Assessing the Role of Input Factors in Harmonious Bilingual Development in Children

Abstract

This study delves into the concept of Harmonious Bilingual Development (HBD), characterizing families where linguistic situations do not adversely affect their well-being. Following a recently proposed holistic framework on Harmonious Bilingual Experience (HBE), this research aims to discern the relationship between input factors and the bilingual children’s social–emotional and behavioral skills (SEBS). While input undoubtedly plays a major role in bilingual development, more insight is needed on the quantity and quality of input necessary to foster HBD. A total of 36 five-year-old children with Russian and French as their first languages participated in our study. Children’s language exposure and input quality, as well as parental attitudes, beliefs and strategies regarding language transmission, were assessed with parent questionnaires and activity journals. Additionally, parents assessed their children’s SEBS using the Strengths and Difficulties Questionnaire. Hierarchical cluster analysis identified family profiles which support HBD. We found that the quantity of Russian input, as measured by current exposure, does not directly correlate with SEBS. However, less than 30% of weekly input in Russian appears insufficient for achieving HBD. Furthermore, our study suggests that engaging in at least one extra-curricular activity in Russian, e.g., Saturday school, coupled with parental impact belief, aligns with a family profile experiencing more HBD. These results are coherent with previous studies on harmonious bilingualism, and emphasize the interplay of quantitative and qualitative input factors, as well as SEBS, for achieving HBD.

1. Introduction

1.1. Harmonious Bilingual Development

Introduced in 2006, the notion of Harmonious Bilingual Development (HBD) characterizes bilingual1 families where linguistic situations do not adversely affect well-being (De Houwer 2006, 2015, 2020), highlighting the significance of overall well-being in the language development of bilingual children. Several studies have already examined the connection between language factors and the well-being of bilingual children and teenagers (e.g., review in Müller et al. 2020). In these studies, ‘well-being’ is often conceptualized as individual characteristics within the child (self-esteem, feelings of happiness, life satisfaction, depression symptoms, social–emotional and behavioral skills (SEBS), i.e., interiorized and exteriorized problems, social competence), or as a social construct (e.g., quality of parental relationship). Results show that proficiency in all languages, societal (SocL) or non-societal (Non-SocL), appears to significantly influence various aspects of children’s well-being (Boutakidis et al. 2011; Han 2010; Portes and Hao 2002; Sun et al. 2021; Tseng and Fuligni 2000). Moreover, better language proficiency, exposure and literacy in the home language are linked to improved social–emotional competence in children (Sun 2022; Sun et al. 2021) and, vice versa, diminished well-being predicted refugee children’s second language (L2) abilities (Paradis et al. 2022). This relationship between language and social–emotional skills is bidirectional (Winsler et al. 2014). Good language proficiency in both languages (SocL and Non-SocL) enables bilingual children to communicate easily within and outside the home, experience family cohesion, bond with peers, achieve better academic results, and, overall, experience HBD (De Houwer 2020). Improved social competence, in turn, leads to more social interactions, which may result in incidental learning and reinforce language skills (Weizman and Snow 2001).

Recently, Sun (2023) proposed a study of Harmonious Bilingual Experience (HBE), a conceptual framework derived from HBD, assessing the relationship between parental language perception and proficiency (input richness), parent–child language use (parental input, child output and their mutual language choices, relying on parental beliefs and child agency), and social–emotional well-being (e.g., child SEBS). Parents’ attitudes towards bilingualism, their ‘impact belief’, and dual language proficiency influence their language use with children. This affects children’s bilingual skills, language use, and literacy activities, which in turn impacts their social–emotional well-being. Thus, parental language use, and therefore input, appear to be important in HBD.

1.2. Input in Bilingual Language Development

Terminology with respect to linguistic stimulation acquisition varies considerably (Schalley and Eisenchlas 2022). The terms ‘exposure’, ‘experience’ and ‘input’ are frequently used alternately and rarely defined. Carroll (2017) clearly explains ‘exposure’ as what is measurable and observable in a particular learning context, such as child-directed speech (CDS), while ‘input’ refers to information available concerning a particular learning problem, e.g., use of direct object pronouns. On the other hand, De Houwer (2009) uses the term ‘input’ to refer to “any speech that children hear, whether addressed to them or not” and ‘regular input’ to designate “the (almost) daily contact with a language through interpersonal interaction or by hearing that language” (p. 4). Input is also considered to be “the linguistic material to which children or adult learners are exposed through their environment” (Brehmer and Kurbangulova 2017, p. 227). Armon-Lotem and Meir (2018) generally use the term ‘exposure’ and the term ‘input’ more specifically when speaking of “the availability and frequency of specific target linguistic units in CDS or gestural production” (p. 4). Following De Houwer (2009), we will adopt here the general notion of ‘input’ to refer to all speech heard, whether addressed directly or indirectly to the child. We will also apply the term ‘exposure’ to refer more specifically to measures of input quantity.

In the present section, we will first discuss the impact of input and then develop more precisely how input can be measured, quantitatively and qualitatively, outlining the aims of the present study.

1.2.1. Impact of Input

The amount of input is known to be a major environmental factor of individual variation in language development in both monolingual (Huttenlocher et al. 2010; Rowe 2012; Rowe and Snow 2020) and bilingual children across different linguistic levels.

With respect to the less documented phonological level (Unsworth 2014), Ruiz-Felter et al. (2016) demonstrated that, in 91 Spanish-English bilingual kindergarteners (5;6 years), current input predicted phonological accuracy for L2 vowels and consonants better than the age of first exposure to English. Parental input also qualitatively influences phonological development in bilingual children (Brehmer and Kurbangulova 2017; Stoehr et al. 2019).

