Engaging in and creatively reproducing translanguaging practices with peers: a longitudinal study with three-year-olds in Luxembourg

ABSTRACT The increasing societal and linguistic diversity in schools challenge traditional teaching approaches and call for pedagogies that cater to the growing number of multilingual pupils. Translanguaging pedagogies can offer multilinguals a productive learning environment that helps them leverage their resources for learning. Translanguaging studies in early childhood education are still scant, especially those that involve emergent multilinguals, focus on adult–child and peer interactions, and examine children’s agency. The present paper from multilingual Luxembourg examines the engagement of two three-year-olds in adult-led and child-led activities in two early childhood education settings as well the ways in which they creatively reproduced translingual activities and strategies in peer interactions. The data of the qualitative study stem from 128 excerpts, video-recorded over the course of a year. The findings show that the children’s language use and active engagement depended on the pedagogy and the practitioners’ language-supportive strategies. In peer interactions, they creatively reproduced routine activities, thereby transforming formulaic speech, as well as the practitioners’ strategy use. The findings can guide curriculum developers and practitioners when implementing inclusive translanguaging practices in early childhood education.


Introduction
The increasing societal and linguistic diversity in schools challenge traditional teaching approaches and call for pedagogies that cater to the growing number of multilingual pupils. Translanguaging, the strategic and coordinated deployment of the speakers' entire semiotic repertoire (García and Otheguy 2019), if embedded in a critical pedagogy, can offer multilingual children the necessary productive learning environment and support that helps them leverage their resources for learning.
Translanguaging pedagogies have begun to be implemented in early childhood education and care (ECEC) but research on them is still underexplored. Existing studies tend to focus on teachers and document their pedagogical practice while also showing how children aged three to six use features of several languages flexibly and naturally (Gort and Sembiante 2015;Kirsch 2020a;Sanders-Smith and Dávila 2019). Studies that focus on children and analyse their translingual interactions with peers and teachers are rare. Such studies are important because peers, like teachers, are mediators of language learning (Lantolf and Poehner 2008). Children have been shown to creatively reproduce their teachers' verbal and non-verbal behavior in peer interactions, which can propel language learning (Corsaro 2005;Erdemir and Brutt-Griffler 2020).
The present study investigates the adult-child as well as peer translingual interactions involving two three-year-old children in two ECEC settings in Luxembourg. This small country located at the center of Europe has three official languages, Luxembourgish, German and French, but many more are spoken on account of its diverse population. In the academic year 2016/2017, 64% of the four-to six-year-olds spoke a first language other than Luxembourgish (MENJE 2018). The children presented in this study were developing competences in their home language, Luxembourgish, and French. (They will be referred to as emergent multilinguals thereafter.) One child was enrolled in early education (the first optional year of preschool), the other one was in a day-care center. The ECEC professionals communicated mainly in Luxembourgish but gave children opportunities to use home languages other than Luxembourgish with them and between peers. These practitioners had taken part in a professional development (PD) course on multilingual pedagogies which was part of the research project 'Developing multilingual pedagogies in early childhood education' (MuLiPEC). The project aimed to analyze the influence of the PD on the practitioners' attitudes, knowledge, and practices, and the children's languaging practices (Kirsch 2020a(Kirsch , 2020b. To this effect, Mortini examined the interactions of eight emergent multilinguals in two preschools and two day-care centers over one academic year. The present paper examines two of the children's engagement in adult-led and child-led activities and the ways in which they creatively reproduced the translingual language activities and instructional support of the adults. It addresses several research gaps: it focuses on young children in ECEC, explores their translanguaging practices, and combines research in the domains of translanguaging and agency. The findings may guide ECEC curriculum developers in promoting translanguaging practices and encourage early years practitioners to implement child-centered multilingual pedagogies.

