Discrimination and exclusion of Roma in Europe

The history of Roma people in Europe is marked by racial discrimination and persecution, including periods of enslavement and genocide, as an estimated 200,000 to 500,000 Roma died in Nazi concentration camps during World War II. Roma continue to experience discrimination today, especially in Central and Eastern European countries, where social exclusion has often led to conditions of systemic poverty and where many Roma face harassment from individuals and vigilante groups, as well as discrimination by police and other authorities (Guglielmo Reference Guglielmo1996; Human Rights Watch n.d.). Across Europe, anti-Roma resentment is often rationalized as a response to alleged criminal activity, of which Roma are widely suspected (see e.g. Goodman & Rowe Reference Goodman and Rowe2014). In Central and Eastern Europe, anti-Roma prejudice has regularly led to incidents of harassment and violence. Hungary in particular has been the site of several high-profile violent anti-Roma crimes, including several murders and two pogrom-like incidents where Roma people were shot to death after fleeing from their fire-bombed homes (Amnesty International 2009; European Roma Rights Centre 2012). In 2012, the Hungarian Red Cross evacuated 277 Roma women and children from their hometown of Gyöngyöspata after they had been subjected to threats and harassment from vigilante groups while the police failed to intervene (Westhead Reference Westhead2012). Meanwhile, an openly anti-Roma extreme right party received 20% of the vote in 2014 parliamentary elections.

Over the past two decades, some Roma from Hungary and other Central or Eastern European countries have sought to migrate to Canada, often taking advantage of visa waiver policies for European tourists, and then requesting refugee status upon arrival (Tóth Reference Tóth, Bigo, Carrera and Guild2013). Since 2009, approximately 10,000 Hungarian citizens have requested asylum in Canada, making Hungary the most common country of origin of refugees to Canada between 2010 and 2012.Footnote ² In December of 2012, the Harper government changed its refugee policies by instituting a list of ‘Designated countries of origin’ which are deemed to respect human rights, so that they ‘do not normally produce refugees’. The Czech Republic, Hungary, and Slovakia are included on this list, and refugee claims by their citizens are now subject to a streamlined process that presumes that the claims are unfounded.Footnote ³ To further discourage Roma migration, the Canadian government advertised this changed asylum policy in early 2013 with large billboards in the eastern Hungarian city of Miskolc, from where many refugee claimants originated (Keung Reference Keung2013). The billboards displayed text in Hungarian that can be translated as ‘The Canadian Government informs: Canada's refugee system has been changed to deter abuse. People with unfounded asylum requests are quickly sent home’ (emphasis in the original through larger white font on red banner). While not explicitly mentioning Roma people, the billboard addresses would-be asylum applicants, and thus it is clear in the local context that the addressees are Roma (Keung Reference Keung2013). The billboard clearly states the government's belief that refugee claims by Hungarian citizens are ‘unfounded’, ignoring the fact that some claims have in fact been granted by the refugee board. Moreover, the city of Miskolc itself has been the site of violent attacks on Roma by right-wing extremists, and its police chief has publicly claimed that Roma are responsible for all burglaries that occur in the city (Westhead Reference Westhead2012). Finally, the billboard is likely to further stoke anti-Roma sentiments among those Hungarians who feel that Roma are giving Hungary a bad name by asking for asylum in Canada (Keung Reference Keung2013).

The Canadian government's billboard can be interpreted as a warning sign to deter Roma people from migrating to Canada, following a long tradition of such excluding messages in the Central European linguistic landscape. For example, it has a historical precursor in signs that were erected at the borders of the Habsburg Empire in the eighteenth century, barring Roma people from entering the land. As described by Crowe (Reference Crowe2007:72–73),

Contemporary European governments generally refrain from such overt, threatening messages of exclusion. Instead, the desired exclusion of Roma may be manifested in the linguistic landscape through racist graffiti, such as the one documented in Figure 1. This bilingual graffiti, photographed by the author in Prague in 1995, combines Czech cikáni ‘gypsies’ with German raus! ‘out (with)!’, echoing the German Nazi slogan Juden raus! ‘Out with the Jews!’ and its Neo-Nazi counterpart Ausländer raus! ‘Out with the foreigners!’. In the Czech context, this use of German can be understood as an allusion to the Nazi genocide, and the graffiti thus constitutes a clear threat of anti-Roma violence, in addition to its literal call for the exclusion of Roma people. As such, it illustrates the violent nature that anti-Roma sentiment may take in Europe, and which continues to drive many Roma people to seek a better life elsewhere.

Figure 1.

