Language choice and direct speech presentation in Kinyarwanda news articles

Introduction

As a Rwandan in the diaspora, thanks to the internet, I am able to access Rwandan news articles any time of day. Many of the Rwandan news outlets are currently accessible online. While reading primarily for content, I have regularly come across a puzzling language choice practice. An illustration of the practice is the following:

Extract 1 (Igihe, 3 April 2016) ¹

Yagize ati: “bapimye umwe gusa, bapima umusarani basanga nta kibazo, bapima ‘ globules blancs ’ nabyo biba uko, noneho bamubwira ko bishobora guterwa n’udukoko ( infection salmonellose ), bishobora kuva ku biryo umuntu yariye …”

(He said: “they only tested one (person), they took stool samples and did not see anything abnormal, they tested white cells and nothing unusual was found, then they told him that one can develop it (the problem) from bacteria (salmonellosis infection) in the food they ate.”)

The extract comes from a news story reporting the eruption of a food poisoning infection on a building site in Kigali, Rwanda. This language choice practice is interesting and puzzling for a number of reasons: Firstly, it occurs in the type of texts Sebba (2002, 2013) describes as “most highly regulated.” These texts, according to Sebba, are characterised by “hegemonic monolingualism.” Therefore, in theory, as the “medium” (Gafaranga, 2007) of these texts is Kinyarwanda, code-switching (hereafter CS), in this case the use of French, would have been anticipated to be impossible (ethnomethodological sense). Secondly, the noun phrase (NP) “globules blancs,” occurring in a “direct speech” presentation (hereafter DS) (Semino and Short (2004), hereafter S&S), is included in (single) quotation marks. This suggests that it is an instance of “embedded quotation” (S&S). According to S&S, embedded quotation is impossible inside DS (Semino, Short, & Culperper, 1997, p. 37). Thirdly, given the fact that “infection salmonellose” is also CS, would it too be a case of embedded quotation? If it is, why is it not included in quotation marks? Fourthly, according to Short and Semino (2002), the degree of “faithfulness” in DS is very high, especially in factual texts such as news articles. Therefore, given the occurrence of the French items, a question arises as to whether original speech was French or whether it was mixed French and Kinyarwanda. Fifthly, as the example shows, CS in DS is diverse. In the first case, the use of French is signaled as quotation within quotation (see quotation marks) while it is not in the second. In the second case, the switched item participates in the structure Gafaranga (2015) refers to as “translinguistic apposition” while in the first, it does not. Further evidence of diversity within CS in DS can be found in the following:

Extract 2 (Umuseke, 4 May 2016)

Mulindwa agira ati: “Mu bigo bimwe na bimwe bari bafite index value yo hejuru, bari ku mafaranga ibihumbi 400 na 500, kuri za minisiteri bari kuri index value yo hasi …”

(Mulindwa said: “In some institutions they had a high index value, at 400 to 500 thousand Rwandan Francs, while in ministerial departments they had a lower index value …”)

Here, as in the above two instances, CS occurs in DS. However, unlike in the above, CS seems to be part of a “bilingual medium” (Gafaranga, 2007). Briefly, the practice raises the question of the orderliness of language choice in DS in news texts and it is this question that the present paper sets out to address.

To be sure, the issue of language choice in DS arises on the assumption that the medium of the “posterior discourse” and that of the “anterior discourse” (Short & Semino, 2002) are different. Without such a difference, there is no language choice to talk of, except the choice to keep the same medium. The practice I propose to investigate has been observed in Kinyarwanda news texts. Therefore, in this paper, only reports of anterior speeches that can be assumed not to have adopted Kinyarwanda as the medium will be included. Given the Rwandan context (see below), anterior speeches meeting this criterion could have used any of French, English and Kinyarwanda–French/English language alternation as the medium. Thus, the paper will investigate the following specific research question.

What medium does DS take if the medium of the posterior text is different from the medium of the original speech and, more importantly, how can language choice at this level, whatever it is, be accounted for?

In their discussion of “faithfulness” in discourse presentation, Short and Semino (2002, p. 351) list “posterior discourse accessibility” as a factor affecting faithfulness. This means that an anterior speech may not be reproduced verbatim if doing so would hinder the recipient’s access to the posterior discourse. Applied to language choice, this would mean that the original medium may not be adopted as the medium of DS if it is likely to affect recipients’ access to the posterior text. Conversely, it would mean that, in DS, the original medium may be adopted so long as it does not affect the recipient’s access. Although this principle has some appeal, especially when considered within a monolingual context, it is not clear how it could apply in multilingual contexts such as Rwanda. As argued by Gafaranga (2015), in interaction, it is possible to assume that Rwandans are plurilingual in their country’s official languages, namely Kinyarwanda, French, and English (Constitution of the Republic of Rwanda, 2003 as revised in 2015, art. 8). Therefore, each of these and even language alternation involving two or all three can be adopted as the medium of an interaction. In this context, the posterior discourse accessibility principle would lead to the wrong prediction that, in news texts in Kinyarwanda, DS may normatively adopt as its medium any of Kinyarwanda, French, English, and even language alternation itself so long as it was the medium of the speech presented. Observation of Kinyarwanda news texts does not support this predication. So, if language choice cannot be accounted for in terms of posterior discourse accessibility, how can it be accounted for? Secondly, studies of language choice in bilingual conversation have amply demonstrated that, once a medium has been adopted for an interactional episode, it can be deviated from for specific functional effects. Therefore, the second question this paper will investigate is whether, once a particular medium for DS has been adopted, it can be deviated from and, if it is, what shapes deviance can take and what functions it can serve. In addressing these two questions, the paper will contribute to two of the three central concerns in the study of reported speech in general and of DS in particular, namely “concern with forms of reported speech, concern with its authenticity and with what it does” (Clift & Holt, 2007, p. 3).

