First-language skills of bilingual Turkish immigrant children growing up in a Dutch submersion context

ABSTRACT

The interdependence between the first and second language of bilingual immigrant children has not received sufficient attention in research. Most studies concentrate on mainstream language skills of immigrant pupils. In some studies, the gaps in the language development of immigrant children are documented by comparing mainstream pupils with immigrant children. The competence in the first language receives very little attention. In order to show the role of first language competence in second-language acquisition and lower school achievement, we present the findings of our empirical study by comparing bilingual Turkish immigrant children (n = 30, mean age = 67.27 months) in the Netherlands and monolingual Turkish speakers (n = 30, mean age = 66.97 months) in Turkey. The evidence presented in this paper shows that compared to monolingual Turkish speakers, Turkish immigrant children lag behind in their first language cognitive concepts, lexical, syntactic, and textual skills. Mothers’ education level turns out to be an important factor in explaining the performance differences of immigrant children. This would have an effect on their concept and cognitive development as well as on their second-language acquisition.

KEYWORDS:

• Bilingual development
• first-language skills
• submersion education
• Turkish immigrant children

Introduction

The focus of this Special Issue is on various aspects of early bilingual education and on the advantages as well as challenges for children’s linguistic, emotional and social development within various settings and contexts. The focus of our paper is on the challenges faced by bilingual Turkish immigrant children in Dutch submersion pre-schools. Baker ([2006] 2011) differentiates between various forms of bilingual education in line with their educational and linguistic aims. There are monolingual forms of education for minority children where the aim is linguistic assimilation of bilingual children. Submersion is one such approach that ethnic minority children are submerged in the mainstream language and receive instruction only in the societal language. There are also transitional bilingual programmes in which minority children are temporarily allowed to use their home languages. In strong forms of bilingual education the aim is additive bilingualism and minority children add the societal language to their linguistic repertoire (for an in-depth coverage and formal definitions of bilingualism and bilingual education terminology, see Baker [2006] 2011).

The primary focus of our paper is on the first-language skills of Turkish pre-schoolers in the Netherlands. Educational specialists and policy-makers increasingly focus on immigrant minority children’s development of second-language proficiencies in the European context. Children start school at the age of four in the Netherlands. The first two years constitute the preschool period during which they play and learn. Bilingual instruction in this initial period is crucial for immigrant children (Baker 2000; Cummins 2000) but the primary focus of preschools is on the acquisition of Dutch in a submersion educational context (Extra and Yagmur 2010). Teachers believe that the more the children are exposed to Dutch, the better it is for acquisition of Dutch (Leseman 2000). Parents are strongly advised not to speak in any other language than Dutch in their homes (Driessen and Merry 2011; Rijkschroeff et al. 2005). Speaking in a language different from Dutch is seen as an obstacle before the learning of Dutch.

Bilingualism and bilingual education have received increasing attention in the scientific literature in recent decades; however, the focus has mostly been on the mainstream language skills of minority children (Dagevos, Gijsberts, and van Praag 2003; Driessen and Merry 2011; Tesser and Iedema 2001). Given the social status differences between languages, formally and informally, a differentiation is created among various languages as national, foreign language, regional minority and immigrant. Given the socially stigmatized position of immigrant languages, they are at the bottom of the ‘social prestige hierarchy’ (Schalk-Soekar, Van de Vijver, and Hoogsteder 2004). Depending on the ideological approach taken, immigrant bilingualism is mostly seen as a deficit but rarely as a resource (Extra and Yagmur 2010). Immigrant minority languages are most often associated with problems of poverty, underachievement in schools, social and cultural problems, as well as lack of integration into the society of residence. A number of research studies have shown over and over that bilingual immigrant children have a lower proficiency in Dutch than monolingual peers have (Aarts and Verhoeven 1999; Driessen and Merry 2011; Leseman 2000). Lexical depth both in Dutch and Turkish were found to be very limited among Turkish immigrant children (Verhallen and Schoonen 1998). Given the value attributed to lexical development, receptive and productive skills of Turkish immigrant children in Dutch vocabulary have been compared with native Dutch children in many studies. The gaps between the performance of native Dutch children and Turkish immigrant children were reported widely (Aarts and Verhoeven 1999; Dagevos, Gijsberts, and van Praag 2003; Tesser and Iedema 2001) and in almost all studies Turkish immigrant children were shown to have much lower competence compared to native Dutch children. Most researchers only measure immigrant minority children’s mainstream language skills rather than their home language skills and claim language deficiency on behalf of immigrant children. On the basis of proficiency levels in the second language, sizeable numbers of immigrant children are considered to be language-impaired and are sent to special classes (Yagmur and Nap-Kolhoff 2010). Only a handful of researchers investigate the skills in both the heritage language and mainstream language (Leseman 2000; Schaufeli 1992; Verhoeven 1994, 1999, 2007).