Supporting young multilingual children in their language development
Scholars working within the framework of sociocultural theories hold that children learn languages through appropriate comprehensible input and opportunities to use languages in interactive and culture-specific ways. Learning occurs when children are cognitively, socially and emotionally involved and supported by teachers (and peers) who offer a productive learning environment, monitor their own and the children's language, and act as models (Lantolf and Poehner 2008;Swain, Kinnear, and Steinman 2010). The following sections outline the roles of the adults and children with a focus on early childhood education. could hamper the creation of equitable learning spaces (Arreguín-Anderson, Salinas-Gonzalez, and Alanis 2018; Axelrod and Cole 2018;Kirsch and Seele 2020a).
Research on translanguaging in the field of ECEC is still underexplored. The majority of the studies have documented the ways in which ECEC professionals created translanguaging spaces and used their languages strategically to leverage children's semiotic repertoire for learning. Studies in bilingual settings include those by Arreguín-Anderson, Salinas-Gonzalez, and Alanis (2018) and Gort and Sembiante (2015), in monolingual countries such as France those by Mary and Young (2017) or in the multilingual environments of Hong-Kong and Luxembourg those by Sanders-Smith and Dávila (2019) and Kirsch (2020aKirsch ( , 2020b. Findings show that teachers design structured play-based learning environments and classroom activities that provide opportunities for multilingual and multimodal communication. Typical routines and language activities studied are 'show and tell' (Gort and Sembiante 2015), play situations (Arreguín-Anderson, Salinas-Gonzalez, and Alanis 2018) and interactive storybook reading and readalouds (Gort and Pontier 2013). Dialogic interactions around books propel language learning in contrast to the hierarchical initiation-response-feedback pattern (IRF) frequently observed in teacher-child interactions (Sinclair and Coulthard 1975). To help children develop language, teachers modeled language and transformed children's utterances by correcting or rephrasing them and adding richer vocabulary or a more complex syntax (Gort and Pontier 2013;Gort and Sembiante 2015). Corrective feedback and expanding utterances are so called languagemodeling strategies (Girolametto, Weitzman, and Greenberg 2004). Like interaction-promoting strategies (e.g. asking questions, repeating), they support language learning. Kirsch (2021) analyzed the interaction-promoting and language-modeling strategies, thereafter language-supportive strategies, of the practitioners in the MuLiPEC project and found that they deployed a wealth of strategies in translingual conversations.
Other research studies focused on children and showed that they use translanguaging to communicate and participate (Alamillo, Yun, and Bennett 2017;Kirsch 2017Kirsch , 2020aGort and Sembiante 2015), construct meaning (Axelrod 2017;García 2011;Gort and Pontier 2013;Gort and Sembiante 2015) and foster language learning (Alamillo, Yun, and Bennett 2017;Kirsch 2017;Sanders-Smith and Dávila 2019). Others indicate that it helps children regulate behavior (Arreguín-Anderson, Salinas-Gonzalez, and Alanis 2018), socialize (Sanders-Smith and Dávila 2019), include and exclude others (García 2011), and mark their linguistic identity (Alamillo, Yun, and Bennett 2017;Gort and Sembiante 2015). While many researchers hold that translanguaging can be transformative because children have some choice and agency in their language practices, few studies investigated the explicit connection between children's translanguaging and their (language-based) agency (Schwartz, Kirsch, and Mortini 2020;Rajendram 2019;Toth and Paulrud 2017). These studies showed children in preschools and primary schools practice their agency by translanguaging to make meaning and advance their own learning.
Overall, few translanguaging studies examine teacher-child interactions (Alamillo, Yun, and Bennett 2017), children themselves or on peer interactions. Those that do, document the children's performances in play (Arreguín-Anderson, Salinas-Gonzalez, and Alanis 2018) or literacy tasks. For example, Axelrod and Cole (2018) and Velasco and Fialais (2016) show how preschool children draw on skills and strategies acquired in one language to produce speech in a second one. The present study focuses both on translanguaging and children's active role in learning and supporting peers.

Children as active language producers
Most studies grounded in Vygotskian theory that investigated the role of peers as mediators for second language learning, analyzed the interactions between the so-called experts and novices as well as their scaffolding strategies. Studies with bilingual students in Years 1, 4 and 5 in dual language classes in the United States (Angelova, Gunawardena, and Volk 2006; Esquinca, Araujo, and de la Piedra 2014; Martin-Beltrán 2010) and with emergent multilingual learners in Year 1 in Luxembourg (Kirsch and Bes Izuel 2017) found similar results. The 'experts' would, for example, offer lexical and grammatical input, rephrase, translate, clarify, extend and transform sentences whereas the 'novices' would echo words, practice by repeating, transform expressions and combine phrases. During such peer interactions, children tended to listen attentively to each other, build on each other's phrases, negotiate meaning, and provide input and modified output. These strategies are very similar to the teachers' language-supportive strategies (Girolametto, Weitzman, and Greenberg 2004;Kirsch 2021).
According to Vygotskyan theories, children are socialized and learn when they actively appropriate, reinvent and reproduce the sociocultural activities and routines in which they participate (Vygotsky 1978). The American childhood sociologist Corsaro proposed the notions of 'peer culture' and 'interpretative reproduction ' (2005) to capture children's innovative participation in practices. He defined peer culture as 'a stable set of activities or routines, artefacts, values, and concerns' which children produce and share during peer interactions (Corsaro 2005, 18). Through this production of and participation in their own peer cultures, children have agency (232). Having studied four-year-olds in preschools in Berkeley, he showed how they creatively reproduced routinized behavior and transformed formulaic speech in self-talk and interactions with peers. The children also repeated and creatively reproduced language like the bilingual children in a Head Start program in New York observed by Axelrod (2017). She observed Spanish and English-speaking children create non-conventional words and grammatical structures when interacting with peers.
Some recent studies in ECEC indicate that children are influenced by the languaging practices of adults. Given that ECEC curricula are not as strict as those in primary school, children have more opportunities to observe, role-play and model others. Langeloo et al. (2019) indicate that some children mirrored the simplified language of their teachers, produced simple language and gave oneword answers. Alamillo, Yun, and Bennett (2017) show that teacher Christina adapted the complexity and speed of Spanish when addressing three-to five-year-old speakers of Spanish or English. The children similarly assessed the levels of comprehension of their peers and adapted their speech accordingly. A clear example of reproduction, in line with Corsaro (2005), comes from Erdemir and Brutt-Griffler (2020). The researchers studied the incidental vocabulary learning of a Turkish boy in a preschool in the United States. The teacher had exposed the class to thematic vocabulary which the monolingual children had picked up more quickly than the Turkish child. The 'experts' exposed the boy playfully to this vocabulary, thereby labeling objects and actions, demonstrating word meanings, correcting phrases, and extending the meaning of words by using them in new contexts. They made these words salient through intonation and verbal and non-verbal cues. The children thereby reproduced the teacher's language and instructional strategies. A different example stems from two -to four-year-old emergent multilingual children from Israel and Luxembourg. Schwartz, Kirsch, and Mortini (2020) show that the children were aware of their own and others' languages. They demonstrated agentic behavior by using their entire semiotic repertoire, making deliberate language choices, repeating expressions of peers and teachers, and shaping the language use in the classrooms.
In sum, this literature identified a shortage of research studies on translanguaging in ECEC. It concluded that existing studies tend to examine the teachers' role in planning a language rich environment and supporting learning, and emphasize the functions of translanguaging. Few studies focus on children, in particular on emergent multilinguals, and investigate children's agency within translingual practices. Based on these research gaps, the present paper is based on the following research questions: 1. In what ways do practitioners and two emergent multilinguals engage in typical language activities in two ECEC settings in Luxembourg? 2. In what ways do these children transform and implement the typical language activities and the adults' language-supportive strategies in peer interactions?

