Is Segregated Language Support Fit for Purpose? Insights From German Language Support Classes in Austria

In Austria, segregated German language support classes (GLSC) were introduced in the school year 2018/19 to intensively support students who had previously little or no contact with German, the official language of instruction. These classes have been widely criticised; however, a formal evaluation of their effects has yet to be published. In absence of this evaluation, this article describes the language support model as it currently exists in Austria and reviews existing evidence about its efficacy. The literature review synthesises findings from educational research undertaken in other contexts that offer insight into features of ‘good practice’ in language support models. The article then explores the extent to which GLSC comply with these features. As such, this review allows insights into ways of ensuring students’ language and socio-emotional development – all central aspects of academic success – in language support models. It therefore allows research-informed understanding of the effects of the newly implemented model of German support classes in Austria and makes recommendations for further development.


Introduction
Across the European Union (EU) within the context of inclusive education, schools have formally embraced policy initiatives to ensure that all students receive the best possible opportunities for academic as well as socio-emotional development (see Schwab, 2020). Ensuring inclusion in schooling has been additionally challenged since 2015, when millions of refugees and migrants fled to Europe. The Organisation for Economic Co-operation and Development (OECD, 2018b) reported that almost one in four 15 year-old students in OECD and EU countries is foreign-born or has at least one foreign-born parent. These students also typically speak other languages than the language of schooling at home. Research continually emphasizes that many of these students face significant disadvantages in the educational system in terms of enrolment in school type, retention, progression, dropout rates and types of school diploma attained (European Commission, 2008;OECD, 2018a;Suchań et al., 2019). In Austria, students from a migration background -both those who immigrated after birth and were born in Austria and are often multilingual -have significantly lower reading skills in the language of education than students without a migration background. The country is ranked amongst the fourteen lowest performing OECD countries in this area (Suchań et al., 2019).
Due to the increase of students from refugee and migrant backgrounds and the ongoing challenge to enhance their educational outcomes, various educational models have been developed (see, for example, Brüggemann & Nikolai, 2016, for Germany; Ziomas et al., 2017, for Greece;Moinolnolki & Han, 2017, for the US). European countries have tried to develop strategies for integrating 'newcomer' refugees and immigrants into mainstream education, including programmes to ensure that students develop proficiency in the language of education. Segregated language support classes which occur outside regular classroom teaching have been implemented in a number of EU member states as 'the solution' to the challenges of including students from migration backgrounds in education (European Commission, 2019). Since this increased challenge is relatively recent, there is little research that investigates the long-term benefits and/or weaknesses of the various language support models that have been developed so far in Europe (Koehler & Schneider, 2019). In Austria, segregated German language support classes (GLSC) called Deutschförderklassen were introduced in the school year 2018/19. These classes aim to intensively support students who have had little or no previous contact with German, the language of instruction, outside of school. The introduction of these classes has been widely criticised by language learning specialists and educational researchers, as this model goes against research-informed notions of language education as well as the promotion of inclusion and emotional wellbeing in education (e.g., Döll, 2019;Müller & Schweiger, 2018;. In Austria, there has been a dearth of research dedicated to understanding the needs of multilingual students and the effectiveness of language support models. While a study on the former model was conducted (Opriessnig et al., 2019), it was published after the launch of the new support model and therefore not taken into account when reforming language support. Until now, no formal and thorough evaluation of the current language support classes has been published.
Further, as only one (untested) model of GLSC was implemented nationally, it has not been possible to evaluate its effects on students' academic and socio-emotional development using suitable research methods such as a control group design. In absence of an evaluation of the current GLSC, an alternative means of determining their effectiveness is urgently required. In order to fill this gap, this article sets out to answer the following research question: What do insights from existing evidence tell us about the efficacy of GLSC?
In order to answer this question, a review has been undertaken that draws on two main types of evidence: 1) existing research on the German support model in Austria and in other contexts where language support models for newcomers have been implemented, and 2) meta-studies, systematic reviews and highly respected international publications investigating features of well-designed and implemented language support programmes and findings from the decades of influential research evaluating structured immersion and dual language education models in a range of international contexts (e.g., Genesee et al., 2006;.
Following this, the article identifies features of 'good practice' in language support models which can be gleaned from state-of-the-art research in language and inclusive education, and explores the extent to which GLSC, as currently implemented, comply with these features. As such, this review allows insights into established means of ensuring students' language and socio-emotional development -all central aspects of academic success. It therefore allows an unprecedent understanding of the potential effects of the current model of GLSC in Austria, which need to be urgently understood in absence of empirical research being devoted to this issue.

