Keywords

1 The Immigrant Turkish Dialect as a Heritage Language in Germany

Germany has always been a country with several bilectal and diglossic contexts (Rash, 2002; Földes, 2005; Koneva & Gural, 2015). In the last decades, dialect use is continuously decreasing, whereas the empowerment and the legal and societal acceptance of minority languages and their speakers, such as “Low German (Plattdeutsch)”, “Lower Sorbian”, or Danish, increases.

Despite Germany’s long history of immigration and experience with heritage speaking and refugee children in the educational system, the languages of migrant communities, such as Turkish, Russian, Kurdish, Syrian Arabic or Bosnian, however, are not addressed with the status of minority languages legally, even though most citizens in Germany acquire one or more of these languages additionally to German. In 2018, 64% of families with children under 18 years of age had a migrant background (Autorengruppe Bildungsberichterstattung, 2018). The number of children speaking more than only German oral language at home increases constantly (Autorengruppe Bildungsberichterstattung, 2018), leading to an increase of heritage language speakers (i.e. Fishman, 2001; Gagarina, 2014).

Turkish is spoken in Europe and other countries, and since the 1960s, many states in Western Europe host large Turkish immigrant communities (e.g. Backus et al., 2010). Importantly, language loss is remarkably rare in the Turkish communities, since immigration is a continuous process. Today, Germany has the biggest Turkish-origin population in Western Europe. An estimated population of 4 million people, of full or partial Turkish origin live in the country (Feltes et al., 2013: 93); that is approximately 5% of Germany's total population of 82 million inhabitants.

Note, however, that, even after four generations Turkish-origin minority populations, people with a Turkish background tend to occupy the lower end of the socio-economic spectrum, as it is not untypical of immigrant communities with roots in labour migration (cf. Backus, 2010; Riphahn et al., 2010). In 2010, Immigrants of Turkish origin were least successful in the German labour market, 30% of adolescents did not finish school, many were jobless, and only one third of Turkish women in Germany were employed. The Turkish communities often live in city centres, and, in cities like Berlin, Hamburg or Mannheim, where seem to be city quarters almost exclusively populated by people of Turkish origin.

The German school system and the educational policies in the Federal Countries of Germany, however, hold specific obstacles for students with a heritage language background. The segregated system of schooling leading to the early tracking of children into higher and lower types of secondary education is particularly disadvantageous for children who grow up speaking non-standard varieties of the majority language, and local dialects, ethnolects, some youth style, or a mixture of these (i.e. Backus, 2010). Moreover, national education reports continuously state the additional disadvantage of children from families with a low socioeconomic status and a history of migration (“migrant background”) (i.e. Autorengruppe Bildungsberichterstattung, 2016, 2018). Such children need more time for acquiring the standard academic variety of the language (“Bildungssprache”). This is – alongside the stereotypes of teachers against the performance of children with a migrant background (Berliner Institut für empirische Integrations- und Migrationsforschung, 2017) – the most relevant drawback on school attainment (i.e. Gogolin, 1994). In spite of the efforts of many scholars in educational science to establish translanguaging as method of teaching and language education already in the 1990s (i.e. Gogolin, 1994), knowledge of the German language is considered a necessary condition for academic and later professional success.

Turkish as a heritage language in Germany, however, is a peculiar case of language acquisition in a bilectal situation. The term heritage language defines the first/family language of minority language children in Germany, being “languages spoken by the children of immigrants or by those who immigrated to a country when young” (Cho et al., 2004: 23). Children acquire the heritage language particularly at home and among the extended family. Exposure to the societally dominant (majority) language may start in the family, but it is more dominant outside home, and especially at school (Polinsky, 2018). The heritage (language) speakers can be successive or simultaneous bilinguals (Bennamoun et al., 2013). A heritage language is acquired incompletely, since the individual uses another (i.e. the majority) language. Secondly, heritage language implies a continuity of proficiencies, reflecting the heterogeneity in heritage language proficiencies observed by several researchers (see Polinsky & Kagan, 2007). Considering linguistic characteristics in detail, there is systematic change in the heritage language of young adults, e.g., of third/fourth generation immigrants. If compared to the standard variety of the L1, the heritage varieties of the L1 show, for example, reduced morphological and syntactic structures (Valdés, 2000; Fishman, 2001; Cummins, 2005; Polinsky & Kagan, 2007; Montrul et al., 2010). At the same time, heritage language speakers seem to have advantages in pronunciation, phonology and spontaneous speech production in comparison to the learners of the second language (Au et al., 2002; Montrul et al., 2008). The reduced input and effects of the second on the first language (Cook, 2003) can result in incomplete acquisition or language attrition (Montrul, 2009; Rothmann, 2009).

