Indexical borders: the sociolinguistic scales of the shibboleth


 This paper engages with the notion of the shibboleth, an indexically loaded, usually referentially indifferent set of (ideologically constructed) minimal pairs that is used in order to mark and perform social differentiation. We argue that the shibboleth is to be considered an interpretive (metapragmatic) phenomenon that operates on different sociolinguistic scales, notably the discursive scale (ideologies of communication), the performative scale (performance and metapragmatic stance-taking), and the subjective scale (lived experience). We propose a scalar metapragmatic theory of the shibboleth as an “indexical border” that takes into account how shibboleths emerge (are enregistered) and how they depend on contextualisation (or the indexical field). As a case in point, we present analyses of biographical construals of sociolinguistic displacement in the context of remigration from German-speaking countries to Bosnia and Herzegovina. Specifically, we focus on construals of displacement that are connected with (mis-)performances of phonologically rather subtle but indexically highly salient Bosnian/Croatian/Serbian affricate shibboleths ( and ).

1 No pasarán: maintaining the "invisible border" Then said they unto him, Say now Shibboleth: and he said Sibboleth: for he could not frame to pronounce it right. Then they took him, and slew him at the passages of Jordan: and there fell at that time of the Ephraimites forty and two thousand. 2 The word shibbolethor, for that matter, the phonetic minimal pair shibboleth/ sibboleththus serves "as a password, a mark of belonging, the manifestation of an alliance" (Derrida 2005(Derrida [1986: 20). The phonological variable (s), as variationists would have it, is "linguistically insignificant but socially significant" (Chambers 2003: 3) in the most existential way. The two "alternative ways" (Labov 1972: 322) of pronouncing the initial sound of the word, [ʃ] versus [s], are divided by an "invisible border" (Derrida 2005(Derrida [1986: 22), a border that makes visible (and/or audible) a difference that is immediately turned into a difference between ¡Pasarán! and ¡No pasarán!which in the Biblical (as well as in Celan's) case equals to nothing less than life or death.
The Ephraimites' fate has become an allegory for all sorts of language-based discrimination, sociolinguistic differentiation (Gal 2016) or linguistic profiling (Baugh 2003), and shibboleth is used in sociolinguistics as a pars pro toto, denoting all uses of intra-language variants to determine user's social position, be it in the context of political conflict (McNamara 2005), of forensic linguistics (Kniffka 2007: 86-109), of language testing (McNamara 2005), purism (Hill and Hill 1980) or in everyday encounter (Gal and Irvine 2019: 59, 73, 105;Johnstone et al. 2006).
In this paper, we are going to discuss one such case: the case of the Serbian/ Bosnian/Croatian affricate pairs <č/dž> and <ć/đ> that were construed by remigrants to Bosnia and Herzegovina as shibboleths. Using this case in point, we discuss the metapragmatic phenomenon of the shibboletha genuinely indexical, and hence ideological phenomenonmore generally and argue that it needs to be regarded on multiple layers, as the shibboleth is an ideological/discursive, a performative and a subjective phenomenon at the same time. We thereby take on the notion of sociolinguistic scales (as proposed by Blommaert 2007) in order to stress (1.) that these layers are to be located on a fluid continuum between "momentary and situated discursive positions" and "categorical, collective, and trans-contextual positions" (Blommaert 2007: 8-9) rather than within a micromacro binary, (2.) that they are inter-connected and inter-dependent (mutually determining), and (3.) that the layers are themselves scalable (i.e. variable in scope and salience), depending on context and actors.
In what follows, we will first expand on the three selected sociolinguistic scales of the shibboleth from a theoretical point of view (Section 2) before we discuss the example case (Section 3). Section 4 provides some conclusions.
2 Three sociolinguistic scales of the shibboleth Derrida suggest that the shibboleth is essentially performed knowledge of difference: The difference [between shi and si] has no meaning in and of itself, but it becomes what one must know how to recognize and above all to mark if one is to make the step, to step across the border of a place […], to see oneself granted the right of asylum or the legitimate habitation of a language.
[…] It is not enough to know the difference; one must be capable of it, must be able to do it, or know how to do itand here doing means marking (Derrida 2005(Derrida [1986: 26; emphasis in original).
The shibboleth thereby at the same time constructs ("marks") difference and demonstrates both (lacking) awareness of, and knowing-how to enact, difference. Differenceor sociolinguistic differentiation, for that matteris thus conceived as the main "meaning" of the shibboleth, but this meaning is, as Derrida reminds us, not just there. Difference needs to be enacted and ratified by subjects (in performance) and it needs to be metapragmatically related to shared knowledge about expected performance of specific types of actors, "legitimate" performance, and social boundaries (in discourse).
Drawing on this, we propose to conceive of the shibboleth as an indexical differentiator that is ideologically rooted (enregistered) in discourse, but that needs to be enacted in local practice and is embodied by subjects through the (emotionally) lived experience of language (Busch 2017). We will discuss these three scales in turn.