As indicated above, the phenomenon at hand was observed in Rwandan online news stories. Currently, there are over 31 online media outlets in Rwanda (http://mhc.gov.rw/index.php?id=63, 20 April 2016). Although the examples above come from two sources, the phenomenon is not limited to these two. Extract 3 below, similar in every respect to the second instance in Extract 1, comes from a different news outlet.

Extract 3 (Umuryango, 23 March 2016)

Yagize ati: “Mu minsi iri imbere turi gushaka uko twakibungabunga, kuko turi gushaka kuhubaka agahotel k’ubukerarugendo (hotel touristique) no kuzashyiraho umucanga…”

(He said: “In the coming days we are going to preserve it as we are planning to build a tourist hotel and to bring in sand …”)

Likewise, Extract 4 comes from yet another source.

Extract 4 (Izuba Rirashe, 4 April 2016)

Yabwiye ikinyamakuru Izuba Rirashe ati: “Hakurikijwe contrat (amasezerano) ubusanzwe tuba tugomba kumwishyura nawe amaze kwishyura abakozi amezi abiri.”

(He told Izuba Rirashe newspaper: “In accordance with the contract, we normally pay him after he has already paid his employees two-month wages.”)

Briefly, CS in DS, and therefore the issue of language choice in this interactional site, is a common phenomenon in Kinyarwanda news texts and it can be found in all news outlets. However, for reasons of presentational consistency, all the examples used in the following sections come from two online media outlets, namely Igihe and Umuseke. ² These media outlets publish in Kinyarwanda, English, and French in the case of Igihe and Kinyarwanda and English in the case of Umuseke. For the purposes of this paper, only the Kinyarwanda versions were monitored. Previous studies have demonstrated that speech presentation is a very recurrent feature of news texts and that this has implications for data collection. For example, S&S collected a corpus of 2271 instances of speech presentation in news texts published over 8 days (2004, p. 24), while Harry (2014) reports having collected 740 instances from articles published over two days (31 May–1 June 2010) on the same topic. There is no reason to assume that speech presentation is less common in Kinyarwanda news texts and in the blogs I have investigated. As the two blogs continue to be accessible, as their archives can be searched and since no quantitative analysis is intended, technically there is no limit to what counts as data for this paper. Instead, following Gafaranga (2017), an inductive perspective has been adopted. That is to say, the sites of the news outlets have been (and continue to be) monitored backward (archives) and forward for instances of DS starting from the month of March 2016 and countless instances have been identified following the criterion that original speech can be anticipated to have used a medium other than Kinyarwanda. Each new instance has been examined in the light of a running hypothesis as it was encountered until no new observations could be made. In addition, two online sources of recordings of speech, namely YouTube and Igihe TV, have been used in order to compare original speeches (where they are available) and their reports in news articles. This paper systematises and reports the results of that observational work.

Sociolinguistic background

The macro “declared language policy” (Shohamy, 2006) of Rwanda can be read in the country’s constitution where it is stated:

The national language is Ikinyarwanda

The official languages are Ikinyarwanda, English and French

An organic law may add or remove an official language

Official documents may be either in one, two or all official languages

(Constitution of the Republic of Rwanda, 2003 as revised in 2015, art. 8)

For the purposes of this paper, we can begin by noting the privileged position of Kinyarwanda: Kinyarwanda is the sole national language and one of the official languages. On one hand, this privileged position at the level of the declared policy reflects the reality on the ground at the level of acquisition and competence. The vast majority of Rwandans acquire Kinyarwanda as a native language and, indeed, the vast majority of them are monolingual in Kinyarwanda (Rosendal, 2009; Samuelson & Freedman, 2010). The few Rwandans who are bilingual in the official languages are so differentially: Kinyarwanda–French bilinguals, Kinyarwanda–English bilinguals and, even less frequently, Kinyarwanda–French–English trilinguals. On the other hand, the privileged position of Kinyarwanda, as found in the declared policy, is reflected in the “perceived policy” (Shohamy, 2006). In Rwanda, Kinyarwanda is associated with “Rwandan-ness,” with the national identity (Gafaranga, Niyomugabo, & Uwizeyimana, 2013).

At the level of practices, a common feature of Rwandan multilingualism is the prevalence of language alternation practices. In Rwanda, monolingual discourse in any of the official languages tends to be consigned to formal written discourse (whether delivered in writing or orally) and to speech addressed to non-Kinyarwanda speakers. CS, on the other hand, is observed both in formal and in informal situations, in private and in public speeches, especially by members of the educated elite. In Myers-Scotton’s (1988) terminology, CS itself is the “unmarked choice” in spoken interaction among Rwandans. At the linguistic level, CS is of the insertional type (Muysken, 2000), with Kinyarwanda as the “matrix language” (Myers-Scotton, 1993). Such insertions often consist of single words either as “bare forms” or as “EL embedded” items, but they may also consist of longer “EL islands” (Myers-Scotton, 1993). In addition, at the sociolinguistic level, language alternation is perceived, not as a mixture of two separate codes, but as one variety, which Gafaranga (1998) has referred to as “Kinyarwanda-for-all-practical-purposes.” An example of CS in a public speech is the transcript below, an extract from an address by a Cabinet Minister to the Parliament on 11 February 2016. Note that the speech was recorded live on a local radio station.