Monolingual Turkish speakers immigrated to Germany with the first generation of migrant workers from the 1960s. Importantly, the German labour market recruited people with little education and supported the intention to return to Turkey after a few years, hardly offering opportunities for learning the majority language. Recent generations, as children under the age of 3–4, might be monolingual speakers of Turkish, as immigration continues and the community members actively chose Turkish as family language at home. However, self- reported survey data in France and Germany show that many families use the national languages increasingly alongside Turkish (Akıncı, 2008; Akıncı et al., 2013). Intra-community variation in language use and family language practice is a relevant factor for sociolinguistic research and language assessment in children with Turkish heritage language in Germany. Even though the ethnolinguistic vitality of Turkish is documented (Yagmur & Akinci, 2003; Extra & Yağmur, 2004), it has to be stated that heritage language acquisition often reduces to the spoken language variety of Turkish. The family language use is mostly restricted to the oral varieties, and literacy or academic use of Turkish is limited to some children participating in secondary education (i.e. Turkish as a subject in secondary schools in Hamburg) or (private) afternoon classes, but the general development in the last decade has been toward the abolition of forms of bilingual education. Though contexts for writing in Turkish exist, the degree to which Turks in Western Europe are used to writing in Turkish varies enormously. Consequently, studies of the written Turkish of the immigrant communities has increased only recently (but see Schroeder, 2007; Akıncı, 2008; Dirim, 2009; Akıncı et al., 2013). Moreover, Schroeder (2009) illustrates that Turkish language education in German schools aims at teaching the written language in a very norm-orientated way, emphasising a dichotomy between the standard variety of “anadil” (mother tongue) on the one and “Türkçemiz” (our Turkish) on the other hand.

The notion of the cultural and linguistic differentiation between the standard (written) and the spoken language is of grave importance for heritage language acquisition in Germany, since the Turkish used in Germany is subject to language change, resulting in a new dialect. Large- scale research projects in France, the Netherlands and Germany compared samples of immigrant speech or texts and samples of speech collected in the regions in Turkey from which the original immigrants came (i.e. Doğruöz & Backus, 2007, 2009; Pfaff, 1991; Rehbein, 2001; Herkenrath et al., 2003; Rehbein & Karakoç, 2004; Baumgarten et al., 2007; Herkenrath, 2007; Karakoç, 2007; Banaz, 2002; Johanson, 2002; Uzuntaş, 2008; Şimşek & Schroeder, 2011; Schellhardt & Schroeder, 2013; Schroeder & Dollnick, 2013). They focused on the changes of spoken varietes of Turkish grammar and written language competencies of bilingual children with Turkish as their heritage language.

The changes to Turkish are systematic, and were defined by Johanson as a New Variety of Turkish. The “Immigrant Turkish” dialect (Backus, 2004) differs from Standard Turkish in several aspects (cf. examples below). Importantly, these changes are not entirely based on language contact phenomena, such as cross-linguistic influence in the lexical domain that leads to almost literal translation of multiword units in the majority language. Importantly, “Immigrant Turkish” as a branching term conceals the specific language-induced contact phenomena in different countries as well as the influence of migrant waves, leading to unique ways of dialect levelling. Syntactic variation, for example, between the Turkey-Turkish norms and the Immigrant Turkish dialect were very few in the Netherlands. Neither were entire subsystems, nor were constructions especially sensitive to Dutch influence, that is <1% of “unconventional” structures (Doğruöz & Backus, 2009).

In Germany, however, the Hamburg project focused on structures above clause level, such as subordination, discourse connectivity, and discourse marking in retelling the Snow White fairy tale. Several differences between Immigrant Turkish dialect (IT) and the data from Turkey were different use of finite verb inflection, the use of a smaller range of forms, limitations to one tense marking in narratives (substitution of the evidential form of –mIş, Pfaff, 1994), and the overuse of deictic temporal adverbs in retelling. While monolingual Turkish children acquire both complement and relative clauses at the age of approximately 5 years or older (Aksu-Koç, 1994), Turkish-German bilingual children between the ages of 4 and 9 prefer finite clauses over subordination. Deviations between the standard variety of Turkish in Turkey and the Immigrant Turkish dialect are especially found in the avoidance of using “complex structures” simple juxtaposition instead of complex structures (Sarı, 2006; Treffers-Daller et al., 2006; Dollnick, 2013; Herkenrath, 2014; Bayram, 2013; Onar Valk, 2015; Schroeder, 2016) (cf. example 1). It has also been reported for Immigrant Turkish dialect speakers that they interchange dative and accusative, and use unconventional forms of plural markings (“iki adamlar” instead of “iki adam” in the standard variety).

(1)

Finite instead of non-finite clauses

 

Immigrant Turkish dialect:

 

Çocuk

sineği

vurmak

istiyor.

 
 

Child

fly-ACC

hit-INF

want-PROG-3SG

 
 

Babası

 

diyor

  
 

Father- POSS-3SG

 

say-PROG-3SG

  
 

“Hayır, yapma!”

