THEMES IN LANGUAGE SHIFT RESEARCH

As García (Reference García2003:37) notes in a recent review article, an exhaustive review of the literature on language shift and maintenance is not humanly possible given “the amount of activity in the field,” and I will not even attempt it. Here I can offer only a very cursory thematic overview. A common and general feature of all language shift research is as a more or less explicit contribution to language maintenance (García Reference García2003:23). Language shift and language maintenance are seen as two sides of the same coin, such that a good understanding of language shift processes is seen, if not as a prerequisite for a language maintenance effort, at least as improving its chances of success. It is also in this spirit that this article on medium request in the Rwandan community in Belgium is written.

One of the most frequently recurring issues in research on language shift is to ascertain that language shift is taking place, and, if it is, how it can be seen (e.g., Bills, Hudson & Hernandez-Chavez Reference Bills, Hudson and Hernández-Chávez2000, Slavik Reference Slavik2001). Researchers have highlighted the relationship between language shift and sociological categories, in particular gender (e.g., Gal Reference Gal1979, Herbert 2002). Research has also focused on the “determinants” (Gal Reference Gal1979) of language shift. Factors most often cited include economic factors, institutional support, and ideological factors. For example, the shift from Hungarian to German reported by Gal Reference Gal1979 was due mainly to economic changes that had taken place in the small village of Oberwart. In Gal's (Reference Gal1978 [1997]:378) own words, the shift coincided with “the familiar post-war process of urbanization and industrialization of the countryside often reported in the literature on the transformation of peasant Europe.” The case for institutional support for language maintenance seems obvious and has led, among other things, to the many programs of bilingual education (see Beardsmore Reference Beardsmore and García2009 and García & Beardsmore Reference García, Beardsmore and García2009 for a discussion). A number of studies have investigated the importance of language attitudes and ideologies in language shift. In his study of language shift in the village of Gapun, Kulick Reference Kulick1992 shows that it took place in the context of, and was therefore facilitated by, ongoing redefinition of self taking place in the community. Of particular interest here are the many studies that report language shift despite community members' claims of attachment to their language and cultural identity (Slavik Reference Slavik2001, Papapavlou & Pavlos Reference Papapavlou and Pavlou2001, Lanza & Svendsen Reference Lanza and Svendsen2007). Notions of language as a core value (Smolicz Reference Smolicz1981) and of the pragmatic or utilitarian view of language (Lanza & Svendsen Reference Lanza and Svendsen2007) have been used to explain such situations. However, of all the factors, speakers' social networks have received most attention and are believed to be most important (e.g., Gal Reference Gal1979, Wei Reference Wei1994, Lanza & Svendsen Reference Lanza and Svendsen2007, Ihemere Reference Ihemere2006).

Although research addressing these themes is without doubt important, it nevertheless falls short of responding to Fishman's call for direct action at the level of face-to-face interaction. Closer to this agenda is research carried out under the general framework of language socialization. Commenting on this framework, Paugh (Reference Paugh and Cohen2005:1809) significantly writes:

Here again, an exhaustive review of studies is beyond the scope of this article (see Baquedano-López & Kattan Reference Baquedano-López, Katta, Auer and Wei2007). I will only mention some studies and their findings by way of illustration. Paugh Reference Paugh and Cohen2005 examined language socialization practices in the village of Penville (Dominica), where language shift was taking place from Patwa, a French-based creole, to English despite official policies aimed at revitalizing the creole. At the level of face-to-face interaction, Paugh observed that adult members of the community, parents and teachers in particular, systematically rejected and repaired children's use of Patwa in interaction both at school and at home. Bani-Shoraka focused on “Azerbaijani families (in Teheran) experiencing language shift from bilingualism in Azerbaijani and Persian to monolingualism in Persian” (2009:105). In these families, Bani-Shoraka observed, among other things, an interactional practice she referred to as “inclusion.” This consisted of parents offering

Rindstedt & Aronsson Reference Rindstedt and Aronsson2002 observed adult–child interaction in San Antonio (Peru), where there was ongoing shift from Quesha to Spanish. They observed two interesting language-choice practices. In interaction among themselves, adult members used either Spanish or Quesha as the medium, but in interaction with children, they “generally spoke one language only, Spanish” (2002:732). In addition, the authors observed a general lack of medium repair (Gafaranga Reference Gafaranga2000), the only two exceptions to this being those where Quesha (and never Spanish) was repaired.

Studies such as those mentioned above that have identified specific interactional practices through which language shift is achieved take the research agenda close to Fishman's program for reversing language shift. However, these studies still suffer a problem both Luykx Reference Luykx and Cohen2005 and Lanza Reference Lanza, Auer and Wei2007, among others, refer to as that of traditional language socialization research. In this methodology, children are viewed as “objects or recipients of socialization, rather than as agents” (Luykx Reference Luykx and Cohen2005:1407), “as something to be molded and guided by society in order to become a fully-fledged member” (Lanza Reference Lanza, Auer and Wei2007:47). In more recent language socialization research, children are seen as “active and creative social agents” (Lanza Reference Lanza, Auer and Wei2007:47) and as “socializing agents” themselves (Luykx Reference Luykx and Cohen2005). In other words, there is a need for studies of language shift to recognize the role of children as interactional partners more than seems to have been the case so far. Second, with respect to language shift, researchers seem to have limited themselves to noticing interactional practices. If studies are to meaningfully inform language maintenance efforts, there is a need to go beyond mere noticing to actually describing specific practices in detail. The present article lies in this tradition of research, which focuses on specific practices in adult–child interaction in language shift situations, while at the same time striving to avoid the abovementioned problems.

