Legitimizing multilingual practices in the classroom: the role of the ‘practiced language policy’

ABSTRACT This paper revisits the notion of ‘legitimate language’ [e.g. Bourdieu 1977. “The Economics of Linguistic Exchange.” Social Science Information 16 (6): 645–668] as it relates to multilingualism in educational contexts. Since Heller [1996. “Legitimate Language in a Multilingual School.” Linguistics and Education 8: 139–157] developed the notion of ‘legitimate language’ to encompass issues of language choice, there has been a consensus that a legitimate language is a language that is appropriate in a given situation. However, a crucial issue remains to be addressed, namely that of knowing what benchmark do classroom participants use to know when a language is appropriate, that is, legitimate or not. To address this issue, this paper takes as an example the case of an induction classroom for newly-arrived immigrant children in France where multiple languages have been observed. A Conversation Analysis of a set of audio-recorded interactions reveals that whilst languages other than French are not legitimised by top-down language policies and ideologies held at the societal and institutional levels, they are nevertheless seen as legitimate according to the local ‘practiced language policy’ [Bonacina-Pugh 2012. “Researching ‘Practiced Language Policies’: Insights from Conversation Analysis.” Language Policy 11 (3): 213–234]. This paper thus argues for a multi-layered understanding of legitimacy and shows how in the classroom under study, and possibly in other multilingual classrooms, practiced language policies play a key role in the legitimisation of multilingual language practices.

simply, that no language can be said to be linguistically superior or inferior to anothersome languages continue to be legitimised whilst others are not and that for political, economic and ideological reasons so that speakers of a legitimate language remain in power whilst speakers of nonlegitimate languages remain marginalised. In this initial and fundamental understanding of the concept of 'legitimate language', emphasis is given, as can be seen in the quote above, on 'legitimate phonological and syntactic forms' (Bourdieu 1977, 650, my emphasis). This means that, initially, the notion of 'legitimate language' was applied to account for varieties of languages that were given more legitimacy over others, as in the case for instance of Standard Modern Greek over Greek Cypriot Dialect in Ioannidou's (2009) study of Greek Cypriot primary classrooms. However, the concept has since then been enlarged to encompass issues of not only language varieties but also of language choice. This is what Heller (1996) initiated in her study of a French-language minority high school in Ontario, Canada. Her focus was not simply on identifying when and how would pupils in that school use standard French or not but also on identifying when and how languages other than French such as English and Somali would be used and legitimised, or not as the case may be. Research conducted in complex linguistic markets has also pointed to the need to further extend the concept of 'legitimate language' in order to be able to account for situations where multiple languages may coexist as legitimate languages (Søvik 2010, 6). This is the case, for instance, of Søvik's (2010) research in the Eastern Ukrainian city of Kharkiv, where he observes that the standard varieties of both Ukrainian and Russian are given legitimacy.
Interestingly, whether the notion of 'legitimate language' is conceptualised as encompassing language varieties or language choice, a single or multiple languages, a consensus arises on the idea that a legitimate language is a language that is appropriate in a given situation. As Heller would put it, 'form is not everything; form must be used appropriately ' (1996, 155). This is echoed by many of the researchers mentioned above. Costa, for example, defines legitimacy as being 'the ability to utter the right linguistic forms at the right linguistic moments in the right situations, and to comply with the type of discourse that society expects one to produce ' (2015, 129, my emphasis). Similarly, in her reflection on legitimate talk, Copland argues that 'legitimacy refers to the acceptability of practices within particular contexts' (2012, 4, emphasis added). Crucially, this raises an important question, namely: according to what benchmark is a language (or are multiple languages) seen as right/appropriate/acceptable or in other words, legitimate? And relatedly, how is a language legitimised? In their study of a Gujarati and Chinese complementary schools, Creese and Blackledge (2010) had been asking similar questions. Following their observations of pedagogic bilingual practices in English and a heritage language being seen as legitimate, they concluded their paper by inviting for further research to be conducted on classroom language ecologies to show 'how and why pedagogic bilingual practices come to be legitimated and accepted by participants ' (2010, 113, my emphasis).