Context: German Language Support Classes (GLSC, Deutschförderklassen) in Austria
In the 2019/20 school year, almost 27% of the 1,135,519 students enrolled in schools in Austria had German as an additional language (GAL) -in the city of Vienna the proportion is about 53% (Statistik Austria, 2021). Students with GAL and from migration backgrounds face significant disadvantages in the Austrian educational system (Erling et al., 2020;Herzog-Punzenberger, 2017;Schreiner et al., 2020;Suchań et al., 2019). Based on Austria's low results in PISA 2018, the Austrian Court of Audit released a critique of the education system. In response, policymakers introduced a new model of GLSC (Deutschförderklassen) for students with beginning and emergent German language competences. According to the Government's website, the goal of the GLSC is: the early and intensive learning of the language of instruction, German, so that these students can be taught together in the class as quickly as possible according to the curriculum of the respective school type and grade. (Federal Ministry of Education, Science and Research of Austria, 2019, authors' translation) Since the 2018/19 school year, this model has been rolled out across Austria. It requires that head teachers interview newly enrolled students to find whether they have to take a standardised assessment (the MIKA-D † ). While newcomer students are commonly selected for taking the test, it can be given to any new student, regardless of how long they have been living in Austria. In the assessment, students are introduced to varying stimuli (e.g., pictures, a finger puppet) which are designed to evoke specific spoken language responses. These responses are then evaluated according to indicators. Based on the results of this test, one of the following three scenarios ensues: 1. Students deemed to have sufficient language competences attend mainstream classes.
2. Students deemed to have emergent German language competences attend mainstream classes and receive additional language support for six hours per week in segregated "pull-out courses".
3. Students deemed to have no or beginning German language competences attend segregated GLSC for 15 hours per week in primary school or 20 hours per week in lower secondary school.
The recommended scenario can be extended for up to four semesters (i.e., two school years), depending on the student's progress. After the end of each semester, students receiving additional German support are given the same standardised assessment, with the results again determining which one of the possible scenarios will ensue: While students following scenario 2 who improve their test scores transfer into mainstream classes, students following scenario 3 either transfer into pull-out courses or mainstream classes, depending on their results. If students' test results following either scenario do not improve, they remain in the designated support model for at least another semester. Transitions into the mainstream classroom can only occur by the end of the winter term. Students who do not make the transition by then have to repeat the grade, which means that they will have a different teacher and classmates the following year, and will have to deal with the additional stigma that often accompanies grade repetition (Lessard et al., 2014).

Methodology
Because a full evaluation of the current GLSC has yet to be undertaken, an alternative means of determining their effectiveness was required for this study. As a first step in compiling insights about the efficacy of GLSC, we undertook a literature review of existing research in this area, critically analysing and examining the main findings and themes central to this topic. We first reviewed the few studies that had already been conducted on the German support model in Austria (e.g., Opriessnig et al., 2019;Schwab et al., 2020), and also examined primary data giving insight into students' success within this model (e.g., Statistik Austria, 2019). As only a limited amount of research has been conducted in this area to date, the review was extended to include studies conducted in similar European contexts where language support models for newcomers have been implemented (e.g., Karakayalı et al., 2016;Nilsson & Bunar, 2016) as well as international reports and evidence-based policy briefs which compared different school arrangements and approaches to pedagogy for newcomer students (e.g., Crul, 2017;European Commission, 2019;Koehler & Schneider, 2019;UNESCO, 2019).
The second type of literature reviewed explored features of well-designed and implemented language support programmes synthesised from a number of meta-studies and systematic reviews in this area. Literature included in this part of the review had to be published fairly recently (i.e., since 2010) in highly respected channels and had to feature evidence-based recommendations (e.g., Faulkner-Bond et al., 2012). Falling into this category were seminal publications from the decades of research dedicated to evaluating structured immersion and dual language education models in a range of international contexts (e.g., . Insights from studies on dual language education undertaken in Austria and other German-speaking contexts are also included. Such studies indicate that, while far less commonly implemented than structured immersion, dual language education models offer much more potential not only with regard to language learning and academic attainment but also to inclusion and socio-emotional development. The review thus synthesises aspects of the dual education model that illustrate features of 'good practice' in any type of language support programme, relying on recommendations from state-of-the-art research and, where available, metastudies and systematic reviews on language-sensitive, inclusive education.

