Language Ideology and Space: A center-diffusion model of perceived linguistic variation among Tzotzil Maya of southern Mexico 1. Introduction In this paper, we revisit a debate on the relationship between ethnicity, language, and space by exploring how the Tzotzil Maya of Chenalhó, Mexico perceive and locate within-group language differences. We focus on the production of identity as it emerges through the construction of withincommunity language differences. Our findings are twofold. First, we show that local understandings of space and place can shape the construction of identity. This finding complicates traditional models of ethnicity that describe identity formation as the product of dichotomizing and partitioning processes in which notions of „other‟ are built in opposition to „self‟ (Gal and Irvine 1995; Bucholtz and Hall 2003). Instead, we propose a model in which categorization and ranking of social groups coexist with more continuous notions of otherness; in other words, people construct identities not only constructed by placing individuals or groups in binary schemas, but also by estimating degrees of difference. Spatial knowledge provides one possible tool people use to reason about such differences. Second, our interlocutors use perceived within-group language differences as social commentary, telling us a story about the perception of ethnicity rather than about language. Refusing simple dichotomies such as indigenous vs. non-indigenous, or state vs. subaltern, we find that the distribution of perceived language differences constitutes socio-historical commentaries that can be described both as native historiography and an act of resistance. While difficult to observe directly, we show that these historical and social commentaries can be detected through the use of indirect elicitation methods combined with ethnographic observation. Constructed language differences present an idiom for rationalizing and imagining social difference, hence they provide us with a window into identity construction, ethnicity, and social history. Consequently, we take the multiple ideas about language (Hymes 1974, 31) not as simple representations of language, but as complex representations of people, their relations, and histories. The way people think about and imagine their language is thus inextricably tied to the formation and maintenance of social 1 © 2016. This manuscript version is made available under the Elsevier user license http://www.elsevier.com/open-access/userlicense/1.0/ identity, and perceived language differences are an idiom to express perceived social differences. Folk ideas and theories about language have been known as linguistic ideologies, a concept which highlights the partial, socially constructed, and interest-laden aspects of beliefs about language (Woolard and Schieffelin 1994). Linguistic ideologies encompass and interrelate representations of linguistic features and their distribution, social categories, and moral and aesthetic values. As such, they cannot be seen a neutral representations of linguistic phenomena (i.e. as existing independently of other cultural/cognitive domains), but rather are associated with social hierarchies and stereotypes and evoked by interested groups or individuals (Kroskrity 2004). Though language ideology can relate to any aspect of a cultural system, most works on the subject have emphasized its role in mediating between forms of talk and social stereotypes. Irvine and Gal have conceptualized this phenomenon as iconization, a semiotic process by which certain linguistic features become associated to – and seen as iconic of – social identities. Iconization, they argue, is accompanied by fractal recursivity – the notion that certain oppositions between groups can generate divisions within those groups at smaller scales – as well as by erasure, a process by which desirable social oppositions and groups are made more salient while others are rendered invisible or ignored (2000). Research on language ideology emerged out of the recognition that social processes such as these can influence language change, thus posing a challenge to theoretical frameworks that explained change based on endogenous factors. For example, through iconization the status of groups can be transferred to the linguistic traits associated with them, creating a „linguistic economy‟ in which certain ways of speaking are seen by speakers as more prestigious and desirable than others, shaping incentives that determine processes of language acquisition and change (Bourdieu 1977; Silverstein 1979; Irvine 1989). An example of how social hierarchies and linguistic ideologies interact can be found in Tzotzil Maya origin myths (recorded by Author in San Andrés Larráinzar, 1991). Here, the characteristics of the mythical founders of respective villages are sometimes evoked to explain the inter-community linguistic variation encountered. For example, regional mythology tells us that SanMiguel sat on a stone singing and playing the guitar while his older brother SanAndrés worked hard to establish their new home, San 2 Andrés Larraínzar. SanAndrés eventually grew tired of this and sent his younger brother away, who then went on to found San Miguel Mitontic, and for this reason people in San Miguel today speak in a singing voice1. Both, the singing voice (and the implicit “laziness” expressed by it) and the older / younger brother dichotomy inscribe a social relation between members of the two communities. ‘Its’inal and bankilal, “younger” and “older” brother [of a male individual], are terms that establish hierarchical relations between Tzotzil men. The two kinship terms are used beyond kinship relations when two men of different social status address one another, establishing a common framework for the ongoing interaction. Clearly, myth-based explanations of linguistic differences do much more than simply explain or describe variations in speech patterns. This example shows how people describe and rationalize existing linguistic differences. The inverse is the case when people imagine linguistic differences based on the existence of real or imaginary social differences. Still, in both cases a clear relation between social groups and language differences is established, and as a result the two are hardly separable. In this paper we examine how people construct within group differences by focusing on how perceptions of linguistic differences and similarities interact with political and geographic boundaries. We are not concerned with the content of differences (how people talk), but with how people perceive dialectal distances and classify variation. We use quantitative methods to elicit perceptions of linguistic and geographic distances within a Tzotzil Maya municipality, combined with GIS analyses to explore how space and place interact with linguistic/social knowledge; we complement these methods with ethnographic observation and ethnohistoric research. Several of our findings are noteworthy: first, we show that notions of within-group linguistic differences are strongly influenced by people‟s perceptions of geographic distances between communities. This finding adds a new dimension to the study of language ideology, as it suggests a strong interrelation between folk knowledge of geography and perceptions of linguistic difference. Second, and contrary to our initial expectations, our data show that the center for these social/linguistic distance estimations is not a speaker‟s speech community (Ego‟s 1 Similarly, Irvine (1989) documented how Wolof villagers of West Africa explain linguistic variation by way of invented migration histories, seemingly also adhering to the one people one language theory. 3 position), but the (presumably) pre-colonial center of the municipality. It is in this area that people locate the best and purest Tzotzil spoken, and it is to this area that they ascribe the heart of their ethnic identity. Given the post-colonial relocation of the center town, the cabecera, by non-indigenous people (Mestizos), we argue that this reflects both the construction of a local identity as well as a social criticism on past and present ethnic relations. Identifying the community identity with a pre-colonial center rather than the contemporary political center of the municipality, the cabecera, people undermine the symbolic function of the latter. Third, we show that social identity is not only conceptualized in dichotomous terms (as in „we vs. them‟), but also be understood in continuous terms, radiating from a central place. This model is historically contingent in that it includes interspersed groups of people as different, reflecting the incursion of past state-sponsored population shifts, land tenure restructuring, as well as contemporary religious and political processes of group formation. By mapping linguistic differences onto space, people create places and place-based identities. Hence our research adds to our understanding of space, place and the construction of social identity. Fourth and finally, we introduce new ways of connecting local ethnographic data with experiments from the cognitive sciences and GIS data. It is this combination of methods and theories that allows us a better understanding of processes of ethnic identity formation, both in the specific location of our research as well as in more abstract terms. More specifically, combining findings from the cognitive sciences on how people conceptualize places and regions (McNamara, 1986) with our GIS methods, allows us to explore the role of spatial regions for the development and conceptualization of social differences. 