact, the reverse is true. The MAR database has a separate six-point scale for level of political protest, going from “none reported” to “demonstrations of greater than 100,000 people.” Table 13.3 contains the result of a regression equation that is precisely the same as in Table 13.2, except that here the dependent variable is PROTEST, the maximum degree of protest in the years 1990–1995. Here MAXLANG is powerfully and positively TABLE 13.3 Protest and Cultural Grievances—Dependent Variable: PROTEST; Linear Regression: OLS Independent Variable Coefficient (B) Standard Error MAXLANG* .284 .121 MAXRELGR .061 .110 LGXRG –.059 .074 DEMOCRACY 1989 .055 .036 GDPCHANGE 1960–1980 .144 .181 RURBASE** .924 .288 LANGSIM –.155 .165 LOGGDP60 .130 .161 Constant .312 1.19 R2=.177. *Significant at p<.05; **Significant at p<.01

OCR for page 531
International Conflict Resolution After the Cold War related to PROTEST, but this is not the case for MAXRELGR nor the interaction term of language and religious grievances (LGxRG). In a point I develop in the following section, language grievances seem to draw angry citizens into the realm of political protest but not into guerrilla armies. Turning the Question on Its Head: Explaining the Modulating Effects of Linguistic Conflict How to interpret the data so far? Language issues themselves do not cause group violence. In fact, language conflicts, under conditions where religious grievances are powerful, are associated with lower levels of ethnic violence than under conditions where religious grievances are weak. We might therefore turn the usual question on its head: Why does language conflict moderate ethnic violence? HOW CAN A STRAW STRENGTHEN THE CAMEL’S BACK? It is a stretch (not justified by the data) to claim that language grievances reduce violence; but since the bivariate relationship with REBELLION is negative (and in interaction with religious grievances it is significantly negative when regressed on REBELLION), it is useful to ask theoretically why the effect of language grievances on ethnic rebellion is not strongly positive and what ameliorative influence language grievance might have on ethnic relations. In this section I present a stylized model of an “official language” game. Traveling down its strategic steps will suggest three ameliorative mechanisms that reduce the probability of official language policy turning into violent intergroup conflict. The first centers on the potential subversion of the oppressive language laws by educated members of the dominant group, which makes it more difficult for a state to implement a new official language. The second centers on the general bureaucratic problem, even if there is substantial support from government and business elites in the dominant group, of changing language norms. These problems ironically enable the government to make credible commitments in bargaining with the discriminated-against minority language groups. The third centers on the problem of collective action that is faced when language entrepreneurs of the minority language groups seek to recruit warriors to fight on their behalf. Consider the stylized “official language game” represented in Figure 13.1. Suppose a popular postcolonial government is being pressed by its ethnic constituency to pass language laws in favor of the dominant national group. It can either accept the status quo (say, continued use of the

OCR for page 531
International Conflict Resolution After the Cold War make rules for the use of Sinhalese, and to make provisions for both non-Sinhalese and Sinhalese to use it in official domains, created a vast administrative task. The Department of Official Languages was beset with pressures from a variety of interests, and it required the work of many Tamils to write translations of official terms. Meanwhile, setting standards for Sinhala writing competence too high could backfire, as it could have jeopardized the tenure of many Sinhalese. Language politics, if implementation were to occur, moved into the realm of pluralistic give and take rather than symbolic pronouncement. Furthermore, because of its inability to implement the Sinhala-Only Act in 24 hours, the government found itself able to commit to Tamils in the bureaucracy that they would not be out in the streets jobless by decree. The regulations for the fulfillment of Sinhala-Only gave assurances to Tamils in the civil service that their jobs and promotions were secure. This helped defuse the anxieties and anger of the Tamil elite. It might also help explain why there was never an alliance of northern autonomists in Jaffna with Tamil professionals in Colombo. The latter group preferred cosmopolitan life in Colombo or emigration to the West over migration to the Northern Province to give intellectual leadership to Tamil Eelam. In 1988, with the Indian government’s intervention, a North Eastern Provincial Government was constituted, and during its honeymoon period it was able to recruit leading Tamil civil servants. But there is no indication in my sources that Tamil officials from the south were moving to Jaffna.45 