Ignoring the possible interdependency between first and second-language skills, most researchers rarely reflect on the first-language skills of immigrant pupils. Besides, some researchers ignore the possible effects of contextual factors on immigrant pupils’ language development and school achievement (Driessen and Merry 2011). The possibility that the acquisition of the mainstream language may benefit from a rich concept development in the first language so that the acquisition of L2 becomes much easier is ignored in most research conducted in the Netherlands. Nevertheless, there is research examining conceptual knowledge transfer from first to the second language in the vocabulary domain (Ordóñez et al. 2002; Snow et al. 1991). Because of ethnically segregated schools and subtractive bilingual environment in the Dutch context, cognitive skills of immigrant minority students do not develop sufficiently compared to mainstream children (Leseman and van Tuijl 2001). Subtractive bilingualism refers to the phenomenon that acquisition and use of a second language (mostly the mainstream societal language) often is at the expense of the development and use of the first language (the immigrant minority language; Baker [2006] 2011). Most educational specialists compare the linguistic skills and educational achievement of immigrant children and native Dutch children without taking background characteristics into serious consideration (Dagevos, Gijsberts, and van Praag 2003). Frequently reported achievement differences between immigrant and mainstream children only serve to further stigmatize immigrant pupils. The stereotype of underachievers is further strengthened in the mainstream public opinion by establishing a causal relationship between low school achievement and ethnicity. As a result, actual causes of educational problems are overshadowed.

As is the case in most communities, there are social and linguistic diversity and performance differences among Turkish immigrant children (Klatter-Folmer 1996). Showing the Turkish immigrant group as a homogenous entity is wrong and would lead to inaccurate research outcomes. Besides, socio-educational background factors have been shown to have more explanatory power than ethnicity in accounting for linguistic differences among immigrant children (Schwartz, Kozminsky, and Leikin 2009). Some research findings have shown that children’s early language skills are related to their experiences with language input in the home context (Aarts, Demir, and Vallen 2011). There is some research evidence that socioeconomic status of parents, material conditions in the home context, mother’s education and the amount of verbal interaction with the child have considerable effect in language acquisition because high quality linguistic input makes a difference in children’s language learning (see e.g. Scheele, Leseman, and Mayo 2010; Schwartz, Kozminsky, and Leikin 2009).

Given the context of situation as briefly described here, our study has two major aims; first of all, we want to present a review of the academic and public discourse regarding language development of immigrant children in general and Turkish immigrant children in particular; secondly, in order to show the role of first-language competence in second-language acquisition and lower school achievement, we present the findings of our empirical study comparing bilingual Turkish immigrant children and monolingual Turkish speakers in Turkey. In the next section, we review the current research on the topic, followed by our Methodology, the Results and finally Discussion and conclusions.

Language acquisition in submersion environments

With a population of over 400,000, Turkish speakers constitute the largest immigrant group in the Netherlands. Each year around 5000 students with Turkish heritage start primary school. In the preschool years Turkish is the most important language (Leseman 2000) but as they grow older Dutch becomes dominant in their lives (Extra and Yağmur 2004). Large numbers of research papers and dissertations have been written on Turkish immigrant children. Most of those studies focus on immigrant children’s second-language acquisition, school achievement and other signs of their integration into the host society (Backus 2011). Large-scale studies have shown that Turkish pupils in Dutch primary schools lag far behind their native Dutch peers in school achievement (Dagevos, Gijsberts, and van Praag 2003; Extra and Yagmur 2010). Based on semi-longitudinal comparative studies, achievement gaps between native Dutch and immigrant children are reported (Driessen and Merry 2011) but the factors leading to those gaps are usually not discussed. As reported by Leseman (2000) disadvantages of immigrant children are already manifest upon entering the primary school. Immigrant children’s first- (L1) and second-language (L2) skills, in particular vocabulary, are less than the language skills of their monolingual peers. Nevertheless, Turkish immigrant children have to acquire a substantial vocabulary in the mainstream language to succeed in school, while they need to maintain and expand their L1 skills for all kinds of communicative purposes in the context of the family and wider cultural community (Scheele, Leseman, and Mayo 2010). An extensive body of research with monolingual children has established that children’s early language skills are strongly related to their experiences with language input in the home context (Scheele, Leseman, and Mayo 2010; Schwartz, Kozminsky, and Leikin 2009; Snow 1972). In the Dutch context, there is no societal and institutional support for first-language development of immigrant groups. Submersion education undermines the use of first language, which leads to subtractive bilingualism. Turkish immigrant children who grow up in low socio-economic status (SES) families mostly do not receive rich and elaborated language input to develop their L1 skills further (Leseman and Van den Boom 1999).

The assumption of interdependence in first and second-language development originates from research on the language development of Finnish immigrant children in Sweden (Skutnabb-Kangas and Toukomaa 1976). Cummins (1979) further developed the idea of interdependence in first and second-language development. He developed a number of hypotheses regarding the relationship between first and second-language skills. The threshold hypothesis claims that bilingual children need a minimum level of linguistic competence in their L1 so that they can achieve similar levels in their L2. If the level in L1 is low, the predicted level for L2 is also low, which is called the interdependence hypothesis by Cummins. Assuming the validity of the interdependency between first and second-language skills (Cummins 2000), delays in first-language development might cause delays in second-language acquisition. Particularly negative circumstances of subtractive bilingual environment might be a better explanation for the reported language delays and lower school achievement of Turkish immigrant children. Nevertheless, the rich linguistic and cultural diversity among Turkish immigrant children as well as the role of first-language development in the acquisition of mainstream language is under-researched (Extra and Yagmur 2010). Moreover, the studies investigating language acquisition and social integration of immigrant children disregard the macro-sociolinguistic circumstances surrounding these children.

The development of first-language skills is not always considered to be beneficial for ethnic minority children in mainstream public opinion. However as discussed by Schwartz (2014), bilingual preschool education of language-minority children from immigrant backgrounds promotes rather than hampers their L2 vocabulary development. Sufficient and meaningful exposure to the majority language is the first condition for successful second-language acquisition (Krashen 1994). However, given the ethnically segregated Dutch schools, immigrant children are exposed to a minimum of native Dutch language use. In order to fully understand the delays in second-language acquisition and widely reported school failure of immigrant children, it is helpful to examine the educational context and schooling process. Without understanding the spatial segregation leading to the formation of black versus white schools, delays in language acquisition cannot be fully evaluated. The term black school refers to those schools whose student population is 70% or more non-Western allochthonous background, mainly Antillean, Surinamese, Moroccan and Turkish immigrants.