2. Linguistic Ideologies: ‘Local’ vs. ‘Ethnolinguistic’ Early work on language ideology focused on ideologies emerging from contexts of nation-state building or colonialism in which the hegemonic discourses of dominant groups are imposed on linguistic minorities, driving processes of language erasure and change (cf. review by Woolard and Schieffelin 1994). A more recent and growing body of research has shown that ideologies can also be articulated by minority groups as a means of contesting or resisting language shift (Woolard 1989; Gal 1993; Urla 2012). An example of the later are the language revitalization movements emerging in Latin America 4 since the late 1980s. Many of these movements seek to reverse language loss by promoting language standardization and grammatical prescription of native languages. Language revitalization activists recognize that, in critical cases, processes of loss and shift can only be averted through a combination of formal linguistic documentation and native language education, which often are only possible with the institutional support from governmental and nongovernmental organizations. Many scholars have noticed, however, that the use of linguistics and formal schooling by revitalization movements is rooted in contradiction and tension since these movements “are often structured around the same notions of language that have led to their oppression and/or suppression” (Woolard and Schieffelin 1994, 60–61) – that is, revitalization activists rely on purist conceptions of standard languages that associate literacy, eloquence, and prestige, and which emerged in the 18th century rise of modern states and academic linguistics (Joseph 1987). The Pan-Maya Movement of Guatemala, for instance, has sought to reverse language shift brought about by decades of assimilationist policies by the Guatemalan government by advocating the idea of a unified (pan) Maya ethnic identity (Fischer and Brown 1996). Pan-Maya linguists discourage the use of localisms, borrowing, and code-switching, instead favoring standardized forms and native neologisms. Although the movement has generally been praised and was in some cases successful at halting language shift (e.g. Collins 2005; Barrett 2008), some Maya communities have been reluctant to join, regarding the movement‟s ideology of unification as a threat to local identities. For instance, the decision by linguists to classify Achi as a dialect of K'iche' rather than an independent language (as its speakers see it) brought to light centuries old animosities between the Achi speakers of Rabinal, and the larger K‟iche‟ community (England 2003). As Reynolds (2009) and French (2010) argue, tensions between lay Mayans and Pan-Maya linguists stem from differences between „local‟ (or place-based) and „ethnolinguistic‟ ideologies shaping each group‟s understandings of linguistic variation. Maya groups have traditionally based their identities not so much on shared language or customs, but on place and territory, with the pueblo or municipality being the locus from which identity is contrasted to adjacent groups (Tax 1937). Pan-Maya activists trace these „local‟ ideologies back to arbitrary administrative divisions imposed by the Spaniards and argue that dialect 5 leveling and the adoption of supralocal (standard) varieties can be a means to recover the „original‟, preColumbian community boundaries (Cojtí Cuxil 1996). While most works on language ideology have focused on „ethnolinguistic‟ ideologies advanced by revitalization movements, few have paid attention to the „local‟, more unsystematic notions of language as held by lay Mayan speakers. Contrary to revitalization movements, for those lay speakers language variation is usually regarded as unproblematic and natural. For instance, Kroskrity (2009) relates that Western Mono speakers regard linguistic variation and syncretism as the normal outcome of family groups, individual differences, and exchanges between them. Similarly, Mannheim (1991, 94–108) notices that bilingual Quechua-Spanish speakers hold conflicting views regarding heterogeneity in both languages: while variation in Quechua is perceived as natural and non-hierarchical, Spanish is associated with hypercorrectness, as speakers understand that competence in Spanish can be translated into increased access to the Peruvian state‟s bureaucracy. „Local‟, non-standardizing ideologies have been overlooked for two reasons: first, in contexts in which metalinguistic awareness is low, ideas about language can be seen self-evident to speakers, which makes it difficult for linguists to elicit them directly (Woolard and Schieffelin 1994, 54). Second, linguistic ideologies in which variation is seen as unproblematic are unlikely to drive language change, which in turn makes them less noticeable by linguists. For example, Mannheim (1991:96-97) notices that Quechua speakers have no incentive to copy or borrow from different varieties since they are considered to have equal status. Though ethnocentrism toward one‟s own variety exists, speakers only prescribe local varieties to outsiders who migrate to their villages. In the absence of an agreed upon notion of „standard Quechua‟, speakers refrain from prescribing varieties beyond their corresponding geographical boundaries, thus maintaining variation between communities relatively stable. Our goal here is not to argue in favor of either „local‟ or „ethnolinguistic‟ ideologies; rather, we focus on the first and examine the mechanisms through which notions of place and space and the 6 perception of linguistic differences are bound together by speakers2. Research in the field of perceptual dialectology has explored the identification of dialects with imaginary geopolitical boundaries (Niedzielski and Preston 1999; Williams, Garrett, and Coupland 1999; Diercks 2002), showing that language and territory are weld together as important markers of national or regional affiliation. Following this line of studies, our research shows that this relation of place and social identity also determines within-group language differences.3 Importantly, we find these effects even in the absence of clearly demarcated boundaries, hence incorporating more localized processes of social order, regional history, changing power structures and resulting processes of identity formation. We focus on how people envision language differences of different local communities within their own municipality. We argue that such perceived differences do not constitute value-neutral observations, but represent commentaries on an existing social order placed within a specific historical and geographical setting. This is particularly true as we elicited comments from our informants about places with which they often had limited familiarity and hence the ascribed language variations do not represent actual observations; our interest is not about perceived language differences per se, but about the social construction of within community differences as expressed by perceived language differences. Hence our focus is on the underlying structure and ideology that leads individuals to identify or imagine such differences, the mechanics behind the ideological production. This includes both abstract (spatial distance) and more specific principles (social organization) that link people‟s reasoning about language and social life. We argue that one cannot be understood without the other. 3. Chenalhó and the Highlands of Chiapas Chenalhó is a predominantly Tzotzil Maya municipality of about 36,000 inhabitants in the Highlands of Chiapas (Mexico). The center town (cabecera or teklum) is home to approximately 3,000 As Field and Kroskrity (2009) argue, a better understanding of „local‟ linguistic ideologies can actually help revitalization movements by improving the dialogue between naïve and expert speakers. 3 The Tzotzil-Maya of this study relate to the macro level of this relation when they refer to their language as bats‟i k‟op, “the real or autochthonous language,” a glossonym related to bats‟i vinik, or “real or autochthonous men”. 2 7 people, located in the base of a valley connecting the hinterland as well as other municipalities to San Cristóbal de Las Casas, the largest urban center in the area. A former colonial center, San Cristóbal today represents one of the major tourist attractions of the area. For the last (almost) 500 years it also constituted the main marketplace both for selling and purchasing goods for the surrounding rural communities. Chenalhó‟s cabecera was founded in the late 19th century by a small group of nonindigenous Mestizos (kaxlan4, for the Tzotzil Maya), setting up home and trading posts in the area, exploiting their position as go-between the city and rural indigenous communities (see Colby and van den Berghe 1961; Köhler 1980; Ross 1997). Today a small group of Mestizos share the cabecera as home with mostly bilingual Tzotzil Maya speakers, who moved there mainly for employment opportunities and to be closer to the municipality‟s only preparatoria (High school). Outside the cabecera, settlements disperse into small hamlets (parajes or comunidades), with households spread out between agricultural fields. Here, Tzotzil is the primary, when not the only, language spoken. Chenalhó includes over 100 parajes in an area of approximately 13,000 ha. The first linguistic map of the Highlands was drawn in the 1960s, dividing Tzotzil into four dialects (Hopkins 1970). In this map the Tzotzil of Chenalhó was grouped with the dialects of Chamula, Mitontic, Chalchihuitán, and Zinacantán as “Central Tzotzil”, while the Tzotzil spoken in Santa Martha, was grouped with Aldama and Simojovel as “Western Tzotzil”. Today Santa Martha is, administratively, part of the municipality of Chenalhó, yet people usually do not regard the inhabitants of Santa Martha as “Pedranos” (named after San Pedro Chenalhó), because they maintain their own „cargo system‟, fiesta cycle, and traditional dress (Brockmann 1992). Not surprisingly, Pedranos comment that the Tzotzil spoken in Santa Martha is different from the one spoken in Chenalhó “proper”. It is interesting to note, however, that a more recent dialectal map – done by the Mexican National Indigenous Languages Institute – no longer distinguishes the two localities (INALI 2009). We surmise that this change in the linguistic registry does not reflect actual language shifts, but represents changes in perceived linguistic variations, here the perception of the 4 Unlike the better documented Tzotzil variety spoken in Zinacantán, Chenalhó speakers do not use the agentive prefix j-; hence our spelling of kaxlan rather than jkaxlan (Haviland 1981, 76). 