IntroductionFootnote *
Since the existence of ethnic violence there is much scholarly debate about its origins and persistence. In this respect Basque political violence is no exception. There may be some communis opinio that long-term processes of nation-state formation and economic integration are at the background of radical ethnic movements. Yet a confrontation of leading theories with empirical studies on protracted conflict often reveals a lack of explanatory power. The resulting plea for meso or middle-range theory as an antidote against meaningless grand theories is however not always a panacea, as illustrates Mees who questions the applicability of new social movement approaches to the Basque case (Mees Reference Mees2004). A middle-range theory that might accommodate Mees’ criticism is the model relating Basque language activism with violence, proposed by David Laitin (1995; republished in Reference Laitin1999). Laitin's model, though not discussed by Mees, will be revisited in this article. Following Laitin we exclude the French Basque realm from our analysis.
In the Basque case nationalism and violence are connected with language in the political discourse and mass mobilization. The Basque language threatened by cultural homogenization induced by state and market forces, has become central to both radical and non-radical nationalist discourse during the 1960s and 1970s (Conversi Reference Conversi1990). Mass mobilization around territorial claims and language demands emerged at the end of the Franco dictatorship, when the actors and agencies of language activism and radical nationalism were formally overlapping but in reality often interwoven (Tejerina Montaña Reference Tejerina Montaña1992, Reference Tejerina Montaña, Mar-Molinero and Smit1996; Conversi Reference Conversi1997; Mees Reference Mees2003). Moreover surveys and electoral statistics show a high correlation between mother tongue and radicalism (Llera Reference Llera Ramo1984, Reference Llera Ramo1994; Mansvelt Beck Reference Mansvelt Beck1999). Euskara, the Basque language has thus become prominent both in the nationalist discourse and in the scholarly debate. Perhaps the most elegant approach connecting language activism, social mobilization and violence is the model developed by David Laitin. The model is promising because it theorizes on Basque empirical findings and it has the pretension of extrapolation because the author argues that it is valid to ethnic violence in Georgia. In addition parts of the model are fit for empirical verification. The Laitin model has had direct and indirect scholarly impact. Directly the model has been used to develop ethnic-conflict theory and to serve as a building block for a theoretical framework applied in case studies of ethnic and other types of collective violence (see for instance Loveman Reference Loveman1998; Cetinyan Reference Cetinyan2002; Mueller Reference Mees2000; Hislope Reference Hislope1998; Spinner-Halev and Theiss-Morse Reference Spinner-Halev and Theiss-Morse2003). Perhaps the indirect impact is even bigger through Laitin's publications with well-known co-authors in which his original piece has been a key reference (Brubaker and Laitin Reference Brubaker and Laitin1998; and Fearon and Laitin Reference Fearon and Laitin2000). However, to our knowledge these publications have left the empirical base of Laitin's model unquestioned, with the exception of a short critical comment by Mansvelt Beck (Reference Mansvelt Beck2005, p. 204-205, 214). In this paper we will test the empirical foundations of the Laitin model.
The paper is structured as follows. The first section contains a presentation and discussion of the Laitin model. The second one concerns an empirical test of the model in the Basque Country, for which the rise of violence will be compared with non-violent situations. In the final section the research findings are confronted with the model. Our argument will throw serious doubts on the rational-choice assumption because the diffusion of language activism and radicalism does not follow the lines predicted. Conversely a geo-historical approach of radical and non-radical Basque nationalism may provide more insight into the occurrence of violence and its legitimization.
The Laitin Model
The model is fundamentally based on the following three building blocks, which will subsequently be described: (1) a specific human ecology (2) a game-theoretic model relating language activism and violence and (3) a random event and rise of a culture of violence.
Human ecology
By comparing the Basque and Catalan cases within Spain, Laitin argues that nationalist violence can only be explained by a particular human ecology, which he labels “micro foundations of nationalist violence” (Laitin Reference Laitin1995, p. 14). In the Basque Country these micro foundations are to be found in localized social groups in small towns and smaller settlements with a subsistence base in nearby factories. The groups concern so-called cuadrillas, which are small groups of up to a dozen boys of similar age who frequent jointly the same bars, often entertain the same activities and share similar views having “norms of honor” (Laitin Reference Laitin1995, p. 15). In addition climbing associations or mendigoitzaleak and other types of local associations usually show overlaps with cuadrillas. These small gangs of youngsters under tight social control are well represented in the industrialized small-town area south of San Sebastian (Donostia), an area well connected with the outside world through a good transportation network. This area forms the seedbed of nationalist violence (Laitin Reference Laitin1995, p. 19-21; see also Gurruchaga Reference Gurruchaga1985 and Gurruchaga Abad et al. Reference Gurruchaga, Pérez-Agote and Unceta1984). Laitin calls this fertile breeding ground for terrorism a “dense rural social structure”. However, with respect to the occupational structure he is aware of the non-rural features of the local economy. In his model a dense social structure is considered as a necessary condition for the rise and continuation of collective violence.
Implicitly, however, he also applies a second necessary condition, which is the existence of a regional language threatened by the state's official language. The coexistence of two languages, in this case euskara and Castilian, Spain's official language is by no means omnipresent in the Spanish-Basque realm. Some parts like the southern territory of Alava (Araba) and the Encartaciones District west of Bilbao are castilianized for centuries. The areas that are potential candidates for the rise and institutionalization nationalist violence are thus characterized by
- the smallness of the settlements,
- a non-agrarian occupational structure, and
- mixed language geography.