Concerning the lexicon, vocabulary skills in monolingual children are closely linked to the lexical diversity of their input (Huttenlocher et al. 2010; Rowe 2012). In bilinguals, De Houwer et al. (2014) assumed that any difference observed in lexical development between bilinguals and monolinguals can be attributed to variations in the quantity and quality of input in each language. This is supported by various studies on bilingual children (Barnes and Garcia 2013; Cattani et al. 2014; Armon-Lotem and Meir 2018; Scheele et al. 2010; Sun 2022; Sun and Ng 2021; Thordardottir 2011, 2014).

Regarding morphosyntactic development in monolinguals, input variations in verb frequency and diversity, as well as the average number of noun phrases, predict the same patterns in children’s production (Huttenlocher et al. 2010). In bilinguals, studies either assess overall grammatical skills (e.g., Chondrogianni and Marinis 2011) or focus on more specific domains, like verbal morphology (e.g., Blom 2010), nominal morphology (Gagarina and Klassert 2018), or grammatical gender (Gathercole and Thomas 2009). They all document links between parental language use and morphosyntactic development. However, input impacts different grammatical features variably. More complex and less transparent morphosyntactic features seem to be particularly sensitive to input frequency (Unsworth 2016). For example, Unsworth (2013) found Dutch–English bilingual children struggled with gender marking in definite determiners and gender detection for neuter nouns, but not for adjectives. Thomas et al. (2014) observed that Welsh children with reduced input in Welsh had not acquired plural marking by age 11. Given the lack of transparency of Welsh plural morphology, it is likely that higher amounts of input are needed in order to reach the critical mass allowing acquisition of this complex feature (Gathercole 2007).

With respect to narrative skills, a recent study by Rodina et al. (2023) comparing 143 Russian-speaking bilingual children (3–11 years) in Norway, Germany and the UK showed that differences in story length (total number of words) and lexical diversity between languages increase significantly with age, the majority language developing faster. Significant differences between countries were attributed to variations in schooling systems and input. Utterance fluency, investigated in only one study so far (Cohen and Mazur-Palandre 2018), showed significant links with cumulative exposure to English in both speech rate in discourse (i.e., number of words per minute) and reduced repetitions and reformulations among first graders at a bilingual school in France.

Overall, input seems to impact the development of linguistic skills at all linguistic levels. However, it remains unclear what amount or quality of input are required to reach the critical mass for acquisition, raising also questions on how to measure input quantity and quality, discussed here below.

1.2.2. Measures of Input

While the impact of input on language development is widely acknowledged, the methods for measuring input and interpreting results vary significantly. Most research has focused on quantifying input, but recent studies highlight the importance of input quality, suggesting that this may be more crucial than quantity (see Rowe and Snow 2020; Schalley and Eisenchlas 2022). The following sections will explore both types of input measures and their interpretations.

Quantitative Measures of Input

Quantitative measures attempt to quantify the amount of current input a child is exposed to during a typical day or week or over time in the case of cumulative exposure2 Unsworth (2013). They are usually indicated in hours per day or percentage of waking hours and established by using parental ad hoc questionnaires (e.g., MacLeod et al. 2013; Unsworth 2013) or language use diaries (e.g., De Houwer and Bornstein 2003, as cited in De Houwer 2018b). Both tools may be difficult and time-consuming for parents and open to interpretation. Given these flaws, comparisons between studies should be considered with caution.

One of the issues researchers have sought to determine through quantitative measures is whether there was a critical threshold of the input amount necessary for (full) minority language development (MinL). Several authors have proposed that less than 20% of input in MinL results in its limited use in production (e.g., Pearson et al. 1997) or in a ‘functionally monolingual’ condition (Peña et al. 2011). A literature review by Pearson and Amaral (2014) concludes that at least 40% input in MinL is necessary for monolingual-like proficiency. Such indications should, however, be considered as very global and depend on whether we discuss language use in reception or production, and, perhaps most importantly, they also depend on the age of the children. Other authors have suggested that, in early childhood, higher MinL input is crucial, as the majority language will necessarily dominate later when the children are growing up. They indicate percentages reaching from 60% of input in the Non-SocL (Thordardottir 2014) to 100% (Barnes and Garcia 2013), as is the case in sequential bilingualism. Thus, Pearson and Amaral (2014) recommend 60% of MinL input during preschool, with immersion and literacy development at school, in order to ensure satisfying long-term progress in that language.

However, once a critical mass is reached, the exact amount of input seems less important (Gathercole 2007), not necessarily reached within one stretch, allowing bilingual children to benefit from discontinued input over the years while still attaining balanced proficiency. Considering the dynamics of language exposure along the developmental span is probably one of the main challenges for future studies in this field.

The focus on input’s quantitative aspects has also led to a debate about possible limitations of bilingual language development due to reduced input in one or both of the languages (Hyams 1991; see also De Houwer 2018b). Without developing this debate here, one point emerging from it is the finding that bilinguals generally thrive with less input in each language than monolinguals. It may be worthwhile to recall the supposed role of input. Ultimately, input should help children understand different linguistic structures and how they allow us to express meaning. It should also allow the children to obtain examples of how to perform diverse communication acts and meet communication needs in various settings. In many cases, more functions will be developed in the majority language than in the MinL. But this largely depends on the input provided in the family. So, if we truly want to understand the process, qualitative aspects of input need to be taken into account.

Qualitative Measures of Input

Input quality refers to the richness of the child’s linguistic environment, but also to its adequateness with respect to their age and interests (Carroll 2017). This mainly involves the quality of child-directed speech (i.e., vocabulary richness and diversity of grammatical structures), and the variety of input sources (diversity of interlocutors, language varieties, activities, acquisition settings, media). Many authors linked quality of speech to nativeness, suggesting that non-native speakers’ input may be less beneficial for the child’s lexical and grammatical proficiency than very proficient native speakers’ speech (Hoff et al. 2020). Immigrant parents may also experience attrition (Brehmer and Kurbangulova 2017).