It is not possible in this context to prove a counterfactual, but it seems at least plausible to argue that with the politicized Sinhala Buddhists in alliance with the rural Sinhalese elites there would easily have been induced pogroms against both Sri Lankan and Indian Tamils as both religious and economic threats. The populating of the Eastern Province with Sinhalese peasants going back to the colonial period continued to threaten the Tamils’ hope for a majority in both the north and east of the island. Reduction in homeland space has been conducive to a national separatist movement, with or without a language issue. Meanwhile, in the arena of language politics, there was a considerable amount of political interaction between Sinhalese and Tamils, with common interests in the development of a reasonable language policy that served both communities’ interests. If the Tamils were willing to negotiate peacefully over language, we should ask, what explains the systematic refusal by Sinhalese to abjure violence in the 1970s and negotiate a fair language policy? This refusal to negotiate, in Tambiah’s judgment, was a principal source of the violence.46 The standard answer to this question is that the Sinhalese were themselves divided into two parties, representing different dynastic families. Each sought dominance by promising the same electoral base the job

OCR for page 531
International Conflict Resolution After the Cold War opportunities that would become theirs should the Tamils be excluded from the white-collar job market. Either party in power that sought reconciliation was challenged by the other party, waiting in the wings. To establish their anti-Tamil bona fides, leaders of both parties underwrote young thugs to victimize innocent Tamils. Here we can say that the Sinhalese language policy of 1956 was an instrument of oppression. It is also the case that the job and university quotas based on language increased the number of disaffected youth who had no better alternatives than to join the guerrilla forces. But the economic downturn in the postcolonial period was a cause of the language laws in the first place and was sufficient to marginalize young men (both Sinhalese and Tamils) in the modern economy. More important, however, is that the language policy in itself drew many Tamils into the political arena, a point not recognized by all too many analysts looking for the mechanisms by which language policy pushed Sri Lanka into ethnic war. Although I concede that the language issue sparked some of the early rioting by Tamils (and the defacing of some Sinhalese signs), a comparativist’s perspective leads me to hold that over the past 40 years its impact has been more ameliorative than exacerbating. The critical piece of evidence, overlooked by several area experts, is that violence was not initiated by those who were aggrieved by an unfair language policy but rather by the group in whose interests the law was passed. The grievances themselves cannot therefore be held to motivate Sri Lankan interethnic violence. A closer look shows that many aggrieved Tamils were drawn by the language policy to bureaucratic insurgency and political protest, peacefully. Here is an example where even the best and most informed case studies may be wrong on the very sign of an important independent variable. POLICY ANALYSIS The thesis of this paper, even with the historical reinterpretations of India and Sri Lanka, is so counterintuitive as to leave any reader with a sense of deep skepticism. Many readers will have cases in mind where language decrees fomented popular demonstrations, which brought in the police and spilled over into violence. The Russification decrees in Poland in the 1870s, the promotion of Afrikaans in South Africa’s township schools in 1976, and the law on language in Moldova in 1989 are all associated with riots and revolution by the oppressed. I submit, however, that a careful reconstruction of these cases, similar to what I have done with Sri Lanka, will give support to my thesis. In Poland the revolution against Tsarist Russia preceded the language decrees rather than resulted from them; in South Africa the riots in Soweto brought accommodation on the linguistic front, and the subsequent war was fueled by the denial of

OCR for page 531
International Conflict Resolution After the Cold War political rights to Africans; and in Moldova the law on language that was said to provoke the rebellion in Transdniester was indistinguishable from the language laws in the 13 other non-Russian republics, with none of the others bringing the Russian-speaking populations into armed conflict. The comparative speculations that have informed this section, however counterintuitive, should be sufficient to undermine the claim that language grievances are a spark that can all too easily set off incendiary ethnic wars. Good policy cannot ignore this finding. In this section I analyze official language policies of states to see if there is a clue as to which policies are associated with the lowest levels of violence. To do so I coded all countries in the