This paper aims to address exactly these questions. It aims to understand what benchmarks do speakers use to know what language(s) is legitimate, i.e. appropriate, or not and when. In this paper, I argue that although legitimacy is often understood from a critical perspective, it can also be understood from a practical perspective. I further argue that legitimacy is therefore a multi-layered phenomenon and that, to fully understand how languages are legitimised, these multiple layers or perspectives need to be taken into account. To support this argument, I take the case of an induction classroom for newly-arrived immigrant children in France, where multilingual practices are seen as legitimate despite the fact that only French is legitimised by top-down language policies and ideologies. I show how understanding legitimacy from a 'critical' perspective only accounts for some, but not all, of these language choice and alternation practices and that for a full account, a 'practical' understanding of legitimacy is needed. Indeed, a 'practical' approach to legitimacy shows that to know whether a language(s) is legitimate or not in this classroomand possibly in other education contextsclassroom participants refer to the 'practiced language policy' (Bonacina-Pugh 2012; see also Papageorgiou 2015), that is, a set of interactional norms of language choice.

Context: an induction classroom in France
The linguistic ideology in France and its educational system is strictly French monolingual. At the societal level, monolingualism in French is considered to be the key to successful social cohesion and the use of any other languages is thought of as a threat to that cohesion. At the level of the educational system, the language policy is also strictly monolingual as evidenced by the Code de l'Education where it is stated that French should be the only medium of instruction: The language of teaching, exams, entrance examinations, as well as theses and dissertations in public and private education institutions is French, except for the teaching of foreign and regional languages and cultures or, when teachers are visiting professors or foreign guests. (Code de l'Education, Article L-121-3: II, my translation) Monolingualism in French is considered to be the key to successful learning and bilingualism in French and another language (especially in a migrant language) goes either unnoticed (e.g. Hélot 2003) or is thought to lead to potential educational failure (e.g. Young 2014) to the extent that, in practice, pupils' multilingual resources are rarely drawn upon.
This monolingual language-in-education policy sits at odds with the increasing multilingual student population in the French education system. This contrast is all the more salient in the context of induction classrooms for newly-arrived immigrant children, where all pupils are speakers of at least one language other than French. These classes are found at primary level and host newlyarrived immigrant children upon their arrival for a maximum of twelve months in order to teach them French before they join a mainstream classroom on a full-time basis (e.g. MEN 2002). In the classroom under study, the twelve pupils spoke a total of eight languages; namely French, English, Spanish, Peul, Japanese, Polish, Lithuanian, and Arabic. Only English, Spanish and Peul (a language from Senegal) were shared between two or more children. More specifically, five children spoke Spanish (these have been given the pseudonyms of Talia, Leila, Maia, Cristina, and Andrea), four spoke English (namely, Karen, Talia, Maia and Matilda) and two spoke Peul (Amkoulel and Samba). It should also be added that some children who had attended this classroom in the previous year came back to attend specific lessons during the week as a way of getting further language support. This was the case of Anika, a Romanian-speaking child. The induction teacher, Miss Lo (a pseudonym), was a white French female in her mid-forties who spoke French as her first language. She also had some knowledge of Hindu and English but could not understand any other languages spoken by her pupils. Contrary to other classrooms I have observed in the past in France , in the classroom under study pupils' multilingual repertoires were used in classroom talk as a resource.

Methods and data
The interactional data analysed in this paper is drawn from an ethnographically informed study of an induction classroom in France at primary level. While a variety of data ranging from interviews to photographs of the school walls have been collected during a period of six months, this paper focuses on a corpus of 30 h of verbal interaction audio-recorded in the induction classroom under study. The focus of the analysis will be on the multilingual practices observed in teacher-led talk in this corpus.