Methodology
The two case-studies reported in this paper stem from the longitudinal mixed-method project MuLiPEC which investigated the development of multilingual pedagogies in formal and nonformal ECEC settings in Luxembourg. While Kirsch and her colleagues observed and interviewed the practitioners in four settings over one academic year, Mortini focused on eight children in these settings. This paper foregrounds Mortini's qualitative data on two three-year-olds' engagement in translanguaging practices in one early education class and one day-care center. The data is framed by the findings from the larger project MuLiPEC.

Research context
The complex early childhood education system in Luxembourg is split into the formal sector (i.e. early education, preschool) and the non-formal one (e.g. day-care centers) (Kirsch and Seele 2020a;Kirsch et al. 2020). Children aged three can enroll in an optional first year of preschool called éducation précoce (early education) or attend a fee-paying day-care center. Where a municipality offers early education on half-days, children may attend early education in the morning and a day-care center in the afternoon. The formal and non-formal sectors differ on account of their language policies, educational practices and staffing. Until October 2017, thus before the data of this study were collected, the language-in-education-policy in the formal sector called for Luxembourgish and was firmly rooted in the national curriculum to be adopted by the teachers. In early education, teachers collaborate with caregivers. The non-formal sector employs mainly caregivers of which some may have little training in child (or language) development. As there were no formal policy guidelines on language use, the practitioners could decide on their linguistic profiles and pedagogical approaches (Kirsch and Seele 2020a). This led to a high degree of diversification of this sector. Over the years and owing to the increasing language diversity in Luxembourg, some multilingual practices could be observed in both sectors ). The move away from monolingual ideologies and practices culminated in the program of multilingual education implemented in the whole ECEC sector from Autumn 2017 (Kirsch and Seele 2020a). The need to help professionals develop methods and strategies to capitalize on language and cultural diversity led to the project MuLiPEC and its professional development (PD) course on multilingual pedagogies. Forty-six teachers and caregivers attended and seven self-selected to continue in the observational study from 2016 to 2017. The present paper focuses on four practitioners with a working experience of ten years who worked with three-year-olds, two in early education and two in a day-care center. The combination of these differing settings offers insights to national and international readers as few researchers in Luxembourg examined both contexts jointly.