Discourse: the shibboleth as an enregistered sociolinguistic (b)order
As any other perceivable phenomenon that is allocated to social value and difference, shibboleths primarily serve to reduce complexity and contingency. This is achieved by constructing mutually exclusive binary categories (such as "Gileadites" vs. "Ephraimites" or "Gileadite speech" vs. "Ephraimite speech") that differ in respects that can be "seen", "heard", "smelt" or "tasted". Shibboleths, thus, emerge in processes of institutionalisation and are part of the taken-for-granted, naturalised life-worlds (Schütz and Luckmann 1973: 3-8) of social actors, i.e. part of their processes to make sense and order the social. As such, shibboleths indeed serve an important social function. They generate order by drawing borders between ways of speaking and groups of speakers and thereby generate sociolinguistic identity and group-applicability. This, however, comes at the cost of exclusion (McNamara 2012: 570). Derrida also highlights this "terrifying ambiguity of the shibboleth, sign of belonging, and threat of discrimination, indiscernible discernment between alliance and war" (Derrida 2005(Derrida [1986: 27). The shibboleth that excludes and condemns the Ephraimites, after all, simultaneously, and by means of this, constitutes the Gileadites' own groupness.
Such border-drawing, however, is not done in social vacuum. The question as to where the border is drawn, and which border is drawn, is subject to social negotiation, and the result is ultimately bound to power. In the Biblical case, it is, not coincidentally, the victorious party in the battle who decides on which minimal pair to serve as social discriminator. Shibboleths result from, and are means of, social and sociolinguistic hegemony which gets manifest in the power to construct and define social difference and to map that onto linguistic difference.
Such sociolinguistic differentiation is always formally arbitrary (Guy and Cutler 2011: 139-140) and subject to processes in which similarities between the differentiated and differences between the similarised are "erased" (Gal and Irvine 2019: 20-21). The difference thus seems to have "no meaning in and of itself", as Derrida (2005Derrida ( [1986: 26) remarks. On a closer look, however, choice and differentiation are not completely random.
To begin with, it is arguably not by coincidence that shibboleths often draw on subtle differences and particularly on what variationist sociolinguistics conceives of as "alternative ways of 'saying the same thing'" (Labov 1972: 322), i.e. referencesemantic variants. What seems to apply here is Bühler's (2011Bühler's ( [1934: 50-54) principle of abstractive relevance, which holds that the stronger a given form is ascribed to a certain semiotic function the more its use for other semiotic functions is limited; and vice versa, the less functional a difference is in one semiotic dimension (such as reference) the more it lends itself to other (e.g. social) functions. Bühler exemplifies the idea by means of prosody: C'est le ton qui fait la musique, it's the tone that makes the music; this holds in the Indo-European languages largely (but not completely) in the sense that the tone is free for appeal and expression and it is irrelevant for representation (Bühler 2011(Bühler [1934: 54). Hodge and Kress (1988: 84) allude to this idea in their discussion of the shibboleth case: The lower down in the table [representing "semiotic hierarchy"], the less significant the difference seems to be, the less concerned with mimetic (referential) meaning, but this allows them a more exclusive concern with semiosic/ideological reference. This quality is precisely why markers with low mimetic value are so widely used for this social purpose (Hodge and Kress 1988: 84). Furthermore, the shibboleth connects to a set of values and beliefs about particular values and beliefs concerning the affected groups ("Gileadites", "Ephraimites") and the way they are supposed to speak ("Gileadite speech", "Ephraimite speech"), as well as to more general values and beliefs about the connection of sociality and language (such as the idea that speech style "reveals" origin or, more generally, that there is something such as "authentic speech" and "authentic identity"ideas also current in sociolinguistics; for a critical discussion see Bucholtz (2003) and Spitzmüller (2019)).
These values and beliefs are known in sociolinguistics as language ideologies or, more generally, ideologies of communication (Spitzmüller forthcoming). Ideologies generally provide "naturalnesses" (Verschueren 2012: 10-14). In the case of language and communication, this includes assumptions of "natural" ways to communicate (in specific contexts) held by a particular group of speakers, as well as (assumptions of) "natural" links between ways of communicating and particular communicators, that is, "culture-internal models of personhood linked to speech forms" (Agha 2007: 135;cf. Park this issue). Such links are, as Agha (2007) terms it, enregistered in registers, i.e. discursively constructed, negotiated and distributed. Since shibboleths draw on such registers, their choice might appear "natural" to the involved social actors due to the naturalisation or rhematisation (Gal and Irvine 2019: 19) process that accompanies enregisterment (a process within which "qualities of the registers […] [are projected] onto the personae indexed in this axis of differentiation" [Gal 2016: 127]).
Yet shibboleths have a specific relation to registers as they relate to (at least) two registers at once. Shibboleths are enregistered sociolinguistic borders, i.e. they represent register boundaries. Thus, they do not index a group of communicators (by their typified ways to communicate), but group difference and hierarchy (by typified differences in ways to communicate). In order to fulfil this function, shibboleths need to be, in Silverstein's (2003) terms, enregistered on a higher indexical order, i.e. they need to be able to reflexively refer to the registers the boundaries of which they represent and to the social values (discriminatingly) associated with them. Therefore, shibboleths are, in Labov's (1972: 314) terms, not mere markers (linguistic forms associated with social groups in context), but "meta-markers" or stereotypes (socially marked forms stripped out of context that carry their context with them). More precisely they are what Agha (2007: 148) terms (without reference to Labov 3 ) metapragmatic stereotypes, "culture-internal models of utterance indexicality associated with speech variants". Using a metaphor proposed by Hodge and Kress (1988: 86) which draws on Vološinov (1986Vološinov ( [1929: 81), we can also say that shibboleths are "accent[s] of an accent […], a recuperation of the deviancy of the accent by reducing it to something simple, manageable and under the control of people outside the accent-community".
The "accent" metaphor helps to remind us that shibboleths are not mere linguistic signs but metasigns of difference between signs (or of valeurs, both in Saussure's [1916: 155-166] and in Bourdieu's [1977] sense), like accents are not mere "ways of speaking", but "differences in ways of speaking" (as compared with other "accents" or, if this ideological construct gets raised, "un-accented speech"). A shibboleth, thus, is not to be equated with a single linguistic item: shibboleth is not a shibboleth without sibboleth. It is the (construed) difference between the "correct" item and a(ny) "deviant" variant that is constitutive for the phenomenon. Thus at least two variants are required. Often, there are arguably more than two. Note, however, thatas many "deviant" and "valid" variants there might bethe shibboleth always constructs and foregrounds binarity, difference between "the valid" and "the invalid" performance.
We maintain that the power of the shibboleth resists on this ability to simplify the sociolinguistic range of variation to the binarism of "valid" and "invalid" form. We therefore conceive of the shibboleth as a (set of) minimal pair(s). Of course such pairs are ideological constructs that contrast with the infinite scale of linguistic variation. Yet their ideological power is the very promise to reduce such uncontrollable infinity to controllable binaries and thereby "fixate" the social (Laclau 1990: 89-92).
The metapragmatic reference of the shibboleth is (register) difference: a shibboleth is a difference that designates difference. Its form, hovewer, is a (constructed) sharp contrast between two linguistic items (or performances) under evaluation. Arguably, differences are themselves scalable (Agha 2007: 192): the lexical difference between Paradeiser and Tomate ('tomato') might serve as a shibboleth to discriminate Austrian from German (and Swiss) speakers, but also to discriminate Eastern from Western Austrians. Likewise, the phonetic difference between [toˈmaːtə] and [ˈtoˌmaˑtə] might serve as a shibboleth to discriminate Swiss from German speakers, but also Southern from Northern Germans (for the American English diastratical shibboleth [təˈmɛy:təʊ]/[təˈmɑ:təʊ], see Silverstein, this issue). In practice, however, such scalability is rendered invisible: differences, both in form and in social meaning, occur as "natural". This is the prerequisite for the shibboleth to expand to social power.

