Exploring family language policy in action: Child agency and the lived experiences of multilingual Ethiopian and Eritrean families in Sweden

Existing family language policy (FLP) scholarship has been criticised for insufficiently addressing children’s voices and perspectives on their multilingual experiences, as well as lacking representation and heterogeneity in terms of studies involving multilingual families from diverse family types, languages, and contexts outside the experiences of Western middle-class bilingual families. Against this backdrop, this paper examines the multilingual familial experiences of three Ethiopian and Eritrean migrant families in Sweden by paying particular attention to children’s agency and caregiver-children dynamics in FLP making. The study draws on multimodal biographic data obtained from children and parents through language portrait methods of body and space mapping activities, post-mapping narration, and semi-structured interviews. The data are analysed in light of Smith-Christmas’s (2020) framework, which views child agency in FLP at the intersection of compliance regime, linguistic competence, linguistic norms, and power dynamics. The findings reveal that the process of FLP making is characterised as a process that is (1) filled with language choice dilemmas triggered by competing linguistic demands, (2) in part shaped by the family constellation via power dynamics between family members, and (3) mediated by family members’ varied linguistic proficiencies in majority and minority languages. Additionally, children’s agency about which language they choose to use impacts the language practices of the home, as they tend to establish their own linguistic norms within the home by overrunning the negotiated language policy set by caregivers.

Exploring family language policy in action © Degu and CMDR.2021 DIKE CALLED OUT from the bathroom, where he had been sent to brush his teeth before bed."Dike, I mechago?"Ifemelu asked."Please don't speak Igbo to him," Aunty Uju said."Two languages will confuse him.""What are you talking about, Aunty?We spoke two languages growing up." "This is America.It is different."Ifemelu held her tongue.Americanah - Adichie (2013)

InTRoDuCTIon
In the novel Americanah, Chimamanda Ngozi Adichie beautifully captures a long-standing quandary experienced by migrant families -namely, which language caregivers should use to speak to their children in the countries to which they moved.These may be contexts characterised by a monolingual norm that devalues the concomitant acquisition of different languages, as Aunty points out in the above extract.The decisions caregivers make about which language should be transmitted to their children have been studied at length in the extensive literature on bilingual parenting (see e.g.(King andFogle 2006, Lanza 1998).However, it is only recently that family language policy (FLP) has emerged as a more comprehensive framework that brings together insights from a variety of interrelated research strands such as child language acquisition, language socialisation, and language policy and planning (Curdt-Christiansen 2018, Lomeu Gomes 2018, King, Fogle, and Logan-Terry 2008, King and Fogle 2013, Smith-Christmas 2020).While I delve into FLP in more detail in the next section, suffice to say for now that FLP seeks to capture the interrelation of the following elements: (1) family members' language ideologies, (2) language use within the home, and (3) the observable efforts put into encouraging desirable language behaviour among family members (King, Fogle, and Logan-Terry 2008, Curdt-Christiansen and Lanza 2018, Curdt-Christiansen 2018).
As Lomeu Gomes (2018) points out in a critical overview, FLP research has concentrated primarily on bilingual parenting and language socialisation processes among Western middle-class bilingual couples, their implicit and explicit home language policies, and the underlying ideologies that inform their policy decisions and practices (King and Fogle 2013, Smith-Christmas 2017, King 2016).Moreover, in such research, adult caregivers have been taken as the main source of FLP research data, whereas children's views and perspectives on their multilingual experiences have remained largely overlooked (Palviainen 2020).
To partly redress the focus on Western middle-class families and the privilege given to parents in previous research, the present study focuses on three Ethiopian and Eritrean migrant families in Sweden and discusses children's perspectives in relation to what their caregivers say.More specifically, drawing on data elicited through multimodal research methods, the article illustrates the multilingually mediated interactional experiences of children and their parents in the process of FLP-making by paying particular attention to children's agency and the overall caregiver-children dynamics.
The article is organised as follows: The next section of the paper gives a brief overview of FLP scholarship and seeks to problematise some of the constraints in the literature about issues of multilingual families with immigrant backgrounds, including children's viewpoints.This is followed by a presentation of the