Insights From Existing Research on GLSC
Insights into effectiveness of the GLSC can be drawn from the only existing study on the former language support models as it existed in Austria before 2019 (Opriessnig et al., 2019). This study, which focused on the school years 2016/17 and 2017/18, compared the two language support models that existed at that time: Either German support took place inclusively, i.e., within mainstream classrooms in so-called language support courses (Sprachförderkurse), or within segregated language starting groups (Sprachstartgruppen). Within both models, students received language support for 11 hours per week for a duration of a maximum of two years. This study did not find significant differences between the two models regarding their effects on students' German language abilities, indicating that separating students in pull-out courses is not required for successful language learning.
Further insights into the effectiveness of the German support courses can be gleaned from a large-scale study that explored primary and secondary school teachers' perceptions of the effects of German support courses during the first year of their implementation in 2019/20 Resch et al., 2022). A total of 1,267 teachers from different school types (primary and secondary schools) in eight federal states in Austria who worked in schools that had introduced at least one German support class took part in the survey. Results showed that more than 80% of the participants would prefer an inclusive language support model to a segregated model. The new model of GLSC was rated (rather) negatively by more than half of the teachers surveyed. Among the reasons for teachers' negative perceptions of these classes were that they were under-resourced and that class sizes were too large. Currently, there is only one teacher per class and usually around 22 students. More than 96% stated that at least two teachers needed to be in the classroom. Further, teachers reported that, ideally, such classes should comprise a maximum of 16 students (see Gitschthaler et al., in press). Such findings corroborate results from interviews with teachers and school principals (Schweiger & Müller, 2021), which reveal the difficult circumstances that are linked to GLSC. Schools do not have adequate space to accommodate separate classes; the change between regular and support class brings unrest into the daily routine; and the heterogeneity of students' ages and German competences requires above-average teacher preparation and expertise. However, teachers also perceived advantages of the GLSC and some felt that students can be integrated more easily into mainstream classes after attending them because of the intensive German learning opportunities offered. Some also perceived segregated classes as a 'shelter' in which students are protected from being stigmatized as 'problem children'. Moreover, there were teachers who perceive positive effects of the model for students in mainstream classes, who are separated from students who are perceived to be 'hindering' their academic growth (Resch et al., 2022).
Another means of assessing the effectiveness of the current GLSC is by examining the number of students who transition from GLSC into mainstream classes. An analysis of official data published by Statistik Austria (2019) shows that at the end of the school year 2018/19 (one year after the current model was introduced), only around 32% of the students in such classes made the transition to mainstream classes (Scenario 1). About half of the students (48%) were transferred to additional language support "pull-out courses" (Scenario 2). Around 16% continued in the GLSC (Scenario 3) and 4% dropped out of the school system all together (Statistik Austria, 2019). This suggests that the current model is not managing to support the transition to the mainstream classroom within one school year for the majority of students.