8 linguist. This reflects just the kind of language ideology we are interested in: changes in the political geography (the formalization of Santa Martha as a “sector” of Chenalhó) led to changes in the linguistic differentiation, which, of course, reflects the map-makers‟ ideologies. As discussed above, indigenous municipalities in the area consist of dispersed settlements unified administratively by a town administrative center (cabecera). In the past these centers have been described as “vacant towns”, which indigenous people visited to barter goods and services or to fulfill a cargo (Vogt 1961). Life certainly has changed: today the cabecera of Chenalhó is a small rural town undergoing rapid modernization (Shenton et al. 2011). Chenalhó‟s first mention in colonial archives dates back to 1571 (Guiteras-Holmes 1961; Calnek 1988). While this established a community that more or less overlaps with the contemporary municipality (as indicated by parochial letters), nothing is known in terms of its actual size or sociopolitical organization, and the exact boundaries of the town have shifted several times, including the already-mentioned incorporation of Santa Martha in the 1930s (Garza 2002) as well as the re-municipalization of Magdalena, which had historically been a part of Chenalhó until 1998 (Burguete Cal y Mayor 2007). Following Mexican independence, all vacant lands were declared national property and were normally sold. Whether the land was in fact vacant was frequently ignored, and land auctions often took place in the absence of actual inhabitants (see Author 1997). Thus in the 19th century large farms owned by Mestizo and foreign farmers emerged; these landowners recruited the now landless and unemployed indigenous peasants as work force. During land reform in the 1930s, several of these fincas were expropriated and transformed into ejidos (communal lands) (Rus 1995), some of which would later be converted into some of the communities included in the present study. Communities that make up Chenalhó today differ in age as well as the history of their incorporation. Older communities were usually founded as kin groups around waterholes and, as a result, kinship plays an important role there. Newer communities are typically the consequence of population growth and recent political fissures. All communities have equal legal and political status, each being led by a locally appointed agente municipal. Communities enjoy a significant degree of autonomy from the 9 town‟s authorities: most decisions about land or resource management take place locally. The municipal authorities usually only interfere in local affairs upon request of community residents. Travel between them is mainly for visits to the market, religious rituals (for various denominations), to visit distant family, or for participation in political meetings (more common among men). This limits traveling to nearby communities, although the center town and, increasingly, San Cristóbal also constitute targets of local and regional travel. The cabecera was established around the mid-19th century when a group of Mestizo merchants settled there and developed it into a semi-urban center (Arias 1990). Notably, this previously uninhabited area is located at the municipality‟s southern border, where its location allows for the control of access to and from San Cristóbal (see map). Today, the cabecera constitutes the primary destination of internal migration for individuals in search of wage-labor or better schools for their children. This process accelerated considerably since the 1950s, when the Mexican government, through the Instituto Nacional Indigenista (INI)5, began to build roads, schools, and medical clinics in the area. While today people are used to traveling within the municipality, it is not uncommon to find individuals who seldom leave the town or even their community. Only a minority of individuals (mainly men who have held public office at one point) have traveled to all the communities within the municipality. Hence, at the municipal level Chenalhó functions as an imagined community bound together by notions of collective political membership that are realized through service in the civil-religious offices. For instance, the over 100 agentes municipales meet sporadically in the cabecera to hear reports or vote on matters regarding the municipality. Meetings like these work to create a sense of common purpose and to disseminate information between communities. Thus in spite of a general lack of direct personal experience, individuals can have a good understanding of the location and economic basis of most of the communities of their municipality. Dialectal differences between communities exist, but associations tend to be covert – i.e., although people acknowledge the existence of variation within the town, they cannot necessarily specify 5 The National Indian Institute is the Mexican correlate to the Bureau of Indian Affairs in the US. 10 what the content of those differences is. For instance, some speakers from the Santa Martha region substitute /v/ for /h/ in some words, a characteristic associated with the Tzotzil spoken in San Andrés Larraínzar (Hopkins 1970; de Delgaty and Sánchez 1978). Since phonological awareness tends to be low, differences like this do not index group identity or prestige, at least on a conscious level6. Awareness of lexical differences is greater since they are more likely to affect comprehension. Still, these differences are usually associated with varieties spoken by adjacent ethnic groups rather than communities within Chenalhó. For instance, when we ask people about different ways of speaking Tzotzil, they often respond with anecdotes about how the neighboring Chamulas have different words (e.g. for „foot‟ or „uphill‟), or how their lexicon is more ancient/primitive. As we noticed earlier in the mythology of Larraínzar, speech from different municipalities is often associated with prosodic features that correspond ethnic stereotypes. This level of specificity, however, is never reached when people are asked to describe dialectal variation between Chenalhó‟s communities. In addition to low levels of awareness of within-municipality differences, it is considered impolite to talk about the subject, either in private or public settings. Since Chenalhó has a history of internal conflict (see section 4), the mere allusion to differences of any kind could bring to surface latent political tensions. Not only this makes the direct elicitation of perceived differences difficult but also shows that Chenalhó as an imagined political entity produces a process of erasure that discourages people from elaborating metalinguistic discourses on internal differences. Revitalization movements such as the Pan-Maya are unheard of to most Pedranos7. Whereas Collins (2005) finds that Mam teachers of Guatemala are less likely to code-switch, here we find the opposite pattern: code-switching increases with educational attainment. Only few people comment on the use of Spanish borrowings, and by and large competence in Spanish marks social status. As the town‟s 6 Cargoholders from Santa Martha usually reacted with surprise when the author called their attention to phonemic differences, which suggests that phonological awareness is low. Still, Marteños have no problem in acknowledging that they speak a different Tzotzil dialect – since logically, “they are Marteños” – even when they fail to specify the content of these differences. This shows that social categories are more likely to determine perceptions of linguistic difference than the opposite. Future research may, of course, reveal covert associations between some registers and prestige, though this is not our goal here. 7 Following the Zapatista uprising, indigenous organizations of Chiapas (such as Las Abejas) have taken the form of „civil societies‟ that define their identities as syncretic, rather than „Maya‟ (Tavanti 2003). As a result, Chiapas‟ indigenous movements have been less focused on cultural or linguistic revitalization than their Guatemalan counterparts. 11 written bureaucracy functions entirely in Spanish, being able to speak and write documents in Spanish is a necessary skill for climbing the social ladder. Conversely, very few individuals write in Tzotzil, which most people (excepting bilingual teachers) consider a useless or unattainable skill. Most primary and secondary schools follow the Mexican “intercultural bilingual education” model. This model, however, does not aim at aiding language preservation, but its goal is simply to create a smooth transition into teaching in Spanish (by approx. 6th grade). In fact, Tzotzil Maya is still referred by many (mestizo and indigenous people alike) as „dialect‟, or in other words not a complete language (del Carpio 2012). Moreover, teachers tend to be mobile: many come from Spanish or Tzeltal speaking areas, with Tzotzil being their second language; others speak drastically different Tzotzil dialects, hindering at times mutual understanding8. In spite of those problems, the number of Tzotzil speakers in the town has been increasing steadily over the past decades due to demographic growth and since bilingual speakers tend to move to Spanish speaking urban centers. Nevertheless, Tzotzil remains a „vernacular‟ to most people. 4. Spatial organization of settlement groups in Chenalhó Through interviews, as well as existing lists and maps provided by the municipality, we sampled a list of 46 communities located within Chenalhó. These communities vary in size (from 50 to about 3,000 inhabitants), location (lowland / highland), their degree of urbanization, and the time of and the reason for their establishment. They are located in an area of 251 km² with altitudes ranging from 722 to 2032 meters. Although most are hamlets dedicated to agriculture, raising cattle has become an important activity. Corn, beans, squash, and chili peppers are the main cultigens intended for local consumption. Since the 1970s coffee has become the most important cash crop, followed by corn and some fruit trees. Given the rugged geography, agricultural production is adapted to different climatic zones and altitudes (Mariaca M ndez 2007). While this makes awareness of the landscape extremely important for farmers, a 8 For instance, the author once saw a meeting between office holders of a community (most monolingual Tzotzil) and a bilingual teacher from the municipality of Huixtán. As the officers struggled to understand the teacher‟s Tzotzil variety, they frequently had to shift to Spanish as a lingua franca, even in spite of their limited knowledge of it. 12 general k’ixin ‘osil and sikil ‘osil (hot and cold land) distinction also serves as heuristic for reasoning about the social status of communities (e.g. how rich or industrious they are; see Author, n.d.). To explore the spatial organization of Chenalhó‟s communities, we used a GIS-based map to calculate distances9 separating all communities from one another. Distances between communities represent only one aspect of spatial organization. People do not inhabit abstract homogenous space, and hence settlement discontinuities are often inflated by geographical features that impede easy communication (e.g. mountain ranges). We explored the existence and role of such geographical regions, investigating whether and to what extend they inform peoples‟ construction of social identities. In effect, we were interested in whether geographical regions and settlement discontinuities also represent conceptual regions, that is, discontinuities that mark social differentiation. To explore discontinuities of actual settlement patterns, we employed a K-means cluster analysis on a community by community distance matrix. This analysis compares individual distances between settlements in terms of continuities and discontinuities to construct possible settlement clusters that form areas or regions by way of their locations. The analysis reveals the existence of three clusters/groups of communities. A visual inspection of the map of Chenalhó confirmed the relevance of these clusters (see map1), which we will briefly describe next. Fig 2: Map of the municipality borders, roads, study communities, and settlement clusters 4.1 Group 1: Western Region. Some of these communities have once been part of other municipalities and only formally joined Chenalhó during the 20th century. The area in the northwest surrounds Santa Martha, encompassing the communities of Atzamilhó, Saclum, Yocventana, Macuxtetic, San José Tepeyac. Santa Martha and the neighboring town of Aldama (Magdalena) were Jesuit 9 We calculated both Euclidean and actual road distances. While cars are increasingly common, a network of pathways is still more important than roads in connecting communities. Still, the two measures are highly correlated. We used maps published by Mexican National Institute of Statistics and Geography (INEGI). 