Language activism and violence: the tipping game model
Rational choice is at the base of the model, which departs from investments in adopting the regional language. The benefits for those who learn the regional language are economic or social, respectively reflected in better jobs and status gains. The changes in economic gains and social prestige are “[…] based on the percentage of citizens who have already invested in the regional language […]” (Laitin Reference Laitin1995, p. 16). Laitin's hypothesis assumes initial heroic pay-offs for choosing the regional language as a language of instruction when the share of those who have done this is still low (fig. 1). However, advantages are declining when these shares are rising. Only when the percentage of citizens who have opted for the regional language tends to become a majority, do benefits begin to increase again. Below this turning point, a rising share for the regional language makes the alternative choice for the language of the state even more advantageous and creates a widening benefit gap between choosing the language of the state versus the regional language. The only rationality left to local language activists, the “language vigilantes”, in order to increase the proportion of people choosing for the regional language is resorting to violence. The desired result will be to reach a tipping point when the final economic benefits to choose the regional language will become higher than the returns for choosing the state's language. Laitin expects the critical moment for ethnic violence when 40 % has switched to the regional language (Laitin Reference Laitin1995, p. 18). Although in the Basque case there is ample evidence of a connection between language activism and nationalist radicalism, some intervening phenomena complicate the parsimonious features of the model. Firstly, there have been earlier waves of language activism, which did not result in violence, as reveals the establishment of privately financed Basque-taught schools (ikastolas) before the Civil War (López Goñi Reference López-Goñi2003). Secondly, within the locale there have been many initiatives of the clergy to protect and foster the Basque language. Small-town priests were well represented among the founders of the early ikastolas of the 1960s. Although former seminarists have had key positions in the early ETA and part of the lower clergy might have shown solidarity with radicalized language activists, this does not imply that everywhere in the Basque-speaking area the lower clergy automatically sympathized with those who resorted to violence as is suggested by, for instance, Ezkerra (Reference Ezkerra2002). Thirdly, during the 1990s up to date there has been a gradual shift in the geography of residence of ETA perpetrators from the original heartland towards the edges of the urban areas of San Sebastian and to a lesser degree Pamplona and Bilbao. This shift has been accompanied by a change from a commando-organized violence directed against military, security and political targets and kidnappings to finance ETA, toward street violence against softer targets like cashing machines, buses, trains and public amenities. Kale borroka, as the street violence is called, is the outcome of a deliberate change in ETA's strategy (Van den Broek Reference Van den Broek2004; Mansvelt Beck Reference Mansvelt Beck2005; De la Calle Robles Reference De la Calle Robles2007). For the origin of collective violence however, Laitin makes a strong case as the early actions were
[...] designed at reconfiguring the pay-off function of fellow Basques in their assessments of “Spanish” life style. The difficulty of the tip, rather than the probability of it, provides an incentive for violent tactics [his italics].
(Laitin Reference Laitin1995, p. 23)
Fig. 1 The Laitin Model
As we will show in the next section there are some, perhaps more serious methodological limitations to the model.
A random event and rise of a culture of violence
Violence may originate as a random event leading to escalation. The probability that out of a few violent actions by rebels or by the state's security forces may arise a culture of violence will be higher if (1) the regional population perceives the victories but is blind to strategic losses; (2) if the cost of leaving the terrorist organization are high; and (3) if a “culture” of violence becomes institutionalized (Laitin Reference Laitin1995, p. 18). In particular the first and third points have changed. Regarding the first point this may have been true up to the end of the 1980s. However the increasing successes of police actions together with violent actions by ETA that shocked many citizens in the Basque Country have certainly weakened ETA's support base (Mees Reference Mees2003). Atrocities, like the killing of the young municipal councilor Miguel Ángel Blanco in 1997, have lead to aversion to radicalism among many Basques. The aversion, also exacerbated by the emerging street violence and a more generalized discrediting of violent strategies as a consequence of the September 11 and March 11 in Madrid has resulted in a decline of radical Basque nationalism from 1997 onwards (Muro Reference Muro2008, ch. 7). The clearest indicator of the loss of popularity of the radicals has come from the first elections for the Basque parliament after the 1998-1999 ceasefire and the subsequent return to violence. Euskal Herritarrok, the political party representing nationalist radicalism, saw their votes almost halved. Laitin is aware that cultures of violence may finally disappear as they “are not eternal” (p. 25). These newer developments illustrate the open-ended character of the model since explanations for the erosion of cultures that keep the process of collective violence going are missing. A different comment regards the randomness of violent mobilization. We do not think that the rise of nationalist violence in a multi-cultural centralized state under a regime that is liberalizing and democratizing is such a random event. This is because at a macro level Gurr (Reference Gurr1993) and Snyder (Reference Snyder1994) have argued from different angles that cultural minorities tend to develop rebellious movements when democratization is underway.
Retesting the model in the Basque setting
Although Laitin has based his model on the Basque case providing quantitative expectations, both conceptualization and expectations show some methodological limits. The model contains timely and spatial dimensions. Particularly the time dimension lacks conceptual clarity and is therefore not ready for empirical testing. The weakest element is the lack of a time horizon because Laitin does not approximate, let alone specify, the time span to be covered after the suggested decline of heroic pay-offs. The only explicit mention he makes for the rise of nationalist violence is the suggestion that it occurs when 40 % has switched to Basque-language education. Therefore, our main task will be to explore the suggested time sequence of respective phases suggested, namely a phase of semi-effective (due to attraction state's language) language activism, followed by violence-loaded effective language activism and ending up in a culture of violence and an entire switch to the regional language. Observations should be interpreted very cautiously because after the granting of Autonomy to the Basque Country in 1979 the regional government has been very successful in Basquisizing the educational system. During the last two decades local intimidation is no more the strongest incentive for language switch. In addition regional language-planning policies have created an impressive educational offer in Basque as well as many new job opportunities for which command of euskara is required (Azurmendi et al. Reference Azurmendi, Bachoc, Zabaleta and Fishman2001; Gardner Reference Gardner2002; Mateo Reference Mateo2005). As a consequence the assumed causality between violence and language choice may have been replaced by a causal relationship between language-based opportunity structure and basquization of education.