Situated practice: the shibboleth as performed differentiation
Then said they unto him, Say now Shibboleth: and he said Sibboleth 4 Shibboleths are enregistered, ideological phenomena, and thus rely on translocal, discursive knowledge about sociolinguistic differences. But in order to function, they still need to be locally enacted, contextualised and linked with (and by) social actors within concrete communicative practices and settings.
In the Biblical case, the practice is the shibboleth test, initiated by a directive (the "task", "Say now Shibboleth") that itself responds to a preceding directive (the "request", "Let me go over"), and conducted at a specific place (the passage of the Jordan) in a specific social setting (an armed conflict) with specific social hierarchies: the Gileadites as victorious, powerful gatekeepers (the "tester") and the Ephraimites as defeated passage requesters (the "tested"). Obviously, the powerful testers do not only occupy the gate (and likely the arms), they also (and thereby) define the rules and conditions of the test ("Say now …" is a command). If the tested want to pass, they need to respond to the directive ("and he said Sibboleth") and thus respond to the interpellation (Althusser 2014(Althusser [1970: 264) of the testers ("Art thou an Ephraimite?"). The testers furthermore hold the exclusive right to evaluate the test and to decide whether or not the tested "could […] frame to pronounce it right", based on their own notion of "correctness". To take up a concept of Inoue's (2006: 66) that draws on Althusser: the interpellation creates "speaking" and "listening subjects" which correspond to the tested and the testers. The testers' "experience of hearing" (Inoue 2006: 66) thereby reconstructs and constructs differences, and as Flores and Rosa (2015: 152) argue in a paper on raciolinguistic stigmatising of language learners, the differences often only emerge through the ideological ear of the listening subject. That is, the testers do not only decidepossibly case by casewhere the dividing line in the shi-si continuum is drawn. Even more, it is upon the testers' ideological construal to decide whether there is actually a (relevant) difference to be heard. And it is, consequently, also the testers who adjudicate upon the performance ("correct"/"incorrect", "authentic"/ "inauthentic" etc.) and decide upon the consequences ("Then they took him, and slew him" or ¡Pasarán! -¡No pasarán!).
So, the script of a shibboleth test boils down to these steps: 1. Request for "admittance" ("Let me go over") by the tested (Gileadites) 2. The "identity test" a. Social interpellation ("Art thou an Ephraimite?") of the tested by the testers (Ephraimites, i.e., "are you one of us?") b. Selection, and request, of test performance ("Say now Shibboleth") by the testers c. Actual performance of the task ("and he said Sibboleth") by the tested 3. Evaluation of the test response by the testers, based on the testers' norms of "correctness" 4. Judgement and metapragmatic sentence by the testers ("he could not frame to pronounce it right") 5. Consequences for the tested, conducted by the testers ("Then they took him, and slew him") The Biblical case is, of course, a model. In practice, performances might differ significantly, and most steps can be taken less explicitly. The shibboleth does not need to be evoked within a formal test. None of the directives need to be explicitly uttered, neither the request for admittance nor the interpellation and the task. The judgement does not need to be spelled out either, and the consequence can be less immediate (and definite). The only thing that is to be explicitly done in any case is (2c), the shibboleth performance. The general principles, however, hold: shibboleths need to be enacted in specific (asymmetric) social constellations. Metapragmatically, they both rely onindexically "presuppose"and result inindexically "entail" (Silverstein 2003: 195)disambiguation and (binary) differentiation. The shibboleth performance is interpreted by the testers with regard to their "appropriateness-to-context" (Silverstein 2003: 194), with regard to "differential identity-indexing features of [the performer's] utterance" (Silverstein 2016: 52), whereby the indexing vector points towards the register either on this or on the other side of the respective sociolinguistic border. This interpretation of the performance determines its "effectiveness-in-context" (Silverstein 2003: 194), the way subsequent contextualisation and interpretation unfolds.
Furthermore, some (ideological) presuppositions that justify specific steps in the procedure seem to generally hold: 1. The right to "admittance" [a] depends on social identification [b] ("Let me go over" → "Art thou an Ephraimite?") 2. We maintain that shibboleth practice and language ideologies are mutually dependent. Although the shibboleth test itself needs to be enacted in situ, the practice draws on ideologies of communication and enregistered assumptions about ways of communication and associated communicators. Vice versa, the practices also constitute, maintain, and change the order, since ideologies of communication emerge from linguistic practices, they are maintained by, and can change in the wake of, linguistic practices, particularly recurring and socially salient practices (in a way like Celan's Februar/Feber, which represents "all Februaries of violence", is the result of, and still co-refers to, any singular violent February). Enregistered difference is the result of iterated enacted differentiation (Agha 2007: 205). Within such processes of enregisterment, communicative "difference" gets densely linked with social subjects, and often also projected onto them when "some perceptible contrast of quality in indexical signs [is taken] to depictnot only to indexa contrast in the conditions under which the signs were produced" (rhematisation; Gal and Irvine 2019: 19). Consequently, particular forms of communication might be construed as symptoms of bodily dispositions (Bucholtz and Hall 2016). This brings us to our third and last scale of the shibboleth: subjectivity and lived experience.