13 reductions10 established over the preexisting settlements of Chupic and Tanjoveltic (de Vos 1994, 207). By the early 20th century most of the land of Santa Martha and Aldama consisted of fincas11, expropriated and incorporated to Chenalhó‟s territory as agencias municipales in the 1930s (Garza 2002). The same happened to communities of San Antonio Caridad and Jolxic. As mentioned above, people from Santa Martha are still not regarded as proper Pedranos; Similarly, anthropologists have usually regarded them as a distinct ethnic group existing within the political boundaries of Chenalhó (Guiteras-Holmes 1961, 341; Brockmann 1992). The same goes for Belisario, a former finca dismantled in 1935 and incorporated into Chenalhó‟s territory in 1968 (Cruz Jiménez 1996). Staffed with workers from neighboring Chamula, those people and their families today form a “colony” of Chamulas within Chenalhó. The Cabecera is the largest urban area within the municipality and the only settlement founded by Mestizos and with a Mestizo population, while Natividad, a former finca founded by immigrants from diverse areas is today classified by the Mexican government as a Tzeltal-speaking community (INALI 2009). Group 2: Central area. This group corresponds to the surroundings of the large settlement of Yabteclum (population: approx. 2000). Yabteclum translates as “the place of the (future or past) town/cabecera.” Oral history suggests that this may indeed have been the previous administrative and ceremonial center of the community (Pérez Enríquez 1998, 41); Several observations are interesting in this regard: First, our central feature GIS analysis revealed Yabteclum as the most central community in Chenalhó (shortest average distance to all other communities). Second, within this general area no finca has ever been established, suggesting a long history of fairly large continuous settlements impeding the establishment of fincas; this in turn points at a relatively high local importance of this location in previous eras (Garza 2002). Together these points establish the importance Yabteclum must have had in the past. (We will come back to this issue in the next section.) Group 3. Eastern region. As in group 1, some of the communities in this cluster were former fincas that became ejidos during the 1930s. The best-known examples are Puebla, Majomut, and Los 10 Loosely speaking, reducciones are areas where people were (mostly forcefully) gathered to live. Bringing people into closer proximity allowed for easier control, conversion and taxation. 11 Large landholdings for agricultural production. 14 Chorros. Polhó is the largest and probably best known Zapatista enclave12 within the boundaries of Chenalhó. Outside observers recognize Acteal as the stage of a massacre of 45 members of the Las Abejas organization in 1997 by a paramilitary group (see Leyva Solano and Gracia Castillo Ramírez 2012). Los Chorros (Miguel Utrilla) and Puebla are large communities, which until the 1930s used to be fincas (Los Chorros, Tanaté, and Los Ángeles) that drew indigenous workers from different municipalities. Today Los Chorros is well known as harboring a neighborhood (barrio) of Tzeltal speakers (Köhler and López Comate 2001; Arias, n.d.). The communities in the northeastern boundary of Chenalhó are the least known among our participants and the least populated in the town. Letters written in the first half of the 19th century show that once the region was the site of a land dispute between residents of Pantelhó and Chenalhó. The area is well known for higher temperatures and a resulting emphasis on coffee production. It is not uncommon for people in the highlands to own land in this area or to visit it sporadically to work on coffee farms. As can be seen from the discussion and the map, these regions emerged from a confluence of both geomorphological (mountain ranges and rivers) and historical factors (time of settlement or the emergence of fincas) making up today‟s landscape of the municipality. These data provide the background of our data analysis and the interpretation of its results. We turn next to the exploration of perceived distances and linguistic differences and the role of regions in the construction of these perceptions. 5. Mapping Perceived Geographic Distance and Perceived Linguistic Variation. A large body of literature exists exploring the relation of language, space, and social identity. Methods used in „perceptual dialectology‟ studies are often borrowed from cultural geography: people are shown maps and asked to circumscribe dialectal boundaries (for a review, see Preston 2010a; for alternatives, see Inoue 1996; Diercks 2002; Tamasi 2003). We approach these methods critically. Given that maps represent territories it seems hard to avoid these territories and not simply ascribe respective 12 With Zapatista we refer to followers of the rebellion that started in Chiapas in 1994. 15 language differences to them. Furthermore, maps have been the main heuristic tool in Western nations to convey geopolitical information; hence it is not surprising that dialectal boundaries drawn during interviews often coincide with the very same geopolitical boundaries acquired through maps (a common finding in perceptual dialectology). Moreover, the scale of the areas used for such studies is often so large that a person could only have learned about it through maps: large-scale regions such as “Western Europe” or “Southern United States” can only acquire linguistic relevance through the use of modern cartography and mass communication means. As Preston recognized, “it has perhaps been inappropriate to provide respondents with even minimal outlines of areas on which to draw their mental maps of linguistic differences” (2010:196). To avoid these problems we measure perceived linguistic and geographic distances employing a triad-task following Author (2005). Triad tasks allow eliciting conceptual distances indirectly and without having to represent concepts visually. The downside of the method is that a complete triad design (with each possible constellation of triads explored) is almost impossible to obtain due to the high number of combinations involved. To keep the interview length manageable we split the overall interview into three sections.13 For each of the triads the participants were asked the following questions: 1) Perceived Geographic Distances (PGD): Which of the three communities is the farthest away from the others? Which are the two communities that are closest to one another? 2) Perceived Linguistic Differences (PLD): In which community do people speak more differently from the others? Which are the two communities where people speak more similarly14? 13 For this we divided the 46 communities in three random groups. Two communities, Cabecera and Los Chorros were included in all three groups. Not only are they among the largest communities in the municipality, but also they are usually ascribed to opposite poles of within-municipality comparisons. We employed a non-random balanced incomplete block design with lambda 3 and 17 items per questionnaire (Burton and Nerlove 1976). Lambda 3 indicates that each pair of locations is compared three times, each time with a different third location. This design yielded three sets of 136 triads or interview questions per sitting. We later combined data from the three triad tasks, generating 46 x 46 distance matrices. This is was possible by using distances to and from cabecera and Los Chorros as common points of reference to estimate the missing distances with an algorithm by Makarenkov and Lapointe (2004). 14 In Tzotzil, we phrased this as: “Li talem oxox sbi parajetik [referring to the paper forms], k’u cha’al [interviewer reads the names of three hamlets]. Ta yoxibalik, boch'o mas ko'ol xlo'ilajik? Boch’o mas jelel xlo’ilajik?” When 16 PGD and PLD interviews were conducted on separate occasions. Answers were coded according to the community chosen as different; “I don’t know” (or if a participant was not familiar with a community) was coded as missing data15. Interviews ranged from 20 minutes to an hour; all were conducted at people‟s homes and in Tzotzil Maya by the author, a collaborator, and two local research assistants. 5.1 Participants. We conducted a total of 251 interviews across the three triad-tasks. While the PLD task was conducted in two different sites, the PGD sample comes only from the cabecera. The sites sampled with the PLD task were (1) Linda Vista, a small rural hamlet with a population of approximately 300 people located in cluster 2, and (2) the cabecera. To get from the cabecera to Linda Vista by car takes approximately one hour (see map). Including participants from two locations allowed us to explore whether people conceive of linguistic variation as simple diversion from their own speech community. Within each community sampling was random (among adult members) and restricted only by an individual‟s willingness to participate and her basic familiarity with the 46 communities. 