Gurr dataset on the basis of their language policies, in a variable I call LANGREGIME, short for Language Regime. There are five values for this variable, and they are characterized in the left column of Table 13.4. Examples of “1,” where there is a single official language corresponding to the ethnic majority or dominant settler group, include English in the United States, German in Austria, Hungarian in Hungary, Malay in Malaysia, and Spanish in Argentina. Examples of “2,” where there is a single (or sometimes a second) official language corresponding to a language not associated with a major ethnic group in the country, include Bahasa in Indonesia, French and English in Cameroon, and English in Kenya. This language is usually referred to as a lingua franca. Examples of “3,” where both an indigenous language and a nonindigenous language are official, include English, Hindi and state languages in India, and Hebrew, English, and Arabic in Israel. The value “3” TABLE 13.4 Language Regimes and Rebellion—Mean Scores for Rebellion (number of cases) Language Regime All Cases Where RURBASE =1 Where RURBASE =1 and Independent After 1944 All cases 1.68 (n=218) 2.10 (n=171) 2.51 (n=95) 1. Rationalization 1.32 (n=92) 1.74 (n=68) 2.92 (n=26) 2. Official lingua franca 1.60 (n=45) 1.82 (n=39) 1.86 (n=36) 3. 3±1 language formula 2.88 (n=41) 3.14 (n=37) 2.78 (n=32) 4. More than one domestic official language with no lingua franca 1.80 (n=10) 2.25 (n=8) NA (n=0) 5. Rationalization + recognition of regional language(s) 1.20 (n=30) 1.89 (n=19) 6.0 (n=1) NA=not applicable

OCR for page 531
International Conflict Resolution After the Cold War corresponds broadly to the 3±1 model discussed in this paper. Examples of “4,” where there is more than one official language corresponding to the leading ethnic groups of the country, include German, French, Italian, and Rhaeto-Romance in Switzerland and Pashto and Persian in Afghanistan. Examples of “5,” where there is but one official language but with recognized regional languages, include cases such as Spanish with Catalan, Basque, and Galician in Spain and Arabic with Kurdish in Iraq. I then compared in Table 13.4 the mean REBELLION scores for each policy, under all cases and under cases where RURBASE=1. It is clear that the 3±1 model, however attractive it is from a welfare or identity point of view, is associated with higher levels of violence. This result should be taken with skepticism. For one thing, groups that have been successfully incorporated into states under the 3±1 formula (e.g., the Tamils, Gujaratis, Kannada speakers in India; the Romansch speakers in Switzerland) are not in the dataset, as they were not considered “at risk.”47 Second, minority violence could well have impelled states to accept such language regimes, which would explain the association but have the causal arrow in the wrong direction. Third, when a dummy for 3±1 is constructed and entered into the equation displayed in Table 13.2, although the coefficient is positive, it is not anywhere close to being significant.48 Nonetheless, it would be foolhardy from the point of view of policy prescription to advertise this policy as a model for other states. Meanwhile, rationalization with concessions to minorities for regional languages (a policy increasingly apparent in Western democracies) has the lowest mean score for REBELLION in the entire sample, and perhaps this is a clue as to how best to handle language grievances when they become heavily politicized (although this is what Sri Lanka did eventually but unsuccessfully). Finally, I examined only those minorities with a rural base and in accordance with Gellner’s theory only from countries that entered into the community of states after 1944. These are the states most subject to problems establishing an official language, many having relied on the colonial language as the official language of modern government. From this set the mean score for REBELLION is 2.51. But for those with LANGREGIME=3 (the set with India as paradigm), the mean score is higher at 2.78; those with LANGREGIME=1 (rationalization) had a mean score of 2.68; meanwhile, for those with LANGREGIME=2 (the set with Kenya as paradigm), the mean score is lowest at 1.86. (The Sri Lankan Tamils are the only case where LANGREGIME=5.) This last column suggests that a single neutral lingua franca has been more peaceful than the indigenous-promoting multilingual schemes and more peaceful as well than the rationalization policies. A dummy for cases with a neutral lingua franca added to the specification for Table 13.2 has a negative

OCR for page 531