Subjectivity: the shibboleth as lived experience
We elaborated above how tested persons are "identified" by means of indexical inference, i.e. "recognised" as belonging to either one or the other social category, by the testers within the process of the shibboleth performance. This recognition is also an interpellation in the sense of Althusser (2014aAlthusser ( [1970: 264), i.e. an address that subjectivates the addressed.
"Art thou an Ephraimite?" positions the person so-asked as a potential (speaking) subject of such-and-such-kind, and the test result confirms and ratifies this positioning: the tested person is identified as "Ephraimite" and thereby recognises themselves as "Ephraimite", which implies, in the contextualised differential logic, "non-Gileadite". This double allocation can be described with Althusser (2014Althusser ( [1970: 263) as a dialectic of recognition (connaissance) and "its inverse […] function of misrecognition [méconnaissance]". We will return to this dialectics in the example case. Butler (1997: 153) adds that such processes of interpellation are also inscribed into the socially constituted body. She thereby draws on Foucault's (e.g. 1982) notion of subjectivation and body politics as well as on Bourdieu's (e.g. 1977: 655-656) habitus. Butler combines Foucault's and Bourdieu's concepts insofar as she conceives of the habitus (and hence, the body) as discursively, i.e. performatively, constituted, dynamic constructs: "the [bodily] habitus constitutes a tacit form of performativity, a citational chain lived and believed at the level of the body" (Butler 1997: 155).
Language (ideology) plays a crucial role in such, as we could term it, somatic enregisterments. For, as Bourdieu (1977: 660) reminds us, "[l]anguage is a body technique and specifically linguistic, especially phonetic, competence is a dimension of the body hexis 5 in which one's whole relation to the social world is expressed". Language ideologies are often connected with ideologies of hexis (posture, gesture, proxemics, etc.), i.e. with ideas how particular communicators are supposed to use and move their bodies in communication (Agha 2007: 197-199;Bucholtz and Hall 2016: 178).
The notion of "authentic performance" that is so central to the shibboleth practice does not only draw on language ideologies, but also on ideologies of hexis (and other ideologies of communication). A body that does not move in the expected way, a body that quakes with fear or sweats, and also a body that does not display the expected deference to the observer might be sufficient enough reasons for a tested person to fail the shibboleth test. Thus, as Derrida notes, the shibboleth's differential capability must be inscribed in oneself, that is, in one's own body as much as in the body of one's own language, the one to the same extent as the other. This inscription of difference in the body (e.g. the phonatory aptitude to pronounce this or that) is, nonetheless, not natural; it is in no way an innate, organic faculty. Its very origin presupposes belonging to a cultural and linguistic community, to a milieu of apprenticeship, in sum, an alliance (Derrida 2005(Derrida [1986: 26).
In other words, discursively enregistered sociolinguistic axes of differentiation are inscribed in bodily performance and experiencethe discursive enregisterment resonates with somatic enregistermentas well as in the linguistic repertoire of speaking subjects ("dans son corps propre autant que dans le corps de sa propre langue", as the hardly translatable pun reads in the French original). Both inscriptions are, as Derrida notes (in line with Bourdieu) results of socialisation and habitualisation.
Vice versa, somatic enregisterment feeds back into what Busch (2017) terms the lived experience of language: the way speakers perceive themselves in interaction and assume to be perceived by others. According to this concept, which draws on the phenomenology of Merleau-Ponty, language, and communication in 5 Bourdieu uses the terms hexis and habitus to some extent interchangeably whereby hexis rather foregrounds observable bodily aspects while habitus foregrounds the process of how social power relations become internalised, incorporated, or incarnated mainly beyond the grasp of consciousness and will.

Indexical borders
general, is perceived and experienced physically and emotionally by means of the "living body" (corps vivant) which Merleau-Ponty (2002[1945: 120) distinguishes from the "physical body" (corps physique). In the shibboleth case, such bodily lived experiences particularly include, as we will see, feelings of anger, humiliation, fear, or shame: the perceiving subject-body turns into a self-observed objectbody, as if seen through the eyes, or heard through the ears, of the other, the listening subject.

A case in point: affricate shibboleths and remigration to Bosnia and Herzegovina
In this section, we are going to discuss an empirical case: The emergence of, and dealing with, (Bosnian/Serbian/Croatian) affricate shibboleths in Bosnia and Herzegovina (henceforth BiH), particularly in the context of remigration. We will argue how the three scales of the shibboleth discussed in this section interrelate in this case in point. We will proceed as follows. After we have introduced our data (Section 3.1) and elaborated on our methodology (Section 3.2), we will provide basic information on the socio-political context, the post-war situation in BiH, and the conditions of migration and remigration that concern our informants (Section 3.3). Then we will discuss a number of general sociolinguistic issues that are to be considered in the context of (forced) remigration (Section 3.4). Drawing on that, we will focus on the three scales of the shibboleth in the narratives (Section 3.5). In Section 3.6, we will summarise these findings and discuss the consequences of shibboleth positioning.