School achievement discussion in submersion environments

Most first generation Turkish immigrants in the Netherlands originate from rural areas in Turkey. According to the national statistics ‘Centraal Bureau voor de Statistiek’ (CBS 2013) there were in total of 395,302 Turkish people living in the Netherlands in the year 2013. Around 49.7% (196,203) of Turkish immigrants were from the first generation and 50.3% of them (199,099) were from the second generation. For a detailed overview of Turkish immigrants in the Netherlands see Backus (2011) and Yagmur (2009). Most of the parents have very little or no schooling at all and they are often unable to provide enough educational support for their children at home. Low skills in Dutch as well as sociocultural differences between the mainstream school system and home culture act as barriers to parents’ involvement in their children’s schooling process. Involvement in children’s schooling is mostly dependent on the parents’ level of education rather than on their ethnicity. Yet, despite the innumerable factors involved, it is common to see reports in the media comparing native Dutch pupils to Turkish and Moroccan pupils. Educational specialists and most researchers treat immigrant groups as homogenous entities (Stevens et al. 2011). In most of these reports, Turkish and Moroccan pupils are grouped together, and findings are generalized for this highly heterogeneous population. Even though Turks (including Kurds) and Moroccans (including Berbers) come from totally different linguistic, socio-cultural, ethnic and historical backgrounds in their countries of origin, due to the assumption that they both have Muslim religious background, children coming from Moroccan and Turkish heritage are grouped together. In terms of educational achievement, Turkish and Moroccan pupils show lower achievement levels than native Dutch pupils. National testing and evaluation institution Cito conducts end of primary school exams. Based on the results received, pupils are referred to various types of schools as vocational or higher secondary schools. Students’ scores might vary from as low as 500 to as high as 550. Each year Cito scores for various immigrant groups are published next to native Dutch students. As presented in Table 1, the results for Turkish and Moroccan pupils are lower compared to native-Dutch pupils.

Table 1. End of primary test results across ethnic groups from 1994 till 2005.

Years	Turkish	Moroccan	Surinamese	Native-low SES	Native-high SES
1994/1995	524.1	525.1	527.1	531.9	538
1996/1997	525.2	526.4	527.4	531.2	537.4
1998/1999	526.9	526.9	529.2	530.6	536.9
2000/2001	527.3	527.4	529.8	530.5	537.3
2002/2003	527.3	528.3	528.3	530.6	537.3
2004/2005	527	527.7	527.9	528.9	536.2

Source: Extra and Yagmur (2010).

On the basis of the results presented in Table 1, it is easy to draw the conclusion that immigrant groups perform much less well than the native Dutch students in primary schools. Such group-based comparisons are disseminated in public and scientific publications. These comparisons are based on an ethnicity criterion alone. Because ethnic groups are not homogenous entities, it is necessary to control for socio economic status, parental background and educational level of the parents. When the results are controlled for, for example, level of family income, the differences become more meaningful. Using the data available on the official website of Dutch Central Statistics agency, we, the researchers, requested a distribution of Cito test results based on income levels of parents. As presented in Table 2, when the SES of the parents is controlled for, the differences between native Dutch and immigrant groups on Cito test results no longer reaches statistical significance. Immigrant students who come from high income families achieve as high results as native Dutch children who belong to high SES families. The differences between low-SES immigrant children and high-SES immigrant children are as large as the differences between low-SES native Dutch and high-SES native Dutch children.

Table 2. Distribution of Cito scores across immigrant and native groups based on income levels.

Family income level	Origin	2005/2006	2006/2007	2007/2008	2008/2009	2009/2010	2010/2011
Lowest income level 1st 20%	ND	533	533	533	533	533	533
	N-WI	528	528	529	529	529	530
Second income level 2nd 20%	ND	534	534	534	534	534	534
	N-WI	529	529	530	530	530	531
Average income level 3rd 20%	ND	535	535	535	535	536	536
	N-WI	531	531	531	532	531	532
Above average income level 4th 20%	ND	537	537	537	537	537	537
	N-WI	533	533	533	534	534	535
Highest income level 5th 20%	ND	539	539	539	539	540	539
	N-WI	537	538	538	538	538	538

Note: ND, native Dutch; N-WI, non-western immigrant.

Once the differences are presented along social and economic factors, ‘ethnicity’ becomes much less meaningful. All ethnic groups are heterogeneous. There is huge social, economic and linguistic variation within each immigrant group. By putting them all in the same category and presenting results on group-based scores, the real causes of achievement problems are disguised. As a result, educational specialists, policy-makers and teachers make inaccurate judgements based on inaccurate evaluations. They tend to identify ‘ethnicity’ as a factor in explaining school failure. Because there are a large number of academic reports and scientific papers using similar methodologies based on ethnic comparison, the real differences between autochthonous and allochthonous groups are disguised. The image of ‘low-achieving immigrant minority’ becomes embedded in the social cognition of the mainstream society members, which turns into a social stigma in the long run. Comparisons based on operationalized and relevant factors might better inform educational institutions and general public.