Findings From Research on Language Support Classes in Other European Contexts
While there has yet to be substantial research investigating the impact of GLSC in Austria, similar language support models were evaluated in other European contexts, e.g., in Germany, the Netherlands and Sweden. Studies in this part of the review all deal with the question of whether it is possible to achieve inclusion of language minority students through education models in which students are excluded from the mainstream classroom. While this work provides relevant insight for the Austrian context, the studies focus primarily on the inclusion and language education of 'newcomers', and it is important to note that GLSC in Austria are not reserved for this category of student.
In research about newcomer education in Sweden, it has been found that, overall, including newcomer students in the mainstream classroom is the only way for education to truly reach its objective of inclusion, though it is acknowledged that this is often not the model followed in most schools. When students are segregated into pull-out classes, it is recommended that they have at least several school subjects with their mainstream class so that they maintain a connection to that group (Berglund, 2017). Separative models have been found not to encourage social interaction with majority language students, and to negatively influence students' performance in school since other subjects like mathematics or natural sciences are often neglected or only shallowly taught (Nilsson & Bunar, 2016). However, both inclusive and pull-out models can 'get it right' and 'wrong', depending on the support offered to students (Tajic & Bunar, 2020;cf., Terhart & von Dewitz, 2018). Just as separation can happen within the integrative setting of the mainstream classroom, sometimes it is easier to implement inclusive practices in the protective space of the excluding space of the pull-out class. Based on this, it is recommended that social and pedagogical provision is extended far beyond pull-out classes into the mainstream classroom (Nilsson & Axelsson, 2013;cf., Terhart & von Dewitz, 2018).
Also in the Netherlands, evidence about the effectiveness of language support classes for newcomers is ambiguous (Bilgili, 2019). There is a need for schools to actively promote social cohesion and cultivate a culture in which multiculturalism is valued in schools, as well as to develop expertise in support language development for newcomers. In exceptional situations, pull-out language support units have found to be advantageous; however, this is restricted to well-resourced, short-term selective models (Crul, 2017;Koehler & Schneider, 2019). In such situations, separate welcome or reception classes in which students learn the basics of the language of education before being incorporated into mainstream classes have been found to be beneficial. Short-term segregation can provide a sheltered environment for a limited period of time until students are ready to transition to mainstream classes, avoiding stigmatisation along the way. Once students have achieved a basic knowledge of the language of education, however, being in the mainstream classroom with proficient speakers of the language of education has been found to be essential for speeding up the learning process and reducing the possible negative outcomes of segregation (European Commission, 2019; Koehler & Schneider, 2019).
The success of short-term segregation seems to be the reasoning underpinning the 'Welcome Classes' designed in Germany for newcomer students. A review of this model has found that being placed for one or two years in these classes can be detrimental to school success, due to students often not having enough engagement with communicative speakers of the language of education and being perceived as a distinct group by those in mainstream classes (Karakayalı et al., 2016). Terhart and von Dewitz (2018) also found that segregating newcomer students into different classrooms, sometimes even in different buildings, has an exclusionary effect, having the potential to lead to social marginalisation within the school. Moreover, in this model, teachers are often ill-prepared, do not have appropriate qualifications and work outside the mainstream educational system on temporary contracts; moreover, there is a limited curriculum with no clear pathway of transition to mainstream classes (Gefäller, 2017;Karakayalı et al., 2016). A particular challenge is the ability to respond to students' psychological needs, and supporting them to develop their socio-emotional wellbeing (Koehler & Schneider, 2019).
The overwhelming message coming out of such research is that inclusive education models have the best potential for newcomer students-with short-term pull-out courses being a potential safe space option for newcomers for a limited period of time. However, for inclusion to be successful, there is a need for high quality teachers and support mechanisms to support the continued language and socio-emotional development of students in the mainstream classroom and beyond.