MulTIlInguAl fAMIlY pRoToTYpES AnD ChIlDREn'S REpRESEnTATIon In flp SCholARShIp
In classifying existing FLP studies, Smith-Christmas (2017) outlines three multilingual family prototype contexts: (1) the one-person one-language (OPOL) prototype, (2) the migrant community, and (3) the autochthonous community.OPOL is a typical strategy followed by Western middle-class bilingual couples (Wilson 2020), in which one parent is from the majority language-speaking community of the host society, while the other is a minority language-speaking parent, often from another Western country (Smith-Christmas 2017).A distinctive factor in the OPOL family configuration is that a child's linguistic experience is limited to what happens in the 'nuclear' family.Here, the principal task of maintaining the minority language rests on the minority language-speaking parent, while extended family members play a minimal role in relation to the child's exposure to the minority language (Smith-Christmas 2017).By contrast, in migrant and autochthonous minority language communities, precisely by virtue of being a 'community', children typically have more exposure to minority language speaking interlocutors.However, as Schwartz (2008) notes, the extent of minority language exposure outside the family space depends on the size of the speech community that speaks the language.
Two main critiques have been levelled against existing FLP scholarship.The first concerns the lack of diversity in terms of family types, languages, and contexts under investigation (King 2016).In this regard, most previous studies have been carried out within the context of Western middle-class multilingual families, which focused on children acquiring more than one European language (King 2016, Smith-Christmas 2017, Curdt-Christiansen 2018).The second point of critique pertains to the fact that the empirical data described in these studies consists of caregivers' accounts, whereas children's perspectives on their experiences have been insufficiently addressed (Schwartz 2020) or only indirectly represented (Smith-Christmas 2017, Palviainen 2020).
Since children are the main targets in the process of FLP making, incorporating their views and perspectives would give a more holistic picture and enhance our understanding of children's languagebased agency (Schwartz 2020, Schwartz andVerschik 2013).Children are not passive recipients of the heritage language (Palviainen 2020); rather, they are active participants in language socialisation (Revis 2019), something that has been described as a reciprocal socialisation process (Smith-Christmas 2020) or bidirectional language socialisation (Schwartz 2020).
Moreover, Lomeu Gomes (2018) emphasises that FLP has been largely framed by what he calls 'Westerncentric', 'canonic epistemologies'.The most prominent example of such epistemological trends is the pervasiveness of Spolsky's (Spolsky 2004(Spolsky , 2009) ) generic conceptual framework in Exploring family language policy in action © Degu and CMDR.2021 FLP literature.According to Spolsky, the three components of a speech community's language policy are its language practice-the habitual pattern of selecting the varieties that make up its linguistic repertoire; its language beliefs and ideology-the beliefs about language and language use; and any specific efforts to modify or influence that practice by any kind of language intervention, planning or management (Spolsky 2004: 5).
Part of the critique of Spolsky's framework is that a model that revolves around decision making and management has inadvertently made caregivers the 'default choice' for the collection of data that would illustrate the practices, management strategies, and underlying language ideologies that inform the decision-making process.Despite being the ultimate target of FLP, children's age could be a factor in excluding them as a source of direct data.Yet, little effort has been made to obtain direct data even from older children, not least because-as Smith-Christmas (2017) points out-innovative research methods are needed for the elicitation of children's views on their multiplelanguage experience.
Against the backdrop outlined above, there is a growing call for shifting the focus of FLP research from adultoriented empirical data to studies that represent children's perspectives and views on their multilingual experiences and home language practices as equal co-participants in the process of FLP making (Smith-Christmas 2017, Schwartz 2020, Smith-Christmas 2020, Palviainen 2020).A similar shift is sought from Western middle-class bilingual families to increasing emphasis on migrant communities from the Global South to diversify the family types, languages, and contexts under investigation (King 2016, Smith-Christmas 2017).

CoMplIAnCE REgIME, lInguISTIC CoMpETEnCE, lInguISTIC noRMS, AnD powER DYnAMICS
Given the centrality of the child's role as a subject and an object in the process of FLP making, the data analysis of the present paper draws on Smith-Christmas's (2020) framework of child agency in FLP (figure 1).Smith-Christmas (2020) captures child agency in FLP as located at the intersections of the four following dimensions: the compliance regime, linguistic competence, linguistic norms, and power dynamics.Moreover, these four dimensions of child agency can be understood through the lens of negotiation and change.Accordingly, I use this model to not only analyse the agentive role of children in the process of establishing FLP, but to also make sense of the overall caregiver-children dynamics in the target families' pursuit of FLP development.
A certain level of linguistic competence is a requisite for adherence to the compliance regime because lack of proficiency in a target language makes speakers shy away from choosing and using the given language as the compliant code.Conversely, strict adherence to the compliant language could eventually improve proficiency in the target language.Moreover, as Smith-Christmas (2020: 221) explains, 'both compliance regime and linguistic competence contribute to the formation of linguistic norms within the family'.The relationship between DEGU © Degu and CMDR.2021 competence and compliance, however, is complicated by the power dynamics between children and caregivers and the ensuing negotiation and change.Put simply, the home domain is a site of constant negotiations between family members about which language should be spoken.Such negotiations about language choices are partly mediated by individuals' linguistic competence.Language choice, in turn, illustrates whether compliance to a certain code is observed by members and leads over time to the formation of family linguistic norms.The data analysis below draws on these four concepts as an analytical framework to make sense of the multilingual interactional experiences of the three participating families, paying particular attention to both children's agency and caregiver-children dynamics in the process of establishing FLP.
In analysing children's agency in the process of establishing FLP, this study draws on the conceptualisation of agency as 'considering individuals as actors with the ability to make sense of the environment, initiate change, and make choices' (Kuczynski 2002: 9).By considering child agency in relation to the compliance regime, it is possible to illustrate how 'a child not speaking Language X' can be an act of resistance against his/her caregiver's command to 'speak language X' (Smith-Christmas 2020: 222).