To analyse the legitimisation processes of these multilingual classroom practices, I call for the need to adopt not only a 'critical' but also a 'practical' perspective. The latter involves a focus on practices and implies conceptualising talk as practical social action as devised in the traditions of Ethnomethodology and Conversation Analysis whilst the former involves a focus on the social, economic and political constraints that impede on language practices as investigated in the traditions of postmodernism, poststructuralism and critical social theory. In this paper, I draw on Conversation Analysis (CA) as applied to the study of code-switching (as initially introduced by Auer 1984) and the study of Language Policy (Bonacina-Pugh 2012). CA calls for a sequential analysis of language acts and provides an emic approach to the analysis of multilingual interaction. I adopt the notion of 'medium of classroom interaction' (Bonacina and Gafaranga 2011), which builds on the work of Gafaranga and Torras (2001; further developed in Gafaranga 2012 etc.). This notion implies that talk is conducted in a medium, that is a 'code', rather than a language, and that this medium can either be monolingual or bilingual. This means that rather than investigating what language(s) are used in interaction, I try to discover what 'medium' participants actually orient to in interaction, whether it being made of monolingual or multilingual linguistic resources.
In the classroom under study, classroom participants used five mediums; namely, a French monolingual medium, an English monolingual medium, a Spanish monolingual medium, a French and English bilingual medium or a French and Spanish bilingual medium. Classroom participants could also temporarily deviate from a chosen medium and switch to another language for functional purposes. Of interest here is the fact that classroom participants saw all these language choice practices as legitimate, that is, as being appropriate/right/acceptable. This raises the question as to how did these multilingual practices come to be seen as legitimate in an educational environment where language policies and ideologies legitimize French monolingual practices exclusively?which brings us back to the central aim of this paper, namely to understand what benchmark do classroom participants use to know what language(s) is legitimate or not. 'Critical' and 'practical' understandings to legitimisation will be adopted in turn to understand how these classroom participants legitimize multilingual classroom practices in an otherwise strictly French monolingual environment.

Understanding legitimacy from a critical perspective
Legitimacy has traditionally been understood from a critical perspective, that is, it has been informed by critical social theory, postmodernism and poststructuralism. According to Fairclough (1992, 7), critical language study 'highlights how language conventions and language practices are invested with power relations and ideological processes which people are often unaware of'. From this perspective, a language is said to be 'legitimate' if it is invested with power by dominant discourses and ideologies in the wider society. For instance, a legitimate language can be one that is institutionalised as the official language of a particular domain (e.g. education, government etc.). From this perspective, speakers are said to use these ideologies and top-down policies as reference points to know whether a language is legitimate or not in a given situation. In this section, I show how the language policies and ideologies held in the school and the French educational system at large can account for the legitimacy of some but not all of the language practices observed in the classroom under study and how a multi-layered understanding of legitimacy is needed.
As a first example, let us turn to extract 1 below where a French monolingual medium of classroom interaction is used between Miss Lo and the pupils. Miss Lo is showing a cartoon to the whole class and asks pupils where the main character (i.e. the detective) is sitting (i.e. in his office).
Extract 1: original transcript Bold font is used to transcribe talk uttered in Spanish.
Miss Lo: il est dans son bureau Extract 1: Translation Bold font is used to transcribe talk uttered in Spanish.
Miss Lo: in his office 23.
Miss Lo: he is in his office In this extract, both French and Spanish are used. However, they do not have the same legitimate status. While the choice of French is oriented to by speakers as being appropriate, the use of Spanish is oriented to as being inappropriate. This is evidenced by the fact that, although Leila is obviously having troubles finding the word elicited by Miss Lo, she does not immediately switch to Spanish (line 21). The 'deviant markers' such as the pauses, the delaying device (the elongation of a vowel) and the recycle (line 21) signal that Leila sees the use of Spanish as being deviant from the current medium of classroom interaction. This orientation is confirmed by Miss Lo, who repairs the Spanish label ('oficina') by providing the French equivalent ('bureau') (line 22). Clearly, Speakers orient to French as being the current medium of their interactional episode, while they orient to Spanish as being an instance of deviance that needs repair. In other words, French is seen as the legitimate language choice act while Spanish is seen as illegitimate. These orientations are easily accountable from a 'critical' approach to legitimisation. Participants see French as legitimate because it is a language that is conferred legitimacy beyond the classroom, in the educational system and the French society at large. It is the one and only official language, the language of power and prestige. In this regard, it