Bilingualism of Children in Different Multilingual Contexts

Abstract

Many parents and professionals believe that learning to speak, read and write in two languages can lead to academic deficiencies due to cognitive overload and the risk of confusion linked to handling two language codes. Therefore, some bilinguals abandon or are tempted to abandon one of the two languages, often the first language, in exchanges with their children, in favor of the language of schooling. However, all recent scientific data tend to show that bilingualism is an asset more than a handicap. Nevertheless, these positive results most often concern English-speaking contexts and are not directly transposable to a French-speaking context. Drawing on the results of our work carried out in Oceania and in other territories, this article will deal with bilingual development and the impact of educational systems that promote the heritage or local languages from primary school. More specifically, the oral language of the bilingual, biliteracy and the effects of cross-linguistic transfer will be addressed. Results from both longitudinal studies in New Caledonia and French Polynesia, confirmed by other studies conducted in Sub-Saharan African, show a positive effect of the bilingual education curriculum on local language (Drehu and Tahitian) skills without having negative effects on French. We demonstrated that the expected effects of cross-linguistic transfer are only possible if the pupils learn to read and write in the two languages (local language and French). Additionally, learning to read in one of those local languages makes it easier to learn to read in French, which has a more opaque writing system.

1. Introduction

Across the world, many parents and professionals (teachers, psychologists or speech therapists) think that learning to speak, read and write in two languages can lead to academic deficiencies due to cognitive overload and risks of confusion linked to code switching. Consequently, some bilingual people abandon or are tempted to abandon one of the two languages, often their first language (L1) when interacting with their children, in favor of the second language (L2), when it is the language used at school. However, since the work of Peal and Lambert (1962), scientific data tend to show that bilingualism and bilingual education are an asset rather than a hindrance (Wright et al. 2015). These positive results can generally be found in English speaking contexts where the two languages are held in high regard by society (for instance, English–French bilingualism). As such, these findings cannot be directly transposed in a French context, in which local languages are undervalued in comparison to French, which is the language of instruction. Furthermore, although some educational systems that value regional languages and heritage languages do exist and have existed for a long time on French territories, it is necessary to conduct experimental assessments in various sociolinguistic contexts to demonstrate their impact.

The goal of this article is to present a synthesis of our studies on bilingual development and the impact of educational systems that valorize heritage languages and/or local languages on language development, in various multilingual contexts (Nocus 2022). Our work started in New Caledonia in 2003 and was first extended to French Polynesia and French Guiana, then to other countries: Haiti from 2012 to 2014 (for preschool children) and eight French-speaking countries in Sub-Saharan Africa (Benin, Burundi, Burkina Faso, Cameroon, Mali, Niger, DR Congo and Senegal) from 2013 to 2016 (for primary-school children). Although these countries have different political, educational and sociolinguistic contexts that limit the scope of possible comparisons between them, they share the French language as an official language and one of the main languages of instruction, even though it is not necessarily the language or one of the languages of socialization of children. Faced with the growing social demand for recognition of local cultural identities (from the 1970s onward) and taking into account the findings of international research on the advantages of bilingualism, these countries have implemented bilingual educational systems. The objective of our work was to highlight the cognitive processes that contribute to the pupils’ academic success. Our first sub-goal was to show whether or not cross-linguistic transfer existed: the objective was to check whether training children to develop some oral language skills (phonology, vocabulary and morphosyntax) in their heritage language or local language improved their performances in the same skills in French. Furthermore, the heritage languages targeted in our work have the particularity of having an alphabetical writing system whose orthographic system is more transparent than the French system. So, the second sub-goal was to test whether or not learning to read in a language with a transparent orthographic system (heritage language or local language) facilitates the learning of a language with a more opaque writing system (French).

First, we will review the results of the scientific literature about the oral language development of bilingual children, biliteracy and the effects of cross-linguistic transfer. Next, we will present the bilingual educational programs targeted in our work. The third part will detail two evaluation studies carried out on three programs. Lastly, we will present the conclusions and some research perspectives.

2. Development of Oral Language in Bilingual Children, Biliteracy and Cross-Linguistic Transfer

2.1. Development of Oral Language in Bilingual Children

This section will examine briefly the development of the main components of language in monolingual and bilingual people.

Developing one’s language in one or several languages does not only mean learning vocabulary and syntax rules. This development involves learning to carry out speech acts that take into account the communication context and integrate the rules of the speaker’s linguistic community. It has been demonstrated that newborns have early and innate abilities and skills in the field of language perception and production (de Boysson-Bardies 1996). When exposed to the rhythm of different languages in the womb, they are able to distinguish between sentences from two distinct languages (Byers-Heinlein 2013). At four-months old, children socialized in two languages can distinguish between sentences in each of them, even when they are rhythmically similar. Between 18- and 36-months old, bilingual children show signs of language adaptation, since they are able to modulate their language production according to the language used by their monolingual interlocutor (Byers-Heinlein 2013). According to Bertoncini and de Boysson-Bardies (2000), newborns are highly sensitive to the prosodic variations of speech and rely on prosodic clues to recognize their mother’s voice and their mother language. Bilingual children use the prosodic variations that they know from the two languages (pitch and duration, for instance) to determine the grammatical structure of utterances (Bijeljac-Babic 2017).

In the first months of their life, babies, whether they were born in a monolingual or a bilingual environment, are potentially capable of producing and discriminating every sound that exists in every language in the world. This ability disappears between 8 and 10 months when the child starts babbling (repetitive utterances with a rhythmical alternation of consonants and vowels) and selecting only the sounds associated with their language(s) of socialization. From this moment on, the foreign phonetical contrasts are no longer distinguished as they were, but perceived through the assimilation of the linguistic contrasts (de Boysson-Bardies 1996; Florin 2020). Introducing an L2 in the child’s environment reorganizes the established categories so that other phonetical contrasts can be differentiated. The later the L2 is introduced, the harder it will be to create new phonetical categories. Thus, the language(s) present in the child’s environment start influencing their perceptive organization and phonation very early on. If two languages are equally present from birth, the newborn will create the phonetical categories of the two languages (Bijeljac-Babic 2017).

The development of vocabulary in the two languages also depends on the age of exposure to the two languages and the degree of exposure to each of them (Bylund et al. 2019). Young bilinguals reach the usual thresholds of early language acquisition (first word, first fifty words and first sentence words) at the same time in each of the two languages and according to time indicators that are fundamentally the same as for monolingual children (Bijeljac-Babic 2017). However, with the exception of balanced and simultaneous bilinguals, the lexical resources of bilingual children in each of the two languages are often more limited than those of a monolingual child (Bialystok et al. 2010; Nocus et al. 2018). Comparatively, bilinguals are slower and make more mistakes in picture naming and lexical decision tasks, produce fewer words in verbal fluency tasks and often have difficulties finding their words (“word on the tip of the tongue” phenomenon), including in their dominant language (Bialystok 2009). The reason for which bilingual children experience deficiencies in their access to lexical resources is not clearly established. A first hypothesis would be that bilinguals must distribute their language “learning time” between two languages, which lowers their level of exposure to each of the two languages, a level of exposure that is necessary to acquire vocabulary (Poulin-Dubois et al. 2013). Moreover, it is likely that some words are produced in a context where bilinguals only use one of their two languages, which reduces the number of words they acquire in the other language. Access to words in a specific language is thus more frequent for monolinguals than for bilinguals, which explains monolinguals’ superior level in lexical access tasks. A second hypothesis suggests that, compared to monolinguals, bilinguals need to go through a more complex process to access words in each language due to the necessity of inhibiting the competing language.

Barring any pathology, all children successfully acquire the morphosyntax of their language(s) of socialization without needing to be explicitly taught grammar (Meisel 2008). The frequency of exposure to the morphological characteristics of languages is important: children first acquire the most productive word formation processes of their language(s), i.e., the ones that are the most frequently produced by adults. However, morphological representations are not the same in every language (for instance, English uses more compound words and less prefixed words than French).