School achievement is a complex and multi-layered topic but some scholars try to simplify it by bringing it down to the ethnicity variable alone. Lower school achievement among immigrant minority children is a serious problem in most European countries. Factors leading to underachievement at school are multifaceted and interrelated. In the literature on bilingualism and school success, individual characteristics of minority students are shown to be one of the most influential on school failure. Because of subtractive bilingual environments, cognitive skills of immigrant minority students do not develop sufficiently compared to mainstream children. If a child’s home language is undervalued or banned on the school ground, identity development might also be hampered. As a result, lower self-esteem among minority students might lead to lower achievement. Due to segregated schools, there is insufficient exposure to the majority language which might in turn lead to inadequate proficiency in the mainstream language. It is also common knowledge that there are gaps between home and school culture due to different socialization patterns, which might also have an effect on school achievement of immigrant children. Many immigrant parents have limited proficiency in the mainstream language, which leads to restrictions in parental involvement. The development of first-language skills of immigrant children have been underestimated in many immigration countries. Comparison of immigrant and mainstream students showed large differences but which factors lead to such outcomes has not been researched in any detail. The majority of studies on educational inequalities take a rather static approach by focusing on educational outcomes at one point in time rather than analysing the school career as a whole (Baysu and De Valk 2012).

In some comparative studies researchers identified some factors to have the most power in explaining the major differences between immigrant children and mainstream students. Language spoken at home, religious background, and the extent of ethnic diversity in the school are shown to be the causes of large differences (Driessen and Merry 2011; Dronkers 2010; Levels, Dronkers, and Kraaykamp 2008). Researchers in the German, Austrian and Dutch context claimed that speaking a different language other than the mainstream language leads to lower school achievements among immigrant children. However, as shown in OECD (2012) analysis language used at home does not account for the differences between achievement levels of immigrant students; however, mother’s level of education makes a difference, which points out the quality of linguistic input in the home context regardless of language. Other research findings from educational linguists indicated that parental involvement is crucial in accounting for children’s language development and higher school achievement. Parents who are involved in their children’s lives, who are responsive, who spend time with them and enjoy joint activities such as book reading foster their children’s cognitive as well as their language and socio-emotional development (Levin and Shohamy 2012; Leyendeckera et al. 2011; Schwartz, Kozminsky, and Leikin 2009).

Discussions seem to concentrate on language ability of bilingual children; yet, there is a close relationship between the larger linguistic environment and language competence of immigrant children. First-language skills and cultural heritage of immigrant children are not seen as valuable assets but as problems that schools need to deal with. According to Verhoeven (1999), a higher level of development in the first language paralleled a higher level of development in the second language. This interdependency positively influences the level achieved in Dutch as a second language. Turkish immigrant children achieve less compared to native Dutch children because their first-language development is delayed. By examining the first-language skills of immigrant children, the actual causes of lower school achievement and delays in second-language acquisition may be more accurately established.

Present study

Research questions

In order to document the linguistic variation in first-language skills of Turkish immigrant children, the following research questions are addressed in this study:

1. Is there a gap between first-language skills (lexical, cognitive concepts, textual, syntactic and phonological) of bilingual children growing up in the Netherlands and monolingual Turkish children growing up in Turkey?

2. How homogenous is the bilingual group regarding their lexical, syntactic and textual skills?

3. Are there relations between bilingual children’s L1 language skills and parental language input and education?

We hypothesize that there is a high correlation between lexical skills and cognitive concepts of Turkish immigrant children (Verhoeven 2007; Yagmur and Konak 2009). Secondly, Turkish language skills of immigrant children are going to show variation depending on parental language input and educational background of the parents.

Informants

The bilingual group was recruited in one large city (Tilburg) and two smaller towns (Waalwijk and Heusden) in the Netherlands. Through the Turkish-Dutch Education Foundation, mothers were approached for involvement in the study. The aim of the study was explained to the mothers. The instruments were shown and the aims are discussed in detail with the mothers. Mothers were told that this was not a formal assessment and the results were only going to be used for research purposes. Oral consent was obtained from the parents. Thirty children, 14 males (mean age = 68.64 months) and 16 females (mean age = 66.06 months) took part in the study. All Turkish-Dutch children were born in the Netherlands. Twenty-three mothers were born in Turkey while six mothers were born in the Netherlands and one mother in another European country. According to the mothers the children were all bilingual in Dutch and Turkish.

The monolingual group was recruited from Ankara in four different kindergartens in a lower middle-class neighbourhood. In order to match the social and educational background of parents in the Netherlands, a purposive sampling approach was followed. First of all directors of the preschools were contacted for permission to work in the kindergartens. Class teachers were consulted for potential children to be involved in the study. Children who have no speech or behaviour problems were indicated by the class teachers and all children were eligible. Mothers of those children were approached for permission. The aim was explained and the instruments were shown to them. Similar to the Turkish-Dutch parents, they were ensured that the findings were to be used only for scientific purposes. Consent forms were distributed to the parents. The children whose parents communicated consent were included in the study. After written consents were obtained, the first researcher chose 16 females (Mean age = 67.5 months) and 14 males (Mean age = 66.36 months) for involvement in the study. In order to control for social variation, participants with similar parental background are chosen in the Turkish context. Because most of the immigrant parents originate from central Anatolia, we selected our Turkish participants from a Central Anatolian city as well. Besides, by targeting a lower middle-class suburb we tried to match the SES of parents.