International Conflict Resolution After the Cold War coefficient (i.e., this language regime lowers the expected value for rebellion), but it is not significant.49 More interesting still is the comparison of the 32 cases where LANGREGIME=3 (i.e., the 3±1 model) and examining whether the minority group living under such a language regime has its language recognized in the official language formula. The result is that groups whose languages are recognized (n=17) have a mean REBELLION score of 3.6; those whose languages are not recognized (n=15) have a mean score of 3.5. Thus, there is slightly more violence associated with recognition of a group’s language in a multilingual scheme than in keeping the language group out of the scheme altogether. In those 32 cases only 7.3 percent of the groups had any recorded language grievance. Meanwhile, in the cases where there was acceptance of a lingua franca (i.e., where LANGREGIME =2), 13.3 percent of the groups articulated language grievances. This means that bringing groups into the official language formula is not a prescription for peaceful ethnic relations and that fomenting language grievances (by ignoring a group’s plea for official recognition of its language) is not a prescription for violence.50 SUMMARY AND CONCLUSION The purpose of this paper was to explore the relationship between language-based political conflict and ethnic violence. The standard view in the literature is that language issues, especially when a nationalizing elite in control of a state in the modern period seeks to impose its language as the principal means of state business and social mobility, exacerbate ethnic tensions, with potential for ethnic violence. An analysis of the Gurr dataset does not support this standard view. Instead, the data support a view that language distance (between groups) and language grievances (by minorities) play no causal role in the emergence of rebellion. In fact, there is some support for the counter hypothesis, namely, that language grievances when expressed under conditions of religious tension tend to ameliorate violence. Theoretical considerations—having to do with the bureaucratization of language conflict, the ability of the state to make language commitments, and the difficult collective-action problems faced by minority-language entrepreneurs—help us make sense of the statistical findings. With the surprising statistical results and the new theoretical considerations brought to mind, a reexamination of two well-known cases of language politics—in India and Sri Lanka—gives added support to the data and the theory seeking to make sense of those data. A necessary condition for violence to be averted, it should be emphasized, is that the state must be willing to bargain over demands articulated by language activists in society. In India the central government’s

OCR for page 531
International Conflict Resolution After the Cold War unwillingness to provoke the Telugu speakers into unyielding opposition is key to the peaceful resolution of that conflict. Violent confrontations between Marathis and Gujaratis in Bombay, and between Nagas and Assamese in the northeast, were also defused in part because the central government went to the bargaining table and was willing to make concessions. In Sri Lanka the government’s passage of the Tamil Language Acts of 1966, as well as the constitutional guarantees of 1972 and 1978, helped limit the range of the civil war mostly to the northeast of the island. My argument is not that even with government intransigence language conflicts will reduce the likelihood of large-scale violence; rather I believe that language conflicts allow for extensive and successful bargaining without making it seem as if either side is a traitor to its group’s interests. But absent utter intransigence by government authorities, the hypothesized route from discriminatory language policy to ethnic civil war and state breakdown is not supported by the comparative data. Willingness to bargain with minorities does not imply democracy. The MAR dataset includes 40 cases where DEMOCRACY was of low quality, MAXLANG was substantial, yet REBELLION did not reach a critical threshold. These include the Indian Tamils in Sri Lanka, the Taiwanese in Taiwan, the Berbers in Algeria, and the Roma in Croatia. There are also two cases with high levels of DEMOCRACY where MAXLANG was also substantial, yet REBELLION has been significant.51 These include the Bodos in India and the Kurds in Turkey. Democratic institutions are neither necessary nor sufficient to ensure official-language bargaining. In fact, statistical evidence does not allow us to reject the hypothesis that democratic institutions play no role in ameliorating or exacerbating language-based conflict. Even more surprising, there is no evidence that sensitive language policies are nostrums for ethnic arousals. For countries that received independence after World War II, bureaucracies usually operated with the colonial language. Leaders of many groups expressed firm desires to have their languages recognized as official and relied on as media of instruction in state schools. Yet the sensitive granting of these desires did not lower the likelihood of violent rebellion, and the failure to do so for some groups did not raise that likelihood. Grievances over language were expressed under all forms of policy; they were expressed regardless of whether the group’s language was represented in multilingual official formulas. Yet when those grievances were expressed, the trend was not to enhance the chances of violence but (if anything) to reduce them. The principal policy recommendation of this paper is that governments should be encouraged to allow groups to express language demands on the political stage. In fact, the Indian case shows that the payoff for peace is enhanced if the government rewards groups with a variety of