Lastly, among language skills, metalinguistic abilities1 were studied particularly extensively in bilinguals. Several literature reviews show that studies do not reach a consensus regarding the benefits of bilingualism for their development (Adesope et al. 2010; Besse et al. 2010). For some, bilingualism facilitates it by allowing the child to differentiate very early on between the semantic and formal aspects of the code, to perceive early on the arbitrary relationship between words and their meaning(s) and to develop more flexible cognitive abilities (Besse et al. 2010). Bilinguals thus would have a better awareness of the abstract characteristics of language and of their own learning process, an ability to control and adequately allocate their attention resources and a greater development of symbolic and abstract representations and problem solving. Several studies comparing monolinguals to bilinguals in metaphonological (Besse et al. 2010), metasyntactical (Demont 2001) and metamorphological (Campbell and Sais 1995) tasks demonstrate the precocity and superiority of bilinguals in performing metalinguistic tasks. Others, on the contrary, such as those of Chiappe and Siegel (1999) and of Lesaux and Siegel (2003), fail to highlight this bilingual superiority in performing several metalinguistic tasks, especially in the case of consecutive bilingualism, probably due to a lack of exposure to L2. The lack of a consensus regarding the advantages or disadvantages of bilingualism can likely be explained by the fact that most studies compare bilinguals to monolinguals as if the two groups were homogeneous. However, there are numerous interindividual differences in language acquisition, language rhythm and style, as much for monolinguals as for bilinguals. For those, there are as many profiles as there are personal histories, as many bilingualisms as there are life circumstances. Bilingual profiles differ depending on the level of proficiency reached in the two languages, the age of acquisition of L2, the frequency of use of the two languages, their status in society and the affective value that the bilingual attaches to them (Hélot 2007).

2.2. Biliteracy

The acquisition of written language is essential in order to adapt to one’s environment and, in particular, the labor market. It is largely based on the mastery of oral language (Gombert 1990). In addition, being literate is not enough. It is necessary to develop literacy, which is the “ability to understand, evaluate, use and engage with written texts to participate in society, achieve one’s goals, and develop one’s knowledge and potential”. (OCDE 2000, p. 10). When bilingual children access biliteracy, they can be faced with various constraints.

First of all, the sociolinguistic context is important for the valorization of languages. To put it simply, it is more motivating to learn to read in a recognized and valued language than in an undervalued one.

Furthermore, the similarity of the linguistic characteristics of spoken languages has a largely demonstrated effect, since the development of oral language is a necessary prerequisite to the development of reading and writing proficiency. It is easier to learn a language from the same linguistic family as one’s L1. For instance, it is easier for a French person to learn Spanish than Chinese (Bijeljac-Babic 2017). The greater similarity of the phonological structures between Spanish and French and the easier access to a phonological awareness in Spanish facilitate the phonological analysis of Spanish for bilinguals whose L1 is French, which is not the case with Chinese.

Moreover, the effect of bilingualism on children who learn to read depends on the type of writing system used in each language. Bialystok et al. (2005) compared the performances of four groups of Grade 1 children in vocabulary acquisition, short-term memory, phonological awareness and non-words reading. Three groups were bilingual (English–Spanish, English–Hebrew and English–Cantonese) and one group was monolingual (English). The results show that learning to read in two languages develops phonological awareness in both languages. However, the phonological awareness of English–Spanish and English–Hebrew bilinguals was superior to that of English–Cantonese bilinguals and monolinguals.

Likewise, among alphabetical writing systems, the various alphabets can have some letters in common and others that are different, or even no common spelling at all (this is the case between the Arabic and the French alphabet). Additionally, all writing systems do not have the same reading direction (horizontally from left to right; horizontally from right to left; in columns from top to bottom). Therefore, the fact that the two languages have two different alphabets—sometimes with a different reading direction—can represent extra cognitive work, since the learner does not have any other choice than learning them by heart.

Lastly, the orthographic depth (transparent language vs. opaque language)2 of one language compared to the other constitutes a major constraint for bilingual children learning how to write. Mann and Wimmer’s (2002) and Ziegler et al.’s (2010) work show that learning to read in a language with transparent spelling seems to make it easier to learn to read than in a language with opaque spelling. Consequently, the orthographic regularity of a language would allow for a quick learning of grapheme–phoneme correspondences and facilitate the understanding of the alphabetical principle, since the correspondences are regular.

2.3. Effects of Cross-Linguistic Transfer

International research indicates that the acquisition of two or several languages involves cross-linguistic transfer in bilingual people, which has an impact on cognitive development in general. Kuo et al. (2016) define these effects of cross-linguistic transfer as when the learning of language A facilitates the learning of language B. Kuo and Anderson (2010) proposed a theory of structural sensitivity according to which the basis of the bilingual metalinguistic advantage is the common experience of the two languages, rather than the transfer of processing skills developed in one language to another language. This increased sensitivity to the structural aspects of a language can be explained by the fact that bilingual children constantly need to overcome cross-linguistic interferences, which forces them to focus their attention on the structural characteristics of the language. Moreover, when confronted with several linguistic systems, bilingual children notice the structural similarities and differences between those languages and develop more abstract representations of the structure of language.

The effects of this cross-linguistic transfer via metalinguistic abilities and more particularly via phonological awareness, given its significant role in the acquisition of reading, were demonstrated in some studies. A few of them established correlations in bilingual children between phonological awareness in English and in Spanish (Lindsey et al. 2003) or French (Comeau et al. 1999). As for syntactic awareness, Demont (2001) highlighted the fact that the bilingual experience has a positive influence on children’s performances in judgment tasks and/or the correction of grammatical mistakes. Furthermore, once it has been developed in L1, phonological awareness can be transferred to facilitate the development of reading proficiency in L2, though this transfer depends of course on the link between the writing systems of the two languages (Geva and Siegel 2000). Therefore, the advantage of bilingualism could also be found in the command of the written language (Bialystok et al. 2005; Deacon et al. 2007). However, researchers do not always reach a consensus about the effects of cross-linguistic transfer. For instance, Besse et al. (2019) examined the transfer of the morphological awareness of L1 Arabic to the morphological awareness and reading of L2 French words (FL2). One-hundred-and-six Tunisian pupils enrolled in 3rd (8–9 years old), 4th (9–10 years old) and 5th (10–11 years old) years in state primary schools in Tunis participated in this study. All of them have Arabic as their first language, are neither behind nor ahead at school and have been learning French since their third year of school. Third-year pupils are thus in their first year of French classes, fourth-year pupils are in their second year and fifth-year pupils are in their third year. Their reading and inflectional and derivational morphological awareness levels were measured in the two languages with the help of one-minute word reading tests and odd-one-out tasks (inflectional and derivational). Hierarchical regression analyses show that the contribution of L1 morphological awareness to L2 morphological awareness only appears in the second year and its contribution to reading in FL2 in the third year. According to the authors, the morphological transfer—notably derivational—between orthographically and morphologically distant languages is relatively limited once a minimum level in FL2 has been reached and prior to the intervention of French-specific skills.

Thus, the effects of cross-linguistic transfer are strong arguments justifying the teaching of local languages or heritage languages at school besides French. The next section briefly presents seven educational systems whose effectiveness on language development and academic success were evaluated in our work: (1) “dispositifs langues et culture kanak-LCK” for the Kanak language and culture in New Caledonia; (2) “langues et culture polynésiennes-LCP” for Polynesian languages and culture; (3) “ECOLe Plurilingue Outre-Mer-ECOLPOM”, plurilingual school for overseas French territories; (4) “Reo mā’ohi au Cycle 3-ReoC3” for Tahitian; (5) “Intervenant en Langue Maternelle-ILM” in French Guiana; (6) convergent teaching method in Haiti; and (7) “programme Initiative École et Langues Nationale en Afrique-ELAN-Afrique”, National School and Languages Initiative in Africa).

3. Educational Systems That Valorize Heritage Languages

These bilingual educational systems are implemented in 12 countries (New Caledonia, French Polynesia, French Guiana, Haiti, Benin, Burundi, Burkina Faso, Cameroon, Mali, Niger, Democratic Republic of Congo and Senegal) that have different political, educational and sociolinguistic contexts. Among the differences between these countries, we can mention linguistic diversity (varying from less than a dozen languages to hundreds of languages depending on the country), with local languages presenting varying degrees of vitality3 (ranging from good vitality for African languages, Creole and some French Guiana languages to feeble vitality for Oceanian languages). The experience of teaching these languages at school also varies, from countries that have over thirty years of experience to others that have just started bilingual teaching. Moreover, since language policies are linked to the countries’ political status, several cases can be identified (independent countries and countries in overseas French territories that have the COM (Overseas Territories Collectivity) status). Lastly, the learners’ language profiles can vary (from French monolingualism to non-French monolingualism), which means the goals of these educational systems vary as well (heritage valorization, preservation of endangered local languages, understanding between communities, proficiency in French, etc.).

However, in spite of this heterogeneity, these countries share a number of common traits: the French language is one of the official languages of the country, the period of francization and banning of local languages at school remains relatively recent and left its mark in the communities’ memory. French is one of the main languages of instruction, and the only one in some cases, even though it is not necessarily the only language of socialization of children. Local languages are most often undervalued compared to French, which is considered a prestigious language. While there is now more and more children’s literature available in the local language, at the time of this study, it was relatively rare and few families bought books written in local languages. In addition, the economic situation and schooling conditions are comparatively less favorable than in mainland France, with a higher rate of academic failure. Lastly, the educational systems of these countries are very similar to the system in mainland France.