Features of 'Good Practice' in Language Support Models
The research reviewed so far suggests that GLSC as they currently exist are not achieving their goals, particularly with respect to inclusion. However, there is little information about the efficacy of the model in terms of language learning. Information on good practice in language education can be gleaned from research that has evaluated language education models in a range of international contexts. The two main language support models currently in use in Europe to transfer students into regular schooling are structured immersion and dual language education. Structured immersion best describes the current German support model and is the most common model of language support in the European Union, offered in 24 out of 27 member states (European Commission, 2019, p. 82). In this model, students with beginning skills in the language of instruction attend separate, "pull-out" language support classes (Baker, 2011: pp. 194). Language learners have an immersive experience with language support, so that they can learn the language of education rapidly and benefit from academic content instruction in mainstream classrooms. Lessons are typically based on the curriculum, with teachers using a simpler form of the language of instruction and students receiving language support. This pedagogical approach does not usually support the further development of students' home languages within school settings (Ray-Subramanian, 2011). In contrast is dual language education, which has as its goal that all students become fully bilingual and biliterate in the majority national language of education and a minority language. All students are together in one class and high-quality, curricular instruction is alternated between the languages. In such programmes, students are encouraged to maintain and further develop their home languages.
Seminal research on the effectiveness of different language support models on students' second language acquisition and academic performance in other subjects has been undertaken in the United Statues by Thomas and Collier, who have collected data on various models of language support for more than 30 years (e.g., Thomas & Collier, 2002). The structured immersion model has been found to achieve some success with regard to promoting language and academic development. However, such programmes offer no guarantee that the language taught supports the development of curriculum knowledge. Moreover, structured immersion programmes promote the idea that once students move into the mainstream classroom, language development is complete and no further support is necessary, which has not been found to be the case. Empirical evidence suggests that sustained, integrated language support is needed, ideally from the pre-school to the upper-secondary level Crul, 2017;UNESCO, 2019). Thus, the long-term, integrated dual language education model has been found to be is far more effective than structured immersion for developing proficiency in the language of schooling, and this has been found to be the most powerful predictor of language minority student achievement Hofstetter, 2004). Only in this model have language minority students been found to achieve at the same or higher levels than their majority language peers -a finding replicated in studies and meta-analyses conducted in a range of global contexts (e.g., Baker, 2011;Cummins, 2000;Genesee et al., 2006;Lindholm-Leary & Hernandez, 2011). Importantly, dual language programmes have not only been found to be more beneficial for language minority students, but also for speakers of the dominant language, with students outperforming peers not in dual language programmes, regardless of ethnicity, socioeconomic status and special needs (Genesee et al., 2006;Lindholm-Leary & Genesee, 2014;Thomas & Collier, 2002).
In the German-speaking world, only a few dual language education models have been developed and evaluated (Gogolin et al., 2011;Möller et al., 2017). The evaluations that exist tend to corroborate the findings of Collier (2002, 2017). In Hamburg, for example, in an effort to address increasingly multilingual student population and reduce the educational gap between students with a migration background and German-dominant peers, dual language primary school programmes were implemented in 2000. In these schools, classes are made up of language majority and language minority students and the national curriculum is delivered through both languages, with the goal of full bilingualism and biliteracy (with German-Portuguese, German-Italian, German-Spanish and German-Turkish). When evaluated, it was determined that six years of bilingual instruction were able to provide higher school outcomes independent of students' socioeconomic status, their families' educational attainment level and their individual cognitive abilities (Duarte, 2011). While there are no dual language programmes in Austria in which migration languages are taught, there are dual language schools in the federal state of Carinthia for German and the autochthonous language Slovene. Such programmes were initially developed for families interested in reviving their often 'lost' mother tongue but have also grown popular for students with low or no previous knowledge of Slovene, due to their prestige (Purkarthofer & Mossakowski, 2011). Evaluations have established that German-speaking students achieve a solid basis for Slovene as an additional language, which also supports their learning of further languages without negative impact on their German competence. In addition, Slovene-speaking students maintain or further develop their use of a heritage language and develop proficiency in German. Students with other home languages (e.g., Serbo-Croat-Bosnian) gain from learning Slovene, as it strengthens their language awareness and their path to German. Moreover, students' positive identities as multilinguals are strengthened (Purkarthofer & Mossakowski, 2011).
The overwhelming message emerging from this research synthesis is that dual language education is more effective than structured immersion with regard to language learning, school achievement, the development of positive self concepts and promoting inclusion. Despite this, it is much less likely to be implemented. The research reviewed has also found that, if properly implemented, certain features of the dual language education model can contribute to more effective results in any type of language support programme (European Commission, 2019; Faulkner-Bond et al., 2012). Those features that seem most relevant to GLSC when considering the findings of research undertaken in Austria and in other newcomer education contexts are high quality language-aware teacher education, high quality language-focused instructional practice and promoting inclusion and socio-emotional development in and through language education, all of which are explored in more depth below, using findings from state-of-the-art research on language-supportive education.

High quality, language-aware teacher education
The implementation of high-quality inclusive language support requires extensive preparation of all teachers in initial and further teacher education, with them all having expertise in supporting language development (Wernicke et al., 2021). A full supply of teachers with qualifications to teach second-language learners in multilingual classrooms is required (Brisk & Kaveh, 2019). In addition, all teachers should be "linguistically responsive" (Lucas & Villegas, 2011), not only holding positive beliefs about multilingualism but also being prepared to use students' linguistic repertoires to improve their language development (García & Wei, 2013;Jessner & Allgäuer-Hackl, 2020). For this, teachers need basic knowledge of sociolinguistic and psycholinguistic theory. Further, they need a research-informed understanding of linguistic diversity as the norm (Blommaert, 2010) and a view of languages as dynamic and complex, with fluid boundaries between them and not as separate entities (May, 2013). This includes awareness of the hierarchies of language, the socio-political history of linguistic diversity and multilingualism, immigration history, language loss and/or language shift, and how these play out in educational systems (Duarte & Kirsch, 2020). This critical language awareness has been deemed necessary as, even in programmes where minority languages are promoted, it can be challenging to achieve bilingualism and biliteracy and combat powerful hegemonic ideologies that privilege the dominant language and ethnic groups (Flores & García, 2020;Hamman, 2018). To shift established hierarchies, teacher education programmes also need to actively recruit teachers who are models of multilingualism and represent language minorities (Melo-Pfeifer, 2018).