Perhaps the best chances for a confrontation with empirical reality are in the conditions of human ecology required before mechanisms of the tipping model come into play. Hypothetically there are three situations. The first situation is in line with the Laitin model in which the independent variables (settlement pattern, occupational structure, and language geography) coincide with the dependent one (nationalist violence). The second one would reveal a presence of the dependent variable outside this area. In this case one would need to search for alternative explanations. In the third situation there would be an occurrence of zones in which the three independent variables coincide but where violence is absent. In this case the question why violence has not arisen should be answered.
Concerning the method, the translation from conceptualized variables into operational attributes is complicated because some concepts are not suitable for quantitative research, while for other least-best indicators should be used. For instance, heroic pay-offs may perhaps be measured by fieldwork surveys or anthropological research in situ, there are no data sets available about gains or losses in social status as a result of language activism. The same applies to the expected economic returns for choosing one of the two languages.
In contrast the proportions of people who choose Basque, as a primary medium of instruction seems fit to measurement. At the operational level, however, the problem arises that the Basque government has introduced two types of Basque-taught education, namely monolingual and bilingual education, officially denominated as model D and model B. In principle the switch to B or D implies a choice for the regional language, though for B it is a more ambivalent choice. The choice for the D model is here defined as a switch, not only because it is the most drastic choice, but also because the demand of radical nationalism is for Basque-medium teaching and not for bilingual education. Data on primary education provides the best proxy for language choice as it concerns crucial age groups of children who start reading, writing and socializing in a language setting. The proportion of children attending D-model schools in municipalities sharing the same schools will be used. For Euskadi we have municipal time series for education at our disposal of which we have used data for respectively 1984 and 2005. Data for 1984 is taken because it was the first year numbers of pupils according to the language models appeared in Basque statistics. Because the data concerns school populations and not municipal populations, we have assigned the proportions of children in the D model also to the nearest school-less municipalities. “Nearest” in operational terms is the shortest distance by road. The actual schooling situation (2005) is also taken in order to analyze whether a possible increase in the model-D proportion has been the result of political radicalization. For the Charter Community of Navarre only part of the testing has been done because Basque-medium education was introduced at a later stage than in Euskadi (end 1980s), impeding a time-series based comparison of Euskadi and Navarre for the entire period. The different jurisdictional territoriality of Euskadi and Navarre has resulted in two distinct language regimes complicating rigid comparison. For example, in contrast to Euskadi, in Navarre only the officially denominated Basque and mixed language zones are eligible for Basque-medium education.
The choice for formal Basque-medium education is however not the only way of assessing concrete behavior in favor of Basque. During late Francoism ikastolas (ikastolak) mushroomed, in particular in Gipuzkoa. Here the number of pupils enrolled in ikastolas per 1000 inhabitants had risen from virtually none in 1964 to more than 55 in 1981. In Alava, Biscay and Navarre attendance of ikastolas was more modest, but still considerable with rates between 10 and 22 ikastola pupils per 1000 inhabitants in 1981. (Tejerina Montaña Reference Tejerina Montaña, Mar-Molinero and Smit1996, p. 228, 230; for inh. INE Anuario de Estadístico 1982). The first more or less complete data set collected about ikastolas, dates from 1976 (Siadeco 1977). Although data about numbers of children enrolled is far from complete because particularly in Biscay many ikastolas refused to provide information, the exact location of all ikastolas in the Basque Country and Navarre is reported. Knowing the locations of ikastolas for 1975/1976 they are useful to our analysis because they reflect collective enthusiasm for rebasquization. Moreover at that stage the teachers were Laitin's “language vigilantes” or in the words of the Basque Academy of Language “authentic militants of the Basque culture” (Euzkaltzaindia 1979, p. 173). Small municipalities without an ikastola but within a road distance of 10 km at a maximum from the nearest ikastola are assigned with the label of “accessibility to ikastola”.
The use of violent tactics, the dependent variable, is hardly measurable in a statistical way. Firstly, the location of violence targeted against “interests of the Spanish state” is different from the places of origin of the perpetrators. Secondly, intimidation, extortion and social exclusion as tactics within small towns, usually produce an omertà-like silence instead of reliable statistics. The only variable that to a certain extent reflects a solidarity with E.T.A. is voting for their political affiliates who have operated in regional elections under the following names: Herri Batasuna (HB/United People), Euskal Herritarrok (EH/Basque Citizens) and EHAK (acronym for Communist Party of the Basque Lands). We will use HB and EH election data for regional, Spanish parliamentary and European elections. The complicating factors arise with the interpretation of election data, namely for the Spanish parliament, the izquierda abertzale (Patriotic Left) as the radicals call themselves, have not always fully participated in the political process. In addition regional elections in the Basque Autonomous Community (BAC) and Navarre cannot be compared due to the different timing and in particular because of the distinct regional powers. Despite the problems of comparison, the data on radical voting for various elections show extremely high correlations (e.g. municipal outputs for EH in post-truce regional elections in 2001 for the Basque Autonomous Community show a 0.952 correlation with the 1999 European elections celebrated during the ceasefire, while for the Basque parliament the 1980 and 2001 elections, the radical vote still shows a 0.8 correlation). A complicating issue is that language activism and political violence started two decades before the first elections (1979). Although the heydays of killings and kidnappings were in the early 1980s, it may be possible that a culture of violence in the ETA-controlled local environment had been established previously.