For over twenty years, the authorities in charge of education in these countries have implemented bilingual teaching systems (see, among others, Fillol and Vernaudon 2004 for New Caledonia; Paia 2014 for French Polynesia; Alby and Léglise 2014 for French Guiana; Nocus 2013 for Haiti; Maurer 2016 for French-speaking Sub-Saharan African countries). The various systems we studied are all bilingual teaching programs that introduce or reinforce the L1 or the local language. However, their organization varies in terms of time (from three hours a week to parity with French), the pupils these programs are intended for (all pupils, only the pupils whose heritage language is the language taught, non-French-speaking pupils, etc.), the resource persons in charge of the teaching (external actor, extra teacher or bilingual classroom teacher) and their content (the subjects taught in the local languages). The ILM system in French Guiana and the convergent teaching method in Haiti use the students’ L1 as a temporary resource for the exclusive teaching of the dominant language. Once the children have acquired solid knowledge of French, they are meant to leave the program. The LCK, LCP, ECOLPOM and ReoC3 systems are aimed more specifically at passive L1 speakers and their goal is to help language minority learners to consolidate their skills in their L1. Lastly, the ELAN-Afrique program aims at enriching the learners’ educational experience by strongly supporting bilingualism and biliteracy.

In the present article, we focus on three programs: LCK in New Caledonia, ECOLPOM and ReoC3 in French Polynesia.

LCK in New Caledonia was tested from 2002 to 2004. Eight trainee teachers, holders of a degree in regional languages and cultures, speakers of Drehu, Nengone, Ajië and Xârâcùù4, taught in ten schools, with a total of 210 students, in kindergarten and Grade 15. LCK trainees taught either half the class group or at some level in multi-grade classes. The session, generally lasting one hour a day, for 90 h a year, took place exclusively in the Kanak language, in a place other than the main classroom, in the presence only of the LCK teacher. The teacher stayed in his classroom with the other children. The skills, to be acquired at the end of kindergarten, were the same as in the French kindergarten curriculum (language at the heart of learning, living together, acting and expressing oneself with one’s body, discovering the world, sensitivity, imagination and creation; Ministère de l’Éducation Nationale 2002). Cultural skills were added, for example, making traditional toys (windmill, braided bird, etc.), corresponding to the main customary events of the year.

In French Polynesia, in 2005, the Department of Primary Education (DEP) experimented with a first system “teaching Polynesian languages and culture in the primary schools of French Polynesia” (LCP program), which aimed to strengthen, exclusively orally, the place of Polynesian languages in kindergarten at the rate of five hours per week. About 1500 students from about twenty schools participated in this program in the following languages: Tahitian, Marquesan, Paumotu, Raivavae, Tupuai and Magareva6. Subsequently, the DEP set up for two years (2009–2011) an experimental system, ECOLPOM (for multilingual school Overseas), for Grade 1 (6–7 years old) and Grade 2 (7–8 years old), which aimed to strengthen oral skills and the acquisition of written proficiency in Tahitian. Seven tenured teachers from Tahiti or Moorea provided this teaching in their classrooms, one hour a day, five days a week. They received a six-week training course in the didactics of Oceanic languages, plurilingual development and activities focused on preparation for entry into writing in both languages (development of metalinguistic skills). Then, from Grade 3 (8–9 years old), the DEP continued to strengthen the teaching of Tahitian via the ReoC3 program (2011–2014). The objectives were similar to those of ECOLPOM but adapted to the academic level of the students. A logbook filled in by the teachers made it possible to describe their pedagogical practices in the 11 disciplinary fields likely to be worked on in Tahitian during the five hours per week (oral language and poetry, reading, production of writings, musical education and arts, mathematics, etc.). It showed that, on average, a teacher tackled four different fields each week, regardless of the school level, and that the three fields worked on almost every week were oral Tahitian, reading and written production. Mathematics came in 4th place and the other subjects (science, technology, history, geography, etc.) were little worked on.

The purpose of the next section is to synthesize the methodological principles and main results of the assessments we were in charge of conducting.

4. Methodological Approach and Main Results of the Two Studies

Our methodology was guided by several principles. One or two languages was selected for each country in order to obtain large enough samples of pupils that would make the statistical analyses reliable. Obtaining evidence of the added value of those systems involved comparing pupils who took part in them (experimental group) with pupils who did not and followed a “regular” curriculum (control group). In order to compare the two groups, we needed to make sure they were equivalent, at least at the start of the study, with regard to a number of sociodemographic variables and to their language profiles. A questionnaire was sent to the families and refined throughout the course of the studies in order to collect this information. Moreover, since all the pupils were emergent bilinguals, the effect of the systems was supposed to vary according to the stage they had reached in their language development. It was therefore necessary to a have a longitudinal follow up with at least two measurement times in the year. For the ECOLPOM and ReoC3, we went as far as to include six measurement times over five years. Tests were adapted in local languages with the help of colleagues who spoke those languages. The assessment battery comprised standardized pre-validated French tests that could be easily adapted into the local languages in order to obtain a bilingual assessment method (local languages and French). The psychometric properties of the tests were controlled since the standardizations are only designed for mainland French pupils, while the items are not always adapted to the living contexts of non-mainland French children. The protocols were designed not to overload the children in accordance with the code of ethics of psychologists. The children took the tests over several sessions of 15 to 25 min, depending on their age, and were never evaluated more than once a day.

4.1. Effects of Teaching in the Kanak Language on Oral/Written Skills in French in Kindergarten and Grade 1 in New Caledonia—LCK Program (Nocus et al. 2007)

4.1.1. Objective and Participants

The objective was to show the effect of teaching Kanak languages and culture (LCK program) on the language skills in Drehu (local language) of kindergarten students and on their first acquisitions in the mastery of oral and written French (second language). It should be noted that, at that time, school curricula did not provide for the teaching of reading and writing in local languages. A total of 61 children in the second year of kindergarten (mean age: 4.4 years) and 82 children in the third year of kindergarten (mean age: 5.3 years) followed in Grade 1 were evaluated orally and in writing, either as an experimental group (having participated in the teaching) or as a control group (not having participated). Only 63 of the 82 third year of kindergarten pupils could be followed in Grade 1: 29 (14 girls and 15 boys) belonged to the control group and 34 (18 girls and 16 boys) to the experimental group. At each grade level, the two groups (experimental and control) had statistically equivalent non-verbal cognitive skills. These were measured in the second year of kindergarten using the WPPSI-R square test (Wechsler 1995) and in the third year of kindergarten using the “Perception, Analogies and Spatial Manipulations” (PAMS) test of Khomsi (1992). Data from a questionnaire sent to parents specifying the socio-demographic characteristics of families and their language practices showed that the control and experimental groups had comparable sociological profiles. In these families, Drehu was practiced to varying degrees. However, family language practices did not differ according to the type of group (experimental versus control). The parents of the children of both groups tended to communicate with each other in Drehu and French, and sometimes in a second Kanak language.

4.1.2. Measures and Procedure

Several language tests in French and Drehu were proposed to children of both school levels (second year of kindergarten and third year of kindergarten), at the beginning and at the end of the year.

Language evaluations in French: five tests from the Oral Language Assessment (ELO, Khomsi 2001) were administered individually. These tests covered the receptive and expressive aspects of language and were calibrated from the first year of kindergarten to Grade 5. The tests were as follows: receptive vocabulary (choice of one image among four from a proposed object name; mark out of 20), expressive vocabulary (naming of images corresponding to objects or actions; mark out of 30), production of utterances (completion of a sentence proposed by the evaluator with the correct morphosyntactic form, mark out of 16), repetition of utterances (repetition of sentences respecting the morphological and syntactic forms of the utterance; mark out of 15) and comprehension of utterances (choice of one image among four, which corresponds to a sentence given orally; mark out of 20).

Drehu language assessments: the five previous tests were adapted into Drehu, with the help of Kanak speakers. The objective was not to translate the French exams into Drehu, but rather to obtain equivalent indicators in language production and comprehension. Accordingly, some words used in the lexical and morphosyntactic tests were chosen from those present in the children’s linguistic environment. For example, words such as “snow” or “train”, for which there is no equivalent in Drehu, were avoided and more local names or animals existing in Drehu, such as “hut”, “canoe”, “coconut crab” and “conch”, were added. The items and instructions were built on the ELO model taking into account the age of the children. During the tests, they were presented in Drehu. The marks were the same as in French tasks (receptive vocabulary, mark out of 20; expressive vocabulary, mark out of 30; production of utterances, mark out of 16; repetition of utterances, mark out of 15 and comprehension of utterances, mark out of 20).