Bilingual Czech/German anti-Roma graffiti (‘Out with the gypsies!’), Prague, 1995.

Roma in Toronto

Despite the efforts by the Canadian government to prevent Roma immigration, Toronto has become home to a Roma community, albeit one that has been unstable and partially transient, as some families have returned to Hungary after their refugee claims were denied (Brown Reference Brown2013). Upon arrival, many Roma migrants settled in the downtown Toronto neighborhood of Parkdale, an area with low income housing in large post-war apartment buildings. The average household income of Parkdale is about 70% of the citywide average,Footnote ⁴ and the neighborhood is home to many social-service institutions, including ones that provide services for new immigrants. Parkdale can be characterized as ‘superdiverse’ (Vertovec Reference Vertovec2007), as a ‘polyglot globalized neighborhood’ (Collins & Slembrouck Reference Collins and Slembrouck2007), where languages such as Tibetan, Tamil, Vietnamese, Tagalog, and others are spoken and are visible in the linguistic landscape.

The arrival of Roma has introduced Hungarian into the linguistic landscape, as public institutions and businesses seek to communicate with newcomers who arrive with little or no command of English. Most Hungarian Roma in Toronto do not speak the ancestral language of Romani (an Indo-Iranian language), but speak Hungarian in their families, a consequence of language shift in Hungary and other former areas of the Habsburg Empire (Matras Reference Matras2002:242).Footnote ⁵ To communicate with Roma newcomers, schools and other institutions have relied on the use of Hungarian, either in writing, or by employing interpreters or Hungarian-speaking social workers. To investigate the presence of Hungarian in the linguistic landscape of Parkdale, I conducted ethnographic fieldwork between 2010 and 2014, regularly observing and documenting multilingual signage in public community centers, schools, and a public library, as well as in other spaces such as apartment buildings and offices of various organizations. In addition, I conducted interviews with institutional agents as well as with neighborhood residents to investigate how such signs are produced and how they are understood by different groups of readers. In this I follow studies such as Malinowski (Reference Malinowski2009) and Papen (Reference Papen2012) who draw on interviews to investigate the production and reception of texts in the linguistic landscape.

Of the many multilingual signs that I documented in Parkdale during this period, a total of thirty signs were addressed to Hungarian speakers. All but two of these were produced by and for institutions in which English is the dominant language. As is evident from Figures 2–5 below, these signs varied by medium and permanence, from temporary texts on paper, to professionally produced signs printed on durable surfaces. Five signs were monolingual signs in Hungarian only, ten signs were bilingual in Hungarian and English,Footnote ⁶ and fifteen were multilingual signs that included other languages alongside English and Hungarian, such as Tibetan, Tamil, Vietnamese, Tagalog, Spanish, French, or Chinese. Where two or more languages co-occurred, this was almost always in the form of ‘parallel texts’, as in Sebba's (Reference Sebba2013) typology of multilingual writing (see also Coupland Reference Coupland2012:9). These are parallel monolingual texts ‘where language-spatial relationships are symmetrical [and], language-content relationships are equivalent’ (Sebba Reference Sebba2013:108–9). As we show below, however, such parallel texts may only appear to have equivalent content, especially when machine translation is used.

Figure 2.

Public notice of code of conduct displayed inside public recreation center; parallel texts on parallel signs in English and Hungarian.

Figure 3.

Sign at side entrance to public school; parallel texts on same sign in English, Hungarian, and Tibetan.

Figure 4.

Notice in the lobby of a large apartment building; parallel texts in English and ‘Hungarian’ (produced with Google Translate).

Figure 5.

Monolingual directive sign, produced with Google Translate (edited to remove the name of the institution).

The use of parallel multilingual texts is a routine occurrence in institutional contexts in Toronto and Canada more generally. For one, federal and provincial legislation mandates the display of parallel texts in English and French in many public contexts, such as the postal service or the provincial health service. Furthermore, beyond official bilingualism, multiculturalism and multilingualism can be considered to be part of Canada's state ideology (see e.g. Haque Reference Haque2012), particularly in institutional contexts relating to immigration. The city of Toronto has a multilingual services policy that specifies criteria for the provision of interpreting and the translation of documents into certain languages.Footnote ⁷ Such a routine deployment of multilingual resources was evident in interviews with institutional agents. For example, the director of a community center noted that “if I do (.) notice that we have an influx of residents that speak a particular language (.7) then I just go through (.) maybe local agencies to find out who has translation services (.) and who can provide me (.9) with the signage I need in that language”.Footnote ⁸ Similarly, the manager of an apartment building reported that a letter informing residents about renovations was sent in multiple language versions: “We did that in all the languages that I felt were stronger (.) and then everybody else just (.) got one in English”. Both quotes articulate a language ideology in which multilingual texts arise from a desire to communicate with addressees who are perceived as lacking proficiency in English.