THE RWANDAN COMMUNITY IN BELGIUM AND LANGUAGE SHIFT

Colonization is often thought of as taking the colonizer to the land of the colonized. But the reverse, although less recognized as such, is also true. As a result of Belgium's being the former colonial master of Rwanda, there has always been a Rwandan presence in Belgium, and in Brussels in particular. However, the size of the Rwandan community in Belgium had remained relatively small until the events of the early 1990s in Rwanda. Following the civil war (1990–1994) and genocide against Tutsis (April–June1994), Rwandans fled the country en masse. The majority of these displaced persons temporarily settled in neighboring countries,Footnote ² but a small minority of lucky ones made it to Europe, Belgium being the preferred destination. Today, this new community of Rwandans comprises a few tens of thousands of members.Footnote ³ In terms of settlement, the general pattern can be described as dispersal. The majority of these Rwandans live in the Brussels area, but individual families can also be found in other cities in Belgium. Even in Brussels, however, settlement remains scattered, and nothing like a Rwandan version of “Chinatowns” can be found. Such a scattered settlement pattern has been found to be a strong factor of language shift (Mesthrie Swann, Deumert & Leap Reference Mesthrie, Swann, Deumert and Leap2000).

On arrival, members of this community were bilingual in French and Kinyarwanda. Up until 1995, the Rwandan constitution recognized Kinyarwanda as the national language and both French and Kinyarwanda as official languages.Footnote ⁴ In terms of acquisition, Kinyarwanda was acquired natively by almost every Rwandan, while French could be learned only through formal education, where it was both taught as a subject and used as the medium of instruction. As a result, French enjoyed very positive attitudes and was often seen as the language of the elite. Understandably, the Rwandans in Belgium are mostly members of this former elite, for it takes a lot of resources, financial and otherwise, to enter Fortress Europe. The fact that the Rwandans in Belgium were already bilingual in French and Kinyarwanda might be contributing to the relative speed with which language shift is taking place.Footnote ⁵ In terms of use, standard Kinyarwanda and French could be used in formal domains, but, in informal contexts, Kinyarwanda-for-all-practical-purposes, a variety consisting of frequent alternation between Kinyarwanda and French, was the preferred medium (Gafaranga Reference Gafaranga1998, Gafaranga & Torras Reference Gafaranga and Torras2001). This preference for Kinyarwanda-for-all-practical-purposes can still be observed among adult members of the community. That is, strictly speaking, ongoing language shift is from Kinyarwanda-for-all-practical-purposes to French.Footnote ⁶

In the community, attitudes toward ongoing language shift vary. Some, based on an ideology of language as a core value (see above), feel genuine concern for the future of Kinyarwanda and have even undertaken to counter language shift through what researchers refer to as “bottom-up strategies” (Nettle & Romaine Reference Nettle and Romaine2000). They have initiated a variety of sociocultural organizations (11 such organizations were identified during the research) with the explicit aim of preserving the Rwandan culture and language. The views of this category of Rwandans are summed up in the following statement from one of them:

(2) Interview 2.

However, as we have seen, such positive attitudes are not in themselves enough for language maintenance. Language shift has been observed to take place even when members hold very positive attitudes toward the language (see above).

Unlike the person quoted above, other members of the community, basing their view on a different kind of ideology, are at best indifferent and at worst in favor of the shift, reasoning, as one member put it:

(3) Interview 4

To understand this position, one needs to make reference to the general sociopolitical background. After the genocide that claimed thousands of lives, both the new Rwandan government and the international community at large (see International Criminal Tribunal for Rwanda) are determined to bring the perpetrators to book. In the process, there is an unfortunate amalgamation between being a Rwandan in exile and being a “genocider”.Footnote ⁷ The Rwandans in Belgium live with a constant fear of being accused, rightly or wrongly, of being genociders. Therefore, for the sake of safety, they avoid anything that could identify them as Rwandans. Some have even gone as far as changing their names. It is in this context that Kinyarwanda, the most obvious identity marker, is rejected.

Language shift, as we have seen, is often said to take place out of sight and out of mind. In the case of the Rwandan community in Belgium, as the brief account above shows, this is not necessarily the case. Members are very much aware of the situation and, whenever the occasion arises, they report it most articulately. The following comments from community members are, I believe, highly significant. In extract (4), two parents (E and B) and a visitor (A) are reporting that, when among themselves, Rwandan children use French only.

(4) Interview 7.

A pattern of language choice in which siblings use the non-indigenous language among themselves is common in language shift situations (see, e.g., Rindstedt & Aronsson Reference Rindstedt and Aronsson2002, Kulick Reference Kulick1992).

In extract (5), in contrast, two young adults (D and C), aged 14 and 16, report their language preference (Auer Reference Auer1984) to B, the fieldworker.

(5) Interview 3.