Data and analytic focus
Our data consists of four autoethnographic essays, each written by a female student of Vienna University originating from BiH between 2012 and 2014 in the course of lectures on biographical approaches in multilingualism research. The texts are part of a corpus of 20 language-biographical essays; the other essays relate to different sociolinguistic contexts which do not concern us here. All essays were written in German, the quotes reproduced here were translated by us.
The essays expand on the complex life trajectories of the four authors that, in their childhood (at an age between three and nine), had led them as refugees from the war in the territory of the former Yugoslavia to Austria, Germany, or Switzerland, and, after several years in exile (as adolescents), back to BiH. At the moment of writing the authors were in their twenties and back in the German speaking area, studying at Vienna University.
Of the complex lived experiences narrated within the texts, what specifically concerns us in this paper are reports of experiences at the moment of the authors' return to BiH. This moment was described by all authors as particularly stressful. Language and in particular the construed pronunciation of the affricate pairs <č/ dž> and <ć/đ> played a central role in this context, as we shall see.

Methodology
Since we are primarily interested in individual lived experiences, we employ a qualitative approach that helps us to reveal the subjective sense-making practices and life-worlds of our informants. Consequently we are not going to correlate "independent" social variables such as age, gender, and class with our data as long as these factors are not made relevant, or become salient, in the data themselves. Rather than that we focus on first-person accounts as decidedly 'subjective' attempts to reconstruct the own lived experience retrospectively, according to the particular situational requirements of the context of narration, a university course in Vienna, Austria (not BiH), in a German-speaking (not Bosnian) interactional setting.
Following interactional narrative theory (e.g. Bamberg and Georgakopoulou 2008) we conceive of the narrative of our authors as a positioning practice on multiple layers. Narrators position themselves towards the narrated story and the narrated figures (including "narrated self"), towards the audience (in our case: us as Vienna-based, non-Bosnian sociolinguists), as well as towards metapragmatic discourses and enregistered ideologies (Spitzmüller forthcoming). Specifically the latter "positioning level 3" (Bamberg and Georgakopoulou 2008: 391-392) became salient in the analysis since, notwithstanding the individuality and singularity of the texts, we were struck by some common orientations and experiences articulated by the authorsin particular those concerning the moment of return to postwar BiH and the general experience of sociolinguistic detachment. These experiences also resonate, as we will elaborate (see Section 3.3), with results of remigration research, and are thus identifiable as recurring in young remigrants' discourse.
It was particularly this observation that stimulated us to engage with these four texts in the context of this special issue, since the scalability of the process of sociolinguistic exclusion in the context of remigration from personal to shared experience and from local to discursive practice struck us worth closer investigation.

Sociopolitical context: post-war conflict in BiH, migration and remigration
As of 1990, long-standing conflicts between the republics and constitutive nations of the Socialist Federative Republic of Yugoslavia escalated and led to the break-up of the federation in 1992 and to subsequent military conflicts between and within the then new Balkan states. One major such conflict was the "Bosnian War" (1992)(1993)(1994)(1995), which entailed what has been termed, euphemistically, "ethnic cleansing", i.e. the killing of 100,000 persons (according to UNHCR), drastic violence and human rights infringements, and the displacement of over 2.2 million persons (Toal and Dahlman 2011;Young 2001). The drawing of new borders in the wake of these conflicts resulted in new majority-minority constellations. In this context, also language and language use were ethnicised and became a means of "ethnic" categorisation (Busch 2010;Greenberg 2017). In fact, language turned out to be a particularly pertinent "ethnic" marker due to its bodily dimension as a performed habitus, or as an only perceived one, projected onto the speaking subject.
In Europe, the war in the territory of former Yugoslavia was a turning point in asylum policies. Whereas before, asylum was granted according to the Geneva Convention, the refugees from Bosnia were only granted temporary or de facto asylum without a regular legal procedure (for Austria, see Böse et al. 2001). As a result, instead of a guaranteed and regulated status, the refugees found themselves in a situation of uncertainty and precarity that depended on changing policies and good will. Already shortly after the Dayton Peace Agreement in 1996, countries of exile started large-scale forced "repatriations" or launched programs for voluntary return.
Return, however, meant arriving in a war-ridden country, where the economy was heavily damaged and the political and administrative system strictly ethnicised. Ethnicisation and related difficulties for those returning as minorities to areas from which they were displaced was, besides the economic problems, one of the central concerns of reports by international organisations (e.g. International Crisis Group 2002; Van Metre and Akan 1997).
Refugee studies dealing with how returnees experienced what was labelled "returning home" raise a number of further topics that are recurrently addressed by remigrants (Meyer DeMott 2007;Palmberger 2010;Porobić 2017). Returning "home" is often reported to be a stressful and lengthy process of adapting to what returnees experienced as a shattered and deeply changed social environment. The returnees often did not feel welcome and were facing "discrimination, for example, through insulting statements in the media or from ignorant teachers in school" (Palmberger 2010: 78) implying that those who had left the country had not suffered from the war. The many difficulties encountered by returnees often resulted in transmigration moves or the wish to enable the children to pursue education in the former countries of exile (Porobić 2017).