High quality, language-focused instructional practice and resources
High quality instructional practice is central for determining students' success in any language support model (Faulkner-Bond et al., 2012;Irby et al., 2019). In fact, the success of dual language education can largely be attributed to the high quality instructional practice that scaffold students' learning appropriately . The strongest programmes include both dedicated language instruction and specialized content instruction, for example drawing on pedagogies developed for content and language integrated learning (CLIL) (Ball et al., 2015;Coyle, 2007). In such programmes, learners learn the new language in context, through the learning of content knowledge, through social interaction with teachers and peers and through drawing on their full linguistic repertoire (Cummins, 2015;Erling, Clegg et al., 2021). Students are scaffolded through tasks in which they need to make sense of and form their own ideas about subject content, using new language, while also being introduced to (language) learning strategies so that they gradually become more independent learners who rely increasingly less on the teacher as the main holder of knowledge (van Kampen et al., 2017). Thus, teachers need to develop didactic methods based on constructivism and cognitive theory, learner-centred methods, active learning, cooperative and collaborative learning and communicative language teaching.
Central to high quality instructional practices are also translanguaging pedagogies, in which learners' multilingual repertoires can be used to raise outcomes of students who may not yet have full access to the language of instruction. Languages are used flexibly, so that students can draw on a wide range of linguistic resources to benefit content and language learning, rather than being limited to their emergent competence in the language of education (Duarte, 2019;Erling & Moore, 2021;García & Sylvan, 2011). This has the added benefit of valorising students' identities and positively influencing their socio-emotional development (Conteh, 2018;García-Mateus & Palmer, 2017). Learners whose translanguaging practices are legitimated often have higher learning motivation, are more psychologically centred and are more embedded in their culture and their community (García & Kleyn, 2016). Finally, these language-supportive instructional practices should be promoted not only in all subjects, but throughout primary and secondary level, accompanied by high-quality resources that are relevant to the national curriculum (European Commission, 2019).

Promoting inclusion and socio-emotional development in and through language education
Along with academic development, a further key task of language support models is to support students' socio-emotional development and inclusion at school. The global paradigm shift towards more inclusive education has resulted in increased focus on students' socio-emotional wellbeing (Schwab, 2020). Research conclusively demonstrates that peer relationships and social inclusion at school are key to students' development of a positive academic self-concept (Roseth et al., 2008;Wentzel et al., 2004) and emotional wellbeing (Amholt et al., 2020). Inclusion at school is also essential for students' language development, as it provides an authentic context for communication and identity development (Oxford, 1997).
While the relationship between various language support models and socio-emotional variables has so far not been thoroughly investigated, insights might be drawn from research comparing the outcomes of different support programs (special schools vs. inclusive settings) for students with special educational needs (SEN). Inclusive schools have been found to provide a more enabling environment for the development of students' wellbeing (e.g., Venetz et al., 2010). However, students with SEN in inclusive settings often have a somewhat lower academic self-concept compared to those in special schools (Bear et al., 2002;Kocaj et al., 2014). This finding has been explained by the Big-Fish-Little-Pond effect (see e.g., Marsh et al., 2006), i.e., that the average achievement of peers influences a student's individual self-concept. Moreover, students with SEN within inclusive education are at risk of lower social inclusion (see e.g., Bossaert et al., 2013;Schwab, 2018a).
While at first glance such findings might seem to imply that inclusive settings offer a disadvantage, it should be considered that mechanisms of hidden exclusion, such as support of students with SEN in pull-out courses, might have had negative effects on all the variables mentioned. Moreover, these studies did not take into account the long-term societal benefits of inclusive education. Early integration and contact is essential for successful inclusive education and this is not possible in special schools or segregated classes. Inclusive education is essential for fostering friendship groups amongst students with and without SEN (see Schwab, 2018b) -which is also essential for students learning the language of instruction. The earlier that students learn about diversity and come into contact with peers with different characteristics (e.g., SEN, diverse language backgrounds, etc.), the more realistic their view of society and the more normal 'difference' becomes.
Moreover, higher levels of intergroup conflict and prejudice are more likely to occur in school contexts where there is less contact between ethnic groups (Dajaeghere et al., 2012), and integrated models are more beneficial in terms of avoiding racial or ethnic discrimination and stereotyping (see e.g., Gatti et al., 2017). Such findings provide insights that can also be drawn on for language support models: Although inclusive models need to be carefully planned and monitored, with particular regard for students who are not 'mainstream', early and consistent inclusion holds the greatest potential for long-term integration.