Among other factors contributing to input quality, Schalley and Eisenchlas (2022) highlight parent’s interaction styles and discourse strategies, as well as literacy access in MinL. Similarly, Rowe and Snow (2020) propose a three-dimensional framework for input quality:

-

Interactional characteristics involving activities with shared attention, parental responsiveness, and child-engaging topics;

-

Linguistic complexity and redundancy as reflected in the phonological, lexical and grammatical features of an input adapted to the child’s developmental stage;

-

Conceptual contents providing developmentally appropriate challenges, i.e., focus on contextualized topics for young children, and introduction of more abstract topics (past, present or irrealis) with older children.

According to these authors, each dimension should be maximized, ensuring optimal language acquisition, though their importance may vary across cultural contexts.

1.3. Aims of the Study

While input significantly impacts bilingual development (De Houwer 2018b; Unsworth 2016), the exact quantity and quality of input needed for harmonious bilingualism in multilingual environments remain unclear. In order to enhance our understanding of HBD and HBE, the aim of this research is to explore how various input factors influence children’s SEBS.

2. Method

2.1. Participants

This study focuses on 34 French–Russian mixed families in France, involving 36 children (22 girls and 14 boys) with Russian Non-SocL and French SocL as their two first languages (2L1). Thirty-four five-year-old Bilingual First Language Acquisition (BFLA) children were born in France (M = 5;6 years, or 66 months, SD = 4.02), while two arrived in the country at the age of 0;2 and 3;8, respectively3. Data collection was mainly undertaken during the COVID-19 pandemic in 2020 and 2021, in participants’ homes. All but one parental couple consisted of a Russian-speaking mother and a French-speaking father, using generally the ‘one parent, one language’ strategy. Almost all children attended the French public education system (second year of preschool (4–5 years) = 3, third year of preschool (5–6 years) = 20, first year primary school (~6-years) = 9), with three children in private schools (Montessori = 2, Russian–French bilingual school = 1 child recently arrived in France), and one child was home-schooled4. None had a history of language delay or impairment. Most children were first- (66.7%, n = 24) or second-born (25%), with the remainder being third-born (2.8%) or twins (5.6%). The majority of families had two children (66.7%) or one child (22.2%); however, there were also families with three (8.3%) and four children (5.6%). Generally, Russian-speaking parents (M = 17.4, SD = 1.72) had more years of education than French-speaking parents (M = 15.6, SD = 2.36). Otherwise, all children had at least one parent with a bachelor’s degree. Finally, all French-speaking parents were employed, while some Russian-speaking mothers were not working at the time of data collection (22.2%).

2.2. Procedure

As recruitment of participants began during the first COVID-19 lockdown, we used social networks to invite French–Russian families in France to complete a brief online questionnaire designed to take an initial survey of the population. Families interested in further participation were then contacted for a preliminary telephone/videocall interview, to determine eligibility. Mothers primarily initiated participation, but the commitment and informed written consent of the entire family—both parents and children—were essential. Despite increased pandemic challenges, the investigator visited nearly all children5 individually twice, following safety protocols and complying to all barrier gestures, using individual or disinfected equipment. Children received French and Russian children’s books or magazines as gifts, while parents were offered consultations with researchers on bilingualism and bilingual education.

2.3. Determining Quantitative and Qualitative Input Measures

To evaluate the amount of current exposure (CE) and input quality, we conceived the Child Language Diary (CLD), a bilingual Russian–French paper notebook given to parents during the first visit. Parents completed the diary over the current week and returned it at the second session, or by post if the tests were close together (M = 5.92 days, SD = 4.49), or during the school holidays. However, some information was missing for a few children, even though all the other data had been collected.

For the CLD, we used a detailed parental questionnaire (following the protocol for obtaining information in a BFLA context, (De Houwer 2009); Russian language proficiency test: Questionnaire, (Gagarina et al. 2010)), preceding the PREschool Language context and Activity Record (PRELAR) (De Houwer 2002), a version of the Language Input Diary (De Houwer and Bornstein 2003, as cited in De Houwer 2018b) designed specifically for preschoolers. The activity log, completed over three days—a full school day, a half school day (Wednesday in France), and a weekend day—recorded the child’s activities outside school, child’s contacts, and languages heard (including audio-visual media). PRELAR guided and assisted parents in reporting their child’s daily experiences with temporal blocks and questions. All the reporters were mothers, most of whom completed the questionnaire and diary in Russian. The notebooks were scanned, anonymized, and individually entered and verified for each participant by two Russian-speaking research assistants.

2.3.1. Current Exposure (CE) to Russian

The CE to Russian was coded separately by two Russian-speaking judges, using the information from the CLD and the proposed evaluation grid. To estimate children’s language exposure during one school week, as in MacLeod et al. (2013), we used a standard day of 13.5 waking hours (M = 13.6 h, SD = 0.758), or 94.5 total waking hours for a week, and calculated the proportion of exposure to each language for this time period. Three typical days were evaluated individually and multiplied in the final calculation: Weekday × 4, Wednesday × 1, Weekend Day × 2.

In the evaluation grid, each day is divided into 5 columns: FR, FR+/RU, FR/RU, RU+/FR, RU. Each column has its own Russian exposure coefficient: FR = 0; FR+/RU = 0.25; FR/RU = 0.5; RU+/FR = 0.75; RU = 1, where the following measures were applied:

-

0, if Russian is not present;

-

0.25, if Russian is present but generally less than French (the mother indicates that she speaks French, there are other French speakers present, therefore more French, even if the mother reports that she speaks Russian with the child);

-

0.5, if both languages are generally present and both parents spend time with the child (for example, it is indicated that the parents share reading tasks, bedtime rituals, etc.);

-

0.75, if French is present but generally less than Russian (the French-speaking father is present but not involved in communication with the child).