6. Results 6.1 Analysis of Consensus. Consensus analysis allowed us to both estimate agreement and explore systematic disagreement16. If agreement across individuals is high, we can capture all individual respondents cited the name of a hamlet as being different than others, we asked them whether they thought that the remaining two hamlets resembled each other (Mi ta schi’in sbaik [hamlet1] xchi’uk [hamlet2]? [or] Mi xko’olajik? [or] Mi ko’ol xlo’ilajik?) After about 10 questions we asked people to explain the task back to us, to be sure they understood the instructions. We repeated the previous questions or the instructions (in different terms) whenever necessary. We also intermittently asked people to explain their answers (k’u yu’un?), to be sure they were focused on the task and to collect additional information on respondent’s reasoning patterns. 15 Note that “missing data” or similar distances between all three communities have no impact on the resulting community*community distance matrix. 16 An inter-informant agreement matrix (who agrees with whom, how much) provided the input for both, the consensus analysis and the subsequent analysis of residual agreement (see Author 2004). We used the formal consensus model (CCM) for multiple-choice questionnaires to measure agreement among groups of participants and the extent to which each individual agrees with the overall model elicited (Romney, Weller & Batchelder, 1986). The CCM is a principal component analysis over the inter-participant agreement matrix. Consensus is found to the extent that the data overall conform to a single factor solution (e.g. the ratio of first and second factor eigenvalues is large (>3), the first factor explains a large amount of variance, and all participants‟ first factor loadings are high and positive). In this case each participant‟s first factor score represents her participation in the consensus, her agreement with everyone else (Weller, 2007). 17 responses in a single model, the average response of our participants. While no single individual might completely agree with this aggregate answer, everyone has to be agreeing strongly with the aggregate, or else there wouldn‟t be consensus (see FN 16). In case of agreement, each individual‟s first factor score represents his or her agreement with the aggregate model. The model invites one to think about both systematic agreement and disagreement, as general agreement may be coupled with systematic disagreement, which can be explored with an analysis of residual-agreement. Residual agreement is the agreement not explained by two individuals‟ participation in the overall consensus (Boster and Johnson 1989; Author 2004). Said differently, if two individuals agree with the overall aggregate response key in 90 of 100 answers, then the model predicts that they agree at least in 81% of their answers with one another17. However, often people agree more with one another than what the model would predict, allowing us to explore the existence of submodels and systematic disagreement 18. 6.2 Perceived Geographic Distances: We found consensus for each of the triad tasks, e.g. subsets of communities (table 1), indicating that individuals agree on the geographic distances between different communities. While women have significantly lower first factor scores than men (0.53 vs. 0.66; t = -3.3376; p = 0.001), residual analysis does not reveal any gender related submodels. This suggests that women know less (show a greater likelihood of guessing) than their male peers. This interpretation is confirmed by the fact that women‟s responses correlate lower with the correct answer key (from the GIS model) than men‟s responses (70% correct answers vs. 61%; t = -3.5874; p = 0.000). This, in turn, is not surprising, as women‟s travel is often restricted to their local community. 17 The notion of residual agreement might be best explained with respect to test taking. If two students study hard yet independently for an exam (with yes / no answers), their test scores will predict how much they agree with one another. This agreement is simply an outcome of both knowing the right answers. Any systematic agreement beyond the test scores (residual agreement) cannot be explained as an outcome of their knowledge of the right responses, but must be explained otherwise; e.g. they studied together and both didn‟t understand certain concepts or one copied the exam of the other. The same logic applies to our analysis. Although we are not dealing with right wrong answers, what “everyone knows” – easily available knowledge does not tell us much about specific group processes of knowledge acquisition. However, it is the systematic deviations thereof – not the random mistakes – that give us insights into the social acquisition of knowledge and hence provide us with a window into important processes. 18 Residual agreement is calculated by subtracting two participants‟ predicted agreement (the product of two participants‟ first factor loadings) from their observed agreement. The resulting residual-agreement matrix can be explored with respect to group differences (is within-group residual-agreement higher than between-group residualagreement?). 18 TABLE 1 HERE 6.3 Perceived Linguistic Differences: We found consensus for all three triad tasks both within and across Linda Vista and cabecera (Table 1) and residual analysis did not reveal any systematic group differences. That is surprising given that people in the center town not only come from a broad range of natal places but are also more likely to travel within the municipality. (Hence, we expected a heightened awareness and greater expertise of linguistic differences.) 6.4 Correlation between perceived linguistic and spatial distances. To examine the correlations between perceived linguistic distances, perceived geographic distances, and real geographic distances, (GD: Geographic Distance) we compared the different distance matrices with Mantel tests (Mantel 1967). The above-described consensus across the two populations formally allowed us to aggregate the data across the two research sites. We find the highest correlation between perceived geographic distances and real geographic distances (0.78, p < 0.001). This indicates that people have a very good understanding of relative geographic distances and locations of the different hamlets of Chenalhó19. We furthermore find significant correlations between both actual geographic distances as well as perceived geographic distances with perceived linguistic differences (r = 0.46 and 0.39, p < 0.001 respectively). These correlations show that people‟s estimations of between community linguistic differences are significantly influenced by their perception of geographic distances. We surmise that the prevailing ideology of linguistic differentiation in Chenalhó is based on the idea that linguistic difference increases with increasing geographic distance. 6.5 Perceived Regions. As seen earlier, the cluster analysis over the actual distances resulted in three clusters of communities. While one might be tempted to simply call these clusters regions, we were interested whether they relate to conceptual regions that is, whether these geographic clusters are 19 In this paper we use a simple Euclidean measure of geographic distances. We also calculated the shortest road distances and travel times (adjusted for road type), but these measures had lower correlations with perceived distances (0.58 and 0.34, respectively). 19 conceptualized as abstract categories. Research in spatial cognition has shown that conceptual regions lead to distortions of distances between objects. Objects belonging to a single “region” tend to be perceived as closer to one another, while objects belonging to different “regions” tend to be perceived as further apart from one another (McNamara 1986). To examine this for our clusters, we ran a cluster analysis on the 46 x 46 matrix of perceived geographic distances and compared results with the geographic clusters described above. The analysis (depicted in figure 2) shows that there is an overall consensus between conceptual and actual geographic regions20. The major difference seems to be the existence of a fourth conceptual cluster, which includes some communities situated in Santa Martha. That people single out this area as a spatial subregion within the Western part Chenalhó shows the influence of ethnicity on conceptions of space. Given that perceived and actual settlement clusters coincide, we predicted that people would overestimate distances between and underestimate distances within perceptual clusters. This was indeed the case, as shown by multiple regression models in table 2. We regressed perceived distance scores on actual geographic distances and a dummy variable in which „1‟ coded for distances within clusters and „0‟ for distances between clusters. The models show that overall distances within clusters are consistently underestimated within clusters 1 and 3. Cluster 2, however, stood out in that it had the lowest scores of within cluster underestimation scores21. FIGURE 2 HERE TABLE 2 HERE 20 This 4 cluster solution had the highest consensus scores when compared to the geographic. We used the Rand index (Rand 1971) to compare consensus across different cluster solutions (e.g. with two or three clusters) and chose the ones that had highest scores (Rand I = 0.92. This method is often referred to as „external validation‟ of partitions (e.g. Legendre and Legendre 2012, 24:416). 21 Most of distances within cluster 2 matched actual distances (a ratio close to 1). Some distances between localities in this cluster were overestimated. We attribute this effect to the existence of the two main roads that divide the cluster (see map 1) and hence the existence of potential sub-regions. 20 These models, together with the correlation tests and the cluster analysis of perceived distances, show that people‟s conceptual maps of Chenalhó are remarkably accurate and correspond tightly to the actual map of the municipality: not only are distances correlated, but communities are conceptualized as being part of regions that closely resemble actual settlement clusters. Differences between perceived and actual distances can be explained as distortions stemming from a perceptual regionalization and by the existence of a distinct ethnic group within the municipality (Santa Martha). As we show below, these distortions also have an effect on how people perceive linguistic differences within the municipality. 