can be argued that, in this extract, the classroom participants see French as the only legitimate language because they are referring to the top-down language policy and to surrounding ideologiesin other words, to benchmarks outside interaction itself. Interaction between Miss Lo and the pupils is systematically conducted in a French monolingual medium (while pupil-pupil talk can be conducted in any of the five mediums of classroom interaction). Nevertheless, language alternation occurs often and can even be seen as legitimate. While a 'critical' approach to understanding language legitimisation can explain how the classroom participants see French as legitimate (as in extract 1 above), it fails to explain how some instances of alternation to a language other than French gets to be legitimised since languages other than French are not conferred legitimacy by top-down language policies and wider ideologies. How can these language alternation practices be seen as legitimate? Consider the legitimisation processes taking place in the word searches found in the following extract. Matilda (an English speaking child) is talking with the teacher (Miss Lo) searching for the French title of the movie she watched over the weekend.  In keeping with previous talk, Matilda starts her turn in French (line 77). However, she soon runs into difficulty finding the mot juste, as evidenced by the numerous pauses and self-interruptions. She then says that she is lacking the word in French (line 79) but that she knows it in English (line 80), asking in this sense permission to switch to English. These first few lines of the extract show that Matilda is orienting to a French monolingual medium. Matilda's interlocutors understand that she is missing a word but they cannot guess what that word may be (this is typical of 'self-initiated other-repair' sequences; c.f. Gafaranga 2012, 19). Miss Lo thus licenses a switch to English (line 85) and in line 87, Matilda provides the title of the movie in English. The two-second pause (line 88) following the switch shows that the teacher allows Matilda to self-repair (a preference for self-repair in conversation was noted by Schegloff, Jefferson, and Sacks 1977); which she attempts in the next turn (line 89). However, the numerous trouble markers (euh, pauses, elongation of vowels and reformulations) signal that Matilda is still lacking the French word and is thus unable to self-repair. In the next few turns and beyond the extract, Miss Lo and Matilda will try and repair the English word and find its French equivalent by relying on the expertise of a more fluent English and French bilingual pupil.
Of interest here is that the switch to English is an instance of 'other-language repair initiation' (Gafaranga 2012, 39); that is, it contributes to repair initiation in clarifying exactly what the problem is. This is made possible only because it was licensed by the teacher first. Furthermore, when licensing English, the teacher says that the use of English is 'not a problem for her' (line 85), thus distantiating herself from the institutional language policy. This signals the possibility of a new 'linguistic market' (Bourdieu 1977, 38), that is, a space at the local level of the classroom where language legitimacy may obey different rules. One might wonder if in extract 3 above, English is legitimised in teacher-led talk simply because it is a high-prestige language and a language that the teacher understands (even if only partially). However, other extracts in the corpus show that languages other than English can also be legitimised in teacher-led talk. A further example can be found in extract 3 below where Romanian is used. Anika, a Romanian-speaking child, is looking for the French word for 'garlic' (i.e. 'ail') as she is explaining what her mum cooked for her the previous evening. A French monolingual medium is used throughout.
Extract 3: original transcript Bold font is used to transcribe talk uttered in Romanian.
Anika: no:: In this extract, Anika is having difficulties searching for a French word and her interlocutors are experiencing difficulties guessing what she is referring to, which means they cannot help and provide other-repair. Miss Lo conducts a 'teacher-initiated peer-repair' (Seedhouse 2004, 147) and asks the other pupils in the class to try and help Anika (lines 37-38). As a result, in line 43, Matilda offers a candidate to repair Anika's problem. However, the two long pauses and lack of ratification (lines 42 and 44) indicate that the problem is not solved. As Miss Lo realises that Anika will not be able to give more precise information regarding the nature of the trouble source, she licenses language alternation in Romanian, Anika's first language (line 49). This switch to Romanian (e.g. line 51) is followed by a switch to a non-verbal medium (i.e. drawing; line 57), which will eventually lead participants to understand the word Anika is missing. Interestingly, language alternation to Romanian is licensed to help Anika clarify what word she is missing and to enable her peers to help; licensing language alternation thus contributes to self-initiation of repair. Romanian is not legitimised by the surrounding top-down language policies and ideologies, it is not a language of particular high prestige in France nor is it understood by the teacher. Nevertheless, at the micro level of this classroom it is licensed by the teacher and thus given a legitimate status. We now see the need to adopt a multi-layered approach to understanding legitimacy since clearly, classroom participants do not solely draw on top-down policies and wider ideologies to know whether a language is legitimate or not; they also draw on a sense of legitimacy developed at the local level of interactional practices.