Phonological awareness: this was evaluated, in French, in the third year of kindergarten and Grade 1 by a rhyme identification test (mark out of 5) from a tool bank (Ministère de l’Éducation Nationale 2001). It aimed to understand the students’ skills in a field strongly linked to the mastery of reading and writing (Gombert 1990).

Recognition of letters, words and non-words: the test was intended only for third year of kindergarten students and at the end of the year only. The objective was to assess, in French, the students’ first knowledge of writing, which is known to play an important role in the rest of schooling (Le Bastard and Suchaut 2000). Three subscores were calculated according to the type of items (mark out of 6 for letters, out of 4 for words and out of 4 for non-words).

The assessments proposed at the entrance to Grade 1 were adapted from the national system of evaluation and learning assistance in the last year of kindergarten and Grade 1 (Ministère de l’Éducation Nationale 2001). They included sixteen tests divided into four disciplines (transversal skills, oral language, reading and writing) and eight fields (finding one’s way in time, finding one’s way in space, mastering language in a situation, etc.).

Transversal skills were addressed by two tests, one assessing notions of time (mark out of 7), the other notions of space (mark out of 5). The overall score (out of 12) corresponded to the sum of the scores in these two tests.

Oral language was assessed by two tests of comprehension of instructions (overall mark out of 13), two tests of pragmatics of language (overall mark out of 12) and two tests of evocative language (overall mark out of 5). The overall score (out of 30) included the sum of the performances in the different language sub-tests.

In prereading, five tests were proposed: two tests of phonological awareness (production of rhymes; mark out of 5) and the identification of phonological units (mark out of 8), two tests of the segmentation of the written and oral language (judgement of the length of the sound chain of a word or an utterance; mark out of 7) and a test of familiarization with the world of writing (identification of the pages of a book, groups of written words, etc.; mark out of 19). An overall reading score (out of 39) summarized all the performances.

In writing, the evaluation included a graphic design test (producing a non-random continuous line in a given space; mark out of 10) and two writing tests: production of capital letters (out of 6) and writing of the first name in capital letters and cursive (out of 3). An overall score (out of 19) summarized the writing performance.

In the second year of kindergarten and third year of kindergarten, the pupils were assessed at the beginning of the school year, before the start of classes in the Drehu language and then at the end of the year, after 55 h of LCK teaching. At the beginning and end of the second and third years of kindergarten, all students were evaluated in Drehu by two examiners who spoke this language and in French by a French-speaking examiner. The different tests were administered in different sessions with an average duration of 15 to 20 min each. In Grade 1, the evaluation took place at the beginning of 2005. It was administered in French by the teachers of the class in accordance with the guidelines for the examination.

4.1.3. Results

Statistical Analysis

For each of the language tests and for each grade level (second year of kindergarten versus third year of kindergarten), repeated measures analyses of variance were performed according to the S < G2 > *P2 design (dependent variable: score obtained in the test; independent variables: session P2 [beginning and end of year] and group G2 [control versus experimental]). In Grade 1, analyses of variance were carried out on the scores for the four disciplines (transversal skills, oral language, reading and writing) according to the S < G2 plan > (dependent variable: score obtained in the discipline; independent variables: the G2 group [control versus experimental]). Step-by-step regression analyses were also carried out to demonstrate the transfer effect of the practice of Drehu on French performance in Grade 1, by controlling the non-verbal cognitive level of the students. The variables to be predicted were the scores for the four disciplines assessed in Grade 1 and the predictor variables are the scores on the third year of kindergarten cognitive test entered as the first step and the scores of each of the tests in second year of kindergarten at the end of third year of kindergarten entered as the second step.

Effects of LCK Program on the Language Skills in French and Drehu

Section 2 of

Table A1. Means, standard deviation and pass rate (in %) in second year of kindergarten obtained in the French and Drehu language tests by session (beginning of year vs. end of year) and type of group (control vs. experimental), Fisher’s F value and significance.

	Beginning of the Year		End of Year		ANOVA
Tasks	Control N = 33	Experimental N = 28	Control N = 33	Experimental N = 28	Session	Session * Group
Section 1: French language
Receptive vocabulary (/20)	9.5 (2.3) 47.5%	8 (2.0) 40%	10.8 (2.3) 54%	10.7 (2.8) 53.5%	F(1, 59) = 547.1, p < 0.0001; End > Begin	F(1, 59) = 5.8, p < 0.01; E > C
Expressive vocabulary (/30)	13.4 (4.2) 44.5%	9.6 (6.6) 32%	16.5 (3.7) 55%	13.6 (6) 45.5%	F(1, 59) = 126.7, p < 0.0001; End > Begin	F(1, 80) < 1 Ns
Comprehension of utterances (/20)	13.4 (2.4) 67%	12.54 (3.5) 62.5%	15.6 (2.4) 78%	15.3 (2.7) 76.5%	F(1, 59) = 60.3, p < 0.0001; End > Begin	F(1, 80) < 1 Ns
Production of utterances (/16)	2.5 (2.4) 15.5%	2.3 (2.2) 14.5%	3.36 (2.7) 21%	3.46 (2.9) 21.5%	F(1, 59) = 19.9, p < 0.001; End > Begin	F(1, 80) < 1 Ns
Repetition of utterances (/15)	4.8 (3.9) 32%	4.9 (4.8) 32.5%	5.75 (3.7) 35.5%	5.71 (4.5) 38%	F(1, 59) = 8, p < 0.01; End > Begin	F(1, 80) < 1 Ns
Section 2: Drehu Language
Receptive vocabulary (/20)	9.2 (3.4) 46%	10.7 (3.4) 54%	9.8 (3.5) 49%	12 (2.6) 60%	F(1, 59) = 4.5, p < 0.05; End > Begin	F(1, 80) < 1 Ns
Expressive vocabulary (/28)	3.8 (4.9) 13.5%	6.7 (5.9) 24%	5.6 (5.3) 20%	10.5 (5.5) 37.5%	F(1, 59) = 35, p < 0.0001; End > Begin	F(1,59) = 4.2, p < 0.05; E > C
Comprehension of utterances (/20)	10.9 (3) 54.5%	13 (2.3) 65%	13.6 (3.8) 68%	15.6 (2.9) 78%	F(1, 59) = 46.8, p < 0.0001; End > Begin	F(1, 80) < 1 Ns
Production of utterances (/10)	0.8 (1.3) 8%	1.7 (1.6) 17%	1.3 (2) 13.5%	3.7 (2.5) 37.5%	F(1, 59) = 33, p < 0.0001; End > Begin	F(1, 59) = 11.3, p < 0.001; E > C
Repetition of utterances (/7)	0.6 (0.9) 8.5%	0.9 (1.1) 13.5%	0.5 (0.9) 7%	2.8 (2) 40%	F(1, 59) = 20, p < 0.0001; End > Begin	F(1, 59) = 25.4, p < 0.0001; E > C

Notes: End = end of year; Begin = beginning of the year; C = control group; E = experimental group; Ns = not significant.

and

Table A2. Means, standard deviation and pass rates (in %) in third year of kindergarten obtained in the French and Drehu language tests by session (beginning of year vs. end of year) and type of group (control vs. experimental), Fisher’s F value and significance.