Auer (Reference Auer, Milroy and Muysken1995:125) uses the notion of language preference to refer to any tendency to use a particular language, whether this tendency is due to a higher or lower competence in one of the languages involved or whether it is “based on political considerations”. Torras & Gafaranga Reference Torras and Gafaranga2002 specifically speak of competence-related preference and of ideology-related preference. In the case of the choice between Kinyarwanda and French by Rwandan children in Belgium, preference is strictly competence-related. Research conducted in a variety of settings shows that, in situations of language shift, members' competence in the minority language ranges from full competence to zero competence. In between, subcategories such as passive bilinguals and semi-speakers can be found (Dorian Reference Dorian1981, reported in Mesthrie et al. Reference Mesthrie, Swann, Deumert and Leap2000:259). Although, in this research, we have not measured speakers' competence,Footnote ⁸ it is obvious that in the Rwandan community in Belgium similar categories can be found. Ethnographic evidence, speakers' own reports, and interactional data themselves confirm the existence of a continuum of proficiency in Kinyarwanda. For example, as we have seen, adult members of the community prefer Kinyarwanda. The views in extract (4) seem to point to the categories of passive bilinguals/French monolinguals. And the young adults reporting their preference in extract (5) above can be said to belong to the category of semi-speakers. We can therefore assume that, in their daily interactions, community members orient to these language-based categories in various ways (Gafaranga Reference Gafaranga2001). As I will argue, the practice of medium request is one of the ways in which members orient to language-based categories and, by so doing, talk language shift into being.

DATA AND METHODS

Through inside knowledge as a distal (Baquedano-Lopez & Kattan 2007:73) member and through my previous work in the community, I had become aware of ongoing language shift long before the project on which this paper is based.Footnote ⁹ Thus, the project was specifically designed to document interactional practices through which the shift was taking place. The project was also undertaken with the assumption that language shift is a community phenomenon, rather than an individual or family one (Baquedano-López & Kattan 2007:79). Thus, during the recording process, preference was given to interactions involving members of the wider Rwandan community (referred to in the transcripts used here as “visitors”) in addition to immediate members of the family. This inclusion of both family and non-family members was presumed to lead participants to adopt interactional practices in currency in the wider community. Also, data were collected using the methodology I have referred to elsewhere as “Doing research in and as social interaction” (Gafaranga Reference Gafaranga1998), which had already been tested in the community. In a few words, this is a methodology whereby the fieldworker, rather than artificially standing aloof so as to “observe” interactions, participates in them like any other member. In the transcripts to be used in this article, where the researcher (a distal member of the community) is a participant, he too is referred to as a visitor.

Data collection was conducted over three phases. In the first phase, participant observation was used so that the fieldworker could familiarize himself with the general sociolinguistic situation prevalent in the community (language shift). During this phase, we also identified 25 families we would focus on during later phases of the research. In choosing the families, preference was given to those with children under 10 years of age, although some families with children we refer to as “young adults” were also included. In the second phase, naturally occurring interactions in the target families were recorded. Here again, as far as possible, preference was given to interactions involving children (child–adult and child–child interactions). Most recordings lasted between 30 and 45 minutes, but some went on much longer. And in the third phase, the same target families were interviewed for their views about the sociolinguistic situation and how they experienced it. Both the interview and interactional data have now been transcribed.

In approaching the data with a view to describing medium requests, I have adopted the conversation analytic approach as outlined by Pomerantz & Fehr Reference Pomerantz, Fehr and Van Dijk1997, among others (also see Auer Reference Auer and Heller1988 [2000], Wei Reference Wei2002, Cashman Reference Cashman, Wei and Moyer2008, and Gafaranga Reference Gafaranga, Bullock and Toribio2009 for the application of Conversation Analysis to bilingual interaction). Pomerantz & Fehr (Reference Pomerantz, Fehr and Van Dijk1997:71–74) have identified five steps one can go through in describing interactional practices. These are:

1. 1. Select a sequence.

2. 2. Characterize the actions in the sequence.

3. 3. Consider how the speaker's actions are packaged and options such packaging sets up for the recipient.

4. 4. Consider how the timing and taking of turns provide certain understandings of the actions and the matter talked about.

5. 5. Consider how the ways the actions were accomplished implicate certain identities, roles and/or relationships for the interactants.

This sequential approach, as it is also known, was applied as in the case of extract (6) below. First, I noticed a situation where a child's turn (line 2) and a directly preceding turn by an adult (line 1) did not use the same language. Typically, the child's turn would be in French while the adult's turn was in Kinyarwanda. I would then ascertain whether the two turns could be seen as belonging to the same sequence, a sequence being defined as a stretch of talk starting from the moment one participant “initiated an action and/or topic that was taken up or responded to by co-participants … [through] to the place where the participants were no longer specifically responding to the prior action and/or topic” (Pomerantz & Fehr Reference Pomerantz, Fehr and Van Dijk1997:71). In (6), for example, in line 1 B initiates an action. In line 2, D speaks to it, and the participants remain focused on it until line 5. As the transcript shows, from line 6 onward, the participants are no longer focused on the same action. In fact, we can speak of a new participant constellation along with a new activity. Therefore, talk from line 1 through to line 5 constitutes one sequence, and talk from line 6 onward belongs to a different sequence.

(6) Talk takes place in A's family. B is a visiting adult and C, D are A's children (13 and 6 respectively).