Understanding legitimacy from a 'practical' perspective: the role of the 'practiced language policy' A 'practical' understanding to language legitimacy is grounded in the Ethnomethodological approach to sociology (e.g. Goffman 1959) and the Conversation Analytic (CA) approach to the study of language (e.g. Sacks, Schegloff, and Jefferson 1974). From this perspective, talk is seen as social action, and choosing a language or switching from and to a language is a social act. Recurrent language choice and alternation acts form patterns of interaction, or what is referred to in the CA tradition as 'practices'. In turn, speakers draw on these practices to deduce interactional 'norms', which will inform their interactional choices. A 'practical' understanding to language legitimacy is grounded in the argument that, when speakers see a language as legitimate, they refer not only to the wider language policies and ideologies but also to a set of norms of language choice; that is, to an implicit understanding of what the 'rules' are, or in other words of what is appropriate or not in a given 'community of practice' (Lave and Wenger 1991). The idea that a set of norms is what confers legitimacy to a particular language(s) is not new and is touched upon, sometimes more explicitly than others, in other classroom-based studies already reviewed in this paper. For example, Costa (2015) writes that in the Occitan bilingual primary school he studied in Provence, France, language choice between French and Occitan was strictly regimented at the local level of the school itself, which functioned as a 'closed linguistic market with its own rules' (2015,135). Here, languages are not legitimised exclusively with reference to top-down language policies and wider ideologies but also with reference to a local set of 'rules' or norms of what is acceptable or not in particular situations. Similarly, Creese and Blackledge (2010) suggest that in the Gujarati complementary school they investigated, the teacher and the students are 'finely tuned to the normative pattern of this classroom ecologythat is, they sense the limits of what is acceptable in terms of the use of one language in relation to the other' (2010, 110, my emphasis). We see clearly here how the notion of language legitimacy is articulated in terms of a set of norms of language choice from which speakers draw upon to know when a language is 'acceptable'. A last example, and probably the most explicit example of all, is Søvik's paper where a legitimate language is said to represent 'the perceived norm', that is 'the language one is most often expected to communicate in ' (2010, 22). He further argues that: the term legitimate language may thus not only refer to a language that is endorsed by state institutions, but also to a variety which one is expected to use in certain settings, i.e. the language variety that is seen as constituting the norm. (2010, 9, my emphasis) Building on this idea that a language is seen as legitimate with reference to a set of norms, I argue that these norms are constructed at the very local level of interaction. They are 'interactional norms' (Hymes 1972), which speakers use as a point of reference, a 'grid of interpretation' (Heritage 1984) to know when a language is appropriate or not. From an Ethnomethodological perspective, norms are there to make sense of and engage in practical social action and, from a Conversation Analytic perspective, interactional norms are there to make sense of and engage in talk as practical social action. In multilingual talk, interactional norms are there to be used by speakers to organise and make sense of their language choice and alternation acts, to know what language(s) is appropriate and when. The assumption that norms are used for the interpretation and production of acts is grounded in the Ethnomethodological principles of the 'reciprocity of perspectives' and the 'documentary method of interpretation'. The former implies that social actors hold each other to have the same perspective and the latter means that any social act is a 'document' or an example of a previously known pattern. Thus, for a speaker to know whether a given language is appropriate and therefore legitimate in a particular situation, they would have to recall what language(s) was used and seen as legitimate in a similar situation they encountered in the past and assume that the current interactional episode is another instance of that interactional practice or pattern and that therefore a similar norm applies (this is the Ethnomethodological principle of 'documentary method of interpretation'). Furthermore, they would have to assume that all participants will identify this interactional episode in the same way as them so that everyone follows the same norm of interaction (this is the Ethnomethodological principle of 'reciprocity of perspectives'). In this sense, a practical understanding to legitimacy suggests that speakers draw on interactional norms to know whether a particular language is legitimate or not.