Forced remigration as a double-traumatic sociolinguistic experience
From a sociolinguistic point of view the situation our informants were faced with is highly complex and explosive. As remigrants they have been expelled twice from a linguistic environment that is, or had become, familiar to them: first from the country of origin, then from the country of exile. They arrive twice in an environment where the linguistic repertoire, i.e. their means of expression and sense of linguistic adequacy, does not "fit". Thus they experience twice the loss of linguistic security (i.e., knowledge of linguistic and social rules) and the loss of a socially acknowledged origo. Furthermore, our authors experience twice the pressure to prove their linguistic loyalty: while in the country of exile, they are subject to hegemonic "integration-through-language" discourses and language policies (Flubacher and Yeung 2016), in the country of remigration, they find themselves confronted with heightened sensibility towards language use. This experience is accompanied with the experience of linguistic histories drifting apart. While the remigrants adapted their own linguistic practices and repertoires during exile in line with their biographic trajectory, language ideologies and language regimes in BiH changed as well: new terminology (e.g. in administration) was introduced, linguistic purism strengthened, registers changed (Vojvoda 2015). A recurrent topic in the four narratives is the experience that, at the moment of return, the linguistic repertoire "does not fit" although the expectation is that it "should fit" according to the language ideology that equates language and ("ethnic" or "national") origin. In our reports this experience was specifically connected to the context of the classroom and leisure activities with peers (three of the four authors continued school education in BiH upon their arrival in the immediate aftermath of the war).
Returning to postwar BiH as adolescents, after more than five years in exile, is described by our authors as a "forced return", even by those whose parents actually volunteered to go back. The "return home" was experienced as "double displacement" and "double eviction". After the eviction from BiH (e.g. due to their parents living in an ethnically "mixed marriage" or having names indexing a particular ethnic or religious affiliations), and after having to handle multiple difficulties in the adaptation to a new environment in Germany, Austria, or Switzerland, back in BiH they felt as being displaced once more. All four authors report being shocked when arriving in a devastated country where the traces of the war were still visible everywhere, although the local situations they returned to differed considerably. For one author it was the urban milieu of Sarajevo, for another a middle range town in the Republika Srpska (the part of BiH under Serbian administration), for two others smaller towns in the Federation (under Bosniak and Croat administration).
Although they returned under different personal conditions to different environments, all four authors report that as young returnees they were confronted with strong negative reactions as soon as they stepped out of the safe space of the close family. As noted above, school and gatherings with peers in the neighbourhood are described as the main sites of confrontation. One author, let us call her Selma, 6 reports: When [in school] we for instance discussed a particular topic and I was asked to give my opinion, it was often said: "That's the way somebody thinks who was in Germany and wore brand-name shoes while we had nothing to eat." Similarly, Majda narrates that she was not only looked at askance because of her new school clothes but "just because I had not been in the war", and Edina remembers: My teachers as well as my class mates told me to go back to Germany because there is no place for people such as me in Bosnia.
Here, the authors construe a process of othering through the construction of binary categories. One category comprises those who stayed in the country and suffered during the war and its aftermath, the other, opposing category consists of those who left the country "for a better life" and now come back.
As elaborated above, the power to establish categories is with those who can claim a higher status. In the given case these equal those who stayed in the country and who can morally substantiate and authenticate their claim through the hardship they had to endure during the war years as defenders and as victims. The narrative that established the binary categorisation draws on the reduction of complexity and processes of erasure. It is rendered invisible that those who left "for a better life" actually fled a war situation in order to save their lives, also their difficulties in exile are erased. Of course, the points of view and the blind spots of both, those who stayed and those who returned, are determined by their respective experiences and the position they take or that is assigned to them.
Interestingly, none of the four essays evokes the topic of discrimination on ethnic grounds, although issues of ethnicity dominate in academic literature on the post-war situation in BiH (Palmberger 2010;Porobić 2017). Neither is class or gender an issue that is construed to be decisive by our authors. The reported discriminative categories are "those who stayed" and "those who left". However, as the reports reveal, these categories only become the powerful societal devices they are because they are linked to, and enregistered with, emblematic forms of language use: those in power are able to "hear" whether their interlocutor is a "leaver" or a "stayer".
Experiences of this kind are made central in all four essays. Majda for instance writes about her first day in school: Finally it was my turn, and when I started to read the murmuring set in. My class mates giggled all the time. First I thought that there was a funny word in the story as I did not understand everything. Then I realised that they were laughing about me and my pronunciation.
Similarly, Selma reports in her essay that she was given different nicknames by her classmates that mocked her pronunciation.
Note that it is not the teacher who gets ascribed here to the powerful role of the evaluator, but the classmates. Also, as in the presented extract, evaluation is often done implicitly, by double-voiced performance, giggling, and laughter (see Spitzmüller (forthcoming) on implicit vs. explicit metapragmatic stance-taking).
In all four essays, the ascribed "lack of language skills" crystallises around a particular phonetic issue. The authors, referring to its graphetic realisation, foreground in their narratives "these letters unknown to me" (Majda) or "these funny letters with the haček" 7 (Edina). More accurately, it is the two affricate pairs <č/dž> and <ć/đ> and in particular their pronunciation that the authors construe as the most salient linguistic devices to differentiate Bosnian-speaking "stayers" from "leavers". These are the crucial shibboleths.
Azra describes her experience as follows: In Bosnia when I started to go out and to meet new people, already after a short while I was asked whether I was from Bosnia. One could spot that [I was not] because I could not pronounce and differentiate the different "c" letters the right way. In the beginning my friends laughed about it because they found it so to say cute. But I felt embarrassed because I did not want to be perceived as different. I am in my country, but because of my pronunciation, I am 7 The term haček (Czech háček 'small hook') denotes the inverted circumflex (caron) used in different Slavic spelling systems (including in the spelling of the word haček itself) to designate a range of palatalised phonemes.