   

(Treffers-Daller et al., 2006)

 

no

do-IMP-NEG-2SG

   
 

Standard Turkish:

   

Çocuk sineği

vurmak

istiyor.

 
 

Child

fly-ACC

hit-INF

want-PROG-3SG

 
 

Babası

da

yapmamasını

söylüyor.

 
 

Father-POSS-3SG

so

do-NEG-NOM-POSS-3SG-ACC

tell-PROG-3SG

 
 

‘The child then wants to catch it with a cloth. And his father tells him not to do (that).’

 

Further characteristics of the IT dialect refer to the omission/substitution of genitive markings in modal constructions (Menz, 1991), compounds (Aytemiz, 1990), and with subjects in nominalised subordinated sentences (Sarı, 1995). IT speakers also tend to overuse pronominal subjects and objects (Aytemiz, 1990; Menz, 1991; Pfaff, 1991; Rehbein, 2001). Besides, bilingual speakers and bilingual children acquiring the Immigrant Turkish dialect as heritage language use the general all-purpose verb yapmak extensively by adding it to the German verb stem or the Turkish infinitive form and to avoid the standard progressive form (Boeschoten, 1994) (cf. example 2).

(2)

General All-Purpose Verb yapmak:

 

Immigrant Turkish dialect:

 

Ondan sonra ödevim

 

bitmediyse,

 
 

Later

homework-POSS-1SG

finish-NEG-PAST-COND

 
 

onu

devam

yapıyorum

(Boeschoten, 1994)

 

it-ACC

continuance

make-PROG-1SG

 
 

Standard Turkish:

   
 

Ondan sonra

ödevim

bitmediyse,

 
 

Later

homework-POSS-1SG

finish-NEG-PAST-COND

 
 

ona

devam ediyorum.

  
 

It-DAT

continue-PROG-1SG

  
 

‘Then if my homework hasn’t been finished, I go on with it.’

Moreover, it is not only the German/Turkish contact situation, but also the origin of the first- and second-generation immigrants that features IT as a distinct spoken dialect. Dialect levelling, i.e. levelling of Anatolian dialects spoken especially by the first generation of immigrants (Boeschoten & Broeder, 1999; Schroeder & Stölting, 2005), is a typical feature of IT in Germany. It arises in, for example, an overuse of ablative forms in locative contexts (cf. example 3), or an omission of interrogative particles in yes-no questions (cf. example 4).

(3)

Overextension of the ablative case

 

Immigrant Turkish dialect:

 

Savaştan

rüya

gördüm.

(Backus & Boeschoten, 1998)

 

War-ABL

dream

see-PAST-1SG

 
 

Standard Turkish:

 
 

Rüyamda

savaş gördüm.

  
 

my dream-LOC

war

see-PAST-1SG

 
 

‘I dreamed about the war.’

(4)

Omission of interrogative particle in yes-no questions

 

Immigrant Turkish dialect:

 

Bugün okulda

 

oynadıın?

(Hess & Gabriel, 1979)

 

Today

school-LOC

play-PAST-2SG-Ø

 
 

Standard Turkish:

   
 

Bugün okulda

oynadın

mı?

 
 

Today school-LOC

play-PAST-2SG

INT

 
 

‘Did you play at school today?’

Since the previous examples are documented for the variety of IT over 20 years ago, the current study used the data from the MULTILIT study (Schellhardt and Schroeder 2015) to test the actuality of these IT features for contemporary learners of IT in Germany. The MULTILIT corpus contains oral and written data from bilingual children with Turkish heritage language in Germany and France. The analyses of the MULTILIT data confirm the status of IT as a dialect that shapes the heritage language (L1) input of bilingual Turkish-German children. The characteristics of the Immigrant Turkish dialect consist of dialect-levelling features from East- Anatolia. Boeschoten (2000), and Şimşek and Schroeder (2011) illustrate such features with the the instrumental case suffix: While the standard form is (y)la / (y)le, a different form, len / lan, is typically for the spoken Turkish in Western Europe. Further, dialectal variations on the lexical level, like the use of değmek (touch) instead of çarpmak (hit) (cf. example 5). The omission of genitive markers and other indications of morphological changes and loss (cf. example 6), (Boeshoten, 2000) are revealed. Other phenomena, such as the use of reflexive pronoun kendi- as a focus marker (Schroeder, 2014) or unconventional plural marking (i.e. an increased use of plural markers as language-contact phenomenon between German and Turkish, Johanson, 1993: 214) are documented (cf. example 7).

(5)

Dialect levelling and code-switching

 

Immigrant Turkish dialect:

 

Kafanlan

Stuhle

değiyon. (OGU; 5th grade; 11–12 years old)

 

Head-POSS-2SG-INS

chair-GER-DAT

DIALECT- touch-2SG

 

Standard Turkish:

  
 

Kafanı

sandalyeye çarpıyorsun.