OCR for page 531
International Conflict Resolution After the Cold War concessions (e.g., jobs) for making language-based claims on the state. Essential as well is that the government must assure the minority groups that the dominant group will not take minority expression of language grievances as a pretext for genocide against them. In sum, governments must be willing to engage in political and bureaucratic conflicts over language issues. Meanwhile, the substance of any particular language framework has no bearing on peaceful outcomes. This recommendation has special relevance for the newly independent states of the former Soviet Union. In 1989 all of these republics passed new language laws that reversed the tides, giving greater official status to the republican language vis-à-vis Russian. In some republics, after independence in 1991, these laws became more stringent still. In 1999 Anatol Lieven of the International Institute for Strategic Studies excoriated the Latvian parliament for its measures that, he claims, “would have virtually banished the Russian language from public life.” Holding up the specter of Kosovo, he argued that because the proposed law embodied “gratuitous provocations” the West had a “special duty to help prevent such legislation.” The argument of this paper is not that the Latvian legislation was fair minded. It was not. Rather, the argument is that the “special provocations” were more likely to drive the Russian-speaking population in Latvia to political action rather than guerrilla action. The international gendarmerie must distinguish those policies that are merely unfair from actions that plant the seeds of civil war. The analysis in this paper helps make that distinction.52 Policy makers—for states that face language conflict and for states that provide support for an international gendarmerie when ethnic conflict spills over into ethnic violence—should be made aware that language conflict, even if it is not threatening to states or democratic regimes, can be extremely dangerous for incumbents. Leaders of disaffected language groups have the skills and intellectual resources to mobilize constituencies that are outraged by current language policies. Incumbents on the unpopular side of a language conflict can be ruthlessly thrown out of office. But this does not mean that language conflict is dangerous to democratic governance or civil peace. In fact, language conflict, when not directly and brutally repressed by fearful incumbents, tends to be fought out in translation committees, school boards, and bureaucracies. If language entrepreneurs are given the chance to mobilize their constituencies, incumbents might lose their positions, but partisans of other languages are not likely to lose their lives. To be sure, people have died while participating in language-based riots. But the data in this paper demonstrate that it is far more likely that language grievances will result in political protests than military action. Politics is the realm where intense conflicts can be resolved peacefully.

OCR for page 531
International Conflict Resolution After the Cold War Policy analysts need to recognize that the politicization of language issues is not a danger signal for ethnic war. In fact, the politicization of language ameliorates the violent potential of religion-based conflict. To understand that the politicization of language issues may be the straw that strengthens the camel’s back would be to take an important step in understanding language, politics, and ethnic violence. ACKNOWLEDGMENTS Much of the work that forms the basis of this paper was done in collaboration with James D.Fearon, who helped me think through the implications of the data presented herein. I would also like to thank Kanchan Chandra and Bruce Bueno de Mesquita for comments on a draft and the “Chicago on the Hudson” seminar (John Roemer, Adam Przeworski, Jon Elster, Steven Lukes, John Ferejohn, Brian Barry, and Stephen Holmes) for showing me how to better specify my argument. Finally, Paul Stern coherently conveyed the often contradictory advice given me by the review committee organized by the National Research Council. This chapter also appeared in Archive Européenes de Sociologie 2000 (41):97–137. NOTES 1   Selig S.Harrison, ed. (1957), The Most Dangerous Decades: An Introduction to the Comparative Study of Language Policy in Multi-lingual States (New York: Language and Communication Research Center, Columbia University). 2   The Ocalan quote is from a report in The New York Times, June 24, 1999, by Stephen Kinzer. 3   Clifford Geertz (1973), “The Integrative Revolution” in The Interpretation of Cultures (New York: Basic Books), pp. 256–257. 