In sum the data used is based on municipalities in the Basque Autonomous Community and occasionally in Navarre. The use of municipalities as a substitute for local settlements is not without problems because some municipalities, in for instance the Bilbao agglomeration, are a functional part of its metropolitan area while other municipalities may contain various nuclei. According to the Spanish statistical office we consider municipalities of less than 10,000 inhabitants as “rural” and characteristic for Laitin's human ecology favorable to the rise of culturally inspired violence.
Empirical findings
The ikastola was the informal precursor of formal Basque-medium education. The data reveals that the Basquization of education has its roots in ikastolas. When we compare the category of municipalities that in 1975 had accessibility to a local or nearby ikastola with the category municipalities which had not, we see a striking difference in the average percentage of school aged children enrolled in Model D education. For the category with accessibility to an ikastola the average was 47 % in 1984 and 84 % in 2005, while for the category without accessibility the average percentages were only 10 % in 1984 and 48 % in 2005. The figures show both Basquization and certain continuity in its geographic distribution. The correlation-coefficients between the binary variable of accessibility to an ikastola in 1975 and the percentages of children enrolled in Model D1984 and in 2005 are respectively 0.475 and 0.493. Based on multiple regression analysis, the 2005 pattern of Model D does not show any significant correlations with socioeconomic variables. High proportions of pupils in Model D in municipalities with no own ikastola but one in a nearby location reveals the popularity of Basque-medium education in small settlements with children going to school in bigger villages and towns. The fact that some municipalities with an ikastola hardly were involved in the D model suggests that the ikastola had a central function for the small-settlement area in its vicinity. On the one hand, this might point to a greater affection to Basque culture in smaller communities, while on the other, arguing in line with Laitin, the higher involvement in Basque-medium education in these communities may be a result of social pressure. However the role of settlement size should be qualified in this respect because the correlation between model D in 1984 and the logarithm of the population size is not so high (r=-0.245).
The scatter plot (fig. 2) approximately displays a linear correlation between the proportion of children in model D in 1984 (annotated as ModelID1984) and the percentage of radical votes in the 1999 European elections (annotated as ehper99). However, there are a considerable number of municipalities with a near-to absence of Model D but a high share of radical voting. This is not according to the expectations based on the Laitin model because the model supposes the occurrence of sustained violence above a certain threshold (his 40 % choice level). There is no evidence that reaching a certain level in 1984 has triggered a culture of violence because such a culture would have been established and resulted in significantly higher EH voting. Even with another threshold (for instance 30 or 50 percent), the percentage of EH voting should have been significantly higher than the linear correlation at that value. Laitin's game-based model is confined to non-agrarian “rural” communities, but scatter plots of these “rural” municipalities show similar patterns.
Fig. 2 Scatter plot % voting for EH 1999 and % Model D choice 1984, BAC, municipalities
Through multiple-regression analysis, the linear relationship can be checked for other variables that may explain radical voting, such as language and socioeconomic variables. For the 227 municipalities of the BAC we have conducted a stepwise multiple-regression analysis for which the following independent variables have been used:
% Basque speakers (home language 2005)
% Pupils in Model D 1984
% Pupils in Model D 2005
% Presence or proximity of ikastola in 1975/76
Increase in % Pupils in Model D 1984-2005
% Population born in BAC
Logarithm number of inhabitants (LOGINH)
% Population working in agriculture 1995
% Unemployed 1996
Alava (dummy variable)
Gipuzkoa (dummy variable)
Biscay (dummy variable)
Obviously not all the variables mentioned could be incorporated simultaneously in the regression model. Particularly language variables show too strong correlations, as reveals the correlation coefficient of %ModelD1984 and % Basquespeakers2005 (r = 0.841). Differences according to the territorial context can be integrated in the regression with dummy variables for the former provinces (nowadays “historical territories”). If no account is taken of these differences between the BAC territories, the percentage of Basque speakers is the only significant explanatory variable at a significance level of 95 % (α ≤0.05), explaining 58 % of the variance of the proportion of EH voting (R2= 0.58). The mean of the percentages of Basque speakers in 227 municipalities is 50.
When territorial variables are taken into account, four variables appear to exert a significant influence on radical voting. They jointly explain 68 % of the variance. The results confirm the relevance of the language variable % Basque speakers 2005. Furthermore, it appears that independent of the percentage of Basque speakers, the Gipuzkoa municipalities clearly tend towards higher percentages of EH votes than municipalities in the other two former provinces. Moreover, the correction for difference between Gipuzkoa and the other two territories discloses significant negative influences of population size and proportion of farmers on radical voting. These influences confirm Laitin's model because it predicts a high intensity of radicalism in non-agrarian small communities.
The regression coefficient reveals that in proportion to their shares of Basque speakers, farmers and population size, municipalities in Gipuzkoa have on average 11.63 % more EH votes than those in the other former provinces. The striking inter-provincial differences urge a more detailed analysis per former province. In Navarre, divided into Basque, mixed and Spanish language zones, and with fewer data, we have conducted a multiple-regression analysis for the municipalities of the Basque and mixed zones. Similar tendencies can be observed as in the BAC, but only 29 % of the variance in radical voting can be explained. The larger part of non-explained variance shows that the explanation only partially coincides with the explanation for the BAC. Hence, a territory-specific analysis will provide more information about the different historical growth paths for Model D.