7. Geographic Centrality and Metalinguistic Evaluation As stated above, we initially assumed that study participants would report linguistic differentiation based on a simple algorithm equating geographic distance with language differentiation. The above data illustrate that this is indeed the case. However, further analysis of our data reveal an interesting twist. Rather than using their position as the center of these calculations, our participants (independent of their location) place the perceived linguistic center of the municipality around Yabteclum, a community in cluster 2. Statistically this relation is expressed in a linear regression where mean perceived linguistic difference is predicted by the distance from the central hamlet of Yabteclum (R² = .61, p = 0.000) but not the cabecera (R² = 0.03, p = 0.12). Of course, there is a mathematical relation between geographic centrality and linguistic centrality, given that geographic distance and perceived linguistic difference are related. Being the central place, Yabteclum maintains by definition the lowest average distance to every other community in the set and hence by extension should also show a low average linguistic difference with respect to all other communities. In this scenario, people would use abstract distances as a heuristic of calculating linguistic differences - independent of their personal positioning. Yabteclum‟s status as a linguistic center would be simply a consequence of it‟s location, with no additional ideological significance. 21 To further explore this idea we asked 24 people (14 male) to rankorder cards representing the 46 communities based on how well Tzotzil Maya (bats’i k’op) is spoken in each location22. Given the consensus found aross the two communities, we only interviewed participants from the center town. Results revealed a strong consensus23. An ANOVA showed that hamlets in geographic cluster 2 and specifically Yabteclum were ranked highest, meaning that participants ascribed the best-spoken Tzotzil to inhabitants of these communities (mean for cluster 1 = 1.026, cluster 2 = 1.919, cluster 3 = 1.125, F = 17.34; p < 0.000). The ranking results also correlated significantly with average perceived linguistic difference (r = 0.78, p < 0.000) and as a logical function with geographic distance from Chenalhó‟s center. This is interesting on several counts: rather than conceiving language differences egocentrically or from the viewpoint of one‟s speech community, people in Chenalhó invoke their shared knowledge of geography and history when judging both linguistic diversity as well as the quality of the Tzotzil Maya spoken in their municipality. The presumed linguistic center of the municipality is not today‟s cabecera – the most important community, both politically and economically – but Yabteclum, a community that is located at the actual geographic center of the municipality. This is further confirmed by examining how people classify language groups within Chenalhó. We performed a cluster analysis on a 46 x 46 matrix with the differences between linguistic and perceived geographic distances24. The results are shown on fig. 3. Three groups that are closely related to the geographic clusters detected earlier emerge from this analysis (see map, fig. 1). We found remarkably high consensus between linguistic and geographic clusters (using a 3-cluster solution, Rand index score = 0.94). FIGURE 3 HERE TABLE 3 HERE In Tzotzil: bu mas lek sna’ xlo’ilaj bats’i k’op li ta Chenalhó? Ratio 1st/2nd factor = 7.99, percentage of variance explained by first factor: 57%; average first factor score = 0.74. Here we used the informal version of the cultural consensus model (Romney, Batchelder, and Weller 1987; Weller 2007). The informal model replaces the matching agreement matrix with a correlation matrix, a more adequate way to handle ordinal rankings. 24 We normalized both matrices and then we subtracted PGD from PLD. This was done to control for the influence of perceived distances on how people categorize speech communities within the municipality. The cluster analysis graph should be interpreted as a map of residual linguistic differences (excluding perceived geographic distances) between communities. 22 23 22 The ranking data and the cluster analysis results suggest that our study participants do see the region near Yabteclum as a linguistic region overlapping with the above-established geographic region. Finally, we were interested in exploring whether the above identified perceptual geographic regions would carry over to an understanding of linguistic regions. More specifically, we theorized that only cluster 2 would be considered a linguistic region, from which language change emanates with increasing diatance. In this scenario people would describe communities within cluster 2 linguistically as more similar to each other than predicted by the geographical distance model (applied by them when calculating linguistic diversity). For this we repeated the regression analysis of within/between cluster estimation of distances, exploring the degree to which perceived linguistic distance deviates from perceived geographic distance. We show the results on table 3. Regression models show that within cluster linguistic distances are underestimated for cluster 2 (distances are .20 lower within the cluster) but overestimated (up to .15 higher within clusters) for 1 and 3. This demonstrates that this region is perceived as more linguistically homogenous, whereas the other two sectors of Chenalhó are seen as more heterogeneous (e.g. linguistic distances are overestimated within clusters). The data also show that people exaggerate linguistic distances from cluster 2 to all other clusters, an effect that overlaps with the findings discussed above for perceived geographic distances. Together these data suggest that Tzotzil Maya of Chenalhó employ both their notion of geographic distances as well as their concepts of geographic regions when reasoning about linguistic differences. Of course, the heavy use of spatial cognition when reasoning about language diversity does not imply that people do not also employ other concepts when analyzing how other people speak. We turn now to some cases where the geographic distance model does not predict linguistic variation. 7.1 Fincas as Exceptional Cases. The model emerging from the above data allowed us to explore which communities violated the expected patterns, e.g. for which communities the perceived linguistic differentiation fell outside what the model based on geographic distances would predict. To do 23 so, for each community we created a simple index of spatial and linguistic differences: the ratio of perceived linguistic difference to perceived geographic distance. We were specifically interested in the communities for which the perceived linguistic differences are larger than expected based on the perceived geographic distances, as this indicates that these communities are marked as special, different from the normal pattern of language differentiation. Results were revealing. All communities for which linguistic variation is higher than predicted by the spatial model are either former fincas or part of Santa Martha. As discussed above, people of Santa Martha are generally viewed as representing a separate ethnic group living within the confines of the administrative unit of Chenalhó. With respect to former fincas, their creation as communities usually led to the inclusion of people not native to the area (see for example the above described case of Los Chorros). In fact people in Chenalhó often refer to these communities as “mixed”, or “already mixed” (kapalik xa), referencing the fact that they regard these places as populated by “non-native speakers”, immigrants who speak a different linguistic variety of Tzotzil or Tzeltal25. This assertion was especially common among participants from Linda Vista – itself located in the perceived linguistic and geographic center region of Chenalhó (cluster 2). As discussed above, fincas were indeed a major factor of displacement of peasant families during the early 19th century. However, it is striking that this history remains very much alive for people in Chenalhó. More than a century and a half later these communities are still marked as non-native speakers and seen as “mixed,” despite the fact that many are indeed native Tzotzil Maya speakers. It is important to repeat that this conceptual markedness not only shapes people‟s assumptions of how others speak, but also influences their evaluations of how well they speak. Clearly then, the perceived differences aided in maintaining the conceptualization of these communities as different, and not in a good way. 9. Discussion: Understanding identity in relation to place and space 25 Some of these histories of migration are well documented. See for instance the case of the Finca San Francisco Paula, which by the late 19th recruited immigrants from as far as Guatemala to pick coffee seasonally (Cruz Jiménez 1996, 71). After the the finca was dismantled, residents from many different hamlets of Chamula and Magdalena (Aldama) asked the Mexican government to formalize the status of their settlement as a community and to acquire formal tenure of the land. In 1935 the finca became the Colonia Belisario Dominguez, which three decades later would be annexed to Chenalhó‟s territory. 24 Language ideologies refer to complex associations between metalinguistic representations and beliefs related to the social world into which speakers are enculturated. Although Irvine and Gal (2000) have proposed that iconization of linguistic difference be an important step in the formation of language ideologies, little has been said about 1) how representations of language differences relate to cultural/cognitive domains other than social identities and classification, and 2) what are the underlying cognitive processes aiding their production. Similarly, studies in perceptual dialectology have been less concerned with the actual processes of knowledge production but are usually satisfied with describing differences in how folk linguistic knowledge is represented, mapped, and evaluated. In contrast, our study focused on exploring the algorithms by which people make decisions about such variation. We were interested in identifying heuristic principles that guide the ways humans explore, describe and define the social landscape surrounding them. Work in language ideology has described the production of identity as a dichotomous process in which groups emerge by being contrasted to others. Bucholtz and Hall (following Bourdieu) call this phenomenon distinction, which “operates in a binary fashion, establishing a dichotomy between social identities constructed as oppositional or contrastive. It thus has a tendency to reduce complex social variability to a single dimension: us versus them” (2003, 384). In the case of Tzotzil Maya of Chenalhó, it appears that the heuristic principles driving the production of identity within groups are inextricably linked to people‟s spatial reasoning. Participants of our study use spatial distances as a template upon which they make inferences about linguistic difference and identity, which