In all, 31 relevant sequences were identified. Identifying sequences in this way was felt to be necessary because, as I have argued elsewhere (Gafaranga Reference Gafaranga1998, Reference Gafaranga, Auer and Wei2007a, Reference Gafaranga2007b), language choice is an aspect of the overall order of talk organization. Thus, I speak of the medium of an interactional episode, and, in this case, of the medium of a sequence.

For each sequence I had identified, I proceeded on through the other steps. More specifically, I examined the whole sequence in order to see if and how the divergence in the first two turns was consequential for language choice in subsequent turns. In the extract above, for example, divergence at the level of language in turn 1 and turn 2 is not consequential, in that each of the participants seems to have kept his talk going in the original language. In extract (1), on the other hand, the juxtaposition of languages in turn 1 and turn 2 is consequential for subsequent choices, since the other participants seem to have given in to E's choice of French. It is through these subsequent choices that participants' own understanding of ongoing actions was revealed. In (1), for example, subsequent choices confirm that A and B have understood E's action as a request for a medium switch to French. In (6), on the other hand, subsequent choices reveal that participants have taken each other to be bilingual in Kinyarwanda and French for all practical purposes (see the notion of linguistic identity in Gafaranga Reference Gafaranga2001), and therefore that the option of keeping the interaction going in a parallel mode (Gafaranga & Torras Reference Gafaranga and Torras2001) was available to them. This article reports on the results obtained by applying this analysis to the whole data set.

FROM LANGUAGE/MEDIUM NEGOTIATION TO MEDIUM REQUEST

To begin to understand the phenomenon I'm referring to as “medium request,” a good starting point is Auer's (Reference Auer1984, Reference Auer, Milroy and Muysken1995) notion of language negotiation sequence, later respecified by Gafaranga & Torras Reference Gafaranga and Torras2001 as “medium negotiation.” According to Auer (Reference Auer1984:20–21), a language negotiation sequence “begins with a disagreement between two or more parties about which language to use for interaction, and ends as soon as one of them ‘gives in’ to the other preferred language.” An example of a language negotiation sequence is extract (7) below.

(7) (from Gafaranga & Torras Reference Gafaranga and Torras2001)

In the example, REC opens the encounter using Catalan, but EN2 formulates the service request using Spanish. That is, in the first two turns, participants are not in agreement as to which language to use. However, in turn 3, REC accommodates to EN2 and adopts Spanish, thus concluding the language negotiation sequence.

As discussed in Gafaranga & Torras Reference Gafaranga and Torras2001, such negotiation sequences can take many forms, some being more explicit than others. For example, extract (7) above is a case of implicit negotiation, while extract (8) below is a case of explicit negotiation. In (8), after the exchange of greetings, BBB explicitly requested to use English.

(8) (from Codo Reference Codo1998)

The phenomenon I'm referring to as “medium request” bears significant structural similarity with language/medium negotiation sequence. Consider the small exchange between a mother (A) and her 6-year-old daughter (D) in extract (9):

(9)

In the example, participants start from different perspectives as A uses Kinyarwanda while D uses French. However, by turn 3, A gives in and accommodates to D's choice of French. Now consider the more complex example in extract (10):

(10)

If we put aside the insertion sequence between the two adult participants (turns 3–5), it becomes clear that the pattern of language choice resembles the one in extract (7). In turn 1, B starts an adjacency pair using Kinyarwanda. In turn 2, C produces the second pair part of the adjacency pair using French. In 6, B again initiates an adjacency pair using Kinyarwanda. In 7, C responds using French. In turn 8, B accommodates to C and, from then on, talk is conducted in French. So in this case, one can speak of a successful implicit language negotiation sequence.

Although the phenomenon I am describing is similar to language negotiation in many respects, there are three reasons to see it as a very specific type. First, in language negotiation sequences, either party can initiate the negotiation. In the case at hand, only children initiate the negotiation to switch from Kinyarwanda to French. Typically, when children initiate an interactional sequence, they do so in French. And when this happens, the adult participant normally follows the leader (Zentella 2007) and adopts French. Only one exception to this was found in the whole data set. This is extract (11) below, where a mother (B) has used Kinyarwanda after her daughter (E) had initiated the sequence using French.

(11)

In addition, adult participants may introduce sequences using French, but such cases are not interesting because children respond using French and, therefore, these cases do not involve any negotiation.

Second, the notion of language negotiation, as used by Auer, does not imply any directionality. In language negotiation, the language to be switched from can be either language A or language B. Likewise, the language to be switched to can be either A or B. That is, neither the starting point nor the outcome can be predicted. An oft-cited example from Heller Reference Heller and Gumperz1982 makes this point clearly:

(12)

As the transcript shows, from what participants are doing it is clear that, in this case, the outcome could have been either French or English.