During the mid-1970s only six Alava municipalities had an ikastola. Given the small number of ikastolas it is not surprising that there were no municipalities with nearby ikastolas. In 1984 the municipalities without ikastola did not have any child enrolled in Model D. Conversely in the six municipalities with an ikastola 18.3 % of the pupils went to Basque-immersion schools. The differences between the six are deep since accessibility to an ikastola in 1975 only explains 18 % of the variance of ModelD1984 (r=0.425).
The variance for ModelD2005 cannot be explained by ModelD1984 (r=0.197), which reveals that the pattern of Basque-medium education had deeply changed over two decades. Basque-medium teaching has significantly increased in most municipalities, in particular in those ones that did not have an ikastola and no Basque-medium teaching in 1984. The rise of Model D in mainly Castilian-speaking Alava is thus a recent phenomenon. Nowadays the average of the percentages for Model D is about 40 %. However Alava is the least radicalized territory. The latter suggests that language choice in Alava has changed not because of radical pressure but as a result of language policies of the BAC.
In contrast the present educational pattern in Biscay and Gipuzkoa has an earlier origin compared to Alava, as reveals the stronger correlation between Model D in respectively 1984 and 2005 (Biscay r=0.628; Gipuzkoa r=0.635). In these provinces the rise of Basque-medium education has been particularly strong in the smaller settlements. There negative correlations between LOGINH and ModelD1984 are high (Biscay r=-0.454; Gipuzkoa r=-0.698) with Alava showing a weak positive correlation (r=0.107). The comparatively low percentages for Model D in 1984 for Biscay (average municipal percentage = 33 %) suggest that many municipalities had reached the critical threshold of 40 percent round that year. Conversely in Gipuzkoa, well endowed with ikastolas in the mid-1970s, the percentages for 1984 are already fairly high (average municipal percentage = 61 %). According to Laitin these municipalities would indeed represent the “dense rural structure” that served as a seedbed for a culture of violence.
Obviously by 1984 Alava had not experienced a rise of Basque-medium education that approached threshold values (average municipal percentage = 2 %). Therefore the proportion enrolled in Model D in 2005 offers a better indication (average municipal percentage = 43 %). However, Alava scatter-graphs do not show any evidence of a linear or curvilinear correlation between language choice and radicalism. The scatter plots for Gipuzkoa and Biscay (fig. 3 and 4), show a distinct patterning, but do not confirm the existence of a violence threshold because in that case the pattern would have a “tilted-S” shape. Notwithstanding, Laitin's hypothesis for Gipuzkoa may still be valid because early language mobilization prior to 1984 may have had a follow-up in radicalization. Although the statistical coexistence allows for causality according to Laitin's proposition, in the next section we will suggest the existence of alternative causal relationships.
Fig. 3 Scatter plot % voting for EH 1999 and % Model D choice 1984, Gipuzkoa, municipalities and hypothetical tilted S-curve
Fig. 4 Scatter plot % voting for EH 1999 and % Model D choice 1984, Biscay, municipalities and hypothetical tilted S-curve
For few municipalities, there are time series available which enable tracing a linguistic threshold. Markina-Xemein is rather exceptional in this respect. Markina-Xemein is a municipality of almost 5,000 inhabitants in the Northeast of Biscay. According to its occupational structure, the municipality is not so different from the settlements in the radicalized zone of Gipuzkoa. Half the population is working in factories of which part is located outside Markina-Xemein while 40 percent work in the service sector. Some 84 % of the population uses euskara as the home language. In contrast to many other settlements, data about the numbers of pupils in ikastolas and from 1984 onwards in Model D are known (respectively Arrien Reference Arrien1993, p. 151-155; Eustat 2006). In 1970 the ikastola, like many ikastolas in Gipuzkoa established by local priests existed for six years and had 100 pupils. The numbers of pupils in the local ikastola Zeretxu slightly diminished to 71 on 1984, but in the official local school of Vera Cruz the numbers of pupils studying in the Basque-medium style had risen to about 500 pupils. However during the 1986-1988 the proportion of children in Basque-medium education stabilized around 40 %, precisely the critical percentage suggested by Laitin when language activists start resorting to violence. In this situation one would expect a rise of activities of radicals and over the years a change in voting behavior. Only one violent act is known, namely the killing of two civil guards in 1978. Against the tipping model's prediction, our data series show that the killings did not result in a subsequent radicalization of the locale. Voting behavior did not reflect a shift to radicalism, whereas over the years the choice for the regional language dramatically increased. Nowadays eight out of ten children are enrolled in Basque-medium education, which reflects the absence of a total social control over education. In successive regional elections the non-radical nationalist PNV has always enjoyed a firm support fluctuating around 60 % of the votes. The example of Markina-Xemein shows that a continuous local dominance of moderate nationalism may finally result in a massive shift to Basque-medium education without the assumed need of violence.
The regression results for the three BAC territories provide interesting differences (table I). In Biscay only the proportion of Basque speakers significantly explains radical voting (49 % of the variance). Within Biscay there is no statistical evidence of a tendency towards greater radical support in an urbanized countryside. For Alava the result is remarkable because language shows an extreme skewness with 89 % euskara-speakers in the village of Aramaio in a sea of Spanish-speakers. However, given its overwhelmingly Castilian-speaking population, the 13 % vote for the radical EH is striking. After introducing the logarithm of the percentage of Basque speakers (skewness 0.136) the relative differences between the municipalities become visible (the average of this logarithm of 0.3678 equals a percentage of 2.3 %). Surprisingly, the small differences in Basque speaking between the municipalities together with the comparatively unimportant share of Basque-born explain 61 % of the variance.