ConTExT of ThE STuDY
Sweden actively supports de jure multilingualism and cultural diversity (Milani and Jonsson (2018).As a result of this commitment to multilingualism on the part of the state, over 150 different languages are taught as 'mother tongues' alongside five national minority languages (Finnish, Meänkieli, Romany Chib, Sami, and Yiddish; (Kheirkhah 2016).Moreover, a recent report from Statistics Sweden (2020a) illustrates that over half a million children below the age of 18 have a foreign background, which means that they were either born in another country or were born in Sweden from one or two foreign-born parents.This figure captures the wider sociocultural and linguistic diversity in Swedish society and indicates that there are many multilingual families experiencing the interface of majority and minority languages, a nexus point in which these families are 'open to influences and interests from other broader social forces and institutions' (Canagarajah 2008: 171).Put differently, multilingual families in Sweden and elsewhere are sites of 'a dialogic and ever-evolving co-construction, which is in turn shaped by the dynamic relationship of the family to the wider community' (Smith-Christmas 2017: 21).This means that families with minority language backgrounds in Sweden are entangled in simultaneous processes that involve the competing interests of maintaining a heritage language and the pursuit of meeting the sociolinguistic and socioeconomic demands of the country to which they moved (Canagarajah 2008).
To the description above, it could be added that certain migrant communities -by virtue of being small in terms of population size -tend to have reduced access to resource allocation by public sectors, limited social contact, and community networks that support heritage language maintenance and transmission to second-and thirdgeneration children.For instance, if the number of pupils speaking the same mother tongue is less than five in a school in Sweden, the school is not obliged by law to arrange mother tongue instruction for them (Utbildningsdepartment SFS 2010: 800).Likewise, these groups are less likely to be represented in multilingualism research, which tends to privilege larger migrant groups.Hence, Amharic-speaking families from Ethiopia and Eritrea, which are the focus of this study, can be taken as a 'minority among minorities' in terms of community size compared to Arabic-speaking, Finnishspeaking, Polish-speaking, Somalispeaking, Persian-speaking, and other migrant groups with larger community sizes in Sweden (Statistic Sweden 2020b).

METhoDologY
The study upon which this article is based employs a multimodal data collection method, which means that the data collection process was primarily carried out using visual methods of bodymapping (language portrait) and spacemapping activities that generated visual and textual data.These visual methods were supplemented by data obtained through post-mapping verbal narration and interview techniques.The language portrait method, as Busch (2018: 4) explains, 'is understood as a means of gaining insight into everyday linguistic practices of bodily and emotional language experience, or of ideologically informed ideas about, of attitudes to, and of stance taking towards particular languages or modes of speaking'.

DEGU
© Degu and CMDR.2021 Consequently, the language portrait is increasingly becoming a very useful method in multilingualism research, particularly when the objective of the research is to obtain the 'experiential perspectives' and 'subject positioning' of multilingual speakers.
Generating a language portrait, as a pictorial-linguistic form of text production (Busch 2018), is a manual activity (Purkarthofer 2017) that involves colouring a blank silhouette of a human body using multi-coloured pens to graphically visualise one's linguistic repertoire, language history, language attitude, language practices, linguistic experiences, linguistic disposition, and so on (Busch 2010).Similarly, the bodymapping method has been described as 'an embodied way of knowing and storying the self ' (De Jager et al. (2016: 52).
As a creative method of visual representation and meaning-making process, the language portrait has brought a shift from traditional qualitative verbal methods, such as individual or focus group interviews, to a mode that allows participants to reflect and visually depict their embodied experiences and multilingual repertoire (De Jager et al. 2016, Busch 2018, Prasad 2014).The shift from a verbal mode to a pictorial mode helps express certain aspects of individual experiences that cannot easily be traced through interviews and other traditional qualitative methods (De Jager et al. 2016).Moreover, as Busch (2010: 286) explains, 'The switch in mode of representation from word to image helps to deconstruct the internalised categories, to reflect upon embodied practices and to generate narratives that are less bound to genre expectations'.Another aspect of this visual method is that it takes more time, such that participants tend to 'linger longer and reflecting more deeply' on the subject than they do during verbal response (De Jager et al. 2016).
Visual methods allow participants to decide how they represent their linguistic repertoire and multilingual experiences through visual portraits and post-mapping narratives (Busch 2018, De Jager et al. 2016).Such a heightened role of the participants reduces the power imbalance between them and the researcher (De Jager et al. (2016: 20).Storying the self freely through image, caption, and oral narration allows 'participants to have greater influence on the data created and the initial interpretation of the data via their testimonies' (De Jager et al. 2016: 20).Furthermore, as there are 'associative, metaphorical and symbolic elements' that are embedded in participants' visual representation of their multilingual experiences, combining the visual with verbal narration (which offers other dimensions that cannot be captured visually, including body languages) allows to obtain thicker description of the multilingual selves (Salo and Dufva 2018: 442).Most crucially, visual methodologies are a better choice than other qualitative techniques because of their suitability for children and adolescents.Drawing and colouring are 'less anxiety-provoking to children, who are likely to have been exposed to similar tasks at home or at school' (De Jager et al. 2016: 25).It is also a method that functions well in a situation in which there is limited literacy among participants or whenever a linguistic barrier exists between a researcher and participants; this is because a mapping task depends less on verbal articulation compared to interviews or focus groups (De Jager et al. 2016).Exploring family language policy in action © Degu and CMDR.2021