Of the bilingual or multilingual signs in Parkdale that included Hungarian versions, many were information signs, for example, about opening hours of a business or institution. Some signs with parallel texts, however, could also be classified as public order signs with directive or prohibitory messages, directing the reader to follow instructions about what is or is not allowed in the respective location. This is illustrated in Figures 2 and 3 with signage from inside a public recreation center and from a side entrance to a public elementary school, respectively.

LL studies do not always make distinctions between different types of signs (see e.g. Landry & Bourhis Reference Landry and Bourhis1997; Gorter Reference Gorter and Chapelle2013), or if they do, don't base them on functional criteria (e.g. the distinction between top-down vs. bottom-up, following Shohamy et al. Reference Shohamy, Ben-Rafael and Barni2010). As a consequence, there is no consistent terminology to describe signs such as those shown in Figures 2 and 3. In their typology of signs, Spolsky & Cooper (Reference Spolsky and Cooper1991:76–81) identify a category of warning notices and prohibitions. By contrast, Scollon & Scollon (Reference Scollon and Wong Scollon2003:184) speak of public notices, which ‘inform the public either about conditions or regulations that are present in that place’. More recently, Mautner (Reference Mautner2012:190) uses the term directive signs, and, as noted above, Kroon et al. (Reference Kroon, Dong, Blommaert, Stroud and Prinsloo2015:4) speak of public-order signs, which they define as signs that are ‘manufactured by a public authority with the intention of informing the public about an aspect of public order’. They add ‘such signs are important and consequential: they often specify what is legally appropriate (and consequently sanctionable in the event of transgression) and are thus strictly normative; they are legally binding both for the authority producing them and for the audiences consuming them’. The legal relevance of such signs makes them prime sites for the investigation of power in multilingual settings. As Mautner (Reference Mautner2012:190) notes, such signs ‘speak to us of social inclusion and exclusion’ and of ‘the power to control space’. At the same time, Kallen (Reference Kallen, Jaworski and Thurlow2010:49) notes that they may be ‘a point of entry in the official domain for languages that lack official status’.

Taken together, studies about public-order signs suggest that an institution's desire to exclude particular people from a controlled space may lead to the inclusion of their language in the linguistic landscape. This is illustrated by the monolingual sign in Figure 5 in the following section (the earliest one documented in the corpus). This sign asserts the recreational center's right to control access to its facilities and to exclude unauthorized Roma users. In so doing, it introduces Hungarian into the linguistic landscape where it previously had not been present. This finding contrasts with much LL research on minority languages that assumes that signage in minority languages either results from officially sanctioned language rights, or is authored by speakers of the language itself (Gorter et al. Reference Gorter and Chapelle2013). Instead it suggests that multilingualism may appear specifically in contexts where it is to facilitate the exclusion or punishment of deviant others. Thus, drawing on Coupland's (Reference Coupland2012) frame-analysis approach, multilingual public-order signs as in Figures 2 and 3 can be understood as exemplifying a particular frame, in what could be called punitive multilingualism.

All of these public order signs have in common that they are produced by institutions for their own purposes (as noted by Kroon et al. Reference Kroon, Dong, Blommaert, Stroud and Prinsloo2015), and can thus be characterized as examples of top-down signage (Shohamy et al. Reference Shohamy, Ben-Rafael and Barni2010). In the multilingual context, this generally implies that the institutional representatives who are involved in the production of the sign are not in fact literate in some of the languages that are included. Consequently, the production of the sign involves translational processes, which is addressed in the following section.

Dual indexicality of public-order signs in Parkdale

As noted in the previous section, Kroon et al. (Reference Kroon, Dong, Blommaert, Stroud and Prinsloo2015:4) observe that public order signs specify what behavior counts as ‘sanctionable in the event of transgression’ (see also Mautner Reference Mautner2012). At the same time, the existence of a given public-order sign serves to index the transgression itself. A similar point is made by Kulick (Reference Kulick2003) in his study of the word no in different contexts marked by violence and/or sexual desire. Drawing on performativity theory and Althusser's (Reference Althusser and Althusser1971) notion of interpellation, Kulick (Reference Kulick2003:145) argues for a dual indexicality of no, which on the surface manifests a denial, but simultaneously affirms the existence of that which is denied: ‘A “no” is not just a refusal of that subjection. It is also an acknowledgement of it; a response to the interpellative call that even in disputing it affirms it’. Such a dual indexicality is arguably also present in public order signs, as the signs do not merely index a prohibition, but also index that the transgressive behavior has previously occurred or is expected to occur at the particular location. For example, readers of the ‘Code of conduct’ in Figure 2 may assume that the presence of the sign is motivated by the prior occurrence of incidents in which participants did not ‘show proper care and regard for City property and the property of others’. Moreover, such signs address (or interpellate) persons who may engage in this transgression and need to be instructed not to do so (consider, for example, the use of the imperatives leave and remain in Figure 3).