Instruments

In order to examine the variation in linguistic skills of Turkish immigrant children, 30 bilingual participants (ages 5–6) in the Netherlands and 30 monolingual children growing up in Turkey were tested using a variety of instruments. The Turkish version of a standardized bilingual test (Verhoeven et al. 1990) was used to collect data. The test included six sections: phonological awareness (30 items), cognitive concepts (65 items), passive lexicon (60 items), text comprehension (20 items) in which participants hear four different stories and then respond to five questions for each short story; active vocabulary (40 items), and sentence imitation (20 items), for instance, children hear a sentence in Turkish ‘Orhan’ın büyük ablası hangi okulda okuyor?’ (Which school does the elder sister of Orhan attend?) Or ‘Siz de gelseydiniz çok güzel eğlenirdik’ (If you had come, we would have had a lot of fun), then the participant is expected to repeat the sentence in the same word order. The first four sections of the test were presented to children digitally (on computer screen), while the last two were done traditionally (first researcher conducted the tests with the children orally). In addition, in order to obtain further evidence on the lexical structure of participants, a lexical production task was used. In this task there were 11 items including the definition of a concept or a word for which participants were expected to tell the relevant lexical item or concept. For instance, the participants hear the following description (or definition): ‘What do we call the thing that bees produce and we eat at breakfast?’ The expected response is honey. Another example is ‘what happens to our clothes when we stay outside in the rain?’ The expected answer is ‘wet’.

Sociolinguistic data were collected using a short survey on background variables, such as birth country of children and parents, mother’s education level, number of siblings, the total number of years in the immigration context, amount of TV watching both in Dutch and Turkish, language choice of children with different interlocutors, reading activity with children in Turkish and Dutch, and finally attitudinal questions on the importance of Dutch and Turkish for children’s development. Parents were asked whether it was important for them that their children could speak in Turkish (as well as in Dutch); whether it was important for them that their children could read and write in Turkish (as well as in Dutch). They responded to these questions by using a five-point Likert scale. The parents also assessed language proficiency of their children in each language by using a five-point Likert scale.

Procedures

In order to collect data, both researchers went to community centres and to local public schools to meet the participants in the Netherlands. On Wednesdays and Fridays children were free after 12.30 and data were collected on those two days in Waalwijk and Heusden. Appointments were made with the families so that children would stay in the school to take part in the study. Data were collected in two different sittings with each informant. Because researchers resided in Tilburg, it was possible to make appointments with the parents to carry on the data collection in the evenings as well. Given the number of instruments used, if necessary, data collection was extended to a third session. The main consideration was the cooperation of children. Unless the children did not fully cooperate, a new appointment was made with the parents. Depending on the cooperation level of the child, each session lasted 20–40 minutes. In the Turkish context, data collection went more smoothly because the researcher could work in the schools during the school hours. Also in Ankara, the data were collected in two different sessions. The test results were entered into SPSS program (Version 19) for statistical analyses.

Results

In this section, first of all, some descriptive analyses are presented to provide an overview of the participants and the results. Afterwards based on a number of comparative analyses, differences between groups are documented so that our research questions can be answered. Given the large number of studies not controlling for SES and educational level of the parents, it was important to control for the educational level of the mothers in this research. As shown by the OECD analysis of Programme for International Student Assessment outcomes (2012), mother’s educational level accounted for the differences observed between high and low achieving students. In our research both groups of mothers had comparable educational backgrounds as presented in Table 3. The number of mothers who had primary and lower secondary school diplomas is larger in the Dutch context.

Table 3. Mothers’ educational qualifications (N = 60).

Mother’s education	The Netherlands	Turkey	Total
Primary	6	2	8
Lower secondary	4	3	7
Higher secondary	11	10	21
College	7	10	17
University	1	5	6
Post-graduate	1	0	1
Total	30	30	60

First we want to discuss the findings of the bilingual group. As reported earlier, school managements and teachers are known to advise immigrant parents not to communicate in their first languages with their children. They assume speaking in Dutch in the home context would increase the amount of input and in this way acquisition of Dutch would be easier. In order to see the extent of language use in the private domain, mothers were asked about their children’s language use practices. As documented in Table 4, children mostly use Turkish with different interlocutors in the home. Mothers turn out to be the caretakers of first-language maintenance as 77% of children are reported to be speaking in Turkish with their mothers. Children use more Dutch with their fathers and the amount of Dutch increases in speaking to the siblings and friends. This is a well-established pattern as reported by large scale investigations as well (Extra and Yağmur 2004). In speaking to grandparents Turkish is used almost always. The unexpected outcome of the survey is the language used with the teacher in the school. In some schools there are Turkish-speaking support personnel and apparently in responding to the question, children and parents considered their interaction with them as well. Even though the informant population is small, the sociolinguistic picture emerging here is fully congruent with other findings reported in the Netherlands.

Table 4. Language choice of Turkish-Dutch children with different interlocutors (N = 30; in %) as stated by parents (mostly mothers).

Interlocutors	Always Turkish	More Turkish	Turkish and Dutch equal	More Dutch	Always Dutch	Missing
Mother	46.7	30	23.3	–	–	–
Father	26.7	26.7	40	3.3	3.3	–
Siblings	10	6.7	33.3	20	3.3	26.7
Friends	3.3	26.7	26.7	33.3	10	–
Neighbors	20	23.3	23.3	20	3.3	10
Teacher at school	–	–	6.7	33.3	60	–

Mothers were asked to rate their children’s bilingual skills in Turkish and Dutch. According to the mothers, children’s skills in Dutch and Turkish are comparable but Turkish is assessed on average slightly better than Dutch as shown in Table 5.

Table 5. Assessment of language skills of the children by their mothers (in %).