In the case at hand, as I have already indicated, the starting point is a turn by an adult participant in Kinyarwanda followed by a turn in French by the child. As for the outcome, in principle three possibilities could be said to be available: a monolingual Kinyarwanda medium, a monolingual French medium, and a Kinyarwanda-French parallel mode. In reality, Kinyarwanda is never adopted. That is, after they have initiated the would-be negotiation, children do not switch from French to Kinyarwanda. Rather, the choice, if any, is between adopting a monolingual French medium and adopting a Kinyarwanda-French parallel medium for the subsequent interactional episode. In extracts (9) and (10) above, the outcome of the would-be negotiation is a monolingual French medium. In extract (13) below, on the other hand, the outcome is the parallel medium, with the adult participants maintaining the use of Kinyarwanda while the child consistently uses French:

(13)

This seems to be consistent with other studies that have observed directional language choices in language shift situations (see for example the direction of repair in Paugh Reference Paugh and Cohen2005 and in Rindstedt & Aronsson Reference Rindstedt and Aronsson2002).

Third, according to Auer, the outcome of a language negotiation sequence is the adoption of one of the available languages as the medium. A language negotiation sequence “ends as soon as one of (the participants) ‘gives in’ to the other preferred language” (Auer Reference Auer1984:20–21). In the case at hand, as extracts (6) and (13) show, such need not be the case. Adoption of one of the two available languages, French in this case, is only one option, the other being the possibility for the interaction to carry on in the parallel mode. Thus, there may be a case for a more accurate descriptor for the sequences under investigation, but I leave this issue for another time. Rather, in order to firmly focus attention on the action of children in position 2, here I will speak of medium request and use the term to refer to the whole sequence for all practical purposes. The action of children in position 2 does indeed meet all the felicity conditions of the speech act of request (see Searle Reference Searle1969, reported in Levinson Reference Levinson1983 [2000]:240). The section below examines this particular speech act in detail, focusing on the strategies children use to accomplish it.

MEDIUM REQUEST STRATEGIES

In the data I have worked with, I have been able to identify four main strategies for accomplishing medium request. All four strategies can be thought of as repairs. A review of the literature on repair in conversation is beyond the scope of this essay, but it is important to note that repair can be accomplished explicitly as well as implicitly (Jefferson Reference Jefferson, Button and Lee1987); it can be initiated by current speaker as well as other-initiated; and it can be operated by current speaker (self-repair) as well as by the next speaker (other-repair) (see Liddicoat Reference Liddicoat2007:171–212 for a comprehensive discussion). More important for us, repair can affect any “significant aspect of talk organisation” (Gafaranga Reference Gafaranga1999), including language choice (see medium repair in Gafaranga Reference Gafaranga2000). The four strategies for achieving medium request can be thought of as various forms of medium repair whereby the speaker in position 2 asks the speaker in position 1 to repair his or her choice. That is, the four strategies can all be seen as cases of other-initiated medium repair.

Two further pieces of research that it is useful to keep in mind in order to fully grasp the strategies that children use in medium request are Zentella's (Reference Zentella1997 [2007]) work in El Bloque and Lanza's (Reference Lanza2004, Reference Lanza, Auer and Wei2007) research on language use in a mixed American-Norwegian couple in Norway. Both studies are relevant because, coming in the language socialization framework, they focus on language choice in parent/adult–child bilingual conversations. Zentella's study is further interesting because the sociolinguistic situation in El Bloque is significantly different from the one in the Rwandan community in Belgium. In El Bloque intergenerational transmission of the language (Spanish) was more or less ensured because the “flow of migrants and visitors between Puerto Rico and New York provided the children of El Bloque with regular contact with Spanish” (2007:49). Zentella (2007:84) found that, in this community, there was a norm that required children to “speak the language their addressee could understand best.” In turn, this translated into a conversational principle Zentella (2007:86–88) phrased as follow the leader. This means that children tended to adopt the language that their adult interlocutor had chosen to address them in. Of course, as Zentella notes, there were occasions when this principle was not followed, and participants adopted “non-reciprocal conversations” (2007:89). Zentella (2007:86) further noted a frequent use of What?? / Qué? as a clarification strategy. Maybe as a result of significant differences at the level of the sociolinguistic background, medium request as observed in the Rwandan community in Belgium seems to be the exact opposite of the follow-the-leader principle. In medium request, after the parent/adult has used Kinyarwanda, the child requests the use of French. In fact, in the context of the Rwandans in Belgium, unlike that of El Bloque, the follow-the-leader principle can be said to apply to parents/adult participants, since, as we have seen, one possible outcome of medium request is to adopt a monolingual French medium after the adult participant has switched from Kinyarwanda to accommodate to the child's choice. As for the clarification strategy, its use was also found to be frequent among the Rwandans in Belgium, as were non-reciprocal conversations or those in which the parallel mode was adopted (see above).

As indicated earlier, Lanza has studied language choice in a mixed American-Norwegian couple. The parents were trying to raise their child bilingually and claimed to be using the famous One Person–One Language strategy (Lanza Reference Lanza, Auer and Wei2007:55). Lanza (Reference Lanza, Auer and Wei2007:56) identified five different discourse strategies that the parents were adopting in interaction with their child:

1. 1. Minimal grasp strategy, whereby the “adult indicates no comprehension of the child's language choice”;

2. 2. Expressed guess strategy, whereby the “adult asks a yes-no question using the other language”;

3. 3. Adult repetition, whereby the parent repeats “the content of the child's utterance, using the other language”;

4. 4. Move-on strategy, whereby the adult allows the “conversation to continue without intervening”; and

5. 5. Adult code-switching, whereby the parent himself/herself uses both languages.