participants
The study presented in this article is about three multilingual immigrant families who moved to Sweden from Ethiopia and Eritrea and are speakers of Amharic.The researcher also speaks Amharic, a language spoken by the majority of Ethiopians and a considerable number of Eritreans.Having a common language with the research participants is instrumental in exploring the research questions in depth without a linguistic barrier.The target families were recruited via the researcher's social network and comprised 11 participants, including children.As revealed through the body and space portraits, 12 different languages were mapped as part of the multilingual repertoire of these three families.These languages are Amharic, Tigrigna, Swedish, English, Guragigna, French, Spanish, Geéz, Finnish, Kiswahili, Hebrew, and Arabic.As I illustrate in more detail below, these languages are closely associated with the spatiotemporal life trajectories of each family.The following paragraphs provide a brief overview of participants' profiles in terms of family structure, family size, and linguistic profile.
Family 1 is a single-parent-headed family of three, including a daughter (10), a son (15), and a mother.The mother immigrated to Sweden 20 years ago from Ethiopia.Her mother tongue is Amharic, and she claims to speak Swedish fluently and to have rudimentary proficiency in English.
Family 2 is a coupled family of three, including their five-year-old daughter.Both parents were born and raised (until adolescence) in Ethiopia from Eritrean parents.They speak fluent Amharic alongside their heritage language, Tigrigna.The father, who moved to Sweden seven years ago, had lived in Sudan and Israel, where he learnt some Arabic and Hebrew.The mother immigrated to Sweden 12 years ago.
Family 3 is a coupled family of five, including their three daughters aged 8, 10, and 13 years.The mother immigrated to Sweden 25 years ago from Ethiopia.The father, who is also originally from Ethiopia, moved to Sweden 15 years ago from Finland, where he had attended university.Before moving to Finland to study, he had worked in Tanzania for six years, where he had learnt Kiswahili.Both parents speak Amharic as their first language.The mother also knows Guragigna as a heritage language through her parents.The father has limited Tigrigna, which he acquired through work-related relocation.
It should be noted that the nearly all the children in the study are girls.I am aware that gender may play a key role in socialisation processes.Therefore, the inclusion of boys could have brought additional insights to children's perspectives regarding their multilingual interactional experiences at home and their role in the process of FLP-making.However, gender (femininities and masculinities) did not emerge as a relevant category in the data under investigation herein.

Data collection
Data were collected mainly during six different sessions -I visited each participating family twice (excluding the recruitment visit).In the first session, both parents and children participated in the body-mapping task and postmapping narration.Drawing on recommendations given by Busch (2018) and adapting her techniques (Busch 2010(Busch , 2012)), I provided A4-sized blank body silhouettes for the participants and asked them to visually describe their linguistic repertoire, multilingual DEGU © Degu and CMDR.2021 experiences at home and outside, and any language-related views and perspectives they had towards language(s).I suggested that they should think about language holistically and try to incorporate every instance of language-related experience and views in their portrait, regardless of how proficient they thought they were in each language.No definitive or specific direction was given as to how participants should represent their linguistic repertoire in the self-portrait task.Rather, they were instructed to map and colour as they wished.Participants were also told that there was no right or wrong way of completing the bodymapping tasks.The language portrait activity was followed by a post-mapping narration task where participants narrated their body portraits orally.
In the second session, the parents were interviewed, and the children did the space-mapping task.Interview questions were posed in such a way as to prompt issues that were not covered through the language portrait method, as well as issues triggered by the visualnarrative data obtained during the first session.Family multilingualism and its lived experiences are situated in shared living spaces.Therefore, capturing spatially related multilingual experiences is equally relevant to understanding the embodied linguistic experiences illustrated through the body language portrait method.Thus, children were asked to draw, map, and colour their apartment using multi-coloured pens to present their multilingual experiences in the home and family language practices.A similar post-mapping narration procedure was followed.All the postmapping narrations and interviews were audio taped.During the post-mapping oral narration, the researcher asked the children to talk in the language they wished: Swedish and Amharic were used.
In terms of research ethics, I followed the guidelines recommended by the Swedish Research Council (Hermerén 2011).Adult participants gave their written consent for themselves and their children.Children were also informed about their participation and gave their assent (Ericsson and Boyd 2017) for participation in the study.Children had the right to withdraw from participation.This right was stated in the consent form and was communicated verbally by parents to their children.Moreover, during data collection meetings, I reiterated that they were free to opt out from participating at any time.
Now I turn into the data presentation and analysis section where the multilingual interactional experiences of the participating families, particularly in relation to children's agency and caregiver-children dynamics in FLP-making, are analysed, drawing on Smith-Christmas's (2020) model of child agency in FLP.