Indexical borders
seen as Austrian or rather as bauštelac. 8 […] Gradually I lost confidence in my speaking when I had to speak Bosnian with friends, because I knew that they would immediately realize that I had not lived in Bosnia. I understood that I was seen as different […] I did not meet friends that often any more and spent more time with the family. I was fed up with having to hear from others how good I had it in Vienna and how difficult it was to live in Bosnia.
This excerpt allows us to exemplify the three scales of the shibboleth discussed above. Let us begin with the first scale, the enregisterment of a sociolinguistic border. Azra construes the pronunciation of what she calls "the 'c' letters" as emblematic differentiator between those "from Bosnia" (the "stayers") and those who are not (the "leavers"). As already noted, she is not alone with that construal. All four authors, despite having diverse remigration trajectories, refer to the same phonetic markers (or their graphetic representations). We think this is not by coincidence. Rather than that, we think these affricate pairs lend themselves to be enregistered as indexical differentiators (see Section 3.5 for elaboration). By turning to the second scale, we foreground how the shibboleth is evoked, and the evaluation performed. Azra does not find herself subjected to a formal test setting. Nonetheless, the setting she finds herself in is clearly hierarchical: the classmates and friends are "homelanders", more "native" and "authoritative" than the remigrants. This is not enacted by them merely by pronouncing the affricates "right" (they are not the ones to prove that), but by their patronising approval of Azra's "cute" way of speaking. As we have seen above, other essays report similar implicit forms of evaluation by means of laughter and murmuring. Azra interprets this patronising approval as verdict that she failed the test (as being perceived as different), "seen as Austrian or rather as bauštelac", consequently construes herself as an "incompetent" speaker of Bosnian, and thereby experiences herself through the eyes or ears of the others (the listening subject). The construal of linguistic "incompetence" is linked with the construction of the social persona of the "leaver"someone who had left Bosnia during the war and lived a comfortable life while those staying in the country suffered.
If we finally include the third scale to the analysis, the shibboleth as lived experience, we can see how encounters of being positioned are perceived as ideological interpellations (sensu Althusser), the "recognition" as a "non-Bosnian" subject. We can furthermore see that such positionings as "non-" or "expatriate Bosnian" ("inauthentic speaker") result in grave emotional reactions that remain inscribed as scar into the subjects' bodies: Azra feels "embarrassed", loses confidence in her speaking and avoids meeting her friends.
We observe similar reactions in the other essays. Majda reports that, due to the reaction to her pronunciation, she felt so humiliated that for a long time she refused to speak with her classmates: "The feeling of shame made me seclude myself so that my self-esteem and my marks suffered." Edina also describes strong emotional reactions: "I hated going to school, I hated the new town […] and the people." She connects these with recurrent experiences of devaluation: being laughed at by her classmates every time she dared to say or to ask something; being told by teachers "that I was simply too stupid and that I should be ashamed of myself". And Selma writes that her "school nightmare" started with the first school day and lasted for four years: "When I was called for an oral question, I got so scared that I hardly managed to get a word out. The panic attacks were stronger than me and spoiled my school life.
[…] I was afraid of being laughed at again." Embarrassment, scare, humiliation, shame, nightmares, panic attacksthe words chosen by the authors of the essays indicate that their experience of linguistic ostracism was not lived as simple teasing among peers but as deeply distressing experience with severe consequences. The authors describe their reactions to the systematic language shaming and silencing in terms of loss of selfconfidence, linguistic insecurity, seclusion, and hate. It is the imposed sanctions and the threatening consequences that make what is established as a mere marker of linguistic differentiation into a shibboleth.
Our informants report the enormous efforts they had to undertake to comply to the norms and rules in the country of "return" by adapting and even over-adapting. On the other hand, we also find the more or less constantly present desire to regain the country of exile and a language situation in which they feel less pressurized. Regaining the German-speaking area is what the four students did when they started studying at Vienna University, for them "returning home" turned out as for many others in similar situations as one step in a process of transmigration (Porobić 2017). This ambivalence between adapting and escaping strengthened their experience of constant liminality (in-betweenness) that seems to get ratified by the shibboleth's indexical entailment.

Divided by the "c"
With "the different 'c' letters", Azra is referring to <č> and <ć>, two graphemes that refer in Bosnian/Croatian/Serbian (BCS) to the post-alveolar affricate /t ͡ ṣ/ (<č>) and the alveolo-palatal affricate /t ͡ ɕ/ (<ć>), respectively. Both these affricates are voiced and paired with a voiceless variant, which is represented by separate graphemes, namely the post-alveolar /d ͡ ẓ/ (<dž>) and the alveolo-palatal affricate /d ͡ ʑ/ (<đ>). What matters here is the distinction between the respective voiceless and voiced variants. This contrast is typologically rather rare (Mihajlović and Ćavar 2018: 266) and also highly heterogeneous in the different varieties of BCS within and without the Balkans: some varieties of the homeland speech have merger or merger in progress, and […] there is a lot of variation in the realization of the two posterior places of articulation for speakers of BCS in Europe (Mihajlović and Ćavar 2018: 271).
Furthermore, it has been reported that a sociophonetic change (the "merger in progress" mentioned above) is underway particularly in the major cities all over the Balkans (Liker and Gibbon 2012, and references therein). Also, "in BCS languages the contrast has a relatively low [reference-semantic] functional load" (Mihajlović and Ćavar 2018: 273) in general.
Nonetheless, perception studies indicate that the difference is being perceived with rather high adequacy even by speakers from dialect regions where the variants already merge (such as Zagreb) or who show inconsistent use (Ćavar and Hamann 2011;Liker and Gibbon 2012). Following the principle of abstractive relevance, all this indicates that the affricate pairs lend themselves to serving as a shibboleth, as the referential function seems to be on decline while the social functions seem to increase (possibly as an effect of the former due to conservative language ideologies). This is what seems to have happened here.
As our data indicates, however, the dividing line seems to be not based on ethnic affiliation. There are other linguistic markers that are employed to flag such affiliations, e.g., the distinction between the Ekavian and Ijekavian sub-varieties, particular by means of the Yat reflexes (Busch 2010;Stojiljković 2017), lexical differences and different scripts. Rather, in the immediate aftermath of the war the affricate shibboleths seems to have become a marker of differentiation between those who stayed in the country during the war and those who went into exile. The sociolinguistic (b)order enregistered in this shibboleth thereby seems to run between "authenticity" and "artifice", "legitimate" and "illegitimate" speakers, "authority" and "wannabe" (incidentally the identity relations suggested by Bucholtz and Hall 2005). This assumption (which has been confirmed to us by many informants [personal communication]) is backed by the fact that the affricate differences are considered hard to learn for non-native speakers (Ćavar and Hamann 2011)tongue twisters used in pronunciation training such as Četiri čavčića na čunčiću čučeći cijuču ('four small jackdaws whistle crouching on the boat') illustrate thisand thus also serve to distinguish between learners and "native speakers".
As elaborated, the shibboleth construction does not only rely on social reductionism (erasure of the hybrid trajectories within and without BiH), but also on linguistic reductionism (erasure of sociolinguistic variability and change of the affricates). The affricate difference is, by large degree, an ideological construct, and a metapragmatic stereotypea shibboleth candidate par excellence.