Lanza further noted that these strategies can be ranked along the continuum from monolingual context to bilingual context, with strategy 1 promoting a monolingual context while strategy 5 fosters bilingualism. In other words, the strategies could be ranked according to the degree to which they promoted the maintenance of the minority language, with strategy 1 being the most maintenance-oriented. As will become clear in the following paragraphs, many of the strategies I will describe have significant similarities to Lanza's, even though our analyses start from different perspectives. Lanza looks at language choice in child–parent conversation from the parent's perspective (see above) while, in my case, the focal point is the child's action in position 2. Thus, Lanza's strategies 4 and 5 cut across the ones described below, as they correspond, respectively, to what I have referred to above as the adoption of the parallel medium and of French as the medium.

The first strategy for medium request, which corresponds maximally to implicit language negotiation (see above), can be termed, following Jefferson Reference Jefferson, Button and Lee1987, as embedded medium repair. In this strategy, after the adult participant has proposed to use Kinyarwanda as the medium, the child participant attends to the ideational content of what the first speaker has said, but, at the same time, proposes to use French. That is, language choice being “a significant aspect of talk organisation” (Gafaranga Reference Gafaranga1999), the child implicitly asks the adult participant to repair his or her choice simply by flagging it as the only aspect of talk organization where they are not in line with each other. In turn, this act of flagging language choice is achieved thanks to a local organizational principle Auer Reference Auer1984 refers to as “preference for same language talk.” In some cases, as in extract (9), alignment is achieved as soon as position 3, but, in some other cases, as in extract (10) above and (14) below, it may take longer to achieve alignment. Extract (14) is particularly interesting because it shows B facing a dilemma (use of both French and Kinyarwanda in 3 and of Kinyarwanda in 5) in choosing whether or not to give in to C:

(14)

As we have seen, alignment may never even be achieved, each participant refusing to give in to co-participant's choice (see (13)).

Unlike the above, the next three strategies all imply a problem at the level of content in the adult participant's talk. To that extent, they are all forms of explicit content repair. Let's look at them one at a time. The first of these three consists of claiming not to have heard or understood, but using a different language. Parallelism with Lanza's strategy of minimal grasp and Zentella's clarification strategy is obvious. We will use the term generalized content repair. In this case, the child in position 2 reacts to the adult's talk in position 1, typically using the unspecified question Quoi? In responding to this claim, the adult participant has the choice to align or not with the child at the level of language choice. If there is no alignment, a parallel medium ensues as in extract (6) above. In the example, note that, after D's repair initiator Quoi?, B repeats himself, thus confirming that the strategy can be interpreted as a mere claim of lack of hearing by participants themselves. A second example is (15):

(15)

The relevant sequence is turns 3 to 6. In turn 3, B asks a question of D using Kinyarwanda. Instead of answering it using French, D calls for repair, indicating that he hasn't heard or understood it properly. In turn 5, B reformulates his question, still in Kinyarwanda, and in turn 6, D briefly answers the question, still in French, and goes on with further talk.

Alternatively, the adult participant may respond to the child's repair initiator by attending to content, and, at the same time, aligning at the level of language choice. In this case, a monolingual French medium is adopted. Consider extract (1), reproduced below as (16) for convenience:

(16)

In turn 1, A uses Kinyarwanda. In 2, E calls for repair. In 3, A produces the repair and, using French, aligns with E.

Of course, in this strategy as in the one described above, things might not be so straightforward. Sometimes, participants face a dilemma as to whether or not to align with co-participants. Consider (17):

(17)

In turn 1, A uses Kinyarwanda to instruct D to count. In 2, D calls for a repair, using French. In 3, A produces the repair, repeating her instruction in French. However, this time, she goes on to repeat herself for a second time, using Kinyarwanda. In this case it may well be that the repetition of the instruction in Kinyarwanda is due to the fact that talk is language-focused. Ongoing topic is whether D can or cannot count in Kinyarwanda. However, the fact that the repair is first produced in French is a good indication that the repair initiator is interpretable as an invitation to medium repair.

The third strategy is one whereby the child in position 2 calls for the repair of a specific item in what the adult participant has just said. We can use the term targeted content repair for this strategy. In its simplest form, only the specific item is repaired, as in extract (18):

(18)

In turn 2, using Kinyarwanda, B announces the move she's about to take. E fails to understand the statement fully and picks up one element (icumi) as the cause of the problem and calls for it to be repaired. And in 4, both B and D provide the repair, translating the item into French, and a new sequence begins in 5, involving A and E.