Participants
The class of early education was hosted in a small building with four classrooms on the outskirts of a town in the South of Luxembourg. Eleven children of ethnic minority background and low socioeconomic status attended the class. Not one of the children spoke Luxembourgish at the beginning  of the school year in 2016. Most spoke Portuguese and Cape Verdean Creole (see Table 1). Like in the three other research settings, Mortini selected two talkative focus children eager to engage in activities (Almér 2017). Other selection criteria included the diversity of the language backgrounds and socioeconomic status. For the present paper, one child in each setting has been selected on account of their participation in each of the observed activities (see Table 2). The first focus child in early education has been called Lana (all names are pseudonyms). She was 3 years and 4 months old at the beginning of the data collection. In interviews, the practitioners described her as a talkative and outgoing child who communicated in Cape Verdean Creole at home and may also have been spoken to in French by her father. In the first month at school, Lana communicated mainly in Cape Verdean Creole and sometimes in Portuguese. She also used some words of Luxembourgish in self-talk and during free-play and mouthed recurrent expressions during circle time in Luxembourgish. Lana and her peers were taught by Ms Clara (teacher) and Ms Jane (caregiver) who had a Luxembourgish background and spoke Luxembourgish, German, French and English. Furthermore, Ms Clara had learned some Portuguese and Ms Jane understood some expressions in Portuguese.
The day-care center was located in a small town close to the center of Luxembourg and offered education and care to a maximum of 100 children who came mainly from middle-class families. A group of 21 children aged two to four participated in the study. Altogether, 11 different languages were spoken. This paper foregrounds Gaspard, who was 2 years and 5 months at the beginning of the data collection. According to the practitioners, he spoke French at home and encountered Luxembourgish in the day-care center. While Mr Ken was one of the caregivers responsible for Gaspard and his peers. Mr Ted worked in a different group but regularly collaborated with his colleague. While both spoke Luxembourgish, German, French and English, Mr Ken was also fluent in Spanish.
Previous findings of the project MuLiPEC show that the PD course helped these four practitioners design child-centered activities around books and songs in Luxembourgish, French and, at times, in the children's home languages . Ms Clara and Ms Jane used multiple languages both in conversations and during language activities and storytelling. By contrast, Mr Ken and Mr Ted organized fewer language activities and made use of fewer rituals and songs to structure the children's day and familiarize them with multiple languages Seele 2020a, 2020b). The provision of activities in multiple languages other than Luxembourgish led all practitioners to translanguage. Like the children, they deployed their entire semiotic repertoire to communicate and mediate understanding . Furthermore, they regularly translated from Luxembourgish to a home language other than Luxembourgish and vice versa, to facilitate communication, ensure comprehension, promote language learning, and value the children's home languages. Mr Ken and Mr Ted asked children more frequently for translations than Ms Clara and Ms Jane, and used translations even when children did not show any signs of misunderstanding. They also switched more often to a child's home language within a Luxembourgish conversation for purposes other than translation. In addition, it was found that the practitioners in early education used translanguaging more 'responsively' (Kirsch and Seele 2020b), meaning that Ms Clara and Ms Jane monitored their own and the children's language use and needs, and only switched languages if they felt The star indicates that the activity included only children.
this improved communication. The practitioners in the day-care center at times switched languages with no critical reflection. Finally, findings related to language-supportive strategies used in monolingual and translingual conversations showed that compared to Mr Ken and Mr Ted, Ms Clara and Ms Jane asked more frequently open questions to stimulate talk and often used more languagemodeling strategies such as corrective feedback and expansion (Kirsch 2021;Kirsch and Seele 2020b). These differences in pedagogy are likely to be influenced, among other matters, by the aims of the formal and non-formal education sectors and the practitioners' different qualifications.

Methods of data collection and analysis
Mortini spent a total of 34 days spread over 11 months in the two settings. She took fieldnotes and video-recorded interactions between adults and children, and between children during storytelling activities (e.g. reading and telling stories, sharing books); focused language activities (e.g. games, rhymes); routine activities (e.g. morning circle); art work; songs and free-play. Altogether, 128 videos of between 5-and 20-min amounting to a total of 18 h, have been analyzed (Table 2). This paper draws its data from video-recordings that feature interactions of the focal children with practitioners and peers. All videos were transcribed and relevant information on non-verbal resources such as tone of voice, mime, or gesture were added. Next, all activities were regrouped into categories as shown in Table 2. In order to analyze the practitioners' and the children's engagement in activities (first research question), the authors coded the various languages used in the adult-child interactions and examined the practitioners' and children's talk turn by turn. This micro-analysis was influenced by Seedhouse's (2005) sociocultural perspective on conversation analysis. The process of coding was both inductive and deductive. It was guided by literature on language-supportive strategies (Girolametto, Weitzman, and Greenberg 2004;Kirsch 2021) and children's scaffolding strategies (Angelova, Gunawardena, and Volk 2006; Esquinca, Araujo, and de la Piedra 2014). Codes related to strategies included asking open and closed questions, repeating, expanding utterances, giving corrective feedback, labeling and pointing. Table 3 provides some insights into the coding.
In order to analyze children's reproductions of activities and strategies (second research question) in peer interactions, the authors compared children's engagement with the practitioners (see Table  3) to their engagement with peers without the adults, thereby examining the content, the languages and the strategies. Table 4 illustrates coding across activities.
This analysis enabled the researchers to identify recurrent activities, topics or expressions and examine similarities and differences. For instance, they established if, when and in what ways peers used identical or similar formulaic expressions, sang the same or similar songs or shared books that the practitioners had read to them. Children rarely replicated expressions one-to-one, rather, they transformed them. The codes included imitating (using similar sounds and gestures), replicating (using expressions or gestures one-to-one) and interpretively reproducing (transforming expressions or activities by adding own relevancies). In Table 4, Lana's utterance 'Está frio' has been coded under 'interpretively reproduced' because she appropriated information from the classroom context (Corsaro 2018) and transformed it by translating it into her home language during a peer interaction.