Discussion
In the following, the extent to which GLSC conform to aspects of good practice established through the research synthesis above will be explored.

High quality, language-aware teacher education
As indicated by the studies reviewed above, high quality teacher education is required for teachers to meet the demands they face in GLSC. However, in Austria there is an established lack of focus on preparing teachers to work with multilingual learners in Austrian teacher education programmes (Cataldo-Schwarzl & Erling, 2022;Purkarthofer, 2016).
In the most recent OECD Teaching and Learning International Survey (TALIS), only 31% of teachers in Austria stated that teaching in a multilingual environment was part of their initial teacher education. In particular, 63% of respondents did not feel well prepared at all, 22% felt reasonably prepared, and only 15% felt well prepared to teach multilingual learners . Schwab et al. (2020) found that only 38% of teachers who work in a German support class and 11% of regular class teachers had a specific qualification to teach students with German as second or foreign language. Many teachers (12.3% of teachers in German pull-out courses and 21.5% of teachers in GLSC) reported that they do not feel competent to teach students with emergent language skills (see also Opriessnig et al., 2019). This can be considered as highly problematic as literature shows that teachers' competences and self-efficacy beliefs are linked with students' outcomes (see e.g., Zee & Koomen, 2016). Moreover, linguistic diversity is not seen as a valuable resource for all teachers in GLSC (see e.g., Kast & Schwab, 2020;Resch et al., 2022), but rather as a burden, as has been found in many other educational contexts (Hall & Cunningham, 2020;Walker et al., 2004;Young, 2014). In recent research conducted with language teachers in Austria, none of them reported using multilingual strategies in their classrooms, which implies that this is not common practice . Thus, existing evidence suggests that teachers in GLSC are not sufficiently qualified to enact policy guidelines and to support the development of language as required of such programmes.

High quality, language-focused instructional practice and resources
That GLSC do not feature high quality instructional practice is suggested not only by the low numbers of trained teachers but also by the low number of students (32%) who transit into mainstream education after one year (Statistik Austria, 2019). The ineffectiveness of language support may also be related to the pull-out teaching context in which language instruction tends to be de-contextualised and non-content related  -which goes against researchinformed principles of communicative language teaching (e.g., Savignon, 2002). While the courses are intended to ensure that students have access to curricular knowledge and languages, policy guidance offers that if students' language level is not sufficient for exploring curricular content, teachers should help the students reach this level by teaching basic German (Chancellery of the Republic of Austria, 2019). Findings suggest that this is more the rule than the exception. This is despite widely accepted evidence from decades of language learning research that confirms that languages are best learned in context (Lyster, 2007).
Furthermore, the classes do not allow for students to develop their German in an authentic context using the language communicatively and meaningfully with peers, as recommended in language education research (e.g., Paradis, 2009). Since students in GLSC are segregated from their mainstream classmates for 15 to 20 hours per week (out of a 25-30 hour school week), this prevents them from interacting with more linguistically proficient classmates . Regarding the duration of German language support, in general, it is limited to two years. Once students move into the mainstream classroom, there is little formal language support offered, which increases the chance that they repeat a grade or are diagnosed with SEN. In Austria, students with multilingual backgrounds are overrepresented amongst those diagnosed with SEN (see e.g., Mayrhofer et al., 2019).

Promoting inclusion and socio-emotional development in and through language education
Policy guidelines for GLSC state that learners should be prepared and motivated to engage with their classmates from the regular class (Chancellery of the Republic of Austria, 2019). However, teaching students with emergent competence in German in separate educational settings is argued to bear the risk of further disadvantaging them, limiting opportunities for social inclusion and language learning as well as learning progress , and their stigmatization as being "remedial" (Ovando et al., 2005). Students are nearly completely segregated from peers with higher competences in German and deprived of authentic communicative opportunities to establish meaningful relationships with students in the mainstream classrooms. Therefore, they are often considered outsiders in the mainstream classroom (see e.g., Schwab et al., 2020). Even if students in GLSC are able to make friends in their mainstream classes, it is likely that they will lose touch with them if they have to repeat a grade. They are often older than their classmates, so it can be difficult to fit in. Within GLSC, there is a lot of instability and high turn-over, and heterogeneous groups of students, so students have difficulty making lasting friendships (cf. Mensing, 2021). Overall, this leads to weak cohesion in the groups and students not being able to make deep connections to their teachers or peers. Thus, here too existing evidence suggests that the German support model fails to promote inclusion and students' socioemotional development effectively, which ultimately also limits their language learning and academic success.