 

Head-2SG-ACC

chair-DAT

hit-PROG-2SG

 

‘You hit your head on a chair.’

(6)

Omission of genitive-possessive markers, kendi- as focus marker and use of locative postposition

 

Immigrant Turkish dialect:

 

Burada

bir

kız

kendi

sınıfın

içinde

 

Here

INDEF

girl-Ø

self

class-ACC-Ø

inside-POSS-3SG-LOC

 

dışlanmasıdır.

    

(YON, 12th grade, 17 years old)

 

exclude-PASS-VN-2SG-GM

 

Standard Turkish:

 

Burada (olan)

bir

kızın

sınıfta

dışlanmasıdır.

 
 

Here (AUX-PART)

INDEF

girl-GEN

class-LOC

exclude-PASS-VN-2SG-GM

 
 

‘What happens here is the exclusion of a girl in her own class.’

(7)

Unconventional plural marking

 

Immigrant Turkish dialect:

 

Üç

kızlar

gine

gittiler.

(ILH; 5th grade, 11–12 years old)

 

Three

girl-PL

again

go-PAST-3PL

 
 

Standard Turkish:

    
 

Üç kız yine gitti.

    
 

Three girl again go-PAST-3SG

 
 

‘Three girls went again.’

 

To summarise the results on the Immigrant Turkish dialect in Germany so far, show that IT is a “catalyst” dialect (Rehbein et al., 2009), which may cause bilingual Turkish speakers either to develop new forms or to use existing ones in ways that differ from the Turkish used in Turkey. Thus, the heritage language input of bilingual Turkish-German children is a dialectal one. The bilectal problem is evident with respect to the heterogeneity of the Turkish speaking community (Johanson, 1991; Chilla et al., 2013). In contrast to other bilectal contexts, such as Cypriot Greek, IT dialect children in Germany have only limited access to a “high” variety (Rowe & Grohmann, 2013; Kambanaros et al., 2013) of Turkish. The “discrete bilectalism” of “low variety” IT in Germany is unique, since Turkish children lack a formal register as well as a general access to formal education (i.e. in preschool) and literacy education for standard Turkish in Germany (Küppers et al., 2015).

2 The Assessment of Developmental Language Disorder in Bilingual Contexts

It is alleged that Developmental Language Disorder (DLD) appears with a prevalence rate of approximately 8% (Norbury et al., 2016). Hence, it is very common in children, especially if compared to genetic syndromes, for example. Research further indicates that DLD is a life-long condition characterised by difficulties with understanding and/or using spoken language and is likely a result of a number of biological, genetic and environmental risk factors (Bishop et al., 2016, 2017). Following the CATALISE recommendations, the term “DLD” is used for children whose language disorder does not occur with another biomedical condition, such as a genetic syndrome, a sensorineural hearing loss, neurological disease, Autism Spectrum Disorder or Intellectual Disability (cf. Stothard et al., 1998; Johnson et al., 1999; Tomblin, 2010). For epigenetic studies, Tomblin et al. (2008) proposed the EpiSLI criterion, based on five composite scores representing performance in three domains of language (vocabulary, grammar, and narration) and two modalities (comprehension and production). Children scoring in the lowest 10% on two or more composite scores are identified as having language disorder. Furthermore, Lancaster and Camerata (2019) point out that DLD should be seen as a spectrum condition.

2.1 DLD in Bilinguales and Bilectals

Given the heterogeneity of DLD, language assessment is generally difficult even to the point that clinically interpretable subtypes are unlikely (Lancaster & Caramerata, 2019). With respect to bilingual acquisition, evidence is clear that children acquiring a second language (L2) in childhood differ from monolingual age-matched peers in several aspects. In the area of morphosyntax, for example, certain linguistic patterns deviating from those of typically developing monolingual children are reported for children acquiring their second language, i.e. German or French (Hamann et al., 2013; Marinis & Armon-Lotem, 2015; Tuller et al., 2018). These distinctive patterns often overlap with those known for monolingual children with Developmental Language Disorders (Paradis, 2010). DLD is common among monolinguals and bilinguals (Leonard, 2010; Engel de Abreu et al., 2013). Therefore, DLD and bilingualism are challenging for research and practice to disentangle DLD specific patterns from L2 interlanguage phenomena. Thus, typically developing bilingual children (BiTD) may be misdiagnosed as having DLD. Several studies focusing on different languages have nonetheless shown that the quality and the quantity of errors differ in BiTD and monolingual children with DLD (MoDLD) (e.g. Paradis et al., 2008; Armon-Lotem, 2014; Meir et al., 2016; Tuller et al., 2018). Since DLD should affect all languages of an individual, it was proposed that the assessment of language disorder must respect both the child´s languages to avoid misdiagnosis.