Language skills	Very poor	Below average	Average	Above average	Very good
Turkish	10	6.7	50	10	23.3
Dutch	6.7	6.7	43.3	30	13.3

Test results

Before testing our hypotheses and answering research questions, we subjected each scale to a reliability analysis. We obtained rather high Cronbach Alpha values for each sub-test. Only definition task turned out to be less reliable compared to other sub-test. Because children were expected to produce the lexical item on the basis of the definition provided, it turned out to be a very demanding task for 5- and 6-year-old children. As shown in Table 6, other scales are highly reliable.

Table 6. Reliability scores of the scales (Cronbach alpha values).

Scales	Alpha Coefficient	Number of Items
Active lexicon	.882	40
Syntax	.881	20
Passive lexicon	.874	59
Cognitive concepts	.880	65
Phonological awareness	.815	30
Text comprehension	.769	20
Definition task	.639	11

Instead of presenting a descriptive Table on each subtest, we calculated a scale score for each subtest. On the basis of accumulated points from each sub-test, each informant obtained a scale score. The numbers of correct responses were added up so that a total score for each scale could be obtained. If a participant gave 35 correct responses to 40 questions in the active lexicon test, then 35 becomes the active lexicon scale score for that participant. In order to see the gap in first-language skills of bilingual children growing up in the Netherlands and monolingual Turkish children growing up in Turkey, an ANOVA analysis was done on each scale. As presented in Table 8, there are highly significant differences between monolingual Turkish participants and bilingual Turkish-Dutch participants. Normally, bilinguals are not compared to monolingual speakers; however, such comparisons are frequently done between mainstream monolingual Dutch speakers and bilingual immigrant children. In most comparisons, bilingual immigrant children are shown to lag behind their native Dutch peers. As clearly seen in Table 8, there are significant gaps between monolingual Turkish speakers and bilingual immigrant children. These differences indicate that irrespective of L1 and L2, there are gaps between monolingual and bilingual children in their linguistic skills. Such differences might be caused by many numbers of factors but most basically these bilingual immigrant children are growing up in an environment where they receive limited input in their heritage language and also in the mainstream societal language. Large differences in active and passive lexicon as well as cognitive concepts show the extent of difficulties before native language acquisition in the immigration context. In line with many other studies’ findings, bilingual immigrant children excel in phonological awareness (Verhoeven 1999). Compared to monolingual Turkish speakers they have much higher scores. When we examine the type of lexical items that bilingual immigrant children know less well compared to monolingual children, it appears that infrequent words are not known by bilingual immigrant children. The response to our first research question is that there are major differences in Turkish language skills of immigrant Turkish-Dutch and monolingual Turkish speakers. The differences are very similar to the gaps reported between Dutch skills of Turkish-Dutch and native Dutch pupils, that is, that the bilingual pupils perform less well on tests than monolingual pupils.

On the basis of the group scores as presented in Table 7, it is not always possible to see the variation within the group. No group is a linguistically homogenous group. In order to see the linguistic variation within the groups, we conducted a percentiles analysis on each sub-test, by means of which it was possible to see low and high achievers in both groups. Percentiles analysis enabled us to group the participants into five competence levels. As shown in Table 8, for each sub-test, both bilingual and monolingual Turkish speakers were divided into performance levels. In the bilingual group, there are high achievers and also low achievers. Even though the majority of bilingual participants are at lower competence levels in Turkish than the monolingual participants, there are also high achievers. In such comparative studies, instead of making broad generalizations, it is important to show the in-group variation. Apparently Turkish immigrant children perform much less well compared to monolingual Turkish speakers but there are also high performing bilingual children. On the basis of the findings presented in Table 8, it can easily be claimed that Turkish immigrant group is not a homogenous group at all. There are considerable differences in the first-language skills of different bilingual immigrant children. In the same vein, there are high and low performing children in the Turkish context as well. On the basis of the outcomes presented in Table 8, it is possible to provide an answer to our second research question that bilinguals do not constitute a homogenous group in terms of their lexical, syntactic and textual skills. There can be many reasons for this variation but possibly first-language input provided at home, the quality and intensity of interaction with heritage language speakers, the SES of parents and the educational level of the mothers might play a role in language development.

Table 7. ANOVA differences between Turkish language skills of Turkish-Dutch bilinguals and monolingual Turkish speakers (n = 60).

Tests		N	Mean	SD	F	p
Active lexicon	NL	30	20.03	7.271	25.783	.000
	TR	30	29.00	6.379
Syntax	NL	30	8.47	4.257	35.634	.000
	TR	30	14.63	3.728
Passive lexicon	NL	30	44.03	6.531	20.652	.000
	TR	30	51.10	5.467
Cognitive concepts	NL	30	46.27	8.043	19.851	.000
	TR	30	54.57	6.279
Phonological awareness	NL	30	24.53	5.290	9.326	.003
	TR	30	20.80	4.106
Text comprehension	NL	30	13.30	3.583	7.980	.006
	TR	30	15.90	3.546
Definition task	NL	30	2.73	1.999	17.028	.000
	TR	30	4.77	1.813

Note: NL, the bilingual participants from the Netherlands; TR, Turkish participants from Turkey.

Table 8. Distribution of competence levels across the groups (NL = 30, TR = 30/scores are in percentages).