This dual indexicality points to the fact that signs produce a range of meanings, rather than a simple, general meaning (see Kroon et al. Reference Kroon, Dong, Blommaert, Stroud and Prinsloo2015:5). This point is emphasized by Ochs (Reference Ochs, Gumperz and Levinson1996:418), who argues for the need to distinguish potential from actual indexicality ‘in a particular instance of use (in the mind of any participating interlocutor – speaker, addressee, overhearer, etc.). The indexical potential of a form derives from a history of usage and cultural expectations surrounding that form’. Following Ochs, an investigation of the indexicality of signage in the linguistic landscape thus requires ethnography and attention to different participant roles. Adapting Goffman's (Reference Goffman1981) model of the participation framework, different positions of producer and recipient of signage can be distinguished. Public-order signs are written by specific authors, but also represent the voice of an institution (the principal, in Goffman's terms). Furthermore, they may be read by addressees as well as by unaddressed recipients, ‘overreaders’, so to speak (corresponding to overhearers, as in Ochs’ quote above). This is especially relevant in parallel-text multilingual signs, where each language version has a different set of addressees, while the presumed translation equivalence facilitates the ‘overreading’ of other-language versions. To illustrate, English-speaking readers of the trilingual sign in Figure 3 are addressed by the English text at the top, but are overreaders of the Hungarian and Tibetan versions, whose propositional content they may presume to be equivalent to the English text. This notion of overreaders can also be illustrated with the case of the Hungarian-language billboards that the Canadian government posted in the city of Miskolc (see the earlier section on the discrimination and exclusion of Roma in Europe). While these monolingual billboards were addressed to Hungarian citizens who considered applying for refugee status in Canada, their unaddressed recipients include members of the Canadian public who read about them in the Canadian press.

Indexicalities of ungrammatical ‘Google Hungarian’

As noted above, interviewees who were representatives of institutions or businesses were asked not only about the social meanings of multilingual signage, but also about the production of signs. In the process, I determined that several signs were produced by machine translation, a fact that had also been suggested by several Hungarian-speaking consultants who evaluated and translated the signage. This was typically the case with signs that were printed on a piece of paper and displayed in the windows of offices or store fronts. One example is given in Figure 4, showing a sign posted at the door to the rental office of a large apartment building in Parkdale where several Roma families live. The sign is a bilingual sign with two parallel monolingual texts, with English displayed on top.

1. (3) Grammatical Hungarian translation of English text in Figure 4.

   Nem fogadunk el készpénzt lakbér fizetésre.

   ‘We do not accept cash for rent payment.’

   Csak csekket, pézes utalványt, betéti kártyát és hitel kártyát fogadunk el.

   ‘We only accept cheque, money order, debit card and credit card.’

The English text in Figure 4 was written by the manager of the building's rental office (whom I interviewed), and then used as input for machine translation into Hungarian using Google Translate. The resulting version ‘looks’ Hungarian (cf. Kroon et al. Reference Kroon, Dong, Blommaert, Stroud and Prinsloo2015), but it is arguably not ‘in Hungarian’, because it is ungrammatical to the point of being incomprehensible, according to several native speaker consultants. For comparison, a grammatical and contextually appropriate Hungarian translation of the English source text is given in (3). The ungrammaticality of the machine translation results from incorrect or unidiomatic word choices and from a failure to mark syntactic relations between words in the target text, as inflectional markings are either absent or misused. For example, in both sentences, direct objects fail to be marked with the suffix -t for accusative case (Készpénz ‘cash,’ csekk ‘cheque’, bankártya ‘bank card,’ and megbizások ‘orders’ instead of készpénzt, csekket, bankkártyát, or megbizásokt, respectively). In addition, the noun dijak ‘fees’ is lacking the causal-final case suffix -ert. In the second sentence, inflectional morphology is found in the third-person singular suffix -ja on the verb fogadja ‘accepts’. However, it does not agree in number and person with the subject mi ‘we’, which would require mi fogadunk (see the translation in (3)). Furthermore, the machine translation appears to treat line breaks in the input as syntactic boundaries, as no cash is not translated as a syntactic unit, but as two separate phrases (nincs ‘there is not’ is a plausible translation of English sentential no, but not of the determiner no). The intended meaning would be best rendered in Hungarian as nem fogadunk el készpénzt ‘we don't accept cash,’ as shown in (3). Both sentences also include unidiomatic uses of the definite article a, in a bérleti díjak ‘the rental fees’ and a bankkártya ‘the bank card,’ as well as inappropriate lexical choices (dijak ‘fees’ and átutalási megbizások ‘transfer orders’ instead of pénzes utalvány ‘money order.’) Finally, the placement of the quantifier csak ‘only’ causes it to scope over the verb (as if to restrict the conditions under which cheques are accepted), rather than over the following noun phrases, as intended (compare its placement before the noun phrase in (3) above, or, alternatively, in mi csak csekket fogadunk el ‘we accept only a cheque’).