The typical first language acquisition of Turkish has been in the focus of research for several years now (e.g. Aksu-Koç & Slobin, 1985). Moreover, knowledge on the delayed or disordered acquisition of Turkish, such as different forms of language impairment, phonological disorders, among others, increases constantly (Topbaş, 1997, 1999, 2005, 2007; Babur et al., 2007; Uzuntaş, 2008; Topbaş & Güven, 2008; De Jong et al., 2010; Rothweiler et al., 2010; Topbaş & Yavaş, 2010; Acarlar & Johnston, 2011, among others). These findings lead to the conceptualization and establishment of standard tests for DLD in Turkish (i.e. TELD-3: T, Topbaş & Güven, 2011; see Chapter 3.2 for more details; TİFALDİ, Kazak-Berument & Güven, 2010; T-SALT; Acarlar et al., 2006). Within the COST IS0804 action, Thordardottir (2015), for example, argues for the applicability of standardized assessment tools with a bilingual benefit to Z-scores for simultaneous bilinguals. In the same wake of the COST Action, cross-linguistically valid tools known as the LITMUS tasks (Language Impairment Testing in Multilingual Settings, Armon-Lotem et al., 2015), were developed, also for children with Turkish as heritage language, such as the Multilingual Assessment Tool for Narratives such as MAIN (Gagarina et al., 2012). Those LITMUS tasks aim at identifying Developmental Language Disorder (DLD) in bilingual populations.

2.2 The Assessment of Developmental Language Disorder with Sentence Repetition Tasks

Sentence repetition tasks (SRTs) are widely recognized as tools for the identification of specific language impairment in monolingual and bilingual children (Conti-Ramsden et al., 2001; Vinther, 2002; Klem et al., 2015). SRTs contain of fixed sentences that the participant repeats and thus generate a restricted set of obligatory contexts. They are subtests of most language testing materials and standardized tests for decades, since they are easy to use in clinical settings and have been shown to assess underlying grammatical representations (Polišenská & Kalpaková, 2014) as well as language processing (Archibald & Gathercole, 2006). In addition, evidence shows their applicability in bilingual contexts for distinguishing bilingual children with and without DLD (Meir et al., 2016; de Almeida et al., 2017; Hamann & Abed Ibrahim, 2017). SRTs have been argued to be more reliable than other language-dependent expressive and receptive language tasks, such as (for English) third person singular or past tense tasks for the assessment of DLD (Stothard et al., 1998; Conti-Ramsden et al., 2001). Note, however, that SRTs differ in their conceptualization. The German and the French versions of the LITMUS SRT, for example, focus on morpho-syntactic knowledge, and knowledge of computationally complex structures in particular (Hamann et al., 2013; Marinis & Armon-Lotem, 2015; Fleckstein et al., 2018).

For speech and language practice, SRTs combine several advantages over other testing materials: They aim at grammatical knowledge, are simple and fast to administer and easy to score (identical repetition yes/no). Moreover, they proved to have reasonable to good diagnostic accuracy for children with or without DLD in several language pairs, such as, for example, Turkish-German, Arabic-French, or Portuguese-German children (Hamann & Abed Ibrahim, 2017; Abed Ibrahim & Fekete, 2019; Chilla et al., in press; Hamann et al., in press). As Marinis et al. (2017) point out, LITMUS SR tasks can tease apart BiTD from MoDLD and from BiDLD in several countries and for several language combinations.

Thus, the use of SRTs qualifies as a promising pathway for the assessment of DLD in bilingual populations. Consequently, the use of L1 assessment tools in heritage language contexts, and especially monolingual SRT tasks as a measure for grammatical development is nowadays common practice in research and speech and language assessment. Ertanir et al. (2018), for example, use the SRT subtask of the TELD-3: T (Topbaş & Güven, 2011; see Chapter 3.2 for more details) for the assessment of Turkish kindergarten children in Germany. Their results strengthen the impression that mean performance level in Turkish grammar was below the norming sample mean. The authors argue that bilingual heritage language children show lower L1 grammar skills, if their performance was evaluated with a sentence repetition task. Ertanir et al. (2018) conclude that their results are in line with earlier research observing lower language levels in L1 and L2 (e.g., Caspar & Leyendecker, 2011; Akoğlu & Yağmur, 2016), although the sample in their study consisted of children with well-developed vocabulary skills in their L1, even when compared with monolingual norms.

It is at this point that this study hopes to contribute. Current trends in the assessment of language difficulties and disorders in bilingual children are often unaware of differences between the standard variety and the (emergence of) a dialect in heritage languages. This study aims at filling this gap by focusing on one of the most frequent first languages in Germany, Turkish, showing that the Immigrant Turkish dialect is the major heritage language (L1) input for Turkish-German children. We hypothesize that the bilectal situation of the Immigrant Turkish dialect has an impact on the individual performance of bilingual heritage language children with Turkish as L1 even for sentence repetition tasks. We will show that the appreciation of the Immigrant Turkish spoken dialect has indeed an impact on the construction, scoring, and outcome of standardized language tests. Thus, the study sheds light on the (non-)applicability of SRTs for bilingual children acquiring this specific dialect variety of the standard language Turkish as a minority language in western European countries.