For each column, raters entered the number of hours that the child had been exposed to languages according to their estimate coefficient based on the CLD. The sum of the number of hours was automatically calculated for a control check and had to be equal to 13.5. The number of hours was then multiplied by the Russian exposure coefficient, and the total amount of Russian and French per week was automatically calculated in hours, as well as overall percentage of CE to Russian and French per week.

To facilitate the assessment and the comparison, we assumed that a majority of children have similar routines and proposed the following outline of a typical day (

Table 1. Three typical days with general schedules for all participants.

	Number of Hours	Schedule
Week (full school day)
Morning (getting ready)	1	7:30–8:30 a.m.
School (with 2-h break)	8	8:30 a.m.–4:30 p.m.
Evening (after school, family time/meal/getting ready to sleep)	4.5	4:30–9:00 p.m.
After-school kids’ club/nanny/parent	2	4:30–6:30 p.m.
Evening	2.5	6:30–9:00 p.m.
Wednesday (half school day)
Morning (getting ready)	1	7:30–8:30 a.m.
Half-school	4	8:30 a.m.–11:30 a.m.
Half-time nanny/parent	5	11:30–4:30 a.m.
Full-time nanny/parent	8	8:30 a.m.–4:30 p.m.
Evening	4.5	4:30–9:00 p.m.
Weekend
WE time	11.5	no schedule
Russian school	2 1

¹ for 4 real hours (because of multiplication).

). However, our flexible grid was able to detect that some participants’ routines could vary from those of the majority.

The intra-class correlation coefficient was calculated to assess the agreement between two raters on the assessment of CE to Russian for 36 BFLA participants. There was good absolute agreement between the two raters, using the two-factor random effects model and the single rater unit, ICC (2,1) = 0.84, with 95% CI [0.75, 0.90], p < 0.001. However, cases of disagreement about the estimated coefficient of exposure were discussed by the judges and adjusted accordingly for the final result. See

Table 2. Descriptive statistics of variables used for cluster analysis.

Variable	M	SD	Range (Possible Range)	Data Source
SEBS
1. Prosocial Skills	7.42	1.81	4–10 (0–10)	SDQ
2. Behavioral Difficulties	9.69	5.53	1–23 (0–40)
Child language proficiency
3. Relative Proficiency Rating	−0.111	0.854	−1–1 (−1–1)	BILTALK
4. Russian Task	0.861	0.351	0–1 (0–1)	Mishka
Input quantity
5. CE to Russian	38.1	14.4	13.2–85.2 (0–100)	CLD
Input quality
6. Support in Learning Russian	5.19	1.58	0–7 (0–7)	CLD
7. Visits to Russian-speaking Countries	1.64	1.07	0–4 (0–4)
8. Number of Activities in Russian	1.22	1.46	0–5 (0–5)
9. Relative Time of Activities in Russian	1.64	1.76	0–6 (0–6)
10. Number of Different Russian-speakers	3.53	1.25	1–7 (1–7)
11. Russian-speaking Parents’ Education	17.4	1.72	14–20 (10–20)
Russian parents’ beliefs and attitudes
12. Beliefs—Input	2.06	0.860	1–4 (1–4)	BILPATT
13. Beliefs—Bilingualism	1.36	0.593	1–3 (1–4)
14. Attitudes—Bilingualism	1.39	0.688	1–4 (1–4)
15. Attitudes—Code-mixing	3.17	0.941	1–4 (1–4)
16. Expectations—Child Language Use	3.42	0.806	1–4 (1–4)	BILTALK
17. Feelings—Child Language Use	2.42	1.20	1–4 (1–4)
18. Practices—Code-switching	1.44	0.809	1–4 (1–4)

Note: M = mean, SD = standard deviation. Variable: (1) Prosocial Skills: 10 points max (0–5: very low; 6: low; 7: borderline; 8–10: normal). (2) Behavioral Difficulties: 40 points max (0–13: normal; 14–16: borderline; 17–19: high; 20–40: very high). (3) Relative Proficiency Rating: −1 = better French (n = 15, 41.7%), 0 = no difference (n = 10, 27.8%), 1 = better Russian (n = 11, 30.6%). (4) Russian Task: 0 = incomplete (n = 5, 13.9%), 1 = complete (n = 31, 86.1%). (6) Support in Learning Russian (“How do you support your child in learning Russian?”): sum of points per act: 0 = Nothing, 1 = I read to him, her/I speak a lot with him, her/I play with him, her in Russian/Russian television, videos/Language exercises/I make my child speak Russian/Other. (7) Visits to Russian-speaking Countries: 0 = No, 1 = Rarely, 2 = Less than 1 month a year, 3 = 1–2 months a year, 4 = 3 months or more a year. (8) Number of Activities in Russian: 1 per different extra-curricular activity in Russian (e.g., Russian Saturday school, dance, theater). (9) Relative Time of Activities in Russian: sum of points per Russian activity, 0 = no activity in Russian, 1 = less than 2 h a week, 2 = 2–3 h a week, 3 = more than 3 h a week. (10) Number of Different Russian-speakers: 1 point for each different Russian-speaker with whom the child has contact every week, even if it was through a video call with grandparents. (11) Russian Parents’ Education: 10 years = GCSE, Youth training NQV, 12 years = A-Levels, BETC National Diploma, NVQ level 1, 14 years = BETC Higher National Diploma, 15 years = BA/BS, 16 years = MA/MS 1st yr., 17 years = MA/MS 2nd yr., 20 years = PhD. (12–18) Russian parents’ beliefs and attitudes: 1 = Entirely disagree, 2 = Not quite agree, 3 = More or less agree, 4 = Completely agree. Data source: SDQ (Strengths and Difficulties Questionnaire, Goodman 1997); BILTALK (Talk and Interaction Questions for Parents in Bilingual Settings, De Houwer 1999); Mishka (Picture-story retelling task in Russian context, Tiulkova 2017); BILPATT (Language Attitude Questions for Parents in Bilingual Settings, De Houwer 2018a); CLD (Child Language Diary, see Section 2.3).

for the descriptive statistics of variable CE to Russian.