can aid the production and maintenance of social identities. In a similar fashion, speakers also use conceptual regions to locate language differences and similarities. Such preconceived regions become more coherent and in the process – their differences to other units become exaggerated both with respect to actual spatial distances as well as linguistic differecnes. Our data suggest that difference comes with a value (or degree) attached, as the Tzotzil Maya we interviewed clearly envision a linguistic center from which the spoken Tzotzil not only varies, but actually deteriorates with increasing distance. This spatial way of reasoning is complementary to binary 25 dichotomous thinking. As we saw, linguistic distances from Chenalhó‟s center to fincas tend to be greatly exaggerated, which stems from the fact that some of these communities are marked as being other, in opposition to us, following a dichotomous logic. However, in the absence of such clear-cut, agreed-upon boundaries separating groups, people use spatial knowledge to estimate proximities. We speculate that this kind of reasoning is the first step in the process of identity formation. Hence, in addition to iconization, fractal recursivity, and erasure, future research might examine processes of identity formation that follow a spatial way of reasoning. For instance, through a process of spatial contagion groups or individuals proximal in space can be considered more or less similar (e.g. sharing the common properties), and thus constructed as members of the same (or different) group(s). We expect that these insights will hold up cross-cultural studies as a generic heuristic, after all, direct personal interaction is strongly related to physical distance, though we concede that spatial reasoning may play a bigger role in shaping identity in some contexts than in others (said differently, some linguistic ideologies will tend to be more place-based than others). When theorizing about ethnicity and identity, social scientists have focused on social categories and categorization processes while downplaying the sort of continuous reasoning we are describing here. We believe this discrepancy results from the kinds of data scholars use to identify ethnic groups: emic terms designating groups or languages (demonyms, glossonyms), as well as clusivity paradigms (inclusive/exclusive pronouns and verbs). It is easy to see how this focus on emic systems will lead scholars to regard identify formation as a process of partitioning and dichotomizing of categories. Languages and native taxonomies will, though a processes analogous to grammaticalization, weed out continuous or hybrid forms, preserving dichotomous systems in the long run. The absence of a lexicon for expressing continuity and proximity does not mean that people do not reason in continuous terms. We hope to have shown that continuous reasoning patterns can be detected through indirect elicitation methods like the triad tasks we used here and that we have to be careful not to simply rely on patterns that our interlocutors are able to express verbally. 26 These methods also allow us to explore historical understandings of communities, that otherwise remain non-verbalized as well. As we have seen, the results suggest that in Chenalhó, the perceived linguistic center thus harbors the best form of Tzotzil Maya spoken in the municipality (in a sense the bats’i k’op - the real autochthonous Tzotzil). This view is held by people in different areas of the municipality and with varying degrees of expertise with the town‟s geographic and cultural landscapes. Several other facts make this interesting. First, this imagined linguistic center is located at the geographic center of the municipality. Within the context of a pre-colonial (non-extractive) subsistence economy one would expect – following Christaller‟s central place theory (1933) – the center town of a community to be located more or less at equal distances from the outlying hamlets, e.g. at its geographic center. Second, as mentioned above, this area is also linguistically marked as “the place of the old town”, Yabteclum. Third, the wider area around Yabteclum (cluster 2) constitutes a conceptual region in the municipality emerging both from a GIS cluster analysis as well as from our formal interviews. Informally this area is often referred to as yutil lum (Moksnes 2012, 44), the inside of the land. Interesting here is also the use of lum (as opposed to osil), referring to the land as a homestead of the Tzotzil Maya, the living landscape inhabited by ancestors and spirits (see Author 1997). Fourth and finally, this specific area has remained untouched by fincas, pointing at a relatively long history of fairly dense settlement, in itself a sign of historical importance. Together these different factors suggest that this area has indeed had social importance in precolonial times for the people of Chenalhó. By locating the linguistic center in that specific area, Yabteclum maintains the role of harboring autochthonous identity, safeguarding the “best” Tzotzil Maya of Chenalhó from outside influence represented by a Mestizo settlement (the cabecera), fincas established in the 19th century, and neighboring communities. One should read this as a refusal to completely cede the power of identity formation to the present center town that originated in the wake of a non-indigenousdriven extractive economy. Just like Tzotzil Maya ritual humor (Bricker 1973), it appears that the aspect of language ideology explored in this paper represents a safe haven for social criticism and the refusal of an existing –outside imposed – social order. Tzotzil Maya of Chenalhó maintain and nurture an idea of an 27 ethnic (community) identity where the language is not simply Tzotzil Maya, but the Tzotzil Maya spoken in Chenalhó; Taken together, language ideology constitutes more than merely reporting on perceived or observed language differences. Language ideology has to be analyzed as a social behavior in and of itself. In the presented case, how people think about language differences represents a social commentary of the history of Chenalhó, the resultant distributions of power and knowledge, as well as complex ethnic relations. Much more fine-grained research is needed to understand the interrelated processes we explored in this paper, as well as their particular local expressions. However, we hope to have carved a pathway for future research. 28 References Arias, Jacinto. n.d. Ya’yejal Kolonya Choro Ta Ch’enalo’, Chiapas: Historia de La Colonia de “Los Chorros” Chenalho, Chiapas. Tuxtla Gutiérrez: Subsecretaría de Asuntos Indígenas del Gobierno del Estado de Chiapas. Barrett, Rusty. 2008. “Linguistic Differentiation and Mayan Language Revitalization in Guatemala1.” Journal of Sociolinguistics 12 (3): 275–305. Boster, James S., and Jeffrey C. Johnson. 1989. “Form or Function: A Comparison of Expert and Novice Judgments of Similarity Among Fish.” American Anthropologist 91 (4): 866–89. Bourdieu, Pierre. 1977. “The Economics of Linguistic Exchanges.” Social Science Information 16 (6): 645– 68. Bricker, Victoria Reifler. 1973. Ritual Humor in Highland Chiapas. University of Texas Press. Brockmann, Andreas. 1992. Santa Martha: Untersuchungen Zur Ethnographie Einer Tzotzilgemeinde in Mexiko. Ethnologische Studien, Bd. 16. M nster: Lit. Bucholtz, Mary, and Kira Hall. 2003. “Language and Identity.” In A Companion to Linguistic Anthropology, edited by Alessandro Duranti, 368–94. Oxford: Basil Blackwell. Burguete Cal y Mayor, Aracely. 2007. “Aldama: Disputas Por La Restitución de Una Municipalidad.” In La emunicipali aci n de Chiapas: Lo ol co Y La ol ca n Tiempos de Contrainsur encia, edited by ochitl Leyva Solano and Aracely Burguete Cal y Mayor, 1. ed, 91. M xico: H. C mara de Diputados, LX Legislatura : Centro de Inves gaciones y Estudios Superiores en Antropologl a Social : Miguel Angel Porr a. Burton, Michael L., and Sara B. Nerlove. 1976. “Balanced Designs for Triads Tests: Two Examples from English.” Social Science Research 5 (3): 247–67. Calnek, Edward E. 1988. Highland Chiapas Before the Spanish Conquest. New World Archaeological Foundation, Brigham Young University. Christaller, Walter. 1933. Die zentralen Orte in Süddeutschland. Jena: Gustav Fischer. Cojt Cuxil, Demetrio. 1996. “The Politics of Maya Revindication.” In Maya Cultural Activism in Guatemala, edited by Edward F. Fischer and R. McKenna Brown, 19–50. Austin, TX: University of Texas Press. Colby, Benjamin N., and Pierre L. van den Berghe. 1961. “Ethnic Relations in Southeastern Mexico.” American Anthropologist, New Series, 63 (4): 772–92. Collins, Wesley M. 2005. “Codeswitching Avoidance as a Strategy for Mam (Maya) Linguistic Revitalization.” International Journal of American Linguistics 71 (3): 239–76. Cruz Jiménez, Andrés de la. 1996. Historia de La Comunidad Tzotzil Belisario Domínguez Chenalhó, Chiapas. Colección Letras Mayas Contemporáneas 12. México: Instituto Nacional Indigenista. de Delgaty, Alfa Hurley Vda., and Agustín Ruíz Sánchez. 1978. Diccionario Tzotzil de San Andrés Con Variaciones Dialectalestzotzil-Español, Español-Tzotzil. Vocabularios Y Diccionarios Indígenas Mariano Silva Y Aceves 22. Mexico D. F.: Summer Institute of Linguistics. del Carpio, Karla Berenice. 2012. “Spanish-Indigenous Bilingual Education in Chenalho, Chiapas in Southeast Mexico.” University of Alberta. de Vos, Jan. 1994. Vivir en frontera: la experiencia de los indios de Chiapas. San Cristóbal de Las Casas: CIESAS. Diercks, Willy. 2002. “Mental Maps: Linguistic-Geographic Concepts.” In Handbook of Perceptual Dialectology, edited by Dennis Richard Preston and Daniel Long, 51–70. John Benjamins Publishing Company. 29 England, Nora C. 2003. “Mayan Language Revival and Revitalization Politics: Linguists and Linguistic Ideologies.” American Anthropologist 105 (4): 733–43. Field, Margaret C., and Paul V. Kroskrity. 2009. “Introduction: Revealing Native American Language Ideologies.” Kroskrity Y Field (Eds.), 3–28. Fischer, Edward F., and R. McKenna Brown. 1996. Maya Cultural Activism in Guatemala. University of Texas Press. French, Brigittine M. 2010. Maya thnolin uistic Identity : Violence, Cultural i hts, and Modernity in Highland Guatemala. Tucson: University of Arizona Press. Gal, Susan. 1993. “Diversity and Contestation in Linguistic Ideologies: German Speakers in Hungary.” Language in Society 22 (03): 337–59. Gal, Susan, and Judith T. Irvine. 1995. “The Boundaries of Languages and Disciplines: How Ideologies Construct Difference.” Social Research 62 (4): 967–1001. Garza, Anna María. 2002. Género, interlegalidad y conflicto en San Pedro Chenalhó. UNAM. Guiteras-Holmes, Calixta. 