3 The Immigrant Turkish Dialect as a Test Case for Standardized Assessment Tools in Bilectal Contexts

3.1 Participants, Materials and Methods

In our BiliSAT and BiLaD projects (see below), data of 61 Turkish-German and Turkish-French children, 52 bilingual typically developing (BiTD), and 9 children with DLD (BiDLD), was gathered. Both projects established the clinical status of the bilingual participants by applying standardized tests in both languages of a child and regarding a “child with DLD” if she scored below adjusted norms in two language domains in each of the languages (cf. Tuller et al., 2018). All participants were tested with a broad assessment procedure (cf. Hamann et al., sub.), including standardized tests in the L1 and L2 (Hamann & Abed Ibrahim, 2017; Tuller et al., 2018; Chilla et al., in press), respecting dominance effects on test performance. Adjustment of monolingual norms was performed following Thordardottir’s (2015) recommendations and by carefully establishing language dominance. Relevant background information was collected with the Questionnaire for Parents of Bilingual Children (PaBiQ; Tuller, 2015).

The analysis here is based on the data subset of 52 typically developing Turkish-German children (age range 5;0-12;4, with 32 boys and 20 girls). 52 SRT subtests from TELD-3: T (Topbaş & Güven, 2011) are taken into account. The TELD-3: T is a norm-referenced test for the Turkish competence of children and an adaptation of the English language assessment tool TELD-3 (Hresko et al., 1991). It includes receptive and expressive language performance in children (2;0- 7;11), using two forms (Form A and Form B). The test aims at identifying a child’s strengths and weaknesses in different language areas as morphology, syntax and semantics and is suitable for the assessment of language delays. Scoring covers expressive, receptive, and global language performance, the latter being a composite value.

Further SRT data was taken from 21 data sets of the TÖDIL (Topbaş & Güven, 2017) sentence repetition subtask. The TÖDIL is an adaptation of the English language assessment tool “Test of Language Development-Primary: Fourth Edition” (TOLD-P:4; Hammill & Newcomer 2008), being a norm-referenced and standardized test for the Turkish competence of children between 4;0-8;11. It intends to provide professionals with a measure for examining receptive, expressive, and organizational language skills and comprises of nine sub-tests such as picture vocabulary, syntactic understanding, sentence repetition, morphological completion, grammar and phonology skills. They include three measures each for listening and speaking abilities.. The combination of all nine sub-tests claims to cover general spoken language abilities. Only children without a risk for DLD and who scored above percentile rank 9 (IQ score ≥ 80 according to Wechsler’s IQ scale) were included in the current study.

3.2 Analysis

Both standardized tests were administered as per description. The children’s responses on the TELD 3: T and the TÖDIL were recorded using special dictaphones. Data transcription, verification and coding for errors were done offline by two independent linguistically trained raters (percentage of agreement was at least 90%). For each repetition measure, the percentage of correct responses was used as basis for data analysis (cf. also Abed Ibrahim & Fekete, 2019). The scoring procedure followed the test handbook, with 1/0 for correct/incorrect repetition of a sentence. Further qualitative analysis classified the incorrect repeated sentences into error types in terms of Immigrant Turkish dialect features (ITfeat), error types that pattern monolingual Turkish DLD (DLDfeat), children or neither of both or unclear (UN). Null reactions were counted as errors, unless they were due to technical problems or errors by the investigators (missing data, less than 1% of the overall data).

3.3 Results

A total of 547 sentences from the SRT subtests was analysed (n TELD 3:T (SRT) = 349; n TÖDIL (SRT) = 198), with a correctness rate of 36% (TELD 3: T: 112/349 – 32%; TÖDIL: 40/198 – 20%). 87 sentences were errors of unclear origin. 152 (36%) incorrectly repeated sentences showed features that pattern errors known from monolingual Turkish speaking children with DLD.

The analysis here focuses on the remaining 156 incorrect sentences (TELD 3: T = 100/349 – 29%; TÖDIL: 56/198 – 28%). The children in our study repeated the sentences from the SRT using patterns typical for the Immigrant Turkish dialect.

These features are, for example, omission of possessive markers in genitive-possessive constructions, that appeared in 8% of all incorrect sentences (13/156 – 8%) (cf. example 8).

(8)

Omission of the possessive marker in genitive-possessive constructions

 

Standard Turkish (TÖDIL SRT item number 34):

 

Dün

öfkeli bir

kaplanın

pençesinden

zor

kurtarıldık.