Extract 5

… kwiyadapta kuri local condition. Ikindi ubasha gusellinga skills… ntabwo tuzaguma ngo dusellinge … soft skills. Abanyarwanda ntabwo ari abaswa. We are ingenious. Ariko dukeneye kubinvestingamo…tukabaa tukab- izo high tech …ingen-ingeniosity abandi bafite natwe tukabasha kuzigira kugirango tubashe.. no kuzisellinga kuko the most important resource dufite nk’igihugu ni abantu. so we need really to invest mu bantu bacu… Yes tukabalanca.. technical na scienfitic . kubera y’uko wenda the high level scientific umwe akeneye technicians batanu. Iyo balance tukamenya y’uko tugomba kuyikora ariko tukemera tukinvestinga muri high level .. science. Icyo ngicyo ngirango Minister wa Education yakivuzeho ariko nagirango nanjye nkigarukereho kuko…

(…adapt to the local conditions. The other thing is you can sell skills… We are not going to continue to sell … soft skills. Rwandans are not stupid. We are ingenious. But we need to invest in them…. So we can have those high tech…ingen-ingeniosity other people have and sell them as the most important resource we have as a country is our people. So we need really to invest in our people… Yes we would balance… technical and scientific. as maybe one high level scientific needs five technicians. We need to achieve that balance, but we must invest in high… level science. I think the Education Minister commented on this, but I thought I’d come back to it so that …)

As the transcript shows, in the address, all three official languages ³ were used, with almost all the linguistic features mentioned above. The point made earlier about language alternation being adopted as the default unmarked choice among Rwandan elites is particularly important for the purposes of this article. In Rwanda, as everywhere else, opportunities for one’s speech to be presented in news stories are skewed in favour of the socially, politically, educationally, etc., powerful (Fairclough, 1995, p. 79). Therefore, in the absence of any evidence to the contrary the speeches presented in news texts in Rwanda can be assumed to have been mixed, particularly in those cases where CS occurs in DS.

A word about the media, print media in particular, in Rwanda is in order before concluding this background. The history of news media in Rwanda is relatively short. The first publication of a newspaper dates back to the 1930s (Kinyamateka was first published in 1933). Another milestone was the publication of Imvaho in 1959. Both newspapers published, and still do, in Kinyarwanda only. Except for very minor publications, these two newspapers dominated the scene until the early 1990s. In the early 1990s, Rwanda undertook a move towards political openness (multiparty system) and an intense media activity ensued. Many new titles, some of which can still be seen today, appeared on the market. Today, the number of print media outlets is estimated to be between 30 and 50 (mhc.gov.rw/index.php?id=62, accessed 20 April 2016). With the arrival of the internet, many of these outlets started publishing online versions, and even internet-only news outlets have since emerged. The two blogs I have monitored for this study reflect this reality as Umuseke, launched in January 2011, has a paper version, while Igihe, started in 2009, is published online only. A feature of the media in Rwanda in general is the gaping training needs and there is no reason to assume that the situation is different in the case of Umuseke and Igihe. Commenting on these needs, Thompson (2007: 25) writes that “newsrooms in Rwanda are populated for the most part by young journalists with little or no professional training”. Briefly it is against this background of societal multilingualism with underlying monolingualism, of the prevalence of CS as a mode of speech and of the media with a relatively short history that language choice practices, including language choice in DS environments, as described in this paper, can be understood.

Towards a theoretical framework

Research on language choice and CS in multilingual contexts has traditionally focused on spoken discourse. However, recently, there has been a notable push to redress the situation, as evidenced by the publication of collections of studies of CS in writing by Sebba and Shahrzad Mahooton (2012) and by Gardner-Chloros and Weston (2015a). Yet despite this push, barring very rare exceptions (e.g., Doplon, 2007; Sukia, Cheng, Yee, & Ling, 2011; Xuan, 2010), the genre of the news article and, within it, the practice of DS, remains under-investigated. Two main reasons might account for this lack of research interest in CS in DS in news articles, namely the fact that news articles are, according to Sebba (2013), “most highly regulated texts” and the almost hegemonic prevalence of the “verbatim assumption” (Clark & Gerrig, 1990). According to Sebba, a major feature of most highly regulated texts is “hegemonic monolingualism.” This means that, in such texts, CS is in principle avoided. Logically, one cannot investigate something that is not in the data. However, as Gafaranga (2015) shows, in newspaper articles, what is avoided is, not CS per se, but rather the use of CS itself as the medium. Only CS as the medium “can be seen as a challenge to hegemonic monolingualism” (2015, p. 96). In other words, language choice in newspaper articles should not be seen a priori as outside the scope of investigation. As for the verbatim assumption, it is the assumption that DS is a mere word-for-word reproduction of an anterior discourse and, therefore, uninteresting. For example, Sukia et al. (2011, pp. 10–11) observed that, in presenting the speech of artists from Hong Kong, Chinese medium newspapers in Malaysia use English–Chinese mixing and they interpreted this practice as “inevitable,” that is, uninteresting, because this is the way people from Hong Kong speak. An even stronger evidence of the verbatim assumption as inhibiting the investigation of language choice in DS in newspaper articles can be read in Doplon (2007). In their investigation of “code-switching in Philippine tabloids,” they write, “Direct quotations in English are not considered in this study because it is not the writers who chose the language used” (2007, pp. 6–7).

Shortcomings of the verbatim assumption have already been expressed by scholars such as Clark and Gerrig (1990) and especially by Tannen (1989, p. 99), who sees DS as “primarily a creation of the speaker rather than the party quoted.” A compromise between these two positions is what Short and Semino (2002) refer to as the “faithfulness assumption,” arguing that it is most evident in the case of factual texts such as newspaper articles. That is to say, although DS in news texts cannot be taken as an exact reproduction of an original text, it is not a totally invented text either. Therefore, DS may be examined for the linguistic processes, including at the level of language choice (as I do later in this paper), which have been applied in moving from the anterior text to the posterior one. Briefly, the lack of research interest in language choice in DS in news articles so far, whatever its cause might be, has resulted in the lack of an “off-the-shelf” model I can draw on in answering the questions I raised earlier with respect to news articles in Kinyarwanda. Instead, a theoretical framework has to be developed, from scratch as it were.