1961. Perils of the Soul: The World View of a Tzotzil Indian. Free Press of Glencoe. Haviland, John Beard. 1981. Skʹop Sot ʹleb: l T ot il de San Loren o Zinacantan. Universidad Nacional Autónoma de México, Instituto de Investigaciones Filológicas, Centro de Estudios Mayas. Hirtle, Stephen C., and John Jonides. 1985. “Evidence of Hierarchies in Cognitive Maps.” Memory & Cognition 13 (3): 208–17. Hopkins, Nicholas A. 1970. “Estudio preliminar de los dialectos del Tzeltal y del Tzotzil.” In Ensayos de antropología en la Zona Central de Chiapas, edited by Julian Alfred Pitt-Rivers and Norman McQuown, 185–214. México D. F.: Instituto Nacional Indigenista. Hymes, Dell H. 1974. Foundations in Sociolinguistics; an Ethnographic Approach. Philadelphia: University of Pennsylvania Press. INALI, Instituto Nacional de Lenguas Indígenas (Mexico). 2009. Catálogo de las lenguas indígenas nacionales : variantes lin ü sticas de México con sus autodenominaciones y referencias geoestadísticas. 1a ed. México: Instituto Nacional de Lenguas Indígenas. Inoue, Fumio. 1996. “Subjective Dialect Division in Great Britain.” American Speech 71 (2): 142–61. doi:10.2307/455482. Irvine, Judith T. 1989. “When Talk Isn’t Cheap: Language and Political Economy.” American Ethnologist 16 (2): 248–67. Irvine, Judith T., and Susan Gal. 2000. “Language Ideology and Linguistic Differentiation.” In Regimes of Language: Ideologies, Polities, and Identities, edited by Paul V. Kroskrity, Paper edition, 35–88. SAR Press. Joseph, John Earl. 1987. Eloquence and Power: The Rise of Language Standards and Standard Languages. Burns & Oates. Köhler, Ulrich. 1980. “Patterns of Interethnic Economic Exchange in Southeastern Mexico.” Journal of Anthropological Research 36 (3): 316–37. Köhler, Ulrich, and Miguel López Comate. 2001. “La Reforma Agraria de Lázaro Cárdenas En Los Altos de Chiapas. Un Relato En Tzotzil de San Pedro Chenalhó.” Tlalocan 13 (0). Kroskrity, Paul V. 2004. “Language Ideologies.” A Companion to Linguistic Anthropology, 496–517. ———. 2009. “Embodying the Reversal of Language Shift: Agency, Incorporation, and Language Ideological Change in the Western Mono Community of Central California.” In Native American Language Ideologies. Beliefs, Practices, and Struggles in Indian Country, edited by Kroskrity, Paul V. and Margaret C. Field, 190–210. University of Arizona Press. Legendre, Pierre, and Loic F. J. Legendre. 2012. Numerical Ecology. 3rd ed. Vol. 24. Developments in Environmental Modelling. Elsevier. 30 Leyva Solano, Xochitl, and María Gracia Castillo Ramírez. 2012. Pequeña Bibliografía Comentada Para Reflexionar Sobre La Masacre de Acteal (Chenalhó, Chiapas, México). Pequeñas Bibliografías Comentadas. COMECSO-Región Sureste. Makarenkov, Vladimir, and François-Joseph Lapointe. 2004. “A Weighted Least-Squares Approach for Inferring Phylogenies from Incomplete Distance Matrices.” Bioinformatics 20 (13): 2113–21. Mannheim, Bruce. 1991. The Language of the Inka Since the European Invasion. University of Texas Press. Mantel, Nathan. 1967. “Ranking Procedures for Arbitrarily Restricted Observation.” Biometrics 23 (1): 65–78. doi:10.2307/2528282. Mariaca M ndez, Ramón. 2007. La Milpa Tsotsil de Los ltos de Chiapas Y Sus ecursos ené cos. 1. ed. San Cristóbal de Las Casas, Chiapas: Universidad Intercultural de Chiapas : ECOSUR. McNamara, T. P. 1986. “Mental Representations of Spatial Relations.” Cognitive Psychology 18 (1): 87– 121. Moksnes, Heidi. 2012. Maya Exodus: Indigenous Struggle for Citizenship in Chiapas. Norman: University of Oklahoma Press. Niedzielski, Nancy A., and Dennis Richard Preston. 1999. Folk Linguistics. Trends in Linguistics 122. Berlin ; New York: Mouton de Gruyter. Pérez Enríquez, María Isabel. 1998. El impacto de las migraciones y expulsiones indígenas de Chiapas: San edro Chenalh y San ndrés Sacamch’en de los obres, Larráinzar. Universidad Autónoma de Chiapas, Facultad de Ciencias Sociales. Preston, Dennis Richard. 2010a. “Language, Space, and the Folk.” In Language and Space: Theories and Methods v. 1: An International Handbook of Linguistic Variation, edited by Peter Auer and Jurgen Erich Schmidt, 179–201. Mouton de Gruyter. ———. 2010b. “Language, Space, and the Folk.” In Language and Space: Theories and Methods v. 1: An International Handbook of Linguistic Variation, edited by Peter Auer and Jurgen Erich Schmidt, 179–201. Mouton de Gruyter. Rand, William M. 1971. “Objective Criteria for the Evaluation of Clustering Methods.” Journal of the American Statistical Association 66 (336): 846–50. doi:10.2307/2284239. Reynolds, Jennifer F. 2009. “Shaming the Shift Generation: Intersecting Ideologies of Family and Linguistic Revitalization in Guatemala.” Native American Language Ideologies: Beliefs, Practices, and Struggles in Indian Country, 213–37. Romney, A. Kimball, William H. Batchelder, and Susan C. Weller. 1987. “Recent Applications of Cultural Consensus Theory.” American Behavioral Scientist 31 (2): 163–77. Ross, Norbert. 1997. Nut Lok’el li Kaxlane - die Vertreibung der Ladinos aus San Andres Larrainzar, Chiapas, Mexiko: Von Geschichten, einem Ereignis und Geschichte. Münster: Lit. ———. 2004. Culture & Cognition: Implications for Theory and Method. Thousand Oaks, Calif: SAGE Publications. Ross, Norbert, Tomás Barrientos, and Alberto Esquit-Choy. 2005. “Triad Tasks, a Multipurpose Tool to Elicit Similarity Judgments: The Case of Tzotzil Maya Plant Taxonomy.” Field Methods 17 (3): 269–82. Rus, Jan. 1995. “La comunidad revolucionaria institucional: la subversión del gobierno ind gena en los Altos de Chiapas, 1936-1968.” In Chiapas: los rumbos de otra historia, by Juan Pedro Viqueira Viqueira and Mario Humberto Ruz, 251–77. CIESAS. Shenton, Jeffrey, Norbert Ross, Michael Kohut, and Sandra Waxman. 2011. “Maya Folk Botany and Knowledge Devolution: Modernization and Intra‐Community Variability in the Acquisition of Folkbotanical Knowledge.” Ethos 39 (3): 349–67. 31 Silverstein, Michael. 1979. “Language Structure and Linguistic Ideology.” In The Elements: A Parasession on Linguistic Units and Levels, edited by William F. Hanks, Carol L. Hofbauer, and Paul R. Clyne, 193–247. Chicago Linguistic Society. Tamasi, Susan Louise. 2003. “Cognitive Patterns of Linguistic Perceptions.” Dissertation, University of Georgia. Tavanti, Marco. 2003. Las Abejas: Pacifist Resistance and Syncretic Identities in a Globalizing Chiapas. Routledge. Tax, Sol. 1937. “The Municipios of the Midwestern Highlands of Guatemala.” American Anthropologist 39 (3): 423–44. Urla, Jacqueline. 2012. Reclaiming Basque: Language, Nation, and Cultural Activism. The Basque Series. Reno: University of Nevada Press. Vogt, Evon Z. 1961. “Some Aspects of Zinacantan Settlement Patterns and Ceremonial Organization.” Estudios de Cultura Maya 1: 131–45. Weller, Susan C. 2007. “Cultural Consensus Theory: Applications and Frequently Asked Questions.” Field Methods 19 (4): 339–68. Williams, Angie, Peter Garrett, and Nikolas Coupland. 1999. “Dialect Recognition.” In Handbook of Perceptual Dialectology, Volume 1, edited by Dennis Richard Preston, 345–58. John Benjamins Publishing Company. Woolard, Kathryn A. 1989. Double Talk: Bilingualism and the Politics of Ethnicity in Catalonia. Stanford University Press. Woolard, Kathryn A., and Bambi B. Schieffelin. 1994. “Language Ideology.” Annual Review of Anthropology 23 (January): 55–82. 32 Table Table 1: Consensus analysis for perceived geographic and linguistic distances triad tasks. Samples Consensus Analysis Triad Tasks Site N (m=male) 1st factor scores 2nd factor scores Ratio 1st / 2nd Competence PGD 1 Cabecera 30 (m:15) 12.48 (0.42) 1.23 (0.04) 10.08 0.627 PGD 2 Cabecera 30 (m:18) 9.14 (0.30) 1.41 (0.05) 6.481 0.524 PGD 3 Cabecera 30 (m:18) 14.50 (0.48) 0.98 (0.03) 14.78 0.664 PLD 1 Both 45 (m:29) 16.20 (0.37) 3.00 (0.06) 6.504 0.574 PLD 2 Both 46 (m:30) 16.84 (0.37) 2.04 (0.04) 8.243 0.583 PLD 3 Both 47 (m:28) 19.75 (0.42) 2.76 (0.06) 7.074 0.624 Table Table 2. Effect of geographic clusters on estimation of distances between communities. Cluster All clusters N 392 Variables Geographic distance Within cluster Cluster 1 191 Geographic distance Within cluster Cluster 2 182 Geographic distance Within cluster Cluster 3 284 Geographic distance Within cluster Independent variable: perceived geographic distances Coefficient (ste) 0.742 (0.042) -0.135 (0.021) 0.778 (0.050) -0.196 (0.037) 0.859 (0.083) -0.068 (0.044) 0.612 (0.047) -0.164 (0.023) T-value 17.451 -6.388 15.393 -5.352 10.342 -1.564 12.910 -6.877 p-value < 0.000 < 0.000 < 0.000 < 0.000 < 0.000 0.119 < 0.000 < 0.000 R² 0.648 0.711 0.474 0.650 Table Table 3. Influence of geographic clusters on perceived language differences Cluster All clusters N 392 Explanatory variables Perceived geographic distance Within cluster Cluster 1 191 Perceived geographic distance Within cluster Cluster 2 182 Perceived geographic distance Within cluster Cluster 3 284 Perceived geographic distance Within cluster Independent variable: Perceived linguistic distances Coefficient (SE) T-value R² 0.561 (0.052) 0.051 (0.027) 0.544 (0.055) 0.137 (0.041) 0.437 (0.076) -0.200 (0.050) 0.727 (0.068) 0.153 (0.033) 10.867 1.872 9.846 3.298 5.727 -3.998 10.606 4.542 0.281 0.307 0.305 0.301 p-value < 0.000 < 0.000 < 0.000 = 0.001 < 0.000 < 0.000 < 0.000 < 0.000 Map Figure Cluster Analysis of Perceived Spatial Distances 6 Cluster 2 Xunuch MajomutCluster 3 Las Limas 4 Yabteclum Chojolho Beumpale Chixilton 2 Tzabalho Chimtic Campo los Toros Takiukum Naranjatic Alto Linda Vista Majumpepen Yashjemel Polho La Libertad Cluster 1 Bachen Naranjatic Bajo ● 0 Bajxulum ● La Merced Natividad ● Cabecera ● Belisario Jolxic ● Puebla ● ● −2 Canolal Los Chorros Poconichim Bajoveltic La Esperanza Chimix San Antonio Caridad ● Majumchon ● Tzanembolom Tzajalchen Cruzton Atzamilho ● −4 Las Delicias Cluster 4 Makuxtetic −6 San Jose Tepeyac Saclum Yocventana −8 Component 2 Pechequil Yiebeljoj Acteal −6 −4 −2 0 Component 1 These two components explain 89.27 % of the point variability. 2 4 6 Figure Cluster Analysis of Perceived Linguistic Distances 3 5 2 Poconichim Takiukum Cabecera Bajxulum Chixilton Chimtic Campo los Toros Tzanembolom Bachen Tzabalho Las Limas Acteal Naranjatic Alto Yiebeljoj Linda Vista Chojolho Beumpale Cruzton Naranjatic Bajo La Esperanza Pechequil Majomut Bajoveltic Polho Xunuch La Libertad 0 Yabteclum San Jose 1 Tepeyac ● Las Delicias La Merced ● Chimix Majumchon Jolxic ● Tzajalchen ● Natividad Makuxtetic ● ● Yashjemel San Antonio Caridad Belisario ● Yocventana ● ● −5 Component 2 Majumpepen Saclum Canolal ● Atzamilho ● Puebla Los Chorros −6 −4 −2 0 Component 1 These two components explain 70.71 % of the point variability. 2 4 6