Dilemmas and power dynamics: The case of either-or flp
The experience of multilingualism by families with migrant backgrounds is not always a straightforward life trajectory lived with a positive bilingual advantage narrative (Hua and Wei 2016).It is a complex, chaotic, and continually evolving endeavour filled with language choice dilemmas triggered by multiple yet simultaneous linguistic demands, as well as competing language ideologies held by individual members of a family.What is intricate about the simultaneous linguistic demands of multilingual families with migrant backgrounds is that learning the majority language is often challenging, but at the same time, it is crucial for socioeconomic integration for first-generation migrants; by contrast, learning and maintaining the heritage language in parallel to the majority language is a difficult task for secondgeneration children to accomplish (Hua and Wei 2016).
An example of language dilemmas can be taken from family 2, in which the parents have Amharic and Tigrigna as their first languages.They have a sort of an ongoing disputed home language policy that can be construed as an 'either/or' FLP.With this term, I am referring to parents' stance about what kind of home language policy each finds in relation to their daughter's heritage language learning, as well as their majority language learning in the family space.The following dialogue, which was recorded during an interview with family 2, illustrates the either/or FLP and the dilemmas experienced by this family.(All excerpts are presented first in the original language [Amharic], followed by a translated version.)Family 2 members are given the pseudonyms of Robel (father), Melat (mother), and Liyu (daughter).Then, Robel suggested the benefits of communicating in Swedish at home by emphasising that the family space is the only language ecology that remains at their disposal to improve their muchneeded Swedish skills through familial conversational routines.The parents' lack of linguistic competence in Swedish, and hence, lack of choice, puts them in a home language choice dilemma, with regards to two competing linguistic demands in the FLP-making process.On the one hand, the family/home domain is viewed as a social space where the parents can improve their skills in the majority DEGU © Degu and CMDR.2021 language (extract 3).On the other hand, the realisation of the same social space is viewed as a language ecology in which they can transmit the heritage language to their daughter (extract 2).Drawing on the above extracts, one could deduce that there are power dynamics and negotiations between the parents in the process of establishing a linguistic norm informed by a choice of a particular language as the compliant code of the family (Smith-Christmas 2020).
During the interview, the father insisted on the importance of having consistent language use at home to avoid confusing their daughter with three or more different languages.Such a fear of multilingual confusion was also expressed by the mother, who said the following: Extract 4 Family 2 (Melat): እንደገና ደግሞ ልጃችን ቋንቋ በዝቶባት ልታወጣዉ አልቻለችም፡፡ አማርኛ አለ ትግሬኛ አለ እንግሊዘኛ አለ፡፡ Our daughter is not speaking well in any of these languages [referring to Amharic, Tigrigna, Swedish, and English] because it is too much for her to manage.
As Curdt-Christiansen and Huang (2020: 175) put it, 'the dilemma to raise children bilingually or only in societal language is never a fading issue'.Such a quandary often depends on an empirically unfounded claim about children being confused by exposure to multiple languages, which is then believed to be the cause of speech delay (Piller and Gerber 2018).Such beliefs are at work in the extracts above.They are compounded by the push and pulls created by the simultaneous attempt on the part of the parents to accomplish the following: (1) enhance their own proficiency in the majority language, Swedish; (2) make sure their daughter also acquires a suitable level of proficiency in Swedish; and (3) facilitate effective heritage language transmission to their child.

family constellation, compliance regime, and perceived linguistic competence in flp making
Whenever there is a change in family constellation triggered by a divorce, members living apart, or the arrival of a new family member (sibling or other), the family language ecology changes and affects the power dynamics in the process of FLP making (Caldas 2012, Palviainen 2020).To date, few studies have explored the effect of changes in family structure on FLP (Palviainen 2020).In the case of one of the families under investigation here, a theme that emerged through a closer reading of the parental interviews is that a single-parent family (family 1 in this case) seems to have and run a uniform and consistent FLP and practice.This could be ascribed to one parental voice that does not face competing voices and perspectives regarding the what and how of FLP from another caregiver.The existence of a consistent family language ideology towards heritage language maintenance and home language choice can be seen in extracts 5-7.Family 1 members are given the pseudonyms of Eden (daughter), Ruth (mother), and Eyasu (son).We, my children and I, always speak in Amharic at home.We use Swedish when we do their homework together.Except at homework time, we usually speak in Amharic.
The interview data from the mother, Ruth, and the space-mapping testimonies given by her daughter, Eden, showed a strong and consistent adherence to using the heritage language (Amharic) as a compliant code in familial communication encounters.In light of Smith-Christmas (2020) framework, it could be argued that speaking Amharic has become the linguistic norm of the family.Besides what is shown in the extracts above, the researcher also witnessed family members' use of Amharic as a compliant code during the data collection visits.
By contrast, family 2 (extracts 2 and 3) and family 3 (extracts 8-14) have varied views, perspectives, and preferences between parents regarding their home language choice and use.The excerpts below from interviews and post-mapping narration testify that there are very different views about daily home language choices and uses in parentchild and parent-parent interactions in the same family.Members of family 3 are called Tigist (mother), Henok (father), Feven (oldest daughter), Blen (middle daughter), and Hana (youngest daughter).

Extract 8
Family We do not have a consistent home language policy.We use to say Amharic, but we do not comply with the rule.We often use the language that is comfortable for us a n d suitable for the situation.
That the mother prefers Swedish while the father favours Amharic is confirmed by Feven, their oldest daughter.
A possible explanation for the discrepancies in family members' language choices may be given in light of Canagarajah (2008: 171) observation that 'members of the same family might also come with different orientations to LM [language maintenance]'.In the case of family 3, the difference in parents' language preferences could be attributed to the higher linguistic competence ('The Swedish comes first to my mouth'; extract 9) that Tigist claims to have in Swedish rather than Amharic after two and a half decades of residence in Sweden.Thus, whether the disparity is explained via linguistic competence that overrides individual home language choice or differences in individual language ideology, the two families displayed ambivalent (family 2) and varied (family 3) FLPs.Such variation can be attributed to the twoparent family structure, which can easily host varied family language ideologies, practices, and language management strategies.Hence, a coherent FLP was not established in family 2 (extract 1) and family 3 (extract 12) through joint decision making between parents.
Again, looking back to Henok's description of the situation regarding their home language policy and habitual home language use, Amharic appears to be the negotiated compliant language of family 3 ('we use to say Amharic').That being said, speaking Amharic is hardly adhered to or practised by family members (extract 12).This claim is substantiated by the following extracts taken from parents' interviews in reference to children's home language use: Extract 14 Family 3 (Henok): What emerges in this extract is not only children's use of the non-compliant language (Swedish) between themselves and with their parents as a norm, but also their strategically selective choice of Amharic in certain interactional contexts.The father's narration also gives an account of children's agentive role in determining linguistic norms in the family space, and by extension affecting the overall home language practices and FLP-making process.
However, because of the absence of parental power dynamics, a single parent-headed family in this study was characterised by more uniform FLP and practices.