In working towards a theoretical framework for the present paper, S&S’s work is a good starting point. S&S are believed to have produced the most comprehensive model of speech presentation (Tabbert, 2016, p. 138), covering literary texts, (auto)biographies and, crucially, newspaper articles. Developing an earlier model by Leech and Short (1981) that had been based on literary texts, the new model consists of the following categories: Narrator’s Presentation of Voice (NV); Narrator’s Report of Speech Act (NRSA); Indirect Speech (IS); Free-Indirect Speech (FIS); Direct Speech (DS); and Free-Direct Speech (FDS). Of particular relevance for this paper is the category (free-)direct speech ((F)DS). This category formally differs from the others in that, in it, the writer reproduces verbatim (supposedly) the original discourse and signals this by means of quotation marks. In that sense, it also differs from the others in that, in it, faithfulness claims are very high (Short & Semino, 2002). In turn, DS differs from FDS by the presence/absence of a reporting clause. DS is introduced by a speech presentation clause while FDS is not.

In addition to the six main categories of speech presentation, S&S describe a number of “specific speech presentation phenomena,” notably the phenomenon of categories embedded in other categories. The most notable of those embedded phenomena is the one the authors refer to alternatively as “Q-form” and as “embedded quotation.” Other scholars have referred to the same as “incorporated quotation” (Clark & Gerrig, 1990), “partial quotation” (Thompson, 1996), etc. In this paper, the term “embedded quotation” will be adopted. The phenomenon consists of “stretches of direct quotation which cannot be straightforwardly categorized as DS because of the way in which they occur inside other, non-direct (speech presentation) categories” (S&S, 2004, p. 54). As S&S say, embedded quotation can occur in any non-DS environment and it may consist of single words or of longer stretches. In other words, embedded quotation does not occur inside DS. An important aspect of embedded quotation is its function. According to S&S, embedded quotation is used when the reporter wants to highlight “some particularly important or newsworthy parts” of an original utterance otherwise presented indirectly (2004: 154) or when they “wish to attribute individual words or expressions to participants in their stories” (2004: 159). In this respect, Short and Semino (2002, pp. 349–351) speak of the “importance of the wording of what is being reported” as an important factor in faithfulness determination. In embedded quotation, the reporter claims to be reproducing the reported speakers’ own words and, therefore, distances themselves from them (Vis et al., 2015). Briefly, there are two types of DS, namely what I might call standalone presentation ((F)DS) and embedded presentation or embedded quotation.

While no account of language choice in DS in news texts is currently available, CS in DS has long been documented in bilingual conversation (e.g., Alfonzetti, 1998; Alvarez-Caccamo, 1996; Auer, 1984, 1995; Gumperz, 1982; etc.). More importantly, Gafaranga (2007, 2017) has developed a model of language choice in speech presentation in informal conversation and it can be used as a second step towards a theoretical framework for this paper. Gafaranga’s key claim is that, in bilingual conversation, the default medium of DS is that of the surrounding discourse. This is so because of an overarching principle of talk organisation alternatively referred to as “preference for same language talk” by Auer (1988) and as “preference for same medium talk” by Gafaranga (2007). Gafaranga further argues that, in actual interaction, participants may deviate from this principle and adopt a different medium in presenting a previous speech. This occurs in either of two situations. In the first case, language choice itself is the “depictive aspect” (Clark & Gerrig, 1990). In this case, the original medium is kept and is itself presented, hence the term “medium reporting” (Gafaranga, 2007, pp. 164–168). The second situation where the medium of DS may be different from the medium of the surrounding discourse is when language choice is a “supportive element” (Gafaranga, 2007, pp. 168–172). In this case, language choice is used as an additional resource in support of meanings expressed at the propositional level. An example will help clarify this type of language choice in DS. The example was recorded among Rwandans in Belgium. Participants are saying that racism is rife in the wider Belgian community. That is to say, a category contrast is established between us (Rwandans, represented by current participants), who suffer racism, and them (the wider community), who inflict it upon us. In his contribution, participant A reports an interaction he has recently had with members of that wider community. As confirmed in C’s (the listener) turn 4, the purpose of the report was to show that the wider community is indeed racist, because they know every Rwandan in their community. As the transcript shows, while the overall medium of the interaction is Kinyarwanda–French language alternation, the words of the members of the wider community are reported in French. Thus a contrast is established at the level of language choice (our language versus their language) and parallels the contrast at the situational level (us versus them). In other words, language choice in DS is used as an additional resource for category building, that is, for highlighting the distance between current participants and the wider community.

Extract 6 (Gafaranga, 2007: 169)

1. A: nagiye kumusura (.) nsigaje kilometero nk’ icumi ngo ngere iwe (.) uwo mbajije wese (.) ngo ah ngo tu vas chez monsieur (.) le monsieur là le noir

2. B: ((laughter))

3. A: umh tu t’imagines ngo tu vas chez le monsieur là (.) le monsieur le noir

4. C: eh donc bose bamuzi

——————————————

1. A: I went to visit him (.) about ten kilometres before I got to his place everybody I asked would say you are going to see mister (.) mister the black

2. B: ((laughter))

3. A: umh can you imagine say you are going to see mister (.) mister the black

4. C: eh so everybody knows him

Logically, in the case of language choice as a supportive element, because propositional content is important, deviance from the overall medium is allowed only to the extent that it does not hinder participants’ access to that content. Some connection can be seen here between this and Short and Semino’s notion of “posterior discourse accessibility” (2002).