 

Yesterday

angry INDEF

tiger-GEN

paw-POSS-3SG-ABL difficult rescue-PASS-PAST-1PL

 

‘Yesterday we hardly survived the paws of an angry tiger.’ Immigrant Turkish dialect:

 

Dün

kaplanın

pençeden

zor

kurtarıldık. (041432; 12;0)

 

Yesterday

tiger-GEN

paw-Ø-ABL

difficult

rescue-PASS-PAST-1PL

More, substitutions of case markings (22/156 = 14%; DAT for ACC: 8; ACC for DAT: 2; LOC for ACC: 2; DAT for ABL: 3; ABL for DAT: 7)) or the omission of obligatory case markings (20/156= 13%; DAT: 6; ACC: 13), and, especially with genitive (20/156 = 13%), were observable (cf. example 9–11).

(9)

Substitution of dative with accusative; omission of dative

 

Standard Turkish (TEDIL SRT Item Number 27d)

 

Zeynep arkadaşlarına

ve

öğretmenine

hediye

verdi.

 

Zeynep

friend-PL-POSS-3SG-DAT

and

teacher-POSS-3SG-DAT present

give-PAST-3SG

 

‘Zeynep gave present to her friends and teacher.’

 

Immigrant Turkish dialect:

 

Zeynep

ve

arkadaşlarını

birşey

alıyordu. (024332; 9;2)

 

Zeynep

and friend-PL-POSS-3SG-ACC

something

buy-PROG-PAST-3SG

 

Zeynep arkadasi

ve

ögretmenine

hediye

verdi. (BAY; 5;1)

 

Zeynep

friend-3SG-POSS-Ø and teacher-POSS-3SG-DAT

present

give-PAST-3SG

(10)

Substitution of ablative with dative

 

Standard Turkish (TODIL SRT Item number 9)

 

Fabrikadan

çıkınca

çocuklar

arabayı

tamir ettiler.

 

Factory_ABL

come out-SUB

children

car-ACC

repair-PAST-3PL

 

‘When the children went out of factory, they repaired the car.” Immigrant Turkish dialect:

 

Fabrikaya

(04432; 9;3)

çıkarken

çocuklar

arabayı

unuttu.

 

Fabrika-DAT

go-out-SUB

child-PL

car-ACC

forget-PAST-3SG

(11)

Omission of genitive case, omission of accusative case; substitution of dative with accusative

 

Standard Turkish (TÖDIL SRT Item Number 14)

 

Kadın

adamın

kendisini

sevdiğine

inanmadı

 

Woman

man-GEN

self-ACC

love-CV-POSS-3SG-DAT

believe-NEG-PAST-3SG

 

‘The woman did not believe that the man loves her”.

 

Immigrant Turkish dialect:

 

Kadın

ama adamı

kendisi

sevdiğini

 
 

Woman but

man-ACC-Ø

self-Ø

love-CV-POSS-3G-ACC

 
 

inanmadı.

   

(040432; 9;3)

 

believe-NEG-PAST-3SG

   

Further, lexical dialect levelling (10/156 = 6%) was found, as well as blending of omission and substitution in the same sentences (cf. example 11). Note, however, that the avoidance of complexity by using finite clauses (40/156 = 26%) was most prominent among all errors (cf. example 12), and that several errors would appear in the same repeated sentence.

(12)

Finite clause instead of adverbial subordination

 

Standard Turkish (TÖDIL SRT Item number 9):

 

Fabrikadan

çıkınca

çocuklar

arabayı

tamir ettiler.

 

Factory_ABL

come out-SUB

children

car-ACC

repair-PAST-3PL

 

‘When the children went out of factory, they repaired the car.’

 

Immigrant Turkish dialect:

 

Fabrikadan .

çıkmışlar

Çocuklar

arabayı tamir etmişler. (019332; 11;3)

 

Factory-ABL

go out-EVD-3PL

child-PL

car-ACC repair-PAST-EVD-3PL

 

Fabrikadan

çıktı.

Çocuklar

arabayı tamir ettiler. (036432; 11;2)

 

Factory-ABL

go-out-PAST-3SG

child-PL

car-ACC repair- PAST-3PL

Importantly, these sentences are correct by Immigrant Turkish dialect standards. Turkish- German bilingual children make use of the dialectal variety in the sentence repetition task, processing and understanding the sentences in the standard variety of Turkish correctly, and repeating them in their spoken Immigrant Turkish dialect.

4 Discussion: The Immigrant Turkish Dialect as a Heritage Variety and Its Implications for Language Assessment and Education

Immigrant Turkish as heritage language for bilingual children in Germany reflects the necessity of an acknowledgement of dialect input for language assessment. From a sociolinguistic point of view, it is remarkable, how differences between the Immigrant Turkish and standard Turkish have long been unattended as a factor most relevant for the validity of assessment tools in bilingual contexts. This might be due, however, to a lack of systematic investigations with broader populations of bilingual children with and without DLD in several countries.