2.3.2. Qualitative Input Measures in Russian

Based on the responses in the parental questionnaire, we operationalized eight measures of input quality for a typical school week. Only five of them were included in the cluster analysis (see Table 2 for more detailed description):

• Support in Learning Russian,
• Visits to Russian-speaking Countries,
• Number of Activities in Russian,
• Relative Time of Activities in Russian,
• Number of Different Russian-speakers.

Two remaining measures—Time of Media in Russian (in hours), Frequency of Reading Practices in Russian—were not used in cluster analysis. First, almost all parents reported reading to their children either daily or weekly, with only one family reporting no reading. Second, the time spent with media without adult comments was insufficient to be considered as a measure of input quality. However, we included Russian-speaking Parents’ Education as a control variable, as it was identified as a vital factor for input quality (De Cat 2021).

2.4. Parental Beliefs and Attitudes

The bilingual couples (34 mothers and 33 fathers) filled in separately an online survey Attitudes, interactions and feelings in a bilingual context (based on BILPATT, De Houwer 1999, and BILTALK, De Houwer 2018a). We used only Russian-speaking parents’ judgements (4-point Likert scale: 1 = “Entirely disagree”, 2 = “Not quite agree”, 3 = “More or less agree”, 4 = “Completely agree”) of seven statements corresponding to the following variables (see also Table 2):

• Beliefs—Input (“Children learn to talk although they don’t hear other people say much”),
• Beliefs—Bilingualism (“Children who hear two languages have more problems with learning one language well”),
• Attitudes—Bilingualism (“It is confusing for young children to learn two languages at once”),
• Attitudes—Code-mixing (“It is not good to mix two languages in one sentence”),
• Expectations—Child Language Use (“I wish my child would always respond in the language I usually speak to him/her”),
• Feelings—Child Language Use (“It upsets me that my child doesn’t always speak the same language that I usually speak to her/him”),
• Practices—Code-switching (“When I talk to my child, I often use both Russian and French in the same conversation”).

In this study, we focused on the exclusive opinions of Russian-speaking parents, as they are the primary providers of input in the Non-SocL. Parental awareness and mutual agreement on bilingual education are also significative factors in dual language development (Nakamura 2019; Seo 2022; Yamamoto 2001). Our preliminary results indicated that better parental cooperation was related to children’s positive attitudes towards both languages (Tiulkova et al. 2023).

2.5. Social–Emotional and Behavioral Skills (SEBS)

Parents assessed their children’s SEBS using the Strengths and Difficulties Questionnaire (SDQ, Goodman 1997). This clinical tool with multilingual interface has also been used in research in order to consider the relationship between bilingualism, dual language performance and children’s social–emotional outcomes (e.g., McLeod et al. 2015; Sun 2019, 2022). Social competence has been associated with better school performance and healthier well-being, while the lack of SEBS has been linked with personal, social and academic difficulties (Sun et al. 2021). Thus, SEBS are operationalized as a measure of well-being in the HBD/HBE framework.

The SDQ comprises 25 questions6 divided into five dimensions referring to distinct psychosocial difficulties or skills: (D1) emotional difficulties, (D2) conduct difficulties, (D3) inattention and hyperactivity, (D4) peer-relational difficulties, and (D5) prosocial skills (Hazo et al. 2023). These five scores indicate the parents’ assessment of the child’s behavior (“not true”, “somewhat true” and “certainly true”), with a score from 0 to 10 per dimension, with 40 points (normal, 0–13; borderline, 14–16; high, 17–19; very high, 20–40) for Behavioral Difficulties (D1–D4) and 10 points (very low, 0–5; low, 6; borderline, 7; normal, 8–10) for Prosocial Skills (D5). The scores for both parents’ responses were calculated using the SDQscore website (https://sdqscore.org/), but, again, only Russian-speaking parents’ evaluation was included (see Table 2 for descriptive statistics of these variables). Studying Behavioral Difficulties in bilingual context, it is also possible to use externalizing (D2 + D3) and internalizing (D1 + D4) behavior categorization (see Paradis et al. 2022).

2.6. Relative Proficiency Rating

We obtained a Relative proficiency rating7 from parents’ responses to the online survey (BILTALK). This rating concerns parental comparative assessment of which language is better developed. The explicit question “As far as I can be a judge of it, my child speaks much better…” was asked in both directions: (1) “…Russian than French” and (2) “…French than Russian”, with four possible answers “Completely agree” = 4, “More or less agree” = 3, “Not quite agree” = 2, “Entirely disagree” = 1. We coded parental responses, as explained in De Houwer (2023), and proposed a common Relative proficiency rating, based on both Russian-speaking and French-speaking parents’ assessments (Table 2).

2.7. Analyses

In order to describe different participant profiles in relation to quantity and quality of input, parental beliefs, attitudes and strategies for maintaining Russian Non-SocL and the SEBS across different families, we carried out a cluster analysis for 18 variables (see Table 1), using the snowCluster package in the Jamovi statistical tool (The Jamovi Project 2022). We used the ward.D2 method and the Euclidean distance. The one-way analysis of variance (ANOVA) for our 18 variables was applied to determine the factors significant in the construction of the clusters. We compared the results for each cluster according to these variables and confirmed whether they significantly differed using Linear Regression Models. Spearman’s correlation analyses were also run between these variables primarily to see if there were links between all input (quantitative and qualitative) measures, whether they were included to cluster analysis or not, and SEBS scores.