However, the need to repair a specific item may be enough motivation for switching the medium. Consider (19):

(19)

We can approach this rather long extract by noting B's shift from using Kinyarwanda (turns 1 & 3) to using French (turn 11 onward), aligning with C's consistent use of French and by analyzing how this alignment came about. In turn 1, B asks a question, but C fails to provide a relevant answer to it, leading B to reformulate the question in turn 3. At this point, no specific problem has been identified as the reason why the relevant answer has not been produced. After this second attempt, C again provides an answer that is not very relevant, and this leads A to come in to try and sort out the problem. Similar talk facilitation practices by parents have been observed in other settings (e.g.. Stivers Reference Stivers2001). To do so, she attempts to clarify B's question as best she can, using Kinyarwanda. But her clarification is not very helpful. Particularly, C experiences problems understanding the key word umubare and calls for its repair (see below for the specific strategy used here). In turn 7, A shows recognition of where the problem lies and repeats the problem item, hoping that, once the item is taken in isolation, C will be able to understand it. In 8, C calls for its repair and A repairs it in 9, using French. In 10, C proceeds with his list. At this point any hope that C will provide a relevant answer vanishes, and B switches to French. Two things must be noted in this example. The first is that A has interpreted the problem as having to do with a specific item, the word zingahe, and has replaced it with umubare and later on with chiffre. In other words, from A's perspective, the simple form of the strategy discussed above obtains. From B's perspective, however, things are different. After the problematic item has been repaired, B does not go back to using Kinyarwanda. Rather, in switching to French, he seems to indicate that, for him, the problem is not one of individual words as such, but one of the medium as a whole.

The fourth strategy, which parallels Lanza's strategies 2 and 3, consists in offering a tentative interpretation of co-participant's talk and calling for it to be confirmed. We will refer to this strategy as understanding check. In extract (19) above, for example, C has a problem with the word umubare and offers a tentative interpretation of it in turn 7 (soeur), hoping it will be confirmed. At this point, switching to French could have occurred. This is exactly what happens in extract (20):

(20)

In turn 1, A uses Kinyarwanda to ask a question of C. In turn 2, C reformulates the question, using French. As he has not selected who the next speaker should be, E self-selects and, using French, confirms C's understanding, and the conversation between the two continues in French until A comes in again (turn 6).

However, an understanding check need not lead to alignment at the level of language choice. This is the case in extract (21). As can be seen from the transcript, the parents, A and B, are consistently using Kinyarwanda while the children (D and E) are consistently using French, even though, in turn 2, E has produced an understanding check, displaying his tentative understanding of B's talk.

(21)

DISCUSSION AND CONCLUSION: TALKING LANGUAGE SHIFT INTO BEING

In this article I have described a recurrent feature of talk organization in adult–child conversations among the Rwandan diaspora in Belgium: This is the fact that children constantly request parents/adult participants to medium-switch from Kinyarwanda to French. In approaching this feature, I first noted that it shows significant similarities with the feature of bilingual conversation Auer Reference Auer1984 refers to as language negotiation, while Gafaranga & Torras Reference Gafaranga and Torras2001 refer to it as medium negotiation. However, I went on to say that it must be seen as a very specific type for three main reasons: First, only children call for the switch; second, the called-for switch is unidirectional (from Kinyarwanda to French); and third, the request to medium-switch may, but need not, lead to alignment at the level of language choice. While recognizing these facts, I proposed to adopt the term medium request so as to focus on the work that children accomplish in position 2 in the sequence.

After this initial observation, I went on to examine what specific strategies can be used to accomplish medium request. In total, four strategies were described. First, the child in position 2 can attend appropriately to the ideational content of the adult participant in position 1 while using a different language. I have referred to this as embedded medium repair, as nothing is specifically signaled as problematic at the level of content. As we have seen, the contrast created at the level of language choice can be resolved either by the adult participant's aligning with the child or by holding fast to his or her initial choice. In the first possibility, a monolingual French medium obtains, and in the second, a parallel Kinyarwanda-French bilingual medium ensues. The second strategy, which I have referred to as generalized content repair, consists of the child in position 2 indicating that she has failed to understand what the adult in position 1 has said, typically using the unspecified question Quoi? In repairing this general problem of understanding/hearing, the adult can either align with the child at the level of language choice or retain his initial choice. Here again, the first possibility leads to the adoption of a monolingual French medium while the second leads to the parallel medium. Slightly similar to the second strategy, the third strategy consists of the child participant's pointing to a specific item in adult's talk as the source of the problem and thereby asking for it to be repaired. I have spoken of targeted content repair in this case. As we have seen, adult participants may choose to repair only the single item, but they may also choose to repair the whole medium. The fourth and final strategy, which I have called understanding check, consists of the child participant's displaying a tentative understanding of the adult participant's talk, using French, and calling for this understanding to be confirmed. In confirming the child's understanding, the adult participant may align with the child at the level of language choice or may hold onto his initial choice. Briefly, among the Rwandans in Belgium, medium request follows a specifiable conversational order.

At this point two general questions arise: Why is this process occurring as observed? And so what? I would like to comment on them briefly before I conclude this essay. As this discussion has shown, among the Rwandans in Belgium there appears to be an unequal distribution of the power to influence the medium, at least in sequences involving medium requests. In these interactional sites, children have the upper hand in the sense that their language preference always prevails. Whether a medium request leads to the adoption of a monolingual French medium or whether it leads to a Kinyarwanda-French parallel medium, children are allowed to use their preferred language. This is consistent with recent trends described in the language socialization literature, where the child is seen as an active actor in the socialization process (see above). Therefore, the issue is where this particular distribution of power comes from, especially as we know that, as in the case of El Bloque, the normal expectation would be for children to accommodate to adults' language choices, if only out of respect. The answer to the question lies in the relationship between micro-conversational structures and the macro-sociolinguistic context. As argued in Gafaranga Reference Gafaranga2001 and Torras & Gafaranga Reference Torras and Gafaranga2002, language preference is a membership categorization device. In language choice, speakers take account of each other's preferences – that is, they categorize each other in terms of language preference. Language-based categories may be decided locally in a specific conversation, but they may also be part of members' stock of knowledge. In the case of the Rwandan community in Belgium, members know in common that children prefer to speak French. On the other hand, they know that adult members of the community are bilingual in French and Kinyarwanda. The state of language shift lies in this categorization and its effect on members' consequent actions.