4   For a full description of the database, see Ted R.Gurr (1993), Minorities at Risk (Washington, D.C.: United States Institute of Peace). There are several worrisome methodological pitfalls in the construction and coding of this database, discussed in James Fearon and David Laitin’s proposal to remedy them, funded by the National Science Foundation, grant no. 9876530, “Minorities at Risk Database and Explaining Ethnic Violence.” The proposed changes will surely have some impact on the relationships discussed in this paper. The findings herein can therefore only be considered preliminary. 5   Ernest Gellner (1983), Nations and Nationalism (Ithaca, N.Y.: Cornell University Press). 6   Max Weber (1968), Economy and Society (Berkeley: University of California Press), pp. 71, 655, 809–838, and 1108 for discussions of different forms of rationalization. 7   Eugen Weber (1976), Peasants into Frenchmen (Stanford, Calif.: Stanford University Press); Abram de Swaan (1988), In Care of the State (New York: Oxford University Press), chap. 3. 8   This is a summary of David D.Laitin et al. (1994), “Language and the Construction of States: The Case of Catalonia in Spain,” Politics and Society, vol. 22, no. 1 (March), pp. 3–30. 9   Juan Linz (1974), “Politics in a Multilingual Society with a Dominant World Lan

OCR for page 531
International Conflict Resolution After the Cold War     guage,” in Les états multilingues: problèmes et solutions, J.G.Savard and R.Vegneault, eds. (Quebec: Presses de l’Université Laval), pp. 367–444. 10   The Habsburg case is an archetype for the foundational figures in contemporary theories of nationalism, especially those who lived in it at the time of its dissolution: Hans Kohn, Karl Deutsch, Eric Hobsbawm, and Ernest Gellner. 11   REBELLION, unless otherwise specified, is the variable “rebel90x” in the MAR dataset, reflecting a value of the group’s rebellion against the state for the years 1990–1995. 12   See David D.Laitin (2000), “What Is a Language Community?,” American Journal of Political Science, vol. 44, no. 1, pp. 142–155. 13   Barbara F.Grimes, ed. (1996), Ethnologue: Languages of the World, 13th ed. (Dallas: Summer Institute of Linguistics). 14   For a discussion of the methodological problems in using a measure of language distance, see Laitin, What Is a Language Community? 15   Gurr, Minorities at Risk, p. 18. 16   Given the limitations of the MAR dataset, where language grievances are coded only for the 1990s and the latest scores for rebellion are also in the 1990s, I cannot now rule out the interpretation of the forthcoming results as rebellion causing a reduction in language grievances. This is rather implausible, and subsequent updating of the database will allow me to assure myself that the causal arrows as I interpret them are correct. 17   James D.Fearon and David D.Laitin (1999), “Weak States, Rough Terrain, and Large-Scale Ethnic Violence Since 1945,” paper prepared for the Annual Meeting of the American Political Science Association, Atlanta, Ga. 18   Users of the MAR database will want to know: there is no rural base for rebellion (RURBASE=0) if the group is primarily urban (REGS=1), or if the group is widely dispersed (REG6=1), if the group did not migrate to the country until the twentieth century (TRADITN=4 or 5), or if the group members (even if it was primarily rural) were the descendants of slaves or are travelers (Romani). Meanwhile, the group was considered to have a rural base (RURBASE=1) if the minority group could trace its origins in the country to the period before state formation (TRADITN=1) or if the group had at least a majority concentrated in one region of the state (GROUPCON=2 or 3). 19   Other interaction terms—with race and class in particular—would be feasible elaborations of this analysis. Getting an objective measure of race prevents an exploration of its dynamic. I lack data on the class composition of the ethnic groups and cannot explore its impact here. 20   The mirror is also true: religious grievances reduce the rebellious potential of language grievances, but this effect is far less strong statistically than the one reported in the text. 21   See David Laitin (1989), “Language Policy and Political Strategy in India,” Policy Sciences, vol. 22, p. 426. 22   Under such conditions violence is more likely, as James D.Fearon argues in “Ethnic War as a Commitment Problem,” paper presented at the Annual Meeting of the American Political Science Association, New York, 1994. 23   This is the source of humor in Woody Allen’s Bananas, when the leader of a Latin American guerrilla army, at the moment of victory, with cigar in mouth, announces that from that point on Swedish will be the sole language of all communication in the island nation. 24   My research career has been devoted to this dilemma. I focus on the identity aspects of language in David D.Laitin (1977), Politics, Language and Thought (Chicago: University of Chicago Press). I focus on the strategic rationality of defection in David D.Laitin (1988), “Language Games,” Comparative Politics, vol. 20, pp. 289–302. I focus on the “Janus-facedness” of culture, which has both an identity and a strategic component, in David D.