3. Results

3.1. Cluster Analysis

The clustering technique is a type of hierarchical agglomerative clustering used to group objects in clusters based on their similarity. By including several quantitative variables, this procedure attempts to obtain relatively homogeneous participants’ groups based on the selected characteristics. An iterative aggreging algorithm starts from separate participants (at the right of the vertical dendrogram in Figure 1) and combines them until one is left (the left part of the dendrogram).

We obtained the following four clusters: Cluster 1 (n = 10, in blue), Cluster 2 (n = 9, in grey), Cluster 3 (n = 12, in yellow) and Cluster 4 (n = 5, in green), which are presented in the dendrogram (see Figure 1).

ANOVA on the nine variables across the four clusters enables us to check whether these clusters are sufficiently individualized. Analysis provides significant effects: Prosocial Skills (F(3, 13.5) = 16.473, p = < 0.001); Relative Proficiency Rating (F(3, 13.7) = 17.294, p = < 0.001); Behavioral Difficulties (F(3, 14.2) = 9.072, p = 0.001); Practices—Code-switching (F(3, 13.2) = 9.925, p = 0.001); CE to Russian (F(3, 13.9) = 6.024, p = 0.008); Number of Different Russian-speakers (F(3, 14.6) = 4.381, p = 0.022); Visits to Russian-speaking Countries (F(3, 17.2) = 3.962, p = 0.026); Attitudes—Code-mixing (F(3, 15.1) = 3.531, p = 0.041); Attitudes—Bilingualism (F(3, 12.8) = 3.65, p = 0.042). The descriptive statistics of these variables according to clusters are presented in

Table 3. Descriptive statistics for the four clusters.

	Cluster 1 (n = 10)	Cluster 2 (n = 9)	Cluster 3 (n = 12)	Cluster 4 (n = 5)
Prosocial Skills	8.70 (1.25), 6–10	8.67 (1.12), 7–10	5.67 (1.07), 4–8	6.80 (1.48), 5–9
Relative Proficiency Rating	−0.900 (0.316), −1–0	0.333 (0.707), −1–1	0.500 (0.674), −1–1	−0.800 (0.447), −1–0
Behavioral Difficulties	6.60 (3.63), 3–14	6.22 (4.02), 1–13	12.1 (4.83), 4–23	16.4 (4.10), 10–20
Practices—Code-switching	1.50 (1.08), 1–4	1.11 (0.333), 1–2	1.17 (0.389), 1–2	2.60 (0.548), 2–3
CE to Russian	30.0 (7.06), 17.3–39.0	46.4 (10.7), 26.5–62.6	42.7 (17.6), 14.3–85.2	27.9 (10.2), 13.2–41.1
Number of Different Russian-speakers	2.80 (0.919), 1–4	4.67 (1.22), 3–7	3.25 (1.14), 2–6	3.60 (0.894) 3–5
Visits of Russian-speaking Countries	1.80 (1.03), 0–3	1.44 (1.01), 0–3	2.00 (1.21), 0–4	0.800 (0.447), 0–1
Attitudes—Code-mixing	3.30 (0.823), 2–4	3.67 (0.500), 3–4	2.50 (1.09), 1–4	3.60 (0.548), 3–4
Attitudes—Bilingualism	1.20 (0.422), 1–2	1.22 (0.441), 1–2	1.17 (0.389), 1–2	2.60 (0.894), 2–4
Time of Media in Russian (in minutes) 1	294 (265), 28.8—780	156 (168), 0—540	166 (172), 0—540	255 (191), 80.4—563
Frequency of Reading Practices in Russian 1	1.90 (0.316), 1–2	1.89 (0.333), 1–2	1.75 (0.452), 1–2	1.40 (0.894), 0–2

Note: Mean (Standard Deviation), Range: ¹ per week. Frequency of Reading Practices in Russian: “no” = 0, “now and then every week” = 1, “every day” = 2.

.

As clusters’ values in Table 3 seem to reveal particular variables’ profiles, we conducted Spearman’s correlation analysis on the inter-variables’ relationship to check links between input and SEBS parameters. No significant relationship was found. However, parental attitudes and language use are positively correlated with SEBS: negative view of code-mixing (Attitudes—Code-mixing) and negative view of bilingual acquisition (Attitudes—Bilingualism) are correlated with Prosocial Skills (r = 0.376, p < 0.05) and Behavioral Difficulties scores, respectively (r = 0.422, p < 0.05). Parental code-switching or dual language use in the same conversation (Practices—Code-switching) is positively correlated with children’s Behavioral Difficulties score (r = 0.474, p < 0.01) and with negative view of bilingual acquisition (Attitudes—Bilingualism) (r = 0.422, p < 0.05). As for SEBS, Prosocial Skills are negatively correlated with Behavioral Difficulties (r = −0.655, p < 0.001). Without surprise, the total number of different Russian speakers is positively correlated with CE to Russian (r = 0.390, p < 0.05), as well as Relative Proficiency Rating (r = 0.663, p < 0.001): a higher score corresponding to “better Russian” is associated with more CE to Russian.

We then described every cluster by comparing the scores of each of these variables. We ran Linear Regression to compare the scores between clusters and see whether the observed differences were significant. In the following description, we will report the results of the models.