The shibboleth trap
The examples illustrate how our informants construe and rationalise social and linguistic differentiation, on the grounds of their linguistic enactment of particular minute (but uncontrollable) features and the listening subjects' response to this enactment, as a verdict that others them as "leavers", and hence as a denial to access the group of "Bosnians proper". We have argued that such construals rest on processes of rhematisation within which the enregistered belief that a certain pronunciation is characteristic for a group is projected onto the individual communicators within the given axis of differentiation (between "authentic" and "nonauthentic" speakers of Bosnian that is recursively mapped in our case onto "leavers" and "stayers")processes that can also be identified beyond the lived experience of our four informants, on the level of metapragmatic discourse, as Section 3.5 elaborated.
It might be argued that these construals are, notwithstanding the discursive indications, but construals: subjective perceptions and interpretations of individuals resting on ambiguous implicit and explicit metapragmatic reactions they encountered within precarious phases of their lives. It can be argued that we have noand, arguably, neither did our authorssolid evidence for the "stayer's" intentions nor empirical evidence for the assumption that "homelanders" indeed pronounce the affricates significantly different than the "expatriates". Our counter is that this is not relevant here.
First, it does not really matter whether or not the groups in question (the "stayers" as well as the "leavers") indeed converge within their groups and diverge from the respective other in the way so stated (we have seen that empirical evidence about the variability of the affricates within and without the Balkans does not support this). As Silverstein (2003: 220) notes, higher-order indexicals always presuppose the lower orders (the belief that an element is typical for a given group [3rd-order indexicality] presupposes that the element can be observed in that group's usage [2nd-order indexicality], and this presupposes that the group actually uses that element [1st-order indexicality]). So, as long as a convergence/ divergence of registers is ideologically presupposed (as it is in our case), the listening subject will expect to "hear" the "foreign" pronunciation (whether it is recurrently performed or not).
And second, we recall that we conceive of the listening subject as an ideological and interactive construct, a powerful discursive role (Flores and Rosa 2015: 151). Even if a person does not intend to test, or ostracise, an interlocutor, their performance might still be contextualised as a verdict from someone in the position to judge upon "right" or "wrong". From the weaker position remigrants find themselves assigned to, any performance that is interpreted as an "access denial" from "above" effectively matters and is (as we have seen, sometimes severely) consequential.
Our authors found themselves repeatedly addressed as subjects that do not belong to the "stayers" category, as "illegitimate" speakers. And once they noticed on which ground this interpellation was done, the linguistic feature that was defined as shibboleth became part of their body habitus, and one of those obstacles one permanently falls over precisely because one has become, and is constantly made, aware of it.
As we have seen, interpellation does not need to be dressed unfriendly. It can come in a supportive manner ("your accent is so cute!"), or even, as Fanon (1986Fanon ( [1952: 35-36) expands with regard to racialising strategies, as a paternalistic approval of correctness: […] there is nothing more exasperating than to be asked: "How long have you been in France? You speak French so well" (Fanon 1986(Fanon [1952: 35).
Fanon also notes that such positioning cannot be rejected, neither within nor without the othered group. This is the shibboleth trap, a trap from which you can hardly escape.

Conclusions
In this paper we set out to understand how the "invisible border" (Derrida 2005(Derrida [1986: 22) that is drawn by the shibboleththe indexical marker which evaluated actors need to enact "properly in order to be granted the right to pass, indeed the right to live" (Derrida 2005(Derrida [1986: 1)operates in discourse, performance, and on the body. We argued how these three scales of the shibboleth are intertwined, and thus need to be triangulated if one wants to understand the shibboleth's social force. Analysing autoethnographic essays by young women narrating their return to post-war BiH we showed how the pronunciation of a particular minute feature was construed and rationalized as emblematic differentiator between those who stayed in Bosnia during the war and those who left the country as refugees; we showed how the returnees felt being evaluated by their peers as incompetent speakers when performing these features and how they experienced the verdict as wounding ostracism.
We argued that the shibboleth's social force emanates from ideologies of communication that provide "reliable" paths to social identification and differentiation on the grounds of "authentic" or "legitimate" forms of communication and linguistic behaviour. These ideologies are enregistered in, and socially shared through, discourse, and thus subject to negotiation, change, and maintenance, and they are maintained by actors in power and "in the know".
Discourse, however, needs to be enacted and contextualized in performance, within which roles such as the "speaking" and "listening" subject, the "testers" and the "tested", get assigned to people due to their respective sociolinguistic agency in context. Vice versa, performance feeds back to discourse and the enregisterment process, and here we stressed the importance of re-iteration of singular occurrences which makes murmuring of classmates, for instance, part of the chain that constitutes constant méconnaissance.
Finally we stressed that shibboleth discourse and practice is enacted by, and effects on, human bodies. Subjects incorporate discourse (the habitus), and discourse is incarnated by subjects within concrete practices. Thus, subjectivity or the lived experience is another crucial scale on which the shibboleth operates. Shibboleths not only act on subjectsin incomparable ways that range from the classification over the devaluation to the destruction of the subjectthey are also lived (through) by subjects. Lived through, as our example cases demonstrated, not only cognitively, but also somatically.
This often hardly noticeable but mighty "indexical border" that ridicules, alienates, others, stops, depresses, traumatises, and sometimes even kills human beings who "misperform"the shibbolethit is a complex construct. If we attempt to understand its scalable power, we need to understand its multiple scales.