Crucially, and as I have argued elsewhere (e.g. Bonacina-Pugh 2012), these interactional norms of language choice constitute a language policynamely, a 'practiced language policy'in the sense that they inform speakers of what language(s) is appropriate or not. The idea that there is a language policy at the level of practices originates from Spolsky (2004Spolsky ( , 2007 according to whom language policy is made of three elements, namely language management, language beliefs and language practices (Spolsky 2004, 14). Although language practices often feature in language policy studies to see whether a particular language policy has been implemented or resisted to (e.g. Canagarajah 2005), the argument that there is a policy at the level of practices is a different matter insofar as it implies seeing language policy as practice (Bonacina-Pugh 2012). I have proposed to call language policy as practice 'practiced language policy' (Bonacina-Pugh 2012) to refer to exactly thata set of implicit and deducible norms of interaction that speakers use in talk to know what language(s) is appropriate and when. Or to put it differently, practiced language policies are what Spolsky and Shohamy (2000, 29) call 'a set of descriptive and explanatory rules that would somehow capture the idea that members of the community have of appropriate behaviour'. This does not mean, however, that practiced language policies are a fixed set of norms. On the contrary, these interactional norms of language choice can be renegotiated, transformed or created every step of the way, at each language choice and alternation act. Nevertheless, knowing what practiced language policy speakers orient to in a given context will shed light on what language(s) is legitimate and when in that context. In this regard, I argue that practiced language policies play a key role in legitimising languages. They are what speakers refer to, the benchmark speakers use, to know whether a particular language is appropriate and therefore legitimate in a given situation.
The issue remains now to know how to uncover practiced language policies. I have made the case elsewhere that Conversation Analysis (CA) can be an efficient tool to discover practiced language policies (Bonacina-Pugh 2012). To put it briefly, a practiced language policy can be unravelled using CA in its broad sense, that is, including both sequential and categorisation analysis when relevant. A particular useful CA tool is 'deviant cases analysis' (Heritage 1984), which consists in studying cases that are deviant from a previously observed pattern. Deviant cases are identified by looking at participants' reactions to a particular act. If a speaker or a hearer reacts to a particular language choice or alternation act as being a problem or as needing repair, this indicates to the analyst that this instance of language choice or alternation is considered to be deviant by the interactional participants. It is by looking at interactional participants' orientation (i.e. reaction) to deviant cases that the analyst is able to deduce what a 'normative' case would be. In this sense, the study of deviant cases reveals what norms speakers orient to in interaction, that is, they indicate what the practiced language policy is. Once we know what the practiced language policy is, the analyst is then able to understand what interactional norm speakers are referring to in order to know whether a language(s) is legitimate or not and when.
To return to the specific case of the induction classroom under study, from their experience of teaching, learning and simply talking in that classroom, the participants have constructed a practiced language policy, which they use as one possible benchmark to know what language choice or alternation act is legitimate or not in a given situation. By revisiting extracts 1 to 3 above, we can infer that Miss Lo and the pupils are orienting to a practiced language policy at the local level of the classroom that enables them to legitimize multiple languages in teacher-pupil talk. Extract 1 consisted of Miss Lo and a pupil interacting in French and switching temporarily to Spanish. There, French was seen as the legitimate language whilst Spanish was seen as inappropriate. Whilst the legitimacy of French could indeed be accounted for from a critical perspective with reference to the fact that French is the language of power and prestige in the French educational system and the French society at large, it can equally be accounted for from a practical perspective with reference to a norm of language choice that could be initially formulated as follows: in teacher-led talk, French should be used as the medium of classroom interaction. This norm can then be further elaborated by looking at the commonalities between extracts 2 and 3. To recall, in both extracts, languages other than French (namely, English in extract 2 and Romanian in extract 3) were licensed by Miss Lo in order to help clarifying what word pupils were missing in French. Clearly, English and Romanian are not legitimised because Miss Lo and the pupils orient to the top-down language policies and surrounding ideologies but rather because they orient to a norm of language alternation acts according to which languages other than French become legitimate when they are licensed by the teacher. This norm is part and parcel of the multilingual 'practiced language policy' of the classroom under study and is consistently and repeatedly followed by the classroom participants. This can be confirmed by further instances of language alternation acts observed in the corpus. Consider for instance extract 4 below. Here, Miss Lo is introducing the negation system in French and wants to bring pupils' attention to the fact that, in French (and especially in written French), the negation is expressed by the two markers 'ne' and 'pas'.