Table i Multiple regression. Dependent variable = % votes EH elections European Parliament 1999, Standardized regression coefficients (β). Municipalities Provinces Biscay, Alava and Gipuzkoa.
The regression analysis for Gipuzkoa shows that in contrast with Alava and Biscay the percentage of Basque speakers is not the most adequate explanation for the percentage of EH votes. Although the correlation between the proportion of Basque speakers and the logarithm of the number of inhabitants is high (r=-0.817), municipal population size appears to be salient in the explanation of radical voting, rather than language.
The empirical findings reveal that a threshold, which is crucial to Laitin's model, is not recognizable in the data. However in the radical heartland of Gipuzkoa language-choice thresholds may have crossed before the introduction of the D-Model in Basque education. Therefore Laitin's approach may be still valid, particularly because the strong negative correlation between the size of the local community and radicalism may reflect the important role of social control. Albeit we accept the possibility of such validity, particularly in the non-agrarian small settlements, it is not that clear in what sequence language mobilization and radicalization have occurred. In 1984 Basque-medium education in Gipuzkoa had already significantly progressed. Consequently, it is probable that there are many municipalities having already surpassed the critical level in 1984, of which we do not know whether this occurred with or without a prior violent push. Hence, also for Guipuzkoa we cannot confirm a sequence of moderate language mobilization followed by radicalization.
The more general conclusion of our analysis of statistical data must be that causalities of Laitin's model may give an explanation for violence in Guipuzkoa, but for the other territories empirical support for the model is rather weak. Moreover, also for Gipuzkoa our results allow for alternative hypotheses. Therefore some alternative models that explain the localized rise of radicalism will be developed in the next section.
Basque Violence: a Reappraisal of Culturalist Explanations
Statistical data and survey results both confirm a high correlation between language and political radicalism in the Spanish-Basque realm. This has led to numerous explanations about the rise and continuity of Basque collective violence stemming from culture. Weakly defined cultural boundaries in daily life where Basque and Spanish culture are often intermingled, have resulted in a tendency to emphasize discursive differences between Basques and outsiders accompanied by boundary-making practices as Conversi (Reference Conversi1997) has theorized. In fact Laitin's model can be perceived as a middle-range theory that postulates a causal relationship between language activism by self-perceived Basques in order to Basquisize the behavior of fellow Basques in a socially controlled environment. However, some of these postulates should be put into perspective for reasons linked with discourse, institutional presence and questions about time and space, which will be subsequently dealt with.
Firstly, the radical discourse about language revitalization is important, but not uniquely focused on language as a core value. This is because in ETA's ideology class and culture have competed historically with each other while today the radical claims are increasingly centered on self-determination and territoriality. On the other hand a certain messianic tone in Basque radicalism can still be discerned because radical nationalism sees itself as the only real protector and saver of the Basque linguistic heritage. However, the moderate, and most voted nationalist parties Partido Nacionalista Vasco (PNV) and its split-off Eusko Alkartasuna (EA) and some smaller ones, also entertain a discourse in which language plays a prominent role. Finally, given the new opportunities for Basque-medium education in Euskadi, and to a lesser extent in Navarre, the context is becoming less and less meaningful for nationalist discourse with language reversal at the core (Mees Reference Mees2003).
Our second comment concerns the presence of radical nationalists in cultural institutions from and around which, as Laitin has emphasized, the local environment has been forcedly mobilized in order to give the last violent push to language reversal. Indeed radicalism is overrepresented in local institutions like ikastolas, adult literacy associations and in supra-local institutions like TV, radio, and the ikastola federations. However the radical grip on cultural institutions is incomplete because today there is much competition from official supra-local institutions such as the region's Ministry of Education and the official organization for adult literacy. In addition, several authors have discerned multiple spaces of mobilization. This fragmentation has led to at least “two nationalist communities” (Gurrutxaga Abad Reference Gurruchaga1996) to a fragmentation into “micro-societies” (Tejerina Montaña Reference Tejerina Montaña1992) or to a geographical compartmentalization of collective identities (Raento Reference Raento, Kaplan and Häkli2002; Mansvelt Beck Reference Mansvelt Beck2005). In this spatial identity fragmentation, Laitin's “dense rural social structure” is the geographical niche for language vigilantes. However, his model in which space and time are crucial components still raises questions, which we will now clarify.
Let us first follow Laitin. He correctly states that from ETA's cradle in Bilbao, the breeding ground for collective violence shifted towards the smaller settlements of Gipuzkoa. This breeding ground did not only produce etarras, but also a culture of violence. This picture may be valid up to the early 1970s as observes Domínguez Iribarren (Reference Domínguez1998, p. 54-77) in his review of the changing geographical origins of etarras. However, by that time ETA's recruitment areas had already shifted to the urbanized zones of Bilbao and San Sebastian. Did a more urban recruitment of perpetrators of violence imply a concomitant shift of a culture of violence to the cities? The answer is yes and no. According to Domínguez Iribarren (Reference Domínguez1998, p. 59-60), in Bilbao such a culture of violence is connected with radical institutions while in the area of San Sebastian it forms part of the personal life sphere of family, friends and colleagues. The old area of smaller settlements lost its predominance as a space of recruitment, though on the other hand it maintained its prominence as an area of legitimization of violence as can be observed regarding the omnipresence of radical protest rituals. The increasing disconnection between spaces of radical loyalty and production of violence is confirmed by examining the origin of ETA membership, which reveals that during the last decade only 14 out of the 39 most radical-voting municipalities in the Basque Country and Navarre produced violence perpetrators (source El Pais).