Findings
The

Children's different experiences of languaging in literacy activities
As seen in section 3.2, the research team of the project MuLiPEC found that the practitioners in early education and the day-care center had established different translanguaging practices and used language-supportive strategies differently to engage children (Kirsch 2021;Kirsch and Seele 2020b). That data set, however, did not focus on children. The following section presents two excerpts of literacy activities, one of each ECEC setting, from Mortini's data set that centered on children's participation. The analysis exposes each time the translanguaging practices followed by the participants' strategies.

Moving flexibly across languages in early education
Every day after breakfast, children spent 20 min in the book corner, looking alone at books or sharing them with peers and Ms Clara or Ms Jane. The following example from a video-recording in January 2017, is representative of book sharing sessions. On this occasion, Ms Jane shared a book with Cape Verdean Creole-speaking Lana, French-speaking Asima, and Portuguese-speaking Angela, Fred and Silviano. After the children had labeled the title of the book, Ms Jane opened it. The first page presented a picture of six cats. The caregiver encouraged the children to count them collaboratively, looking at each child and nodding to confirm the correct number. Ms Jane then pointed to a small cat, showed 'small' with her fingers and said in Luxembourgish 'dës Kaz hei ass kleng' (this cat here is small). Angela pointed to the cat and repeated 'é pequeno' (is small) in Portuguese. Ms Jane pointed to a big cat, made a gesture for 'big' and uttered 'déi hei Kaz ass grouss' (This this cat here is big). In Portuguese, Silviano told Ms Jane that a cat is extinguishing a fire. Ms Jane Excerpt 1 and the preceding explanation illustrate that the practices were child-centered and meaningful, and promoted language learning. In this translingual conversation, the adults and the children used their entire semiotic repertoire to communicate their thoughts and experiences. Apart from using pictures and pointing gestures (small, big; lines 1, 6, 12, 17) and movements to visualize ideas (i.e. flying) and further understanding, they orchestrated features of different languages. While Ms Jane spoke Luxembourgish, she demonstrated that she understood some Portuguese and French (lines 10, 15), testifying to her multilingual competence. She legitimated the use of home languages by confirming children's answers in their home languages verbally and non-verbally (lines 4, 8, 10, 12), and praising them (lines 10, 15). Ms Clara similarly valued Angela's contribution in Portuguese by translating it into Luxembourgish (line 18), thereby ensuring Ms Jane understood. All actors translated, possibly to ensure comprehension, provide input, and mark their linguistic identity. For instance, Ms Jane's translated the expression 'o ovo de pàscoa', which the three-year-olds may not have known, but did not translate the word 'fish' which children had provided in two languages (lines 13, 14). The three-year old children labeled words in Portuguese, Cape Verdean Creole and Frenchand Ms Jane in Luxembourgishjuxtaposing words in different languages (lines 2, 3, 4, 7, 9, 10, 11, 13, 14). In this way, they had an opportunity to learn that objects have different names in different languages and that names in Portuguese and Cape Verdean Creole are at times the same (lines 7, 9), at times different (lines 2, 3). The analysis of language-stimulating strategies indicates that Ms Jane promoted participation in various ways: she looked at children, listened, built on their contributions, and asked closed questions (lines 1, 12) as well as clarification questions (e.g. are they flying; line 19).
The children took an active part in shaping the conversation. They followed Ms Jane's lead but also created spaces to verbalize connections to personal matters such as the firefighter cat and Angela's toy. They built on contributions from Ms Jane and peers, labeled words in their home languages, repeated words uttered by Ms Jane and peers (lines 5, 7, 9, 16), and expanded utterances (line 7). Silviano, Fred and Angela also expressed themselves in longer stretches using decontextualized language (firefighter cat; line 17). This was possible because they used their home languages.