Other documented signs were similarly produced with Google Translate, and likewise marked by considerable ungrammaticality. Like the sign in Figure 4, these were paper signs posted on walls or windows of institutions, such as two community recreation centers and a legal aid office, and they were either directive signs or information signs (e.g. posting opening hours).Footnote ¹⁰ With the exception of one monolingual sign, they displayed parallel texts in English and ostensible Hungarian, that is, ‘Google-Hungarian’, and sometimes also in other languages. The monolingual sign is shown in Figure 5. It is the earliest documented sign in the corpus, from November 2010, when it was posted at the entrance to a gym inside a public recreation center. While Hungarian-speaking consultants struggled to make sense of the text, the sign was likely intended to inform the reader that access to the gym is restricted to users who have preregistered for a program. The first two words, Nem lép, in particular are again ungrammatical to the point of being incomprehensible, as they could be translated as ‘is not making a step’ (lép is third singular of lépni ‘to move/make a step’, nem ‘not’), but omit a subject.Footnote ¹¹ As in Figure 4, syntactic roles are unmarked, so that it is unclear how a program ‘the program’ relates to the rest of the sentence. Finally, köszönjük ‘we thank you’, would be pragmatically more appropriate than köszönöm ‘I thank you’ (which machine translation programs are likely to treat as the default translation of English thank you). At the same recreation center, another sign posted alongside an English version was intended to inform readers about the hours during which the swimming pool was open to the public, rather than reserved for classes. In English, this was conveyed by the phrase Leisure swim, which had been translated into Hungarian as pihenés úszni ‘rest to swim.’ Several native speakers judged this to be nonsensical, and suggested alternatives such as szabadidős úszás ‘recreational swimming.’

Indices of exclusion and belonging

The preceding sections have addressed the indexicality of multilingual public-order signs and of ungrammatical language use produced with machine translation. Some of the signs that were documented for this study, however, do not fall into either of these categories, but rather provide information in grammatically correct and pragmatically appropriate Hungarian. These were found particularly in the context of public schools attended by Roma children, for example, multilingual wayfinding signs inside school buildings, and bilingual signs with information about school events and procedures, such as registration for kindergarten classes. An example is shown in Figure 6, an advertisement for a school fair that was shown in Hungarian, alongside an English version with parallel content and layout. The poster invites the reader to attend a free party at the school (sponsored by several organizations) and advertises attractions such as music, food, kids’ activities, and a raffle, as well as a number of community services available for parents.

Figure 6.

Advertisement for school fair; Hungarian version displayed alongside English version (edited to remove the name of the institution).

Such signs elicited very different reactions from interviewees than the signs discussed above. When shown the sign in Figure 6, one Roma interviewee stated that the sign made him happy. Two interviewees noted that the text was in ‘proper’ Hungarian, and this fact contributed to their positive assessment of the sign. In contrast to the signs discussed above, they attributed this sign to a Hungarian-speaking author (“I don't think so it's a Google translator”) who took the time to produce a carefully crafted text and who cared about Roma children at the school (“It's made by heart”). As in the case of the signs made with Google Translate, the specific linguistic forms thus clearly contribute to the social meanings that are indexed by the sign. While users of machine translation proclaim to be content with getting a basic meaning ‘across’, and appear to believe that it is possible to do so without concern for grammaticality and pragmatic appropriateness, addressees use both the form and the content to evaluate signs and to ascribe motives to their authors, especially if the evaluations of form and content converge. Thus, as discussed earlier, the ungrammaticality of machine translation is evaluated more negatively on public-order signs than on information signs, whereas the use of grammatically and pragmatically appropriate language on an invitation to a public event has the potential to index acceptance and belonging more strongly than it would on public-order signs.