perceived agentive role of the child in majority language socialisation
The child in family 2 (Liyu), who is five years old, is perceived by her parents as a socialisation agent to the majority language.This could eventually bring a shift in traditional family power structure as the child's proficiency in the majority language (Swedish) increases (see also Revis (2019).During the interview, the parents portrayed themselves as novice learners of Swedish, and they predicted that their daughter would socialise them into the majority language.When I speak to our daughter in Swedish, her mother always complains and tries to stop me, and I do the same when her mother speaks in Swedish with our daughter.Because it is she -our daughter -who can correct our Swedish -not us who can correct her Swedish.Schwartz (2020) describes as reciprocal learning or bidirectional language socialisation, which is an outcome of the linguistic competence asymmetry (Smith-Christmas 2020) often existing between migrant parents and their children.As Hua and Wei (2016: 656) also point out, 'the first-generation migrants find learning the languages of the new resident country is the most important and often challenging task, whilst their local-born children face the challenge of maintaining the home/heritage language'.While we can only speculate about the role Liyu will have in influencing family 2's FLP, it is interesting to see how the parents already foresee the epistemic authority or agency that their daughter will have correcting their Swedish, rather than the other way around.

These extracts are examples of what
Against this backdrop, Robel reported how they (parents) used to police each other to ensure compliance with what seems to have been a no-Swedish norm during parent-child interaction (extract 16).The no-Swedish norm or policy (extract 16) at home in parent-child interaction is not geared to opening up more space to accommodate heritage language use and practice; rather, it is to avoid affecting Liyu's Swedish with what Melat calls 'fel svenska' (the wrong kind of Swedish).As can be seen in the extract above, both parents are convinced of their daughter's agentive role in socialising them into the majority language.

Children's metalinguistic awareness of their multilingualism
One of the initial premises that informed the present study was to research FLP through children's direct engagement and representation as a way of including their views and perspectives on their multilingualism.Such an approach allows to us avoid relying entirely on data collected from caregivers, which was the case in previous FLP scholarship (Curdt-Christiansen 2018, King 2016).
All participating children, except the daughter in family 2, who is five years old, described themselves as functionally multilingual; they rated their levels of proficiency, feelings, and challenges regarding each language they could speak, as well as where, when, and with whom they used them.Referring back to the conceptualisation of agency as described in Kuczynski (2002: 9), making a choice to use a certain language over the other is an act of agency.Hence, children's metalinguistic awareness allows us to see which language choices they say they make and what mediates such choices.Some instances of children's words taken during body-and spacemapping narrations are given below.The second daughter in family 3, Blen -who did the oral narration in Swedish and Amharic -described her proficiency in Amharic as shown in extract 17.
By contrast, the daughter in family 1, Eden -who did the post-mapping narration in Amharic by her choiceclaimed to be proficient in Amharic and said that she conversed with her grandmother in Amharic via telephone.She also reported using Amharic more often than Swedish in daily communication encounters with her mother and her brother.The Amharic alphabet is more difficult than the Swedish one.Writing in Swedish is much easier than writing in Amharic.
Looking at the children's metalinguistic commentaries in light of the reported home language choices, the choices they made to speak Swedish instead of Amharic, particularly in the case of children in family 3 (see extracts 14 and 21) or vice versa in the case of family 1 (extracts 5 and 6), are instances illustrating the agentive role of children in determining the compliance regime and linguistic norm of their respective families, and by extension, influencing the entire process of FLP-making.
Concerning the writing skill challenges that the children described in their metalinguistic commentaries, Amharic language is considered an easier language to speak compared to English but more difficult to write compared to both English and Swedish (extracts 19 and 20).The writing-related challenge could be attributed to the fact that Amharic has quite a different and complex orthography compared to Swedish and English.The alphabetic and writing system used in Amharic and Tigrigna, which is known as 'Geéz script' or 'Ethiopic script', has over 276 different letters, posing a huge challenge to children born, raised, and schooled outside Ethiopia.Moreover, the 40-60 minutes of weekly mother tongue instruction offered to students with a mother tongue other than Swedish, which is mostly arranged outside the normal school time, is not enough or 'too marginal' (Ganuza and Hedman 2019) compared to the many hours of weekly lessons given in Swedish and English subjects.