Low-level engagement of children in the day-care center
On the day of the observation presented in Excerpt 2, Mr Ted sat on the floor with Gaspard, Louise, Jeannette and Nella, who all spoke French and some Luxembourgish. He told the following invented story: a cow meets a tired snail. The cow puts the snail on its back, moves through a tunnel, walks over a stone and arrives at a house where both see playdough. In this excerpt, he showed them a picture of a meadow, asking them what they saw. (Utterances in French are underlined.) Excerpt 2. Video-recording, January 2017. This excerpt illustrates that Mr Ted and the children communicated in Luxembourgish apart from one turn where the adult shifted to French for no apparent reason. The analysis of the strategy use shows that he engaged children in an adult-centered question-answer game rather than telling a story. He asked many closed questions (lines 1, 3, 4, 5, 7, 11, 13, 21) which resulted in both himself labeling objects (lines 4, 5, 21) and the children labeling and describing pictures in one or two-word phrases (lines 2, 6, 8, 19). While Mr Ted asked several follow-up questions (lines 9, 13, 15) based on children's answers, this did not encourage them to develop their answers and engage with him or between themselves in longer exchanges. Rather, it resulted in a lot of repetition of single words in an initiation-response cycle with some feedback. Neither the brief switch to French in line 3, nor the praise for Nella's pointing (line 21) encouraged children to talk. Mr Ted's language-supportive strategies, the low-level task and the quality of his talk did not call for high-level engagement. His simplified language (simple words, few elaborations, short sentences) and the amount of visualization through pictures, objects and gestures (lines 1, 4, 5, 9, 17, 18, 21), seemed exaggerated given children's understanding of Luxembourgish. Mr Ted taught no new words, not even when Gaspar pointed to the udder (line 19). A further missed teaching opportunity arises in the exchange about the speed of cows (lines 14, 16). It is unclear whether Gaspard and his peers were teasing Mr Ted by suggesting that cows can run fast but their comments are left unchallenged.
In the subsequent playdough activity, the children sat with the adults in the activity room. Mr Ted showed the children his playdough creation, asking what it was. Louise answered in French 'un vélo' (a bike) and Gaspard in Luxembourgish 'Auto' (a car). The French words 'vélo' and 'auto' are used in the Luxembourgish language albeit with a different pronunciation. Mr Ted responded by repeating the children's answers in Luxembourgish, phrased as questions. He then asked Gaspard for the word 'auto' in French. Gaspard uttered 'uh' seemingly confused but then answered by providing the French synonym 'voiture'. This testified to his well-developed vocabulary in his home language and his willingness to participate in this exchange. Next, Mr Ted invited Gaspard, who was producing a snake, to count the three plasticine balls the boy had made earlier. He pointed to the first ball, saying 'one'. Gaspard pointed to the same one, saying 'two', then to the third one saying 'three' and finally, pointing to the middle one, saying 'five'. Mr Ted repeated 'five' with the intonation of a question. Gaspard confirmed, saying 'yes'. Without challenging the boy, Mr Ted turned to a different child. Gaspard was able to count to 10, express himself in complete sentences in French and three-word sentences in Luxembourgish as shown in other observations in December 2016, January 2017, and February 2017. This episode testified to the practitioners' tendency to ask closed questions, ask for translations and omit corrective feedback, and to Gaspard's willingness to participate and, at times, challenge.
In sum, this first section has shown that the children experienced learning differently in these two settings owing to the language practices and teaching approaches underpinning literacy activities. The practitioners and the children in early education used several languages naturally. Like the adults, they built on each other's contributions, labeled, repeated and expanded utterances. By contrast, the children in the day-care center spoke little in adult-led conversations and appeared to answer questions in conversations that tended to be mainly in Luxembourgish. This first finding indicates that multilingual education, which all four practitioners aspired to in the MuLiPEC PD, needs to be embedded in inclusive child-centered pedagogies that build on children's cognitive, linguistic and social skills for children to use multiple languages naturally. However, children do not learn languages through engaging in practices alone. Sociocultural theories remind us that they also learn the values and norms that underpin these practices.

Children's creative reproductions in peer interactions
Both children creatively reproduced the translingual routine activities and strategies in child-led activities. For reasons of space, not each section will give examples of both children.

Transforming and creatively reproducing routine activities
Both children were also observed recreating routine activities or shared reading activities (see Gaspard in the following section). While doing so, they creatively used formulaic language and typical expressions of the practitioners which they transformed, for instance by translating them into or adding features of their home languages. This interpretative reproduction was mostly observed in early education where ritualized language activities were practiced daily. Two examples from Lana make the case. In March 2017, Lana and Asima recreated a shared reading activity and replicated the typical verbal and non-verbal behavior of the practitioners. During the initial reading of a book, Ms Jane had labeled the color gray in Luxembourgish, repeatedly used the expression 'ass kal' (is cold) and visualized feeling cold by putting her arms around the upper part of her body (see Table 3). When interacting with Asima, Lana mentioned 'gro' (gray) and ''s kal' ('s cold) while imitating Ms Jane's movement. Prior to sharing the book, the girls had been painting with a piece of ice and Lana had uttered 'está frio' (it's cold) in Cape Verdean Creole (see Table 4). While Lana replicated Ms Jane's languaging in the first situation, she interpretatively reproduced it in the second one when she translated Ms Jane's expression into her home language.
Lana and her peers also drew on the formulaic speech of the morning routines. Every morning, children sang the following song to greet each other: In the day-care center, this pattern was less frequently observed as ritualized language activities and rituals were rare. Nevertheless, in June 2017, the caregivers introduced a German rhyme which children recited while holding hands before lunch. Gaspard seemed to creatively reproduce this new routine during free play, as he held hands with a peer, a doll and a teddy bear while sitting around a table (observation, 26th June 2017).