	Beginning of the Year		End of Year		ANOVA
Tasks	Control N = 45	Experimental N = 37	Control N = 45	Experimental N = 37	Session	Session * Group
Section 1: French language
Receptive vocabulary (/20)	10.6 (2.5) 53%	10.6 (1.7) 53.5%	12 (2.6) 60%	11.4 (1.8) 57%	F(1, 80) = 12.8, p < 0.001; End > Begin	F(1, 80) < 1 Ns
Expressive vocabulary (/30)	15.2 (3.4) 51%	13.4 (4.1) 45%	18.6 (4) 62%	16.5 (3.4) 55%	F(1, 80) = 148.3, p < 0.0001; End > Begin	F(1, 80) < 1 Ns
Comprehension of utterances (/20)	14.1 (2.8) 71%	14.8 (2.8) 74.5%	16.7 (2.7) 83.5%	15.8 (2.4) 79.5%	F(1, 80) = 23.2, p < 0.0001; End > Begin	F(1, 80) = 3.6 Ns
Production of utterances (/16)	4 (2.7) 25%	3.8 (2.3) 24%	5.2 (3.2) 32.5%	3.9 (3.3) 24.5%	F(1, 80) = 4.9, p < 0.05; End > Begin	F(1, 80) = 2.5 Ns
Repetition of utterances (/15)	8.1 (3.6) 54%	7.2 (3.7) 48%	9 (3.4) 60%	8.5 (2.9) 57%	F(1, 80) = 11.9, p < 0.001; End > Begin	F(1, 80) < 1 Ns
Phonological awareness (/5)	2.4 (1) 48%	2.2 (1.1) 45%	2.9 (1.3) 58.5%	3.1 (1) 62%	F(1, 80) = 16.1, p < 0.0001; End > Begin	F(1, 80) < 1 Ns
Section 2: Drehu Language
Receptive vocabulary (/20)	10.7 (4.1) 54%	13 (3.4) 65.5%	11.5 (3.5) 57.5%	14.8 (2.3) 74.5%	F(1, 80) = 16.9, p < 0.0001; End > Begin	F(1, 80) = 1.9 Ns
Expressive vocabulary (/28)	6.4 (6.1) 23%	12.6 (5) 45%	8.2 (6.7) 29.5%	16.3 (4.3) 58%	F(1, 80) = 51.7, p < 0.0001; End > Begin	F(1, 80) = 4.1, p < 0.05; E > C
Comprehension of utterances (/20)	11.6 (3.1) 58.5%	13.2 (1.9) 66%	12.7 (3.5) 64%	14.8 (1.7) 74%	F(1, 80) = 15.2, p < 0.001; End > Begin	F(1, 80) = 0.3 Ns
Production of utterances (/10)	2.5 (2.5) 25%	4 (2.4) 40%	2.7 (3.2) 27.5%	6.2 (2.1) 62%	F(1, 80) = 34.1, p < 0.0001; End > Begin	F(1, 80) = 14.4, p < 0.001; E > C
Repetition of utterances (/7)	1.5 (1.8) 22.5%	2.2 (1.6) 32.5%	1.8 (1.8) 26%	4.4 (1.7) 63.5%	F(1, 80) = 41.7, p < 0.0001; End > Begin	F(1, 80) = 23.5, p < 0.0001; E > C

Notes: End = end of year; Begin = beginning of the year; C = control group; E = experimental group; Ns = not significant.

(Appendix A) shows that both in the second year of kindergarten and in third year of kindergarten the progress of the experimental group is always statistically superior to that of the control group for all the tests of production in Drehu (vocabulary, production of utterances and repetition of utterances). On the other hand, we could not conclude that there was an effect on French, in the second and third years of kindergarten, because the students in the experimental groups obtained, on the whole, language performances equivalent to those of the children in the control groups (See Section 1 of Table A1 and Table A2 in Appendix A). The results (

Table A3. Mean, standard deviation and pass rate (%) in third year of kindergarten obtained in the letter, word and non-word recognition tests by group type (control vs. experimental), Fisher’s F value and significance.

Tasks	Control N = 40	Experimental N = 36	ANOVA
Letters (/6)	2.9 (1.9) 48.5%	4.1 (1.4) 70%	F(1, 74) = 10.4 p < 0.001; E > C
Words (/4)	1 (1.1) 26%	1.7 (1.1) 43.5%	F(1, 74) = 7.5 p < 0.01; E > C
Non-words (/4)	1.2 (1.2) 30.5%	1 (1) 27%	F(1, 74) = 0.2 Ns

Notes: C = control group; E = experimental group; Ns = not significant.

in Appendix A) show the clear superiority of the experimental group compared to the control group for the overall score, the “letter recognition” items and the “word recognition” items. On the other hand, the two groups do not differ significantly in “non-word recognition”, due probably to the inefficiency of the alphabetical procedure, at the end of the third year of kindergarten. Consequently, the pupils who have benefited from the program have, overall, better knowledge in the field of writing than those who have not benefited from it.

Effects of LCK Program on the Grade 1 Assessment

If we consider the overall results for the Grade 1 assessment (

Table A4. Average, standard deviation and success rate (%) achieved in disciplines as a function of group type (control vs. experimental), Fisher’s F value and significance.

Disciplines	Control (N = 29)	Experimental (N = 34)	ANOVA
Transversal skills /12	6.8 (1.9) 57%	9.3 (1.9) 78%	F(1, 61) = 20.48, p < 0.0001; E > C
Oral language /30	14.9 (4.2) 50%	17.4 (4.9) 58%	F(1, 50) = 4, p = 0.0509
Prereading /39	23.2 (6) 60%	28.1 (4.1) 72%	F(1, 61) = 11.167, p = 0.0016; E > C
Writing /19	11.6 (3) 61%	13.2 (3.3) 70%	F (1, 61) = 9.291, p = 0.0037; E > C

Notes: C = control group; E = experimental group.

in Appendix A), the experimental group has superior results to the control group in three of the four areas in French (transversal skills, reading and writing). On the other hand, this difference is only tendential for oral language.

Contribution of the Language Tests in Drehu, in the Third Year of Kindergarten, to the Grade 1 Assessment Tests after Control for the Cognitive Variable

Table A5. Significant contribution of each of the third-year-of-kindergarten Drehu language tasks extracted from the simple regressions to the different tests of the Grade 1 assessment, after control of cognitive skills (R2 in %).

		Variables to Predict Grade 1
	Predictor Variables Third Year of Kindergarten	Transversal Skills	Oral Language	Reading
1er pas:	Score au PAMS	0.0003 ns	0.002 ns	2.6 ns
2ème pas	Receptive vocabulary
	Expressive vocabulary			18.4 **
	Comprehension of utterances			13.7 *
	Production of utterances	9.7 *		15.1 **
	Repetition of utterances			18.8 **

Note: * p < 0.05. ** p < 0.01. In bold: significant contributions to p < 0.01; ns = not significant.

(in Appendix A) shows the absence of a specific contribution of the cognitive test (PAMS) to the various Grade 1 tests. In addition, no third-year-of-kindergarten language test predicts oral language skills in Grade 1. However, only the utterance production test helps to explain the overall score on the tests measuring transversal skills (R² = 9.7%, p < 0.05). On the other hand, as far as reading in Grade 1 is concerned, the explanatory part of each of the third-year-of-kindergarten Drehu language tests is strong, with all the tests, except the receptive vocabulary, making a significant contribution. The repetition tests of utterances and expressive vocabulary are those that best predict reading performance in Grade 1 (respectively, R² = 18.8% and R² = 18.4%, at the threshold of p < 0.01). The tests of immediate comprehension, global comprehension and production of statements make significant contributions of 12.6%, 13.7% and 15.1%, respectively.

4.1.4. Conclusions

At the end of this study, it was possible to conclude that the LCK program had a positive effect on the acquisition of spoken Drehu, both in the second and third years of kindergarten. On the other hand, it could not be concluded that there was a positive effect on French because the pupils in the experimental groups presented, on the whole, language performances equivalent to those of the children in the control groups. Regression analyses showed, however, that in Grade 1, performance in reading in French is strongly predicted by performance in third-year-of-kindergarten Drehu oral language tests. Among the explanations put forward, it should be noted that work on bilingualism shows that second-language proficiency is linked to mastery of the mother tongue, when the latter is well established (Cummins’ (2000) hypothesis).

However, these results were partly replicated in French Polynesia with the LCP program (Nocus et al. 2012) with 125 children divided into an experimental group (partially schooled in Tahitian at a rate of 300 min per week) and a control group (schooled in French) and followed for 3 years from the second year of kindergarten to Grade 1. The methodology was identical to that used in the LCK program, with the addition of reading tests in Tahitian at Grade 1. The results showed that the program appeared to have a positive impact on Tahitian proficiency, without any negative effect on the learning of French (oral and reading). Nevertheless, the bivariate Pearson correlations and the hierarchical regression analyses calculated among all the Tahitian and French oral and written tests for the last test session according to the group showed that there was no cross-linguistic correlation between oral tests and reading tests. Two reasons could explain this phenomenon. Firstly, the program could only be effective in developing superior cognitive skills in French if it was followed over several years. Secondly, we thought that regression analyses revealed no transfer effects, probably because the language tests we used did not measure metalinguistic awareness directly. Therefore, cross-linguistic transfer is only possible when children have access to tasks that require metalinguistic skills, such as reading and writing, in both languages. This is the strong argument that we defended in the context of the ECOLPOM program carried out from Grade 1 to Grade 2 in New Caledonia, French Polynesia and French Guiana. In this program, the children in Grade 1 learned to read and write in local languages. In the next section, we present the results obtained in French Polynesia because it is the only territory where ECOLPOM continued until Grade 5 under the ReoC3 program, which allows us to observe the effects of bilingual teaching in the same students over a period of 5 years.