Represented family multilingualism in a monolingual-dominated social space
In the space-mapping portraits, the children provided a detailed account of language-mediated interactional patterns between family members in terms of who speaks which language to whom and how often.In the spacemapping portraits, the children also gave detailed descriptions of their own DEGU © Degu and CMDR.2021 monolingual and multilingual language practices.Accordingly, the spacemapping data and the follow-up oral narration revealed a rich description of language use by family members in various common spaces and private rooms, particularly the language use pattern in parent-child interaction between children, language use at the individual level, and language choices in the presence of guests and visiting friends.Figure 2 shows one instance from the children's space-mapping portraits, and an excerpt from the post-mapping narration is presented below the figure.In the basement, I coloured it in Swedish and Amharic.Television is in Swedish.When we have guests, we speak either in Swedish or in Amharic.In the bathroom, I use Swedish because I think or contemplate in Swedish.In the kitchen, it is in Amharic and Swedish.When I talk to my dad, we talk in Swedish, but we use Amharic in some instances, like 'give me a glass' or something.When we dine, we talk in Swedish with little Amharic.In the living room, it is Amharic and Swedish, but we [children] do not often sit in the living room.In Hana's room, we always talk in Swedish, but we use Amharic when we pray.In Blen's room, we use English and Swedish.Blen used to sing in English.In mama and papa's room, it is Swedish and Amharic.Mama often talks in Swedish, while Papa talks mostly in Amharic.However, most often they [parents] talk in Swedish.In my room, I use Amharic, Swedish, Spanish, and English.I am learning Spanish, and I do homework in Spanish.
In the above space-mapping portrait, multiple languages are represented through the colours of national flags all over shared and private spaces.
The post-mapping verbal narration elaborates and explains.The spacemapping and the verbal narration clearly indicated that Swedish is a prominent language in a monolingual-dominated multilingual space, where the minority language (Amharic) has a less functional communicative role except as a language of prayer, and it is seldom used in childparent interaction and in child-guest interactions.The use of Amharic also seems less frequent even between parents, even though it is their first language.As narrated by Feven (extract 21) and her two younger sisters, the children seem to have established a monolingual linguistic norm characterised by the sole use of Swedish between themselves and largely with their parents.
Moreover, Feven's space mapping portrait offers a panoramic view of the home language practices in the given family, which, in some respect, is telling of the overall linguistic context in which the process establishing FLP operates.Parallel to this, the children's dominant use of Swedish over Amharic -even though Amharic is explicitly negotiated as a desired compliant language of family 3 (extract 12) and children can explain themselves in Amharic (extract 14) -highlights their agentive role in shaping the process of establishing FLP.Put differently, the above space mapping narration appears to attest that Swedish is the dominant linguistic norm of the given family.
Curdt-Christiansen and Huang (2020) report a similar phenomenon about children's roles in establishing not only a multilingual space where both the minority and majority languages co-exist in the family space but also a monolingual space dominated by majority language practices.As stated in Caldas (2012), language choices and uses between siblings are often out of the © Degu and CMDR.2021 oversight of parents.Seeing this in light of Smith-Christmas' (2020) framework, the monolingual-dominated family space in part reflects members' varied linguistic proficiencies in majority and minority languages, which -as a by-product -pave the way for the establishment of a laissezfaire kind of linguistic norm where the compliance regime is ruled by members' individual choices rather than by jointly established FLP.

hYphEnATED IDEnTITY
The body-mapping portrait of Hana from family 3 (figure 3) -which is decorated by the colour of the Swedish and Ethiopian flags in proportional fashion -represents a sense of her built identities.The affective texts of 'I love Swiden, I love Ethiopien' (sic) written in her body-mapping portrait could index a harmonious embodiment of the two ethnocultural identities.This case is strengthened in her oral testimony, where she said, 'I am Ethiopian but still a little Swedish, half-half '.Although Hana is only eight years old and was born and raised in Sweden, her selfidentification resembles the way in which many youth respondents identify themselves in the United States.As reported in Kagan (2012), based on a survey conducted by the National Heritage Language Resource Center in 2007-2009, many youth respondents identified as hyphenated Americans (e.g.Asian-American, Latino-American) with dual identity.Hana's use of the term half-half perhaps defies the common ethnonational and racial identity categories of 'us' and 'them'; hence, it could be taken as '"we" and "they" at the same time' -a wording that Kagan (2012: 72) uses to describe hyphenated Americans by citing Rudyard Kipling's poem titled 'We and They'.
As can be inferred from the portraits in figure 3, Hana depicted her identity through the languages she speaks and identified herself as half-Ethiopian and half-Swedish.As seen in her body portrait, she chose to map her body in proportionately partitioned fashion using the colours of the Ethiopian and Swedish flags.The brown colour used around her face and arms, and the cross sign shown down her neck are additional signposts of her identity.The body mapping portrait, the metalinguistic commentary given in the form of text, and the oral narration that she gave all speak to the co-existence of multiple identities, which is formed by both the societal context she is in and her familial background.Children's data tell us, in particular, about how they feel, position, and identify themselves within their immediate social and familial context; this contrasts with the caregivers' data, which tend to be about children's heritage language learning outcomes and learning strategies.Although identity positioning is not part of Smith-Christmas's child agency framework in FLP, it is plausible to make a claim that the ways in which children identify themselves is an important point of consideration when studying children's multilingual experiences in the home domain and their role in the formation of FLP.By making sense of the environment she is in, Hana made an identity choice that embraces Ethiopian-ness and Swedishness.While relating Hana's identity positioning to her language choices at home, as depicted by her older sister Feven, -"In Hana's room, we always talk in Swedish, but we use Amharic when we pray" (extract 21) -the use of Amharic as a language of prayer and Swedish as a language of interaction can be taken as the materialisation of her identities.