While it is proposed that the above account of language choice in speech presentation in bilingual conversation can be used as a component of a theoretical framework for language choice in DS in Kinyarwanda news articles, it is also important to acknowledge Sebba’s (2013) caution against importing wholesale models of language choice in spoken interaction for language choice in written texts. A basic difference between spoken and written discourse is the interactive nature of the former and its absence in the latter. Therefore, a model developed for the former might have been developed to reflect that nature. However, Sebba (2013) does not rule out the possibility of drawing on models of spoken discourse either. Some aspects of CS in written texts may be analysed using models developed for CS in spoken discourse (Gafaranga, 2015; Gardner-Chloros & Weston, 2015b). Therefore, it is with this caveat that I propose to flex the muscles of the above model of language choice in DS in everyday conversation by applying it to DS in Kinyarwanda news texts.

Language choice and direct speech presentation in Kinyarwanda news texts

Embedded medium quotation as functional deviance

As we have seen, S&S have identified embedded quotation as a very important tool for journalistic writing. Among other things, embedded quotation allows reporters to attribute “individual words or expressions to participants in their stories.” (2004, p. 159). That is to say, embedded quotation is functional. As DS already attributes the words in it to a participant, it is logically impossible to have embedded quotation within DS (Semino et al., 1997, p. 37). However, this seems to be based on a mono-dimensional view of discourse, namely that only the propositional content of speech can be presented. This mono-dimensional view has been rejected by scholars such as Clark and Gerrig, according to whom “anything from intonation and dialect to non-linguistic actions of all sorts” (1990, p. 782) can be presented. Building on this, Gafaranga (2007, 2017) argues that language choice itself can be presented. Therefore, it becomes possible to envisage embedded quotation at the level of the medium, hence “embedded medium quotation”, occurring in a propositional DS environment.

In turn, embedded medium quotation, as observed in Kinyarwanda news texts, can take many forms. A first structural distinction is between cases where the switched item is a component part of the main DS and cases where it is not. In the latter case, the switched item appears as a second part of a translinguistic apposition structure. Extract 1 contains both types. In the first case, the French item “globules blancs” is a component part of the main DS, while in the second case, the French NP “infection salmonellose,” added as a second part of a translinguistic apposition, is not. Other examples of the first possibility are the use of the French item “diplome” and of the English “mask” in Extracts 14 and 15, respectively.

Extract 14 (Umuseke, 31 March 2015)

Yagize ati: “Ntimwaje gushaka ‘ diplome ’ mwaje gushaka ubumenyi, nimubijyane ejo mutazadusebya kandi mwaratorejwe ku ntebe y’intore.”

(He said: “You didn’t come to seek degrees you came to seek knowledge. Go and do us proud as you were trained to be the best.”)

Extract 15 (Igihe, 20 April 2017)

Yagize ati “Ariko kugeza ubu imyanzuro yafatiwe icyo kintu ni uko izo ‘ mask ’ zishobora kwambarwa igihe ari ngombwa. Ni ukuvuga igihe habaye nk’icyorezo cyagaragariye abaganga.”

(He said: “The decision so far is to wear those masks when it is necessary. That is to say when there is an epidemic doctors have been made aware of.”)

As for the second possibility, further examples are the use of “hotel touristique” in Extract 3 and of “products” in Extract 16 below.

Extract 16 (Umuseke, 30 September 2015)

Yagize ati: “Hakwiye ubufatanye bw’inzego, abaguzi n’abagurisha mu kurwanya ibi bigurishwa (products) byangiza abaturage …”

(She said: “There has to be co-operation between the authorities, businesses and consumers in the fight against these products which put people at risk…”)

In turn, the first possibility, where the switched item is a component part of the main DS, divides into cases, such as those illustrated above, where the element stands alone and cases where it enters in a translinguistic appositive structure. An example is the use of “official” followed, in an appositive structure, by its Kinyarwanda equivalent in Extract 17.

Extract 17 (Igihe, 29 March 2016)

Yagize ati: “Basanzwe bababuza bakongera bakabareka. Wenda keretse niba babigize ‘ official ’ (ku mugaragaro) ariko ubundi si ubwambere.”

(He said: “They have always stopped them and then allowed them. Except if they have made it official otherwise it is not the first time (they do it).”)

Also consider Extract 18.

Extract 18 (Umuseke, 21 September 2015)

Ngo arite mu gutwi, Mugesera yongeye kwaka ijambo, mu mvugo wumva ko yuje uburakari n’umujinya mwishi agira ati “kuvuga ko ntinza urubanza nta n’igikoresho bampaye (Ubushinjacyaha) kandi hari ibyo Umushinjacyaha Mukuru yari yanyemereye ntabimpe ca me fait très mal (bituma mererwa nabi).”

(On hearing this, Mugesera requested to speak and, sounding very angry, he said “to say that I am delaying the case when they (the prosecution) haven’t given me the tools the Prosecutor General had promised me makes me feel bad.”)

Briefly, embedded medium quotation can take any of three forms: standalone alternated item; alternated item as a first part of a translinguistic apposition; and alternated item as a second part of a translinguistic apposition.