4.2. ECOLPOM and ReoC3 in French Polynesia (Nocus et al. 2018)

4.2.1. Objective and Participants

This study examined the specific link between bilingualism and morphological awareness and the cross-linguistic effects of Tahitian–French transfer via the written word. We conducted a five-year longitudinal study on 128 Polynesian pupils we followed during six evaluation sessions (beginning and end of Grade 1, end of Grade 2, end of Grade 3, end of Grade 4 and end of Grade 5) from 2009 to 2014 in Tahiti and Moorea in French Polynesia. A questionnaire survey forwarded to families showed that the dominant language of these pupils was French and Tahitian was their heritage language, which they did not actively speak at the beginning of the study. Among them, 59 had taken part since Grade 1 in two consecutive educational programs (ECOLPOM and ReoC3) offering five hours a week of reinforced teaching of spoken and written Tahitian (experimental group) and 69 others did not take part in it, but could benefit from two hours and forty minutes of Tahitian lessons as provided for by Polynesian school curricula (control group). At the start of the study, the pupils from the two groups had a comparable level in French and Tahitian in addition to the usual sociodemographic variables.

4.2.2. Measures and Procedure

All students were tested each year on their oral language skills in French and Tahitian. As in the previous LCK study, an expressive vocabulary test and a morphosyntactic test from Khomsi’s (2001) Oral Language Assessment (ELO) battery were used (see above). In Tahitian, the vocabulary and morphosyntax tests consisted of 30 and 15 items, respectively. The maximum scores were 30 for the first and 15 for the second. At the end of Grade 3, a morphological awareness assessment in French was introduced (Reder et al. 2013). The three tests developed were:

• An odd-one-out detection test: the student listened to 16 series of three words composed of two complex words (containing a root + an affix) and a pseudo-derived word (simply containing a root, beginning or ending with the same combination of letters as those making up the affix of the other two words). The student had to find the word that looked like a complex word, but was not really derived (e.g., in “grandeur [size], fraîcheur [coolness] et humeur [mood]”, you have to find humeur [mood]”). The maximum score was 16 (1 point per correct answer).
• A test to extract the root of derived words: in this 16-item test, the student had to extract the root of a word presented in a derived form (e.g., “If I say ‘maisonnette’ [little house], what little word from the same family do you hear inside?”). Half of the items were existing words in French and the other half were neologisms (e.g., décopier [neologism] → copier [to copy]). The maximum score was 16.
• A test for the production of derived words: in this 16-item task, the student must produce a derived form from a root (e.g., “What would you call the gentleman who dances? It’s a …”). The first 8 items were words existing in French, the following 8 items were pseudo-words (e.g., «Une petite poire, c’est une … (poirette)» “[a small pear is a …]” in French, “ette” means “small”). The maximum score was 16. Morphological awareness was therefore assessed on a total of 48 points.

These three tasks were adapted into Tahitian, also scored out of a total of 48 points:

• An odd-one-out detection test: The student had to find the odd one out who corresponded either to a “real” odd one out, in the sense that it could not be segmented into smaller morphemes (e.g., “t¯arifa” [price] is not a derivative of “rifa”, by prefixing with “t¯a”). The maximum score was 16 (1 point per correct answer).
• A test to extract the root of derived words: In Tahitian, the child had to extract the root of a Tahitian word presented in a derived form (e.g., in “fa’ainu” means “getting someone to drink” and « inu » means “drink”). Half of the items were existing words in Tahitian, the other half were neologisms. The maximum score was 16 (1 point per correct answer).
• A test for the production of derived words: The child had to produce a form derived from the word presented to him (e.g., “’Ia fa’ariro’ei mea’ino, e …” [When you return in bad condition, it’s …], expected response: fa’a’ino [deteriorate, damage]). The maximum score was 16 (1 point per correct answer).

The identification of the written word was evaluated with two tests according to the school level. In Grade 1 and Grade 2, the pupils were assessed in reading with the Timé2 (Ecalle 2004), which has three tasks of 12 items: recognize a word provided orally by the evaluator (e.g., for the image of a hat, find the written word “chapeau” [hat] among four distractors: “chapeou”, “chapo”, “chameau” [camel] and “cpaheua”); recognize a written word corresponding to an image (e.g., find the word that corresponds to the image “enfant” [child] among the distractors “etnfan”, “enfonce” [push in/drive in], “eufant” and “enfan”) and associate a word that is semantically or pragmatically close (e.g., find the word that corresponds to the written word “branche” [branch] among “arbitre” [referee], “ardre”, “arebre”, “arbre” [tree] and “arrbe”). In the three tasks, the child had to circle the expected word from five items, including four distractors (a homophone, a visually similar pseudo-word, an orthographic neighbor7 and a pseudo-word with an illegal sequence of letters). The maximum score was 36 (1 point per correct answer). The Tahitian version had the same three tasks of 10 items each. The 4 distractors were pseudo-homophone words of the target word (e.g., haere [go]) whose spelling is close to French (e.g., aéré), a visually similar pseudo-word (e.g., haepe), an orthographic neighbor (e.g., haehae [toe]) and a pseudo-word with an illegal sequence of letters (e.g., hraee). The maximum score was 30. From Grade 3, Timé3 (Ecalle 2006) was used in French and adapted into Tahitian. Two tasks consisting of finding the target item in the middle of 4 distractors were proposed: finding the word corresponding to the image and associating two words related semantically. The maximum score was 40 in French and Tahitian.

Reading comprehension was assessed with a test from the D-OR-LEC battery (Lobrot 1980). The children had to silently read 36 isolated sentences and to indicate the missing word from a list of 5 words (e.g., «Prends le panier et va me chercher des … armoires, oranges, ordures, ombres, ordres » [“Take the basket and go get me some … cupboards, oranges, garbage, shadows, orders”]). The maximum score was 36. The Tahitian version has 26 sentences (e.g., “E mea au n¯a′u te meri, e mea … mohimohi, tuhituhi,’ava’ava, monamona,’uo’uo” [“I love honey, it’s … indistinct/cloying/bitter/sweet/white”]) and the maximum score was 26.

Finally, the spelling test was taken from Khomsi’s (1998) ECS3. The test was composed of a 30-item “written image naming” test for lexical spelling (OL, e.g., write the word “bol” [bowl] under the image of a bowl) and a 6-item sentence completion test for grammatical spelling (OG, e.g., complete in writing the second sentence “First a bear sleeps. Then two s …”). The procedure and instructions for taking the test are identical for the Tahitian test (e.g., writing the word “honu” [turtle] under the image of a turtle) and 6 items for OG (e.g., “I’¯o nei, n¯a te pere’o’o uira e t¯ura’i i te pere’o’o uta tauiha’a. I’¯o, n¯a …” [“There, the car pushes the truck. Here is the truck …”]). The maximum score is 42 (OL: 30 and OG: 12) for French and 36 (OL: 30 and OG: 6) for Tahitian.

During the first evaluation session at the beginning of Grade 1 (October–November), the reasoning test was administered in small groups (6 to 8 students), and then all students individually took the expressive vocabulary and statement production tests during a 15–20 min session for each language. At the end of Grade 1 and Grade 3 (April–May), the students were evaluated individually on the same oral tests, then in small groups on the identification of written words (Timé 2) in French and Tahitian in two 20-min sessions. From the end of Grade 4 (April–May), the same oral tests in French and Tahitian and the Timé3 in French and Tahitian were presented in three individual and semi-group sessions of about 15–20 min. Morphological awareness tests in both languages were added over two sessions of 20 min each as an individual test. The reading comprehension and spelling tests in French and Tahitian were administered during semi-group sessions of 15 min each. The fifth and sixth evaluation sessions were carried out at the end of Grade 4 and Grade 5 (April–May) according to the same methodology as that used at the end of Grade 5.

4.2.3. Results

Statistical Analysis

Multivariate analyses of variance (MANOVA) and, when necessary, multivariate analyses of covariance (MANCOVA) were performed on the three measures of oral language in Tahitian (expressive vocabulary, utterance production and morphological awareness), on the three measures of written language in Tahitian (identification of written words, reading comprehension and spelling) and on the same measures in French. Then, cross-linguistic transfer and the contribution of Tahitian to reading and spelling were tested with correlation analyses (not shown here due to lack of space) and multiple regression analyses. The results are presented in Appendix B.

Effect of the ECOLPOM and ReoC3 Programs on Performance in Tahitian

In Tahitian, for the expressive vocabulary and utterance production tasks, the MANOVA results show a significant effect on the group, respectively, F(6, 119) = 17.062, p < 0.001 and F(6, 119) = 6.845, p < 0.001, with the univariate tests for each evaluation session showing a superiority of the experimental group over the control group from the end of Grade 2 (for both tasks, see

Table A6. Mean (%) for language assessments in Tahitian according to sessions and groups.