To introduce the syntax of negation in French, Miss Lo asks pupils to provide a positive sentence and to transform it to its negative. Therefore, Karen suggests the French sentences 'je veux manger' (i.e. I want to eat) and 'je ne veux pas manger' (i.e. I do not want to eat). Once that French example is set, Miss Lo wants to check pupils' understanding and asks if the same negative structure is found in their other languages. This launches a series of 'translation quests' (Bonacina-Pugh 2013), where each pupil attending this lesson get asked to translate these two sentences in their respective other language(s). First, in extract 4a, Miss Lo turns to the Spanish-speaking pupils and asks them to provide the Spanish translation for these two French sentences (line 127).  Piotr: ja nie chcę jeś ć((writes on the board)) 242.
Miss Lo: ok alright (.) and I do not want to eat 241.
((children repeating the Polish sentence)) Here also, because Miss Lo licenses the use of Polish, it becomes legitimate. Miss Lo then turns to English-speaking children, the Japanese-speaking child (extracts not reported here due to space constraints) and the Peul-speaking children (see extract 4c) to elicit respectively English, Japanese and Peul.   Here again, Peul becomes legitimate because it is elicited and therefore licensed by Miss Lo. To summarise, this interactional episode further demonstrates that the classroom participants consistently refer to a language policy devised in practicea 'practiced language policy'whereby languages other than French are appropriate when they are licensed by the teacher. It also illustrates the key role that practiced language policies play in legitimisation processes since it is because the classroom participants draw on this practiced language policy that languages other than French are given a legitimate status.

Concluding remarks
In this paper, I have revisited the concept of 'language legitimacy' and showed how languages can be legitimised at different levels, including that of interactional practices. I have taken the example of multilingual language practices observed in an induction classroom in France at primary level and tried to understand how the classroom participants in that context came to see multilingual language practices as being legitimate while the surrounding top-down language policies and ideologies held in the school as well as in the wider French educational system are strictly French monolingual. Two complementary perspectives to language legitimacy have been discussed, namely a 'critical' and a 'practical' understanding to legitimisation processes. A 'critical' understanding to legitimisation implies that top-down language policies and ideologies are what legitimize classroom language practices. From this perspective, only French is seen as a legitimate language. While this could account for extract 1 where only French is seen as appropriate (whilst Spanish is not), it did not account for the fact that, in extracts 2 and 3, alternations to English and Romanian respectively were seen as acceptable. Consequently, a 'practical' account to legitimisation was discussed in the second half of this paper. According to this alternative approach, legitimisation does not come from speakers' orientation towards top-down language policies and wider ideologies but rather from their orientation towards a 'practiced language policy' (e.g. Bonacina-Pugh 2012) constructed at the local level of classroom language practices. The practiced language policy oriented to in this classroom is definitely multilingual and consists of a set of norms of language choice and alternation acts that legitimize the use of languages other than French. I focused specifically on one norm of language alternation acts that legitimised the use of languages other than French in teacher-led interaction where a French monolingual medium is used. That norm can be summarised as follows: languages other than French become legitimate when they are licensed by the teacher. Ultimately, it is hoped that, in this classroom, languages other than French will eventually gain legitimacy without having to be licensed by the teacher first, so that pupils can draw on their linguistic repertoire freely in interaction.
To summarise, in this paper, language legitimisation is apprehended as a multi-layered process according to which legitimacy is conferred at the societal and institutional levels as well as at the practical level of interaction. The suggested 'practical' understanding of language legitimacy grounded in the Ethnomethodological and Conversation Analytic traditions offers an additional layer to our understanding of legitimacy, namely that of practices. This paper thus argues that what legitimises multilingual practices in the classroom under studyand possibly in other multilingual classroomsis not only wider ideologies and top-down language policies but also (and sometimes, primarily) the 'practiced language policy' (Bonacina-Pugh 2012), that is, the norms of language choice and alternation acts constructed and negotiated in interaction by classroom participants themselves.