Two points must be made before we conclude these structural considerations. First of all, it is important to note the difference between language alternation as embedded medium quotation and language alternation as the medium of DS. In his early contribution to the field of CS, Auer calls for the need to distinguish between language alternation phenomena in which alternation is oriented to and used as such by participants themselves and those in which the other-languageness of the alternated element is not oriented as such (Auer, 1995, p. 117). In view of this, in later studies, the notion of “language alternation itself as the medium” has been adopted for those instances where language alternation is not oriented to as such, keeping the notion of CS for cases where the other-languageness of the relevant elements is oriented to as such (Gafaranga & Torras, 2002). In the case of language alternation in DS, difficulties could potentially arise in the case of alternated items that are component parts of the main DS. In the case of alternated items that are not part of the main DS, that is, which come as second parts of translinguistic apposition structures, the use of brackets ensures that the alternated items are outside the scope of the main sentence structure. As the data shows, in embedded medium quotation, writers themselves orient to the other-languageness of the alternated items that are part of the main DS. In Extracts 1, 14, and 15, the alternated items are signaled as other, that is, as deviance, by means of single quotation marks. However, there are situations, such as Extract 19, where no signalling is used. In such cases, functional analysis (see below), can help settle the question of whether a particular instance is a case of embedded medium quotation or a case of language alternation itself as the medium of DS.

Extract 19 (Umuseke, 30 September 2015)

Yagize ati: “ Comme un grand professeur de l’université , birakwiye ko nandikira ku tuntu nk’utu!!!” (Yerekanaga impapuro zishaje yandikiraho)

(He said: “As a senior academic, is it acceptable that I write on small things like these!!!” (Showing old pieces of paper in which he has scribbled notes)

Secondly, there is a need to confirm that the use of alternated items that are not part of the main DS, that is, which come as second parts in translinguistic structures, can indeed be seen as an embedded medium quotation, that is, as a “faithful” (Short & Semino, 2002) reproduction of items in their original media. As they are not part of the main DS, such items could also be interpreted as the presenter’s own comments. A “checkability” (Short & Semino, 2002) test, consisting of a comparison of the recordings of original speeches and their reports, was conducted on some instances and it gave positive results. Consider Extract 19.

Extract 19 (Igihe, 25 June 2017)

Ati “Jenoside ikirangira, nta maduka yari ahari, nta n’iduka ritoya riciriritse ricuruza isabune ryari rihari, byose byari byarangijwe, ibindi byaratwawe, ubutunzi, ububiko (coffers), nta na kimwe cyari gihari, byose byari byaratwawe.”

(…)

Ati “ (…) Hanyuma turavuga tuti reka tubatinyure turashaka gusangira namwe imbogamizi (to take the risk with you), tugura 51%, bo bari bafite 49% ariko tubaha inshingano zo kuyiyobora.”

(He said: “At the end of the Genocide, there were no shops, there were no small shops selling basic commodities such soap, everything had been destroyed, properties, coffers, there was nothing, everything had been stolen.”

(…)

He said: “(…) then we offered to share the risk and bought 51% of the shares against their 49 and yet allowed them to take control of its (MTN company) administration.”)

The extract comes from a story reporting what President Kagame told journalists in a news conference about the origin of the wealth of the ruling party (RPF). In the extract, there are two relevant instances (highlighted). The recording of the original speech (see https://www.youtube.com/watch?v=oIT_dN3HXkI, accessed 10 July 2017) confirms that both “coffers” and “to take the risk with you” were actually used by President Kagame himself. Also consider the following:

Extract 20 (Umuseke, 19 July 2017)

Yagize ati “Oya. Kagame arakunzwe cyane rwose. Ariko igihugu cyahisemo politike y’ubwumvikane (système consensuel). Uguhangana kwa Politike ntikurimo. Icy’ingenzi gikomeza kuba kongera kubaka igihugu.”

(She said: “No, Kagame is actually very popular. However, the country has adopted a political system based on consensus. Political confrontation is not welcome. The focus remains to the reconstruction of the country.”)

The extract comes from a report of an interview the Rwandan Minister of Foreign Affairs has had with Jeune Afrique, a French medium international news outlet. As the transcript of the interview shows (http://www.jeuneafrique.com/mag/455397/politique/rwanda-louise-mushikiwabo-personne-nest-empeche-de-se-presenter-a-presidentielle/), the Minister actually used the NP “système consensual.”

So far in this section, I have described the forms deviance from the default DS medium can take. Three patterns have been identified, namely standalone alternated item, alternated item as a first part of a translinguistic apposition and alternated item as a second part of translinguistic apposition. I now turn to the functions deviance can serve. Recall that, in monolingual news texts, embedded quotation is functional, as writers use it to highlight “some particularly important or newsworthy parts” of an original utterance (2004, p. 154) or “to attribute individual words or expressions to participants in their stories” (2004, p. 159). This would also appear to be the primary function of embedded medium quotation. In embedded medium quotation, writers attribute the medium of “individual words or expressions to participants in their stories” (S&S, 2004, p. 159). However, a closer examination of the data reveals a more complex picture. In the literature on language choice in bilingual conversation, the notion of function has been used either to refer to the role of language choice in the negotiation of identities and social meanings in interaction or to its contribution to talk organisation (see the socio-functional perspectives on CS in Gafaranga (2007)). In relation with the organisational function of language choice, Auer (1988) notes that language alternation can be addressed to issues having to do with the medium (e.g., lexical gaps in the selected medium) just as it can be addressed to general aspects of talk organisation, such as turn-taking, preference marking, marking repair, etc. In the first case, he speaks of “participant-related” phenomena and, in the second, he speaks of “discourse-related” phenomena (1988/2000, p. 170).