	Groups
Sessions	Control (N = 69)	Experimental (N = 59)	F	p
Section 1: Expressive vocabulary
1. Grade 1, beginning	8	10	1.329	ns
2. Grade 1, end	10.5	11.5	0.250	ns
3. Grade 2, end	11.5	24.5	22.791	0.000
4. Grade 3, end	18	37	37.483	0.000
5. Grade 4, end	23	46.5	46.176	0.000
6. Grade 5, end	23	50	65.526	0.000
Section 2: Production of utterances
1. Grade 1, beginning	8.5	9.5	0.126	ns
2. Grade 1, end	10.5	12	0.273	ns
3. Grade 2, end	8.5	24.5	20.121	0.000
4. Grade 3, end	18	38	20.321	0.000
5. Grade 4, end	18	38	25.614	0.000
6. Grade 5, end	24.5	48.5	27.139	0.000
Section 3: Morphological awareness
5. Grade 4, end	33 a	37.5 a	6.968	0.009
6. Grade 5, end	36.5 a	45.5 a	16.914	0.000
Section 4: Identification of written words (Timé2)
3. Grade 2, end	29 b	31.5 b	1.079	ns
Section 5: Identification of written words (Timé3)
5. Grade 4, end	32 b	31.5 b	0.000	ns
6. Grade 5, end	21 b	30.5 b	13.395	0.000
Section 6: Reading comprehension
4. Grade 3, end	11.5	14	2.054	ns
5. Grade 4, end	17.5	28	10.102	0.002
6. Grade 5, end	16.5	20	2.121	ns
Section 7: Spelling
5. Grade 4, end	14.5 c	13 c	0.504	ns
6. Grade 5, end	18.5 c	35 c	44.043	0.000

ns: Not significant. ^a Mean scores estimated after controlling for morphological awareness level in Tahitian at the end of Grade 3. ^b Mean scores estimated after controlling for the level of identification of words written in Tahitian at the end of Grade 1 (Timé2) and CE2 (Timé3). ^c Mean scores estimated after checking the level of spelling in Tahitian at the end of Grade 3.

, sections 1 and 2). For the morphological awareness test in Tahitian at the end of Grade 3, the MANOVA showed a significant effect of the groups: F(3, 124) = 16.313, p < 0.001. Univariate analyses indicated a significant difference between the end of Grade 3 groups: F(1, 126) = 25.637, p < 0.001. An ANCOVA allowing us to adjust the scores of the end of Grade 4 and the end of Grade 5 groups with the scores of the end of Grade 3 group as a covariate indicated the main effect of the group: F(2, 124) = 9.855, p < 0.001. Univariate analyses showed that the experimental group obtained significantly higher scores for morphological awareness in Tahitian at the end of Grade 4 and the end of Grade 5 than those of the control group (Table A6, section 3). The MANOVA revealed a significant effect of the group on the scores of the Tahitian Timé2: F(2, 125) = 8,680, p < 0.001. But, the two groups differed at the end of Grade 1: F(1, 125) = 8.680, p < 0.001. Once the means were adjusted for the end of Grade 1 scores, the univariate analysis indicated that there was no significant difference between the two groups at the end of Grade 2 (Table A6, section 4, Appendix B). For Tahitian Timé3, the MANOVA showed a significant group effect: F(3, 124) = 25.947, p < 0.001. Univariate analyses indicated a significant difference between the two groups at the end of Grade 3: F(1, 126) = 58.859, p < 0.001. The MANCOVA showed a main effect of F(2, 124) = 6.787, p < 0.001. According to Table A6, section 5, the two groups do not differ at the end of Grade 4, but a significant difference appears in favor of the experimental group at the end of Grade 5. For Tahitian reading comprehension, a significant group effect was observed (Table A6, section 6): F(3, 124) = 4.207, p < 0.001. Univariate analyses indicated no significant difference between the end of Grade 3 and end of Grade 5 groups, but a significant difference in favor of the end of Grade 4 experimental group. In spelling, the MANOVA showed a significant group effect: F(3, 124) = 33.230, p < 0.001. Univariate analyses indicated significant differences between the two groups at the end of Grade 3: F(1, 126) = 34.965, p < 0.001. The MANCOVA showed a significant group effect: F(4, 119) = 25.549, p < 0.001. The two groups did not differ at the end of Grade 4, but a significant difference appeared at the end of Grade 5 in favor of the experimental group (Table A6, section 7).

Effect of the ECOLPOM and ReoC3 Programs on Performance in French

For the two tasks of expressive vocabulary and production of utterances in French, the MANOVA revealed a significant group effect, respectively, F(6, 121) = 4.238, p < 0.001 and F(6, 121) = 4.238, p < 0.001. Univariate analyses (

Table A7. Mean (%) for language assessments in French according to sessions and groups.

	Groups
Sessions	Control (N = 69)	Experimental (N = 59)	F	p
Section 1: Expressive vocabulary
1. Grade 1, beginning	59	57.5	0.676	ns
2. Grade 1, end	70	63.5	11.283	0.001
3. Grade 2, end	71	67	3.833	ns
4. Grade 3, end	79	76.5	1.880	ns
5. Grade 4, end	71	79.5	9.423	0.003
6. Grade 5, end	88	86	2.342	ns
Section 2: Production of utterances
1. Grade 1, beginning	39	36	0.842	ns
2. Grade 1, end	56.5	47	6.959	0.009
3. Grade 2, end	58.5	54	2.234	ns
4. Grade 3, end	73.5	68.5	2.577	ns
5. Grade 4, end	79	73	6.168	0.014
6. Grade 5, end	85	83	0.583	ns
Section 3: Morphological awareness
4. Grade 3, end	69	70.5	0.228	ns
5. Grade 4, end	82	82.5	0.021	ns
6. Grade 5, end	90	86	3.747	ns
Section 4: Identification of written words (Timé2)
2. Grade 1, end	43	45	0.537	ns
3. Grade 2, end	54.5	52.5	0.408	ns
Section 5: Identification of written words (Timé3)
4. Grade 3, end	49.5	47.5	0.507	ns
5. Grade 4, end	62.5	58.5	2.226	ns
6. Grade 5, end	70	65.5	3.244	ns
Section 6: Reading comprehension
4. Grade 3, end	49.5	45.5	1.562	ns
5. Grade 4, end	62	59.5	0.748	ns
6. Grade 5, end	52	69	5.468	0.021
Section 7: Spelling
4. Grade 3, end	26	25.5	0.012	ns
5. Grade 4, end	42.5	37.5	2.442	ns
6. Grade 5, end	52	50	0.237	ns

Note: ns = not significant.

, sections 1 and 2) showed no significant difference between the early Grade 1, late Grade 2, late Grade 3 and late Grade 5 groups. But a significant difference appeared in favor of the control group at the end of Grade 1 and the end of Grade 4. In French, in the morphological awareness and reading identification tasks (Timé2 and Timé3, respectively), the MANOVA indicated no group effect, respectively, F(3, 124) = 2.374, ns; F(2, 125) = 960, ns and F(3, 124) = 1.559, ns. Univariate analyses confirmed the absence of differences between the two groups (Table A7, sections 3–5). In French reading, the MANOVA indicated a trend effect of the group: F(3, 124) = 2.631, p = 0.053. Univariate analyses confirmed that the two groups did not differ in the scores at the end of Grade 2 and the end of Grade 4, but that the control group obtained significantly higher scores at the end of Grade 5 (Table A7, section 6). Finally, in French spelling, the MANOVA did not show a group effect: F(3, 124) = 3.487, ns. Univariate analyses confirmed that there is no difference between the two groups (Table A7, section 7).

Cross-Linguistic Effects of Tahitian–French via the Written Word

Three hierarchical multiple regression models successively tested the impact of the control variables (block 1: non-verbal cognitive level, membership of the ReoC3 program, age), then of the performance in French in Grade 5 (block 2: expressive vocabulary, production of utterances and morphological awareness) and finally of the indicators of mastery of writing in Tahitian in Grade 5 (block 3: morphological awareness, identification of written words, reading–comprehension and spelling) on French performance in the identification of written words (model 1), reading–comprehension (model 2) and spelling (model 3). This procedure makes it possible to measure the specific impact of the indicators of written proficiency in Tahitian on those in French, independently of French skills. Finally, as the control group was familiar with Tahitian through school curricula and family, the regressions were calculated taking into account their scores.

According to

Table A8. Multiple hierarchical regressions models predicting word recognition, reading comprehension and spelling in French from scores in morphological awareness, reading and spelling in Tahitian, after controlling for non-verbal cognitive level, group membership, age and oral performance in French—Grade 5.