3.1.1. SEBS

There is no difference in Prosocial Skills and Behavioral Difficulties scores between Cluster 1 and Cluster 2, nor between Cluster 3 and Cluster 4 (Figure 2a,b). However, a significant difference in Prosocial Skills is found for Cluster 1 and for Cluster 2 compared to Cluster 3 (p < 0.001) and Cluster 4 (p < 0.01). The same founding is observed for Behavioral Difficulties: Cluster 1 and Cluster 2 compared to Cluster 4 (p < 0.001) and Cluster 3 (p < 0.01).

3.1.2. Relative Proficiency Rating

There is no difference in Relative proficiency rating scores between Cluster 1 and Cluster 4, nor between Cluster 2 and Cluster 3 (Figure 2c). However, a significant difference in Relative proficiency rating is found for Cluster 1 compared to Cluster 2 and Cluster 3, as well as for Cluster 4 compared to Cluster 2 and Cluster 3 (

Table 4. Difference in Relative proficiency rating found for the four clusters.

	Cluster 1	Cluster 2	Cluster 3	Cluster 4
Cluster 2	p < 0.001		p = 0.518	p < 0.01
Cluster 3	p < 0.001	p = 0.518		p < 0.001
Cluster 4	p = 0.754	p < 0.01	p < 0.001

).

3.1.3. CE to Russian

There is no difference in CE to Russian scores between Cluster 1 and Cluster 4, nor between Cluster 2 and Cluster 3 (Figure 2d). However, significant difference in CE to Russian is found for Cluster 1 compared to Cluster 2 (p < 0.01) and Cluster 3 (p < 0.05), as well as for Cluster 4 compared to Cluster 2 (p < 0.05) and Cluster 3 (p < 0.05).

3.1.4. Input Quality Measures

There is no difference in Number of Different Russian-speakers scores between Cluster 1, Cluster 3 and Cluster 4 (Figure 3a). However, significant difference in Number of Different Russian-speakers is found for Cluster 2 compared to Cluster 1 (p < 0.001), as well as to Cluster 3 (p < 0.01), but not to Cluster 4. There is no significant difference in Visits of Russian-speaking Countries scores between all clusters (Figure 3b).

3.1.5. Parental Beliefs and Attitudes

There was no difference in Attitudes—Bilingualism scores between Cluster 1, Cluster 2 and Cluster 3 (Figure 4a). However, significant difference is found for Cluster 4 compared to Cluster 1, Cluster 2 and Cluster 3 (p < 0.01).

As for Attitudes—Code-mixing scores, no difference was found between Cluster 1, Cluster 2 and Cluster 4 (Figure 4b). However, a significant difference was found for Cluster 3 compared to Cluster 1 (p < 0.05), as well as to Cluster 2 (p <0.01) and Cluster 4 (p < 0.05).

Finally, there was no difference in Practices—Code-switching scores between Cluster 1, Cluster 2 and Cluster 3 (Figure 4c). However, a significant difference in Practices—Code-switching was found for Cluster 4 compared to Cluster 1 (p < 0.01), as well as to Cluster 2 and Cluster 3 (p < 0.001).

3.2. Spearman’s Correlation Analysis

No significant correlation was found between CE to Russian and SEBS. However, Number of Activities in Russian was negatively correlated with Behavioral Difficulties (r = −0.361, p < 0.05), as well as with Relative Time of Activities in Russian (r = −0.365, p < 0.05).

4. Discussion

While social competence is related to well-being, we used social–emotional and behavioral skills (SEBS) to determine whether children’s bilingual development was harmonious or conflictual. Other factors potentially linked to Harmonious Bilingual Development (HBD) include: (1) predominant use of the same language in parent–child interactions, (2) child’s active use of two languages, and (3) their nearly equal dual proficiency (De Houwer 2015). Following Sun (2019, 2023), we aimed to contribute to the definition of the conceptual framework of HBD/HBE. To achieve this, we described different family profiles using a clustering technique, which has proven to be informative for multidimensional data (Camps 2023; see also Cattani et al. 2014).

We also investigated connections between SEBS and aspects of Non-SocL input, both in quantitative and qualitative terms. Exposure to Non-SocL is typically more frequent in the early years of bilingual children, with input variability increasing once children start preschool, usually in SocL. Thus, we focused on CE, a sufficiently reliable predictor of Non-SocL proficiency (De Cat 2024). Less evidence was found for the need to consider cumulative exposure, the most robust indicator of SocL, but also the most time-consuming. However, we collected brief parental reports on children’s care and language history in and outside the home before preschool, which can be used in further research.

4.1. Family Profiles and HBD

According to De Houwer (2015, 2020), children in Clusters 1, 2, and 3 are likely to experience HBD, albeit with some variations across different aspects (Table 3), developed here further.

First, all children, except one from Cluster 1, use both languages; however, their relative language proficiency rating varies. Thus, in Cluster 1 (n = 10), nine children were estimated as more proficient in French (25% of the total of participants from four clusters) and, for one child, parents reported no differences between the languages (2.8%); in Cluster 3 (n = 12), seven children were estimated as better in Russian (19.4%), and four children showed no difference in 2L1 proficiency (11.1%), with one child having better relative proficiency in French (2.8%). Four children in Cluster 2 (n = 9) were estimated as better in Russian (11.1%), four with the same proficiency in both languages (11.1%), and one child as better in French (2.8%).

Second, Cluster 3 differs from Cluster 1 and Cluster 2 for SEBS and is characterized by the lowest scores for Prosocial Skills and a higher level of Behavioral Difficulties. The children in Clusters 1 and 2 had the highest score for Prosocial Skills and lower Behavioral Difficulties level.

Third, the average CE to Russian for Cluster 1, 2 and 3 was 39.7% (SD = 14.4). However, the bilinguals from Cluster 1 were less exposed to Russian (M = 30%, SD = 7.06), and the children from Cluster 2 most exposed (M = 46.4%, SD = 10.7).