Language skills	Level 1	Level 2	Level 3	Level 4	Level 5
Active lexicon	TR	3.3	13.3	20.0	33.3	30.0
	NL	43.3	23.3	20.0	6.7	6.7
Syntax	TR	6.7	10	26.7	23.3	33.3
	NL	43.3	23.3	23.3	10	0
Passive Lexicon	TR	6.7	10	23.3	30	30
	NL	43.3	23.3	13.3	13.3	6.7
Cognitive concepts	TR	6.7	13.3	23.3	23.3	33.3
	NL	36.7	26.7	20	10	6.7
Phonological awareness	TR	40	16.7	26.7	13.3	3.3
	NL	16.7	6.7	16.7	30	30
Text comprehension	TR	10	20	13.3	40	16.7
	NL	33.3	26.7	16.7	20	3.3
Definition task	TR	13.3	13.3	23.3	20	30
	NL	60	13.3	10	10	6.7

Note: TR, participants from the Turkish context; NL, participants from the Dutch context.

In order to see the effect of the educational level of the mothers, we grouped the mothers into two groups as ‘better educated’ and ‘less educated’. Those who had no school diploma, primary diploma, and lower secondary school diploma were grouped under ‘less’ educated and those who had secondary, college and university degrees were categorized as ‘better’ educated. As seen in Table 9, there are significant correlations between mothers’ level of education and the children’s scores. In line with our hypotheses, there are significant correlations between mothers’ level of education and the test scores of children in active and passive lexicon as well as cognitive concepts. There are also significant correlations between all sub-tests except for the phonological awareness.

Table 9. Pearson correlations between mothers’ education level and test scores.

	1	2	3	4	5	6	7	8
1	1
2	.333**	1
3	.170	.674**	1
4	.328*	.599**	.712**	1
5	.301*	.554**	.669**	.779**	1
6	.075	−.069	.054	.113	.092	1
7	.153	.580**	.609**	.553**	.513**	.152	1
8	.186	.639**	.521**	.420**	.416**	−.140	.354**	1

Note: 1 = mother’s education, 2 = active lexicon level, 3 = syntactic level, 4 = passive lexicon, 5 = cognitive concepts, 6 = phonological awareness, 7 = text comprehension, 8 = definition task.

In order to see the relative effect of other contextual factors and mother’s level of education, we carried out a regression analysis. Taking the scales as dependent variables we wanted to examine the possible predictive relationship between mother’s level of education and country of residence on participants’ test scores. Mother’s level of education played a significant role on children’s active lexicon and text comprehension scores. However, country of residence had highly significant prediction on all test scores except on phonological awareness. Bilingual children had higher scores than monolingual participants for phonological awareness (Table 10).

Table 10. Outcomes of the regression analysis.

Dependent variable	R^2	Standardized regression coefficients
		Mother’s education – β	Country of residence – β
Active lexicon	.379***	.237*	.512***
Passive Lexicon	.311***	.217	.462***
Syntax	.355***	.059	.578***
Cognitive concepts	.267***	.167	.449***
Phonological awareness	.210***	.147	−.473***
Text comprehension	.174***	.255*	.311***

Discussion and conclusions

As shown by a number of linguistic investigations, there is a close correlation between first and second-language skills of immigrant children (Cobo-Lewis et al. 2002; Schwartz 2014). If the first-language skills are underdeveloped, the skills in L2 tend to be similarly underdeveloped (Verhoeven 1994). The evidence presented here shows that compared to monolingual Turkish speaking children, Turkish immigrant children lag behind in their first-language skills. In line with additive bilingualism perspective, we claim that the acquisition of the second language may benefit from a well-developed first language. In order to make use of the positive effects of bilingualism, immigrant minority children should have the opportunity to develop a mature level of proficiency in their first language. As shown in this study, the first-language skills of Turkish immigrant children are underdeveloped, as a result of which they might not have any positive transfer to their second-language skills. As long as the social and political circumstances are not conducive to the development of first-language skills, immigrant children might not benefit from their first-language skills in acquiring the second language. The role of a submersion education context on the development of first-language skills has been shown to be negative in other contexts as well (Schwartz 2014). This would have an effect on the children’s lexical and cognitive development as well as on their second-language acquisition. On the basis of the findings reported in this paper, it is obvious that the skills in the first language need to be taken into consideration in evaluating immigrant children’s linguistic skills in their second language. Language is at the basis of all types of learning. If immigrant children cannot achieve a certain proficiency level in the societal language, they are most likely to perform less well compared to mainstream students.

Educational institutions and teachers’ misconceptions about language and learning lead to serious linguistic, emotional and social problems among immigrant minority students and their parents. As reported by Helot and Young (2002), most teachers still believe that speaking an immigrant language at home delays the acquisition of the national language and consequently integration into the mainstream society. On the basis of a large scale Languages in a Network of European Excellence project, Franceschini (2011) reports that many of the teachers in their research believe that using a home language other than the national language might be an impediment to the students’ learning of the official language because the home language could confuse the learners. Franceschini points out the most important problem by emphasizing the role of teachers in negatively influencing the parents. Most immigrant parents are misguided by teachers in schools by giving inaccurate information on the role of home languages in the learning of school languages. Parents are advised not to speak their native language with their children. This suggestion might have serious cognitive and linguistic implications for language development of immigrant children. If parents are not fully proficient in their L2, they will communicate in a restricted code which might seriously hamper the level and quality of communication between parents and children.