Creatively reproducing strategies
Lana and Gaspard also creatively reproduced the practitioners' instructional strategies. Excerpt 3 stems from the day-care center. Mr Ted had just finished reading a book about a wolf dressing in the morning. Gaspard, French-speaking Jeanette and Luxembourgish-and German-speaking Francis began to retell the story with Mr Ted standing in proximity. Jeannette began the storytelling event in French. Like Mr Ted and Mr Ken, she held the book up, showed it to her peers, pointed to a picture and asked what it was. Francis participated at the beginning but then left, a behavior typically seen in interactions with Mr Ted. He may have left because he did not understand his peers' French. Jeannette continued pointing, asking closed questions and praising Gaspard who repeatedly offered a one-word answer. Gaspard did not use the French word 'loup' but the word 'Mëllef', pronounced in Luxembourgish. There is no such word but it resembles the Luxembourgish 'Mëllech' (milk) and 'Wëllef' (wolves). The French words are underlined.
Excerpt 3. Video-recording, June 2017. The children continued interacting in this way in French with Gaspard offering 'Mëllef' three further times with his 'teacher' praising him. When Gaspard reached the last page, he looked at Mr Ted and pronounced the word a final time. Mr Ted whispered 'Wollef ass richteg' (wolf is correct), offering corrective feedback. Gaspard repeated the word 'Mëllef' while looking at Jeannette who nodded to confirm. Both went over the book twice more, first with Jeannette playing the adult, then Gaspard. Gaspard repeated the one-word answer seven more times before it was Jeannette's turn to say 'Mëllef'. It remains unclear if the children did not know the Luxembourgish (or French) word for wolf, but their reaction to the corrective feedback and the playful situation seems to indicate that Gaspard had invented this word and played with it during this reproduced storytelling event. The children interpretatively reproduced the adults' frequently observed IRF pattern (Sinclair and Coulthard 1975) which they fueled with low-level interaction-promoting strategies and the lack of any language-modeling strategies. Their play was a caricature of their real book-sharing events.

Discussion
The present article examined the ways in which emergent multilinguals engaged in language activities and creatively reproduced language activities and strategies. In relation to the first research question, the findings showed that the two focal children used multiple languages in interactions with adults and peers. They orchestrated their complex linguistic resources in combination with non-verbal resources (e.g. mime, gestures, tone of voice) to communicate and learn. The children thereby adapted their languages to interlocutors and situations, which testified to their pragmatic competence (Almér 2017;Bergroth and Palviainen 2017;Schwartz, Kirsch, and Mortini 2020). While the translanguaging practices of the emergent multilinguals in Luxembourg were reminiscent of the emergent bilinguals studied by Axelrod and Cole (2018), Alamillo, Yun, and Bennett (2017) and Arreguín-Anderson, Salinas-Gonzalez, and Alanis (2018), the children in the presented paper were younger and moved between more than two languages. Furthermore, the findings indicate that their language choices in adult and peer interactions depended on the practitioners' pedagogy. The three-year-olds had learned that the use of home languages was legitimate and that it was important to switch languages to adapt to the interlocutors and situations. They translanguaged in peer interactions, thereby recreating the behavior of the practitioners who served as role models (Björk-Willén and Cromdal 2009). Lana and her peers experienced the importance of bridging home and school languages, an aspect of Ms Clara's and Ms Jane's translanguaging pedagogy (Kirsch 2020a), and created spaces to use Luxembourgish in combination with several home languages (Portuguese, Cape Verdean Creole, French). The children in the day-care center were less open to children with another home language and tended to select peers with the same home language to play with. Finally, the focus children had internalized typical interaction patterns and acted accordingly. Having experienced meaningful dialogue in her early education class, Lana and her peers built on each other's contributions in adult-child and peer interactions. By contrast, Gaspard appeared to have noticed that Mr Ted and Mr Ken used simple language with the children and rarely engaged in long conversations (Kirsch and Seele 2020b). This awareness may have led him to express himself in short phrases when communicating with Mr Ted and Mr Ken. Other studies similarly point out that children adapt the complexity of their speech to that of the teachers (Langeloo et al. 2019). For example, Lara-Alecio et al. (2009) found that the Spanish-speaking children in English-immersion programs in Texas mirrored the low complexity level of the teachers' input and produced short answers. Taken together, these findings show that the children's language and social behavior was influenced by the translanguaging practices in the ECEC settings and corroborate those of previous early childhood studies (Axelrod 2017;García 2011;Gort and Sembiante 2015;Kirsch 2020a;Mary and Young 2017). Furthermore, they provide evidence of the active roles that children play when learning languages (Lantolf and Poehner 2008;Swain, Kinnear, and Steinman 2010;van Lier 2010;Vygotsky 1978).
In relation to the second research question, the focus children interpretatively reproduced typical circle time and shared reading activities, thereby using formulaic speech and recurrent expressions in a creative way. At the same time, they transformed the activity. Playing with expressions, manipulating and modifying them, as shown with the examples of Lana, has also been identified as a typical behavior of preschool children in New York (Axelrod 2017) and in Sweden (Boyd, Huss, and Ottesjö 2017). The present study extends these findings: Gaspard's example shows that even younger children can practice a 'language-based agency' (Schwartz, Kirsch, and Mortini 2020).
The children's ability to identify their teacher's recurrent expressions and use them with peers has similarly been demonstrated by Erdemir and Brutt-Griffler (2020). Apart from using thematic vocabulary with a peer, the preschoolers in that study also 'mirrored' their teacher's instructional support. In the present study, the children likewise creatively reproduced the dominant instructional strategies as shown in the retelling of the story by Gaspard and his peers. The three-year-olds interpretatively reproduced the initiation-response-feedback pattern (Sinclair and Coulthard 1975) typical of Mr Ted as well as his extensive use of closed questions (Kirsch and Seele 2020b). The children thereby represented the most salient features of the interaction patterns: closed and low-level questions, short answers, praise and lack of feedback. This finding is reminiscent of a study with eight-to ten-year-old children in Germany. According to Eckermann and Heinzel (2016), they engaged in 'mimetic action' (266) when they mirrored the teacher's classroom behavior. They did not copy the teacher but produced 'an image' to make visible particular points which did not conform with 'the original' (267).
The children's active role in creating, recreating and shaping practices, activities and strategies, is in line with socialization theories. Expressed in Corsaro's words, the children engaged 'by creatively taking or appropriating information from the adult world to address their own peer concerns' (2018,18). In the present study, it was mostly during free-play that children interpretively reproduced routine activities thereby drawing on formulaic speech and typical adult interaction patterns. This is not surprising as play is a leading learning activity. It offers children opportunities to creatively reproduce the practices in which they engage with adults thereby trying out roles and knowledge (Corsaro 2005;Long, Volk, and Gregory 2007). The present article also shows that play provided children with opportunities to practice language-based agency. The children actively chose which language to use when communicating, making meaning, or including or possibly excluding peers. Taken together, the findings of the present article support previous findings on child agency but further nuance the description of children's languaging behavior by differentiating between children's one-to-one replications and creative reproductions of the practitioners' expressions, strategies and activities.