The studies carried out within the IS0804 and the bi-sli networks, however, allow for new insights to the relevance of dialects for language input and assessment, since they provide research with a broad database and a fair number of participants for linguistic study. Further, earlier studies struggled with the (im)possibility of disentangling bilingual children with DLD from typically developing children (i.e. Paradis, 2008; Armon-Lotem et al., 2015; Tuller et al., 2018), so that reliable data for the evaluation of assessment tools in bilingual and bilectal populations is only emerging (i.e. Marinis et al., 2017; Theodorou et al., 2016; Abed Ibrahim & Fekete, 2019; Leivada et al., 2019; Chilla et al., in press).

Our results confirm former studies, underlying systematic differences between the standard variety or dialects in the country of origin, and the Immigrant Turkish dialect. The omission and/or substitution of case markings (i.e. Cindark & Aslan, 2004), as well as changes in genitive-possessive constructions (Dirim & Auer, 2004), and, most relevant, the use of finite and/or co-ordinated sentences instead of more complex structures continue to be prominent features of Immigrant Turkish dialect. If compared to monolingual speakers of the standard variety of Turkish, bilingual speakers of IT avoid complexity. Non-finite sentences are more prominent among bilingual IT speakers than finite clause coordination, and juxtaposition is more common than complex embedding (Treffers-Daller et al., 2006; Bayram, 2013; Herkenrath, 2014; Schroeder, 2016). However, some features, such as an overuse of ablative forms in locative contexts, or an omission of interrogative particles in yes-no questions, are characteristics of dialect levelling or, as for genitive-possessive without possessive marker, common in informal spoken Turkish and some dialects (i.e. Csató & Johanson, 1998).

The robustness of the Immigrant Turkish dialect as a heritage language for bilingual Turkish- German children is evident. Even if language proficiency was measured by an easy-to- administer and age- and language-appropriate task, IT children tend to repeat the sentences in the dialectal variety. However, sentence repetition tasks should be robust of language change phenomena, if the (in)correctness of answers was based on working memory capacities, only. The sentence repetitions of the IT dialect-speaking children here, though, refer to structural changes and to systematic deviations from the standard variety, and to the necessity and meaningfulness of grammatically motivated SRTs (i.e. Hamann & Abed Ibrahim, 2017). Since the data provided here contains of a homogeneous group of Turkish-German bilingual children without DLD, who took part in a comprehensive assessment procedure (i.e. Abed Ibrahim & Hamann, 2019) the high error rate in a SRT should not result from language disorder, children being under age or on limited cognitive development. The corpus is furthermore representative for the Turkish-German population of heritage children in Germany, since participants from different German Federal Countries (i.e. Baden-Württemberg, Berlin, Hamburg, Hessen) and from different living environments (cities/rural areas) attended. It is also true that the majority of test items (64%) was processed and repeated in the standard Turkish model, as expected.

If the scoring procedure considered IT sentences as correct, the overall correctness rate would increase considerably. Note, however, that the data also provide further evidence for an overlap between Immigrant Turkish and DLD features in bilinguals (i.e. Babur et al., 2007; Rothweiler et al., 2013; Topbaş et al., 2016; Chilla & Şan, 2017), since there are nearly the same number of sentences in the corpus (101 in the TEDIL and 51 in the TÖDIL), which are likewise characteristic for DLD in Turkish monolinguals. These features are, for example, the substitution of case markings (i.e. accusative for dative), or the omission of obligatory elements or suffixes. It is also true that the reduction of syntactic complexity is a distinctive feature of DLD in Turkish.

Thus, to avoid misdiagnosis, scoring of language proficiency in the bilectal context of Immigrant Turkish should not just rely on a transformation of raw scores based on knowledge of the dialectal input of the child, in the sense of adding a “bilectal error bonus”, as it has been proposed for bilinguals. Rather, further systematic study on the qualitative and quantitative differences between the language performances of IT speaking children with and without DLD with sensitive error type analysis could lead to a better understanding of clear patterns of DLD vs IT, respectively. Prospective test design should contain interpretation variability with respect to dialectal and/or DLD outcomes, and scoring (cf. Leivada et al., 2019). First steps have already been explored by, for example, Hamann and Abed-Ibrahim (2017); Theodorou et al. (2017); Abed-Ibrahim and Fekete (2019); Chilla et al. (in press).

Further studies might moreover investigate the specific heritage language situation of the IT dialect: Most IT-speaking children have no access to formal Turkish or literacy education. Education and assessment should further withdraw from the construction of homogeneous groups of “first language” children and adults in diglossia, bilingual and heritage language populations, implying sufficient language testing with assessment tools for monolingual contexts.

Sensitive qualitative research with respect to language attrition vs. IT vs. DLD features at different ages with broader cross-sectional studies would contribute to a dialectal-fair development of testing materials for bilectals, and especially for a population as large as this of Immigrant Turkish as dialect speakers in Germany.