Conclusion
International research on language support models in other contexts specifically show which factors can improve integration and language learning for language minority students: high quality, language-aware teacher education, high quality, language-focused instructional practice and the promotion of inclusion and socio-emotional development. While GLSC in Austria have not been rigorously evaluated, the existing evidence reviewed above suggests that they are not delivering in terms of these factors. Teacher education systems do not properly prepare teachers to work in this context. The classes do not equip students with the required German knowledge, nor do they support students' socio-emotional development or their inclusion at school. They also result in majority language students having decreased contact with students from other linguistic and cultural backgrounds. Though some studies have affirmed that short-term segregated models can achieve success when integrating newcomers into education, this is only the case when limited to a minimum period of time, when there is well-trained personnel, and where support in the language of education is still sustained once the transition to the mainstream classroom has been made (Koehler & Schneider, 2019). None of this is the case in Austria. Moreover, while short-term segregated language courses may be beneficial under these circumstances, GLSC in Austria are not reserved for newcomers: students who are longer term residents in Austria may also get tracked into them and can remain in them for up to two years, having serious negative effects of their academic self-concept and feelings of inclusion. The above review suggests that GLSC are likely to perpetuate rather than minimise disadvantage and exclusion, as it can be difficult for students to transition back into mainstream classes. Though implemented nationwide, the model goes against research-informed approaches in terms of both language support and socioemotional development.
Given the clear disadvantages of this language support model, the question can be asked why it was implemented in the first place. While answers to this question can only be surmised, societal beliefs and political agendas seem to play a key role. Students who are learning the majority language are often seen to be 'slowing' the pace of subject-matter learning. German-speaking parents may perceive a need to 'protect' their children by ensuring that they are in a learning environment where learners of German are excluded (cf. Gefäller, 2017). Indeed, in the current context, teachers in the mainstream classroom do not have the education or resources to individualise adequate support for linguistically diverse students, and so may struggle to support all their students. "Outsourcing" language learning to GLSC is perceived as an acceptable solution. However, as shown above, neither the resources nor the teacher education is currently provided to make this solution fit for purpose.

Recommendations
An appropriate model is required to successfully integrate students with GAL into Austrian schools and support their development of German, curricular knowledge and socio-emotional wellbeing. This model should be evidence-based and draw on good practice in language support and not on political ideologies and popular beliefs. Based on the above review, we make suggestions for recommendations for further research, teacher education and educational policy.
The implementation of GLSC in Austria was not guided by research nor was it preceded by a pilot study to test its appropriateness and effectiveness. Appropriate research and evaluation are required for an understanding of the academic, socio-emotional development of students in non-integrated GLSC. This would include the perspectives of the students and their parents. With regards to teacher education, all teachers should be highly qualified to support linguistically and culturally diverse students to be found in Austrian classrooms, and feel responsible for ensuring educational equity. This requires innovations in teacher education, which -in turn -must be research-informed. Urgently required are insights into what instructional practices -and the extent to which multilingual practices -can support the development of language and curricular content for German learners. Different options should be considered for newcomer students and those with emergent proficiency who are long-term residence. There is a need for research which provides better understandings of how to support language and literacy development, especially for learners with interrupted education, from families with less experience of formal education, and for multilingual learners with special educational needs. Better understandings of appropriate models for supporting socio-emotional and positive academic self-concept for minority language students, and how teachers can be prepared for this, are required.
At the level of policy, schools need to be well resourced, so that teachers are supported by well-trained professionals in the mainstream classroom and students do not have to be segregated to be served with individualised, meaningful instruction. 'Good practice' features of dual language education models could be drawn on in the re-conceptualization and implementation of GLSC. This would also require mother tongue education programmes to be more widely available and integrated into the formal curriculum in order to raise their status and quality. Further consideration of appropriate assessment types that take into account students' multilingualism are required, as well as further evaluation of the MIKA-D as an appropriate assessment instrument for multilingual development (see .

Limitations
This study has the obvious shortcoming that it has not been informed by original empirical investigation of the current German support models. Such research has not yet been made a national priority. In absence of this, this study has sought to provide insights into the efficacy of the current model in order to show that it requires urgent rethinking in order to get closer to achieving social justice in education multilingual students from migration backgrounds.