Observation of Kinyarwanda news texts reveals that embedded medium quotation in its various forms can serve the same two broad functions. As we have seen, in Kinyarwanda news texts, reporters replace non-Kinyarwanda items by their Kinyarwanda equivalents so as to achieve a Kinyarwanda medium. However, this process is far from a straightforward one. As is well-known, in many postcolonial contexts, a challenge for indigenous languages, including Kinyarwanda, is the limited nature of their lexicon (Simpson, 2008). In face-to-face conversation, participants overcame the challenge, among other things, by adopting language alternation itself as the medium. On the other hand, in “most highly regulated” (Sebba, 2013) texts, such as news articles, writers do not have this option. Monolingualism is hegemonic in such texts. Thus, faced with lexical gaps in Kinyarwanda, writers improvise translations of non-Kinyarwanda items. In many cases, these spontaneous translations are successful and pass unnoticed. However, there are cases where they are felt to be either not transparent or not exact. In this case, the original item in its original medium is appended immediately after, this leading to a translinguistic apposition structure with the alternated item as its second component. The term “back-translation” will be used in this case. Examples of this process can be found in Extracts 3, 16, 19, and 20. Alternatively, no translation is attempted and the original item, signaled as other by means of single quotation marks, is inserted as it is. Examples of this process can be found in Extracts 1 and 15. In both cases, corresponding to what Nakvapil (2016) has called “discourse-based language management,” language alternation in DS is medium-related.

On the other hand, as we have seen, there are cases where alternated items are a component part of the main DS. These serve a discourse-related function. Consider Extract 14. The extract comes from a speech delivered during a graduation ceremony in Butare (Rwanda). In the speech, a contrast is established between “diplome” (degree certificate) and “ubumenyi” (knowledge). Likewise, consider Extract 18. The extract comes from the report of the court hearing of Dr Leon Mugesera, a former university lecturer (professeur), who is being tried for genocide-related offences. The former lecturer is complaining about not receiving the resources he needs to prepare his defence and notably that he has not been provided with adequate writing facilities. Therefore, brandishing old pieces of paper on which he had written his notes, he contrasts his academic status and the destitute conditions he has to work in. Here again, language contrast seems to be used in support of this situational contrast. In both examples, language choice contributes to the organisation of discourse and one can speak of language choice as a supportive element (Gafaranga, 2007, pp. 168–172).

This supportive function of embedded medium quotation comes with a constraint. As argued by Gafaranga (2007, 2017), in DS, language choice as a supportive element is conditioned by “posterior discourse accessibility” (Short and Semino, 2002). As we have seen, Rwandans are differentially competent in all three official languages. Therefore, in some cases, the potential for a clash between the function and the posterior discourse accessibility becomes salient. In this case, the strategy of translinguistic apposition is adopted in order to circumvent the clash. The alternated item is used and its Kinyarwanda equivalent is appended. In this case, we can speak of Kinyarwanda annotation. Take Extract 17, for example. The extract comes from a report of the security situation at the border between Burundi and Rwanda. It has been reported that Burundi has closed its border, thus stopping its citizens from taking their produce to markets in Rwanda. Asked about this situation, a local authority in Rwanda said that there is nothing new about this as Burundi frequently prevents its citizens from taking their produce to Rwanda only to allow them again shortly after. The only difference, he says, would be if this time this was an “official” decision, implying therefore that previous actions were informal. Therefore, a contrast is established between what is normally done (unofficial) and what would count as different (official) and language contrast contributes to this underlying contrast at the level of content. To ensure that this use of French does not become a barrier, thereby undermining its very purpose, its Kinyarwanda equivalent is appositioned. Extract 18 comes from the same court case as Extract 19 (see above). One of the most commonly cited functions of CS in bilingual conversation is what Gumperz (1982, p. 79) has referred to as “message qualification.” In this function, among other things, CS allows participants to highlight the main message relative to background information (Sebba & Wootton, 1998, p. 269). Before the extract, prosecution has alleged that Dr Mugesera is delaying the case. In the original situation, Mugesera would have used this strategy to contrast what he feels (his reaction) (“ça me fait très mal”) and what makes him feel so (background). In presenting Mugesera’s discourse, the writer preserved the original contrast by preserving the original language contrast. Here, as in the previous example, the French clause “ça me fait très mal” is annotated in Kinyarwanda as soon as it is produced so as to ensure it does not become a barrier (for some readers at least) to the posterior discourse.

Thus, based on the discussion so far of the forms and functions of language choice, a schematic representation of the practice of language choice in DS in Kinyarwanda news texts is shown in Figure 1.

OPEN IN VIEWER

Figure 1.

Language choice and speech presentation in Kinyarwanda news texts.

In his description of language choice in bilingual conversation, Auer made the important observation that, in language, “category boundaries are fuzzy, and any attempt by the analyst to dissolve this fuzziness in favour of the Procrustean bed of clearly delimited categories will lead to a loss of realism in description” (Auer, 1988/2000, p. 177). Consider Extract 21.

Extract 21 (Umuseke, 22 April 2016)

Ati: “urabyumva nawe impamvu bari batekereje kuyiduha ni uko iyo twari dufite itari ijyanye n’igihe. Ubu turabyariza muri ‘ conditions ’ (uburyo) zitarizo.”

(She said: “You too can understand. The reason they wanted to give us another one is because the one we currently have is not up to scratch. Currently deliveries take place in conditions which are not good.”)

The extract comes from a story on a project, initially meant to build a maternity wing at a local hospital, which has been stagnant for a long time. The Head of Maternity is reported saying to journalists that, currently, women deliver in inappropriate conditions. Formally, given the direction of CS (French + Kinyarwanda), the instance looks like a case of “discourse-related function + annotation.” However, no particular discourse function can be attributed. On the contrary, content-wise, the extract is very similar to cases of “medium-related + back-translation.” Indeed, “uburyo” and “conditions” are not equivalent. “Uburyo” translates better as “ways.” Formally, for the instance to be a typical “medium-related + back-translation,” the direction of CS would have to be reversed. That is to say, the instance is between the two categories, and has the properties of both categories. Also consider Extract 22.

Extract 22 (Igihe, 7 July 2017)

Yagize ati “Imitungo iri kuri ruriya rutonde ntirishyurirwa ibirarane rwose! Turacyakurikiza procedures (inzira zabyo), tuzayigurisha (imitungo) nk’uko amategeko abiteganya.”