OCR for page 531
International Conflict Resolution After the Cold War     Laitin (1986), Hegemony and Culture (Chicago: University of Chicago Press). In this paper the identity aspect of culture plays only a bit part because the dependent variable is “violence” (where strategic action is more important) rather than “assimilation” (where identity issues play a major role). 25   David D.Laitin (1986), Hegemony and Culture (Chicago: University of Chicago Press). 26   For a compilation of the details for this aspect of corpus planning, see Government of Andhra Pradesh (1968), White Paper on Official Language (Telugu): Preparation of Authoritative Texts (Hyderabad: Government Secretariat Press). 27   K.Ravi (1982), “Regional Separatist Agitations in Andhra Pradesh,” in A.Prasanna Kumar, V.Linga Murty, and K.Ravi, eds., Government and Politics in Andhra Pradesh (New Delhi: S.Chand), pp. 54–65. 28   R.V.R.Chandrasekhara Rao (1979), “Conflicting Roles of Language and Regionalism in an Indian State: A Case Study of Andhra Pradesh” and Dagmar Bernstorff (1979), “Region and Nation: The Telengana Movement’s Dual Identity,” in David Taylor and Malcolm Yapp, Political Identity in South Asia (London: Centre of South Asian Studies, SOAS, University of London), pp. 138–150 and 151–169. See also Myron Weiner (1978), Sons of the Soil (Princeton, N.J.: Princeton University Press), chap. 5. 29   Paul R.Brass (1974), Language, Religion and Politics in North India (London: Cambridge University Press), p. 430. 30   Jyotirindra Das Gupta (1970), Language Conflict and National Development (Berkeley: University of California Press), pp. 259, 266, 268, 270. 31   David Laitin (1989), “Language Policy and Political Strategy in India,” Policy Sciences, vol. 22, pp. 415–436. 32   Stanley J.Tambiah (1986), Sri Lanka: Ethnic Fratricide and the Dismantling of Democracy (Chicago: University of Chicago Press), pp. 5–7, and Gannath Obeyesekere in a letter to the New York Times (April 24, 1984) hopefully lay to rest any lingering notion of such a divide in Sri Lanka. 33   Tambiah, Sri Lanka, pp. 58–60. 34   Stanley Tambiah (1992), Buddhism Betrayed? (Chicago: University of Chicago Press), pp. 49–57. 35   Tambiah, Sri Lanka, pp. 20–27; Stanley Tambiah (1996), Leveling Crowds (Berkeley: University of California Press), p. 100. 36   Tambiah, Sri Lanka, pp. 71–78; Tambiah, Leveling Crowds, p. 86. 37   M.R.Narayan Swamy (1994), Tigers of Lanka: From Boys to Guerillas (Delhi: Konark Publishers), p. 21. 38   S.G.Samarasinghe (1996), “Language Policy in Public Administration, 1956–1994” in R.G.G.Olcott Gunasekera, S.G.Samarasinghe, and V.Vamadevan, eds., National Language Policy in Sri Lanka (Kandy: International Centre for Ethnic Studies), p. 98. 39   R.G.G.Olcott Gunasekera (1996), “The Implementation of the Official Language Policy, 1956–1970” in Gunasekera, op. cit, p. 32. 40   Samarasinghe, op. cit., p. 105 41   Gunasekera, op. cit., p. 45. 42   Ibid., pp. 58–62. 43   Samarasinghe, op. cit., pp. 79–91. 44   The Houdini-like leader of the Liberation Tigers, V.Prabhakaran, became a militant largely because of the “standardization” decrees that were designed to advantage Sri Lankans who took official examinations in Sinhalese. In this way, language laws took young Tamils out of the education stream and into the guerrilla river. This relationship is incomplete, however. First, Prabhakaran had a fixation for explosives well before he thought about Eelam. Second, recruitment into militant groups was extremely slow until 1983, when the LTTE had fewer than 50 hard-core members. But when rumors spread in 1983 that the

OCR for page 531
International Conflict Resolution After the Cold War     Indian government was funding and training Tamil guerrillas, recruitment skyrocketed. At that time, however, standardization was hardly an issue. See Swamy, op. cit, chap. 4, and pp. ix, 96. 45   Ibid., p. 294. 46   Stanley Tambiah, comments at the National Research Council seminar to review a draft of this paper, October 22, 1998. 47   James Fearon and I, in the context of our National Science Foundation grant, will include some groups not considered to be at risk in future analyses and in doing so may pick up the violence-decreasing aspects of the 3±1 missed in Table 13.4. 48   B=.475264; SE B=.467966. 49   B=−.438306; SE B=.449604. This is when the dummy for 3±1 is also in the equation. 50   There are insufficient numbers of cases to analyze LANGREGIME=4 or LANGREGIME=5 for groups with RURBASE=1 living in countries that entered the world system after 1945 (YRENTRY>1945). Even under these conditions the bivariate correlation between REBELLION and MAXLANG is weakly negative. 51   Users of the MAR dataset might want to note that the threshold for democracy of high quality is ndem89=8; the threshold for substantial language grievances is MAXLANG >1; and the critical threshold for REBELLION is rebel90x>3. 52   For Lieven’s position, see “No Russian Spoken Here,” The New York Times, July 16, 1999.