Medium request, with the specific packaging I have described, is one of the ways in which members orient to prevalent language-based categories. As we have seen, in the Rwandan community in Belgium, only children accomplish medium requests – and the request is for the use of French. Additionally, independently of the outcome, children hold onto their initial choice of French. Conversely, adult members of the community do not initiate medium requests and respond to child-initiated medium requests either by switching to French or by holding onto Kinyarwanda. All these actions are consequential and accountable given the prevalent language-based categorization. In requesting the use of French (after Kinyarwanda has been used) –– that is, in accomplishing medium requests – children are in a sense reminding their interlocutors to be socially accountable in their language choice acts, to take account of their interlocutors' language preference; hence the notion that medium request is a type of repair. Likewise, after a medium request, Kinyarwanda is never adopted as the medium, because doing so would be contrary to the categorization of children as preferring French. By implication, adult members do not initiate medium request because their preferred language is never going to be adopted. On the other hand, adult members of the community can be asked to accommodate to the children's choice of French because they are known in common to be bilingual in French and Kinyarwanda.

The above explanation, although it is valid at a general level, leaves some of the details of the interactional practice unaccounted for. For example, it does not account for the fact that adult members initiate sequences in Kinyarwanda in the first place, nor does it explain why they can hold onto their choice of Kinyarwanda, leading to the parallel medium. In fact, the explanation, if it stopped here, would strongly imply that adult members are sociolinguistically incompetent since they have to be constantly reminded how to interact with some members of the community. Throughout this article, I have used the expressions language shift in progress and ongoing language shift, and they have their full meaning here. As I have indicated, in the Rwandan community in Belgium there is a continuum of proficiency in Kinyarwanda. If the different subcategories of speakers are interactionally “relevant” (Schegloff Reference Schegloff, Boden and Zimmerman1991), we can easily assume that, in their everyday interaction, speakers orient to them. Medium request appears to be co-selective with the category of passive bilinguals. That is, in actual interaction, adult members of the community hold children to be not monolinguals in French, but rather passive Kinyarwanda-French bilinguals. Thus, they can accountably initiate sequences in Kinyarwanda and continue to use Kinyarwanda after their child interlocutors have used French only because they hold them to be capable of following the conversation. Children themselves, in adopting specific strategies such as embedded medium repair, targeted content repair, and understanding check, confirm to their interlocutors that they have some competence in Kinyarwanda. Briefly, medium request as I have described it here is possible because of the language shift from Kinyarwanda-French bilingualism to monolingualism in French currently in progress in the community. The macro-sociological order occasions the conversational order. Conversely, through medium request language shift actually takes place – is talked into being. The conversational order and the wider macro-sociolinguistic order come together.

As indicated above, the second question that emerges from the analysis is the so-what question. As I have said above, data were collected with a view to documenting interactional practices through which language maintenance is accomplished as well as those that facilitate language shift. To understand the contribution of medium request to the issue, we need to refer to Lanza again. As we have seen, Lanza has identified five strategies that parents can use in interaction with bilingual children. Furthermore, she has ranked those strategies in terms of those that promote the maintenance of the minority language and those that don't (bilingual). The discussion above shows that, with reference to medium request among the Rwandan diaspora in Belgium, similar strategies can be identified from the child's perspective. On the whole, therefore, we can say that medium request, as a conversational strategy, facilitates language shift because it promotes the adoption of a monolingual French medium. However, as we have seen, none of the strategies for medium requests necessarily leads to the adoption of a monolingual French medium. Adult participants may hold onto their choice of Kinyarwanda, in which case a parallel medium is adopted. Elsewhere, I have highlighted the significance of the parallel mode as a practice through which language shift is accomplished (Gafaranga Reference Gafaranga2007b:196). In light of the above analysis, this view has to be nuanced. When it occurs in the context of a medium request, the parallel mode can be seen as a maintenance-oriented strategy, if only a very weak one.

As I noted in the second section, an underlying motivation for studies of language shift is to contribute to language maintenance. In addition, as we have seen, for language maintenance to be possible, “face-to-face, small-scale social life must be pursued in their own right and focused upon directly” (Fishman Reference Fishman1991:4). The aim of this article has been precisely to focus on an interactional practice in a situation of language shift in order to understand it in detail. Drawing on the above analysis, my advice to members of the Rwandan community in Belgium is to practice RLS in face-to-face interaction, in addition to setting up and participating in sociocultural organizations. A specific interactional practice they could adopt in this respect is to refrain as much as possible from giving in to children's medium requests in their different forms, even if this means adopting the parallel medium. However, as important as medium requests are in language shift situations, alone they cannot account for the totality of language shift in a community. Therefore, in line with Fishman Reference Fishman1991, I would like to conclude by calling for further detailed descriptions of interactional practices through which language shift is accomplished in the Rwandan community in Belgium and elsewhere.