According to our analysis Laitin seems to fit in the idea of a proper space of violence. Some doubt can however been thrown about the suggested sequence of events, for example what came first: was it the application of violence or a mass support for language reversal? Moreover, the rational choice or the need to force people to choose for the regional language can be criticized. Laitin has overlooked that even at a micro level spatial fragmentation exists. Albeit the data confirms that his model may be valid in many small settlements in the densely populated and industrialized area south of San Sebastian; in the smallest settlements in this area the establishment of a culture of violence did not follow the same path as in the size category of settlements with some central functions. In these smaller nuclei we suggest the following alternative models of radicalization and/or language revival.
The first alternative model (fig. 5) also connects language activism and radicalism. However the relationship is of a different nature than Laitin suggests. Here language activism has been locally successful from its early beginning. Language activists were almost unanimously welcomed as well as ikastolas and adult literacy programs. Mass support for the Basque language particularly occurred in the smaller settlements with a considerable part of native Basque speakers. This is in line with the survey of the Basque Language Academy of 1976, published in 1979, which revealed an overwhelming enthusiasm of Basque speakers to introduce Basque into education (Euskaltzaindia 1979, p. 196-200) with about half the Castilian-speakers showing sympathy. The occurrence of widely spread enthusiasm throws some doubts about the need to use violence in euskara-speaking environments in order to convince parents to send their children to Basque-medium schools as suggested by Laitin. Therefore, if Laitin is correct one might expect a stronger tendency of violence in zones with a mixed linguistic composition, which we could not confirm. The phenomenon of localized mass support for language revitalization can be traced in the late 1960s and early 1970s. Up to the mid-1970s many locals remained detached from politics. Up to that period “[…] “Basqueness” tended to be viewed as a matter of cultural content and not as a basis for political claims” as Heiberg concluded in her study of Elgeta, a predominantly Basque-speaking locality of about 1200 inhabitants (1985, p. 303). When local youngsters brought radical ideas from the nearby towns to these smaller settlements, there was already a massive support for cultural revival. Thus the need to force the major part of the population reluctant to support a language switch was absent. Local cultural associations like the ikastolas, mountaineers clubs, and folk dance groups accompanied by a culturally shaped social control, served as the “infrastructure for nationalism” (Heiberg Reference Heiberg and Douglass1985). The subsequent establishment of a radical political ideology was thus not a rational choice to foster language reversal, but rather a successful spread of a new “political religion” that according to Casquete (Reference Casquete2003) often developed into ritualized protest that served to reinforce local bonds of the participants. As Casquete has commented, the function of protest rituals in these small settlements is inward looking and not instrumental. Taking his point further, solidarity with radical views and ritualized protest has shaped a culture of violence, a culture that legitimizes violence but does not necessarily produce violence itself because the perpetrators may be from elsewhere.
Fig. 5 Alternative Model: Early mass language mobilization and violence
In the second alternative model (fig. 6) the role of radical nationalists in successful language activism is marginal. The initial conditions in terms of mass enthusiasm for cultural revival in a mainly Basque-speaking environment and a rich set of cultural institutions and practices are similar to the first alternative model. However, in the next phase the protagonists of social control did not belong to the new radical political belief, but remained in the camp of moderate nationalism. In the localities where non-violent nationalism dominates the community, generally the Partido Nacionalista Vasco in Biscay and Eusko Alkartasuna in Gipuzkoa and Navarre, the efforts of radical language vigilantes were not needed to convince the locals of the virtues of language switch while promoters of radical beliefs had only limited success due to the firm establishment of peace-abiding nationalism. This second model is not a refutation of Laitin's model because our analysis reveals that it is mainly confined to the more rural areas or, in other words, areas in which agriculture is still an important means of subsistence. In this respect it is not clear whether the case of Markina-Xemein is an exception to the Laitin model or whether it represents more industrialized small settlements of Biscay, where moderate nationalism dominates to the detriment of radicalism.
Fig. 6 Alternative Model: Early language mobilization and peaceful social control
In the third place we distinguish an alternative model (fig. 7) not based on the role of language vigilantes in cultural revival, but rather on the struggle for political control of local institutions. Cultural institutions like ikastolas and later on schools, play a role which is distinct from Laitin's idea of violently pushing up language reversal. Here our assumption is different from Laitin's basic assumption of an exclusive connection between language activism and a resort to violence because of the existence of a group of language militants who have not chosen intimidating methods. Local priests, for instance were often the initiators of ikastolas, in which it was quite common to hold confirmation classes. Although part of the lower clergy and in particular part of ex-seminarians had become sympathetic to ETA or joined them (Ezkerra Reference Ezkerra2002), founder-priests followed by the local community of practicing Catholics did not always agree with violent methods. Here the ikastola is a battleground consisting of radicals who aim at taking over the ikastolas to the detriment of violence-rejecting language activists. The final aim of the radicals is to take control over all local institutions in order to eradicate competing nationalist views. Here intimidation has a Machiavellian flavor because it is aimed at political control and not primarily at cultural change. Jon Juaristi (2006) who has been an ikastola teacher in Sopelena in 1974, has described the successful radical takeover of the ikastola, which he considers a metaphor of “ethnic and ideological cleansing” everywhere in the Basque Country. In this model cultural institutions for language revival are mini arenas. The model is probabilistic in contrast to Laitin's rather mechanical view because it leaves the possibility of either a complete or a partial social control by non-violent forces.