Fourth, Russian-speaking parents of children in these three clusters responded that they did not mix languages in conversation while speaking to their child. They did not find also that “it is confusing for young children to learn two languages at once”. The majority of parents of these clusters have a master’s degree and apply a wide range of Russian learning support activities.

Finally, the participants from Cluster 1, 2 and 3 visited more or less frequently Russian-speaking countries. Almost all children from these three clusters participated actively in extra-curricular activities in Russian (at least the Russian Saturday school). Children also benefited from regular exposure to books in Russian. Cluster 2 and Cluster 3 spent on average less time watching media in Russian (M = 162 min/per week, SD = 166) than Cluster 1 (M = 294 min/per week) and had a more varied Russian-speaking entourage with three–four different Russian-speakers, compared to two speakers for Cluster 1 (Table 3). Families in Clusters 1, 2 and 3 claim that they read to their children almost every day (Table 3).

On the other hand, children in Cluster 4 showed probably conflictual bilingual development. All of them, except one, did not actively use Russian (although they generally understand the language, they do not speak it themselves). They also had low scores for Prosocial skills and the highest scores for Behavioral Difficulties. Their CE to Russian was low (M = 27.9%), with fewer opportunities for travelling to Russian-speaking countries or taking part in extra-curricular activities in Russian. Parents provided less language support and had lower education levels. These children were also less exposed to books in Russian (Table 3), but frequently watched media in Russian (255 min/week) and had a diverse Russian-speaking entourage, including older siblings (Table 3).

4.2. Quantitative and Qualitative Input and SEBS

The results demonstrated no significant relationship between input quantity (i.e., CE to Russian) and children’s SEBS, including Behavioral Difficulties or Prosocial Skills. However, two closely related measures of input quality—the Number of Activities in Russian and Relative Time of Activities in Russian (both per week)—negatively correlated with Behavioral Difficulties. Children with fewer activities and less time in Russian had higher Behavioral Difficulties reported by their parents. These findings align with Sun (2019), who found home language reading frequency and library visits negatively related to Behavioral Difficulties and home reading positively to Prosocial Skills. According to the author, home reading activities were found to be more important for children’s SEBS than language exposure. The less-frequent shared reading is associated with higher risk of social–emotional problems in young children (Martin et al. 2021), while dialogic reading is a helpful and beneficial strategy (Fettig et al. 2018). We were unable to observe the same thing for literacy activities, given that the majority of families reported reading to their children every evening, either in Russian or in French, and we have not yet been able to detail the measurement in terms of the number of reading days in Russian Non-SocL.

Ours results support previous research on harmonious bilingualism (Han 2010; Sun et al. 2021; Sun 2019, 2022), confirming the importance of input quantity (Schalley and Eisenchlas 2022; Unsworth 2016) and quality (Rowe and Snow 2020) for children’s language proficiency and use. We claim that 30% of CE to Non-SocL was potentially a critical threshold for bilingual 5-year-olds to actively use both languages (a crucial factor of HBD) and experience social–emotional well-being. Furthermore, our results show that there is no direct correlation between the amount of CE to Russian and SEBS, suggesting that French–Russian children living in France have more opportunities and more need to communicate in French SocL, fundamental for their social competence.

Based on our data, 40% CE threshold was optimal for better proficiency in Russian Non-SocL, comparable to proficiency in French SocL. However, given that 60% of all participants had less than 40% of CE to Russian, the diversity of Russian speakers (grandparents, educators), visits of Russian-speaking countries, extra-curricular activities in Russian (e.g., “Russian school”) and various parental strategies for supporting Russian learning in everyday life appeared as significant factors in language development, contributing also to harmonious bilingualism. This can also be explained by the effect of cumulative language exposure, a very solid predictor of proficiency, that requires further study. Hence, a bilingual child will not need twice as much input as a monolingual because input in both languages may allow the bilingual to understand the functioning of all structures and mappings used in both languages. Only very specific aspects exclusive to one language need specific input and may be considerably longer to acquire (see for example the ergative case in Basque–French bilinguals described in Duguine and Köpke 2019). So far, only relative proficiency ratings have been used to evaluate children’s skills. Future research will explore objective language proficiency measures, like utterance fluency, via narratives in both languages.

Nevertheless, we cannot extend these findings to other populations. Historically, France has been a rather Russophile country, where Russian culture and language are valued. Despite the current political context, which could negatively impact its transmission, the Russian-speaking community in France generally strives to maintain the Russian language in daily life and improve its status in the national educational system (e.g., Russian is taught in middle and high public and private schools, as well as at universities). Many Russian-speaking associations offer extra-curricular activities for preschoolers and primary school pupils. Russian-speaking parents usually have a high level of education, are aware of their role in their child’s language development and are well-informed about practices fostering bilingual education. However, it is very challenging to encourage participation from parents who do not transmit their language and are not interested in bilingualism.

In conclusion, our study demonstrates that an inclusive linguistic environment is beneficial for children’s HBD. Formal instruction and various parental practices supporting language learning may provide the richer linguistic experience necessary for developing and maintaining strong competencies in Non-SocL. Input may influence not only children’s language proficiency but also foster their social–emotional and behavioral skills. Large-scale longitudinal studies are needed to verify whether the series of potential effects operate unidirectionally or bidirectionally, and to determine other factors influencing HBE (Sun 2023). Indeed, in order to describe more thoroughly their unique experience, research might consider investigating children’s agency and their feelings about bilingual situation (for an example, see Tiulkova et al. in press). Finally, the findings of this work encourage researchers, parents, educators, therapists, and policymakers to collaborate in advancing children’s bilingualism and their social–emotional well-being.