Language skills and cognitive skills go hand in hand. Offering bilingual education in preschools with an emphasis on the first-language skills is essential for the acquisition of the conceptual-lexical knowledge in both L1 and L2; this knowledge can hardly be developed in the home context alone. If linguistic skills are not developed sufficiently, cognitive skills will be affected as well. In the case of most immigrant children, limited skills in their first language lead to lower skills in mainstream language skills as well (Cummins 1979; Verhoeven 1994). As research findings have shown there is a linguistic interdependency between first and second-language skills (Bialystok 2005; Scheele, Leseman, and Mayo 2010; Yagmur and Konak 2009). Limited linguistic skills in one language lead to limited skills in the second language. If immigrant children’s linguistic and cognitive skills are sufficiently developed in their first language, this will transfer to their second-language skills. By supporting the use and acquisition of first-language skills of immigrant children in mainstream schools, learning and acquisition of the mainstream language might also be supported.

Immigrant languages are often seen as obstacles before the learning of national language in many immigration contexts (Ammermüller 2005; Extra and Gorter 2008). Reflecting on the lower school achievement among immigrant children and in particular among Turkish immigrant children, Ammermüller (2005) argues that the main reason for the low performance of immigrant students in the German context should be sought in their later enrolment in schools and the less favourable home environment for learning. He claims that many immigrant children have lower achievement levels because about 40 % of all immigrant students speak a language other than German at home. According to Ammermüller (2005), differences in parental education and family situation are far less important. Yet, as shown in this study, mothers’ education level turned out to be important at least for children’s development of active lexicon and textual skills. As in many national contexts, also in the German context, students’ home languages are apparently shown to be the cause of low achievement in the schools. Some educational experts and researchers blame multilingualism of immigrant children for lower school achievement (Dagevos, Gijsberts, and van Praag 2003; Dronkers 2010). However the ineffective pedagogical approaches used in the submersion schools are rarely questioned. The main causes of lower school achievement are sought in the cultural and linguistic background of immigrant children. Schools can improve school achievement of immigrant children by abandoning outdated submersion models. By employing teachers and support personnel from linguistic minority backgrounds schools could support first and second-language development of bilingual immigrant children. The school’s language policy, the structure of curriculum, the teachers’ qualifications and experience with language minority children and parental factors account especially for bilingual children’s school achievement (Helot and Young 2002; Klatter-Folmer 1996; Scheele, Leseman, and Mayo 2010). Whether the school has a bilingual approach or a submersion approach would make a huge difference in the language development of minority children. Submersion is the most common educational approach in most West European schools. Bilingual language support of immigrant children should start during the preschool period so that children are exposed to meaningful input as early as possible. As documented in the large scale European multilingualism research project of Extra and Yagmur (2012), The European Commission Green Paper on Migration and Mobility (2008) emphasizes the critical importance of children from an immigration background learning the host language as early as possible while retaining the heritage language and culture of the country of origin. The importance of preschool bilingual education is underlined by Beacco et al. (2010, 45) that

as spaces for discovery and socialisation, pre-primary schools represent a basic stage in plurilingual and intercultural education, particularly for children from underprivileged and migrant backgrounds, whose language practices at home may conflict with the varieties and norms selected and fostered by schools. To that extent, and since the issue here is the right to quality language (and general) education, one of the first desiderata is that schooling of this kind for very young children be guaranteed and provided in optimum conditions for all the groups concerned – both permanently resident natives and recently arrived immigrant families.

Even though there is a general reluctance to refer to immigrant students as bilinguals and to develop bilingual programmes for them, there is widespread support for mainstream students in various bilingual programmes. Bilingual programmes in high-status languages such as English-German or French-German find huge public support but strong negative attitudes surround immigrant children’s bilingualism. In a typical anti-bilingual fashion, many mainstream teachers believe that immigrant children are overloaded by dealing with two languages, which lowers their proficiency in the mainstream language. Apparently, this outdated separate underlying proficiency (Cummins 1979) model can still find many supporters.

Instead of holding immigrant children responsible for the inappropriate language teaching pedagogies in the mainstream classrooms, the solution is offering balanced bilingual programmes where minority children build on their already available first-language skills. Preparing language minority children for more successful school careers ideally requires a balanced bilingual approach in which children’s greater proficiency in the home language is utilized to promote general cognitive development and acquisition of the school language (Leseman and van Tuijl 2001). However, given the widespread use of submersion models in most European schools, immigrant children’s first-language skills cannot be further developed. As reported by Cenoz and Gorter (2010) the idea that non-native speakers are deficient communicators is still widespread in many national contexts. As shown by this study, most Turkish immigrant children have much lower skills compared to monolingual Turkish speakers. Most probably the lower skills in L1 also lead to lower skills in L2. As documented by Yagmur and Konak (2009) in the German context, Turkish immigrant children who had higher skills in Turkish had equally high skills in German.

Limitations and future research

The findings reported in this paper are based on 60 participants and cannot be generalized to the Turkish immigrant population in the Netherlands. Second-language skills of Turkish immigrant children have been researched and compared to monolingual Dutch peers extensively; however, there are not many previous studies comparing first-language skills of Turkish pre-schoolers to monolingual peers growing up in Turkey. This study concentrated only on the first-language skills but we need additional insight into first and second-language skills of Turkish immigrant children in the Netherlands.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Gözde Akoğlu is an Associate Professor of Child Development in the Department of Child Development at Kırıkkale University. Her areas of interest include early literacy, language and concept knowledge in preschoolers with and without language disorders as well as the relationship between working memory and language.

Kutlay Yagmur is Professor of Language, Identity and Education in the Department of Language and Culture Studies, University of Tilburg. In his ongoing project, Yagmur investigates the relationship between the integration ideology of the receiving society and socio-cultural adaptation of immigrants in four national contexts. He has published extensively on multilingualism and language contact issues in Australia, Germany, France and The Netherlands.