(He said: “the properties on the list have not paid the arrears yet! We are following the procedures, we will auction them in accordance with the law.”)

The extract comes from a story on an upcoming auction of some properties whose owners have defaulted on tax payments. An officer of the relevant authority is reported saying that the properties have not been auctioned yet only because procedures are being followed. As the transcript shows, the English word “procedures” is used without any quotation marks as if it was participating in a bilingual medium. However, right after, a Kinyarwanda translation is apposed, as in the structure “discourse-related function + annotation.” Here again boundaries between two categories are blurred. In short, the prototypical categories can be seen as points of reference with respect to which less prototypical instances can begin to be examined. Also, it would not be too farfetched if these cases where category boundaries are blurred were seen as deviance, that is, as yet another evidence of training needs in this aspect of language use in Kinyarwanda news texts.

Summary and conclusion

This paper was triggered by the observation of what I considered to be a puzzling case of language choice in Kinyarwanda news articles. In these articles, CS was observed in DS. Not only CS was anticipated to be impossible in these articles given their nature as “most highly regulated texts” (Sebba, 2013), but it also showed a great deal of diversity. Some instances were signaled, by means of quotation marks, as embedded quotation while others were not; some were used in translinguistic apposition structures while others were not; where translinguistic apposition was observed, the direction of language alternation appeared to vary from one instance to the next, etc. Therefore, I set out to investigate the relationship between language choice and DS in Kinyarwanda news texts. In order to do so, I specified two research questions:

To answer these questions, I conducted qualitative observation on two online news media blogs, namely Igihe and Umuseke. In terms of theoretical background, I drew on S&S’s model of speech presentation. This model has the obvious advantage of having focused on speech presentation in news texts, but it also has the disadvantage that it has limited itself to monolingual news texts. Therefore, while the model provided insights about speech presentation in news texts in general, it did not have much to say about language choice in DS. To overcome this limitation, I drew on Gafaranga’s (2007, 2017) model of language choice in speech presentation. Here again, while the model provided insights about language choice in DS, it was limited to everyday conversation. Therefore, a combination of the two models was felt to be necessary and, I hope, it has led to satisfactory answers to the research questions.

Systematic examination of language choice in DS in Kinyarwanda news texts revealed that the default medium for DS is Kinyarwanda. In Kinyarwanda news texts, Kinyarwanda is used as the medium of DS independently of the medium the original speech could have adopted. In other words, the medium of original speech was found to be incidental in DS in Kinyarwanda news texts. This result is significant if it is viewed against Short and Semino’s (2002) posterior discourse accessibility principle. Applied to language choice, this principle would allow any original medium to be reproduced in DS so long as it does not hinder the recipient’s access to the posterior discourse. In the Rwandan context, any of Kinyarwanda, French, English, and even language alternation would be predicted to be a potential medium of DS. The data I have examined does not confirm this prediction. On the otherhand, if the medium of original speech is seen as incidental in DS, the general language choice principle of preference for the same medium discourse applies. In short, at this level, language choice in DS in news articles is better accounted for in terms of discourse organisation than in psycholinguistic terms, such as posterior discourse accessibility.

Observation also revealed that, once the medium of DS has been adopted, it can be momentarily deviated from. I adopted the term “embedded medium quotation.” In the paper, I demonstrated that, beyond the primary function of attributing parts of discourse to their originators, there are two potential motivations for deviating from the medium. Deviance may be due to issues related to the medium itself (e.g., lexical gaps) and it may be discourse-related (e.g., contrast at the level of language choice supporting contrast at the level of content). This was seen to parallel Gafaranga’s (2007, 2017) distinction between language choice itself as the depictive aspect and language choice as a supportive element. At a lower level, each of the two categories detailed into situations where the switched item entered into an appositive structure with its Kinyarwanda equivalent (translinguistic apposition) and those where it did not. Translinguistic apposition was adopted, among other things, as a response to an anticipated communication problem due either to readers’ differential access to the languages involved or lexical gaps in Kinyarwanda. In turn, the direction of language choice in translinguistic apposition was seen to be category-specific. In the case of discourse-related functions, CS took the direction non-Kinyarwanda–Kinyarwanda, while in the case of medium-related deviance, the reverse direction was most common. Briefly, the paper has shown not only that an embedded medium quotation need not be excluded from DS environment, but also that it is structurally and functionally more complex than a casual notice, along the lines that the function of the embedded quotation is to help attribute “individual words or expressions to participants in their stories,” (S&S, p. 159), would suggest.

Although this investigation has been undertaken as a response to a personal observation and although, as the discussion above shows, the practice described might be context-specific, the study itself has implications beyond the specific case and they are worth mentioning by way of a conclusion. First of all, Gafaranga argues that, in order for the study of multilingualism to continue to be “relevant and interesting,” research should “aim to build an ever-growing collection of detailed descriptions of interactional practices involving the use of two or more languages” (2017, p. 11). This study of language choice in DS in news texts in Kinyarwanda is yet another example of how such studies might be undertaken. Secondly, the case study adds to our understanding of speech presentation in general. Notably, it shows that language choice in DS is better understood in terms of discourse organisation and that embedded quotation, structurally and functionally more complex than previously thought, need not be excluded from DS environments. Lastly, it is well-known that speech presentation is “the basis of all news stories” (Mitchell & West, 1996, p. 51). It is also a fact that, in this era of globalisation, most news stories present speeches that were originally delivered in a different medium. Finally, as we have seen, there are hardly any studies of language choice and speech presentation in news texts. Therefore, this case study serves as a call for further empirical investigations of language choice and speech presentation in news articles in a variety of multilingual settings.