Competence-run flp and home language use
Although the three families are distinctive in terms of language ideologies and power dynamics, a consistent aspect that emerged throughout the visual and verbal data is that language competence or proficiency seems to have a decisive role in the process of FLP-making in these families.Parents' beliefs about the importance of heritage language maintenance seem to have little role in shaping home language practices.Canagarajah ( 2008) also noted such a discrepancy between language ideology and language proficiency in his study on language shift among families of the Sri Lankan Tamil diasporas.Many parents told the researcher that 'rather than their language practices influencing their children, their children's language choices shaped the everyday language use of everyone at home' (Canagarajah 2008: 164).These parents also admitted that their acts did not align with their beliefs.Similarly, in the present study, family members' habitual home language use is found to be a function of competence or proficiency in the target language -not an ideology-driven and value-oriented course of action set to achieve a certain goal.In this regard, the overwhelming majority of the participants tend to choose the language with which they feel more competent and comfortable to converse, regardless of what they believe or wish to have as a home language policy.As reported through interviews with parents and children's spacemapping narratives -for instance, in the case of family 3 -heritage language use is more of a suggestion than a strictly followed compliant code, as Henok, the father in family 3, reported.We often say Amharic as a rule to speak at home, but we -ourselves [he is referring to them as parents] -do not follow it.We speak in a spontaneous manner.However, when we recall the rule in the middle Exploring family language policy in action © Degu and CMDR.2021 agency in FLP, which conceptualises how the compliance regime, linguistic competence, linguistic norm, and power dynamics intersect in child agency (Smith-Christmas 2020).
This study provided additional evidence of the complex, chaotic, and continually evolving nature of multilingual experiences among migrant families in the home domain.First, the FLP-making process is filled with home language choice dilemmas triggered by competing linguistic demands that are at play simultaneously.In this regard, the case of family 2 is a notable example that reflects the above claim.In family 2, the question of balancing or prioritising between the socioeconomic value of learning and practising the majority language (parents) and the sociocultural value of transmitting the heritage language to the child in the home domain remains a perplexing situation in the FLP-making process.
Second, family constellation, where the issue of power dynamics between family members comes into play, appears to be one driver that shapes the process of FLP making.Although it is not possible to generalise based on the data from this study, the single-parent-headed family (Family 1) runs a consistent and uniform FLP compared with the two coupled families.The two coupled families -as a consequence of family members having different language ideologies and varied proficiencies in the majority and minority languages -are either experiencing the home language choice dilemma (family 2) or have a laissez-faire kind of FLP (family 3), in which home language choice is ruled by individually chosen norms rather than a certain compliant language set by the family.The latter two cases will affect parental language input patterns in terms of both the quality and quantity of inputs that children need to acquire the heritage language (De Houwer 2007).Parallel to what family constellation does in the FLP-making process, children's agency manifested through their home language choices was found to be pivotal in determining whether a compliant regime to minority language use is adhered to, thereby contributing to what kind of linguistic norm is established in their respective family.
When it comes to children's views and perspectives about their multilingual experiences in the home domain, the visual and verbal narrative data provided a highly detailed account of their metalinguistic commentaries on their multilingualism, their home language choices, and interactional patterns described in line with interlocutors and spatial-based experiences within their living spaces.These multilingual accounts highlight the centrality of children's agentive role in affecting the whole process of FLP-making.For instance, the three daughters in family 3 have established a monolingual norm in which they practise Swedish entirely between them and largely with their parents.Hence, the heritage language of Amharic had a minimal place and role in everyday familial communication encounters, which clearly contrasts with the case of family 1.Likewise, in family 2, the parents foresaw the agentive role of their five-year-old child, Liyu, in socialising them in the majority language in the future since they expect her to become proficient in Swedish through schooling.
In addition, individual language competence in the majority and heritage languages is partly what determines home language choices and home language practices regardless of family members' beliefs about heritage and other languages.Finally, except for DEGU © Degu and CMDR.2021 a few studies by parent-researchers (Kopeliovich 2013, Caldas 2012), FLP research tends to examine the home language policy and practices of a family at a single point in time (Lanza andGomes 2020, Päivi et al. 2020).Hence, when the family is considered 'a dynamic temporal body' (Lanza and Gomes 2020: 164), the effect of time on the process of FLP-making remains an interesting empirical question in the field that requires longitudinal studies.For instance, how the now five-year-old daughter of family 2, Liyu, will shape the FLP of the given family in the next four, five, or more years is an important area of future research.Similarly, as the present study is based on a small number of migrant families from a similar sociocultural and minority language background, which can be considered the minority of the minority in the Swedish context, more studies involving diverse families from other minority language communities with migration backgrounds are needed to gain further insights into the commonalities and particularities in the FLP-making experiences.
I wish to conclude by going back to the exchange between Aunty Uji and Ifemelu in Americanah, with which I started this article.The quandary of choosing which language to speak to children in the context of migration was not the only part of the puzzle in the Ethiopian and Eritrean families I studied; the issue was also about which language children chose to speak at home, which eventually had a considerable impact on the FLP.

noTES
By using the term linguistic competence -one of the four dimensions in Smith-Christmas' conceptual framework -I am referring to participants' proficiency in a language in question or their perception of their own and other family members' proficiencies in the target language(s).
figure 3: Body-mapping portrait and excerpt from post-mapping narration by hana, the youngest daughter of family 3.