Fig. 7 Alternative Model: Struggle for control between radicals and moderates
Laitin clearly localizes the culture of violence in the urbanized countryside. As we have shown, this culture of violence still exists as far as concerns the loyalty to radicalism, but does not produce the bulk of violence perpetrators. This is because etarras have more and more an urban background. The cities of Bilbao and San Sebastian and particularly the urban fringe of the latter have increasingly been the areas of production and reproduction of collective violence (Domínguez Iribarren Reference Domínguez1998). However, compared to the districts south of San Sebastian with their tight social control, these urban areas allow a rather anonymous life. Therefore the production and reproduction of violence is not part of the entire social environment, but is part of an urban subculture of violence that is confined to specific institutions such as radical political organizations, bars, labor unions, cultural institutions and so on. In these urban zones a radical subculture competes with other political cultures dominated by moderate Basque nationalists and Spain-wide institutions. All radical institutions are ardent defenders of Basque culture, in particular the language. Recruitment of young ETA members is now predominantly an urban phenomenon. When ETA started to organize street violence during the mid 1980s, youngsters who participated often did so by youth associations, which were soon banned, but continued to exist under other names and clandestinely. Jarrai after its banning became Haika, which after its prohibition became Segi, the name the now illegal youth association still bears. Bars, a radical music scene, squatter houses, associations for ETA prisoners are places and institutions where both a culture of violence legitimization and violence perpetrators are produced. The typical career for young etarras nowadays starts with kale borroka as street violence is called. This violence usually takes place against soft targets, like cashing machines that are burned, buildings of non-radical political parties, buses, post offices and whatever objects that can be associated with Spain, moderate Basque nationalism and capitalism. Kale borroka does not exclude intimidation towards specific persons as graffiti against targeted persons witnesses. When a borroka person has been waged as brave and courageous in street violence, recruitment for ETA may initiate the next career stage.
Like violence related with language activism, street violence has a spatial dimension, which is reflected in its bias towards bigger towns where “soft targets” are located (De la Calle Robles Reference De la Calle Robles2007; Raento Reference Raento1997; Mansvelt Beck Reference Mansvelt Beck2005). In addition, polarized settings in terms of language mix and political divide between radical nationalists and other parties statistically show a higher incidence of street violence rather than places where social control is either entirely radical or overwhelmingly non-radical (De la Calle Robles Reference De la Calle Robles2007). Like Laitin, De la Calle Robles considers social ecology and strategic considerations based on rational choice by a militant organization as the basic factors behind the geography of political violence. The present context in the Basque Country, however, has changed to such an extent that the rationality of today's radical violent strategies may be disputed.
Cultures of violence from which rational strategies arise may suffer from decline, like rationality itself may become obsolete due to contextual changes. Laitin, though aware of the finiteness of cultures of violence, has not predicted the circumstances, which may lead to their transformation. We would suggest the following circumstances. Concerning the cultural and political context, in the present-day Basque setting radical grievances of cultural victimization, appeals to a lack of democracy and a surplus of repression directed against Basques are no more in line with daily-life experiences of the constituencies. This is because they are respectively free to enjoy Basque language and culture in all aspects of life, they may vote for their favorite candidates and they can enjoy the virtues of the rule of law. In addition de-legitimization as a consequence of atrocities is gradually eroding the social base of Basque radicalism and internally dividing the izquierda abertzale into violence-condoning and peace-seeking factions.
Conclusion
Laitin's explanation of the rise and continuity of Basque nationalist violence in terms of a rational choice of language activists in a specific local setting cannot be rejected by reanalyzing statistical and qualitative data. However, serious doubts may be thrown on the validity of Laitin's rational choice model in the area where his observations seem valid. This is because the application of nationalist violence has wider aims than exclusively linguistic ones. These aims may be framed in, for instance, class struggle or territorial control by using tactics to take over local cultural and non-cultural institutions.
Conversely, the data suggests that within the industrialized small settlement area, there is a specific group of tiny Basque-speaking settlements where Laitin's explanation should be refuted. In these settlements the establishment of a culture of violence cannot be attributed to language activism. In these settlements language shift was a massively made choice from the early beginning and, as a consequence, rebasquization preceded nationalist radicalism. On the other hand we have demonstrated that outside Laitin's radical seedbed area, successful language shift may occur without a significant pressure of coercive language vigilantes.
Based on our reappraisal, the question arises on the rationality of choices made by language activists. May we assume different types of local conditions in the smaller villages that have produced other types of locally bounded rationalities? Was an early massive language shift a rational choice of the villagers, or a less rational psychological reaction against assimilation policies? Nationalist violence against fellow citizens in the ethnically mixed suburbs of San Sebastian has continued, despite significant changes in language choice. Politically, these suburbs remain divided as there is a long-lasting stalemate between radical and non-radical parties. In these settings language activism has attained their goals but violence continues. For Basque language revival there is no more need of coactive initiatives because the respective institutional contexts of Euskadi and Navarre have produced a favorable environment for language revival in which terrorism has become an irrational choice for language vigilantes. The only rationality is probably embodied in the psychological and social pay-offs within the sectarian environment in which the propagators of collective violence operate.
In theorizing political violence, Laitin's rational choice theory based on the Basque has played a significant role by emphasizing the role of local social control mechanisms. Our analysis, however, demonstrates that the empirical base of his model has some severe shortcomings. Extrapolations of the model to settings outside the Basque realm should therefore be made with extraordinary caution. Future research should take into account that in each nationalisttpdel conflict there may be local settings in which other rationalities based on alternative goals may get rooted and that local social control may work in favor of moderate nationalism instead of radical nationalism. Moreover, local conditions may have changed to such an extent that political violence has become an irrational choice.