Grade 5	Model 1 Identification of Written Words French				Model 2 Reading-Comprehension French				Model 3 Spelling French
Predictors	β	R2	ΔR2	ΔF	β	R2	ΔR2	ΔF	β	R2	ΔR2	ΔF
Bloc 1		0.11				0.15				0.10
PAMS a	0.29 ***				0.33 **				0.23 **
Group b	−0.16				−0.21 *				−0.03
Age	NI				NI				0.21 *
Bloc 2		0.40	0.29	19.97 ***		0.34	0.19	11.67 ***		0.45	0.35	25.31 ***
Expressive vocabulary in French	0.11				0.22 *				0.18
Production of utterances in French	0.20 *				0.12				0.20 *
Morphological awareness in French	0.41 ***				0.27 ***				0.42 ***
Bloc 3		0.55	0.15	10.22 ***		0.45	0.11	6.23 ***		0.53	0.08	4.862 **
Morphological awareness in Tahitian	0.01				−0.19				0.9
Identification of words written in Tahitian	0.59 ***				0.28 **				0.37 ***
Reading-comprehension in Tahitian	−0.08				0.25 **				0.02
Spelling in Tahitian	−0.17				−0.11				−0.06

Notes: * p < 0.05; ** p < 0.01; *** p < 0.001; NI: variable not integrated into the model because of the lack of a correlation between the predictor and the criterion. ^a PAMS: non-verbal cognitive level; ^b control group = 0; experimental group = 1.

, all three models are significant and explain respectively 55%, 45% and 53% of the variance. The non-verbal cognitive level was related to all three indicators (β = 0.29, p < 0.001 for model 1; β = 0.33, p < 0.001 for model 2; β = 0.23, p < 0.001 for model 3). Contrary to what was expected, membership of the ReoC3 system did not predict performances in word reading and spelling in French, and contributed negatively but modestly to reading comprehension (β = −0.21, p < 0.05). In addition, age, which was only included in model 3, contributed modestly to French spelling performance (β = 0.21, p < 0.05). After controlling for non-verbal cognitive performance, the indicators of reading proficiency in French were essentially explained by morphological awareness in French (model 1: β = 0.41, p < 0.001; model 2: β = 0.27, p < 0.001; model 3: β = 0.42, p < 0.001). In addition, the production of utterances in French contributed modestly to word reading and spelling in French (β = 0.20, p < 0.05 in models 1 and 3). Finally, morphological awareness and spelling in Tahitian did not explain the performance associated with the three indicators of proficiency in French in Grade 5. However, after controlling for non-verbal cognitive level, group membership, age and oral level of French and reading words in Tahitian contributed very strongly to the mastery of writing in French (reading words: β = 0.59, p < 0.001; reading–comprehension: β = 0.28, p < 0.001; spelling: β = 0.37, p < 0.001). Reading–comprehension in Tahitian contributed to reading–comprehension in French (β = 0.25, p < 0.01).

4.2.4. Synthesis

As in the previous study conducted in New Caledonia, the results show a positive effect of the bilingual education curriculum on Tahitian skills without having a negative effect on French. Through a multiple regression analysis, we showed that reading and writing proficiency in French could not be directly explained by the morphological awareness of Tahitian but rather by the level reached in the identification of written words in Tahitian, after controlling for the non-verbal cognitive level and oral French level.

5. Discussion

5.1. Synthesis

After conducting both longitudinal studies, we can conclude that the pupils who benefited from these bilingual programs from preschool and at the start of elementary school presented a much higher level in the local language than their peers. Moreover, participating in these programs did not in any way hinder the development of the language of instruction: French. The follow up that we conducted with the same cohort of pupils (ECOLPOM then ReoC3) in French Polynesia confirmed that the results were reliable and stable over the long term (five years). Our data also show the existence of cross-linguistic transfer between two relatively distant languages, in keeping with Kuo and Anderson’s (2010) and Kuo et al.’s (2016) theory of structural sensitivity.

These results are confirmed by the other studies we conducted. The assessment of the ELAN-Afrique (Nocus et al. 2017) system allowed us to confirm on the whole the added value of bilingual programs on a larger scale (2700 pupils), with a higher number of countries (eight in total) and diverse educational contexts (varying levels of experience with bilingual teaching, more or less important language diversity, differences in the implementation of the system according to the countries, etc.).

Additionally, Mann and Wimmer’s (2002) and Ziegler et al.’s (2010) hypothesis is validated: learning to read in one of those local languages makes it easier to learn to read in French, which has a more opaque writing system, by allowing the children to practice the word formation process.

We demonstrated that the expected effects of cross-linguistic transfer are only possible if the pupils learn to read and write in the two languages. In this regard, the assessment of the LCP program in primary schools in Polynesian that we conducted presented us an interesting result (Nocus et al. 2012). It showed that the teaching of Tahitian exclusively focusing on oral language (as was suggested at the time in school curricula) does not allow for any cross-linguistic transfer, and this has limited effects on academic success. It is based on these results that we recommended learning to read in the two languages.

Our most important contribution and one that contradicts many received ideas is the fact that the pupils manage to overcome the constraints that learning two languages as linguistically distant as French and the targeted local languages can present. Our results contradict the conclusions of the studies showing that a relative linguistic proximity between the languages is necessary to allow for the effects of cross-language transfer. In fact, more than the languages and their characteristics, it is the cognitive processes implemented when using two codes that create a benefit. Whatever the languages in question are, bilingualism places people in a double-tasking situation. The bilingual must constantly switch from one code to another, but also inhibit a language in favor of the other according to the languages spoken by their interlocutors. This experience with attention management across two linguistic systems improves the performing metalinguistic abilities indispensable to acquire written language proficiency (see, among others, Bialystok 2009; Bialystok et al. 2005).

5.2. Limitations

Of course, this type of research is hampered by limitations imposed by the practical realities of the research context. We voluntarily set aside the impact of such programs on conative aspects (self-esteem, language self-concept, motivation and children’s positive attitude toward school or toward second-language learning) and cultural (preserving original heritage) and socio-cultural aspects (“empowerment” of minority groups) in order to concentrate more on the cognitive and linguistic aspects. We are convinced of the effects on other dimensions, in particular, on minority rights (see May and Hill 2005, for an example, with Maori-medium education), on the impact of such programs on families (for the role of the home and family, see Fletcher et al. 2009), the implications for school community relations (relationships between children and their families, parents and teachers and teachers and senior management; Tuafuti and McCaffery 2005), the implications of teacher proficiency in bilingual and second-language teaching (Benson and Plüddemann 2010) and the implications for teacher training. However, within the framework of the French school system, these arguments are not sufficient to encourage educational authorities to promote local languages. Only the argument of the added benefits of bilingualism on cognitive and academic aspects enables government authorities to be won over and to modify the official directions and teaching practices in favor of the recognition of local languages.

5.3. Perspectives

In terms of the perspectives, emerging and future research should take into account the stakes of education and training in the 21st century (UNESCO 2014; OCDE 2018), which no longer focus only on academic performance. One of the new key issues at stake in tomorrow’s education and training is the necessity to recognize the creative potential of bilingual children and fostering their well-being, while considering the various languages they can share as being of equal value. On the one hand, the daily use of two languages appears to improve creativity, i.e., the ability to produce original and relevant ideas (Lubart et al. 2019). Furthermore, some authors support the idea that bilingualism and creativity are positively related (Van Dijk et al. 2019; Gulati 2017). However, researchers do not reach a consensus on these relationships between bilingualism and creativity, probably because the mechanisms that underlie creative acts in the bilinguals are influenced by their sociolinguistic context, which can valorize their bilingualism, and by parental language practices (Kharkhurin 2009). Therefore, it would be necessary to conduct further studies to examine whether or not a bilingual educational system can have an impact on children’s creative thinking and thus on their cognitive development (Vincent-Lancrin 2020). On the other hand, the conative aspects of children’s bilingual development, such as their perceived well-being or life satisfaction, defined as the cognitive and affective assessments that every person make of their own lives, as a whole or in the various dimensions it comprises, have not been the focus of many studies. And yet, a child who grows up in contact with several languages can face difficulties related to their bilingual environment, which can affect their well-being (De Houwer 2020). Some authors show that not speaking the minority language of the family is likely to weaken the parent–child relationship and more generally affect children’s well-being. However, there are divergences in the results of the studies according to the aspects linked to the families’ bilingual contexts (e.g., family language practices; children’s and parents’ language skills) when it comes to studying the links between bilingualism and children’s perceived well-being (Sari et al. 2019). Humeau et al. (2023) show that, in 86 ten-year-old children living in France in a bilingual context, and whose language practiced at home is minorated, the parents’ language practices and the children’s use of languages at home are the only two variables that significantly predict the children’s life satisfaction. Even though they are linked to their life satisfaction, the children’s skills in the minority language do not have a significant impact in the tested model.

Studies should thus be conducted in order to explore the relationships between some factors linked to children’s bilingual contexts, their creative potential and their life satisfaction within the framework of the implementation of bilingual educational systems.