Conclusion
Despite the known limitations of any case-study, the present study contributes to literature in a number of ways. Firstly, it confirms that the adults used a range of interaction-promoting and language modeling strategies found in previous studies (Girolametto, Weitzman, and Greenberg 2004;Gort and Pontier 2013;Gort and Sembiante 2015;Kirsch 2021). Furthermore, it provides insights into the active ways in which twoto three-year-old emergent multilinguals engage in language activities with peers, adding to the little literature that exists on peer interaction of preschoolers and emergent multilinguals. The examples also testify to children's agentic behavior when learning languages (Almér 2017;Bergroth and Palviainen 2017;Boyd, Huss, and Ottesjö 2017;Schwartz, Kirsch, and Mortini 2020), an emergent research field. The present study is one of few that has investigated the explicit connection between child agency and translanguaging. Studies on child agency, which often come from the fields of sociology and applied linguistics, rarely focus on the ways in which multilingual children practice their language-based agency. Rooted in the field of education, the present paper applied Corsaro's concept of 'interpretative reproduction' to children's translanguaging practices in a complex multilingual ECEC setting. Future research studies could investigate children's translanguaging and language-based agency across the home and ECEC contexts and add the perspectives of parents and practitioners through interviews. Finally, though not focusing on the teachers' pedagogy, the findings add to the literature on pedagogical translanguaging in ECEC settings. They show that translanguaging needs to be embedded in an inclusive child-cantered pedagogy (García, Johnson, and Seltzer 2017), as was the case in early education, for children to actively participate and use multiple languages to communicate with peers.
The study has some implications for ECEC practitioners, curricula (curriculum) developers, and teacher educators, nationally and internationally. Firstly, the findings of this study that examined educational practices with three-year-olds in two different sectors in Luxembourg, suggest that the caregivers in the non-formal sector may need additional coaching in language-supportive strategies. Since caregivers and teachers may educate the same children, collaborative professional development may benefit the children's experiences in ECEC. Secondly, the children's ability to appropriate, transform and implement strategies, activities and practices and, in this process, reshaping the learning environment, calls for solid child-centered and inclusive pedagogies. Practitioners need to be (or become) aware of their language ideologies which inform their language practices, design meaningful and interactive activities, use their languages flexibly and responsibly, and deploy effective language-supportive strategies. While curricula can guide ECEC practitioners, initial teacher education and professional development courses can deepen the practitioners' understanding of learning multiple languages and help them implement child-centered translanguaging pedagogies that leverage children's entire semiotic repertoires for learning.