From policy to practice: how schools implement German language support policy in Austria

ABSTRACT In 2018/2019, the Government of Austria instituted a new language support model to Austrian schools that included segregated language support measures for non-fluent German-speaking students. The central research aim of the paper is to investigate how the implementation of this new language support model compares to the policy requirement by examining different models of implementation. In the framework of the initial phases of the project ‘A multi-perspective study on German language support’, 12 interviews with German language support teachers and school principals were conducted. Following a prototypical case study analysis, four models of implementation were identified: (1) Segregationmodel, (2) Mixed segregated and integrated model, (3) Individualized language support, and (4) Inclusive support model. Given that thisis the first study in Austria to show the concrete implementation models of German language support in schoolsFindings highlight important implications for school principals and educational policy makers.


Introduction
In the school year 2020/21 more than 25% of children and youth living in Austria were speaking a first language other than German (Statistik Austria, 2022).Taking into account the high percentage of students who speak a first language other than German, Austria could be considered as multilingual.In a country with a monolingual school system (like Austria), proficiency in the (primary) language of instruction is seen as highly important for school success.Language policies tend to follow this monolingual habitus by segregating students based on their German language proficiency.Moreover, according to Herzog-Punzenberger et al. (2020) teachers and school principals from Austria (as well as Ireland, Norway, and Turkey) mainly perceive diversity in classrooms based on proficiency in the language of instruction which may result in minimal efforts to adapt instruction, and assessment measures.
Because of a monolingual instead of a multilingual habitus, schools often do not draw on the diverse language repertoires of students and German is still the dominant language in the school system in Austria (see also Erling et al., 2022).Segregation, discrimination, and a general lack of support for students with migration background is not only an issue for Austria, e.g.international comparative studies repeatedly indicate significant achievement gaps between students with and without migration background in several countries with different support models in place (OECD, 2018).However, not all Austrian schools adhere tothe segregated language support policy, which leads to a mismatch in policy and implementation practice, the main topic of this study.

Educational support models for multilingual students
A wealth of research has shown that 'good practice' in language support is best achieved in the mainstream classroom, through supporting high-quality, language-focused instructional practices that promote inclusion and socio-emotional development (cf.Erling et al., 2022).From a linguistic perspective, a holistic approach to language learning would be the translanguaging approach, where students use their entire language repertoire in learning activities (Cenoz & Gorter, 2011; City University of New York-New York State Initiative on Emergent Bilinguals, 2021).
However, within most countries of the European Union (24 out of 27) newly arrived students who are not yet proficient in the language of instruction are often taught in separate language support classes/ transitional classes/ preparatory classes for a limited period of time (European Commission, EACEA, and Eurydice, 2019).Such an educational practice (Van Laere et al., 2017) belongs to the so-called 'structured immersion programs' (Baker, 2011)where students must learn the new language without utilizing skills in their native language (Van Laere et al., 2017) which purport that intensive training in the not yet proficient language is necessary in order for students to be successful.Within their review of the effectiveness of English language support programs, Collier and Thomas (2017) concluded that English-only and transitional bilingual programs that are implemented for a short period of timewere not highly successful.Instead, the academic success of two-way immersion programs (providing native language instruction for twobalanced groups with distinct languages -50% of the time for each of the two languages) was shown to be most effective.These authors emphasized the importance of students' multilingual abilities and the value of linguistic diversity after reviewing 32 years of research from longitudinal studies.

Second language learning in Austrian schools
In the 2018/19 school year, the Austrian government (see e.g.BMBWF, 2018) established two segregative language support measures, namely the German Language Support Classes (GLSC) and the German Language Support Courses (GLSCO).One argument for the installation of GLSC and GLSCO was thatthe majority of students in Vienna do not develop sufficient reading skills according to a national educational panel survey (BMBWF, 2018).The government introduced these measures with the stated aim of harmonizing and improving the quality of German language support in schools (BMBWF, 2019).The GLSC and GLSCO operated as pull-out classes, in which students who had not yet mastered the language of instruction (sufficiently) were separated from their peers.Pull-out classes like these are often legitimized in international contexts, by the 'inclusion through exclusion' approach (see e.g.Mendenhall et al., 2017; in this context see also Bunar & Juvonen, 2022).International literature (García & Bartlett, 2007) as well as studies on GLSC (Resch et al., 2023) identified that separate language support classes might be a 'safe space' for a short period of time for students who need to learn the language of instruction.However, students who are spatially segregated might feel excluded from the school community (see e.g.Svensson & Eastmond, 2013; for GLSC see: Resch et al., 2023;Schwab & Schweiger, 2023;Spiel et al., 2021).Nilsson and Bunar (2016, p. 406) demonstrated how this language support measure (transitional class) 'is a label that contains a whole narrative concerning a particular child's difficulties'.For Austria, teachers reported that students who attend GLSC may sometimes face exclusion or be subjected to discriminatory practices of othering (Resch et al., 2023) as well as limited access to grade-level content.A further problem of spatial segregation,confirmed for Austrians' GLSC (Erling et al. 2022;Resch et al., 2023) as well as within international research (Hos et al., 2019),is linguistic isolation, which limits interactions with peers who speak the language of instruction.In this context, Alvarez et al. (2022) emphasize the importance of age-appropriate content instruction for language and social development.However, students in GLSC lack language role models (peers) and occasions to use the language they are learning (Resch et al., 2023).
Language-tests which serve as high-stakes gatekeepers for mainstream classes are widely used across international contexts (see Oikonomidoy et al., 2019).For example, in Austria, to determine if and which of the two measures (GLSC/GLSCO) students need to attend, students take the MIKA-D language screening instrument, which was designed especially for this procedure.This screening is performed regardless of whether or not the student was born in Austria and is up to the discretion of the principal in their observation if the child is at risk of not meeting the language requirements (BMBWF, 2019).Given this high-stakes assessment, school leaders are faced with increased pressures and responsibilities as they decide which students have to undergo MIKA-D.As a result of the MIKA-D scoring, three categorizations are deduced:a) Students labeled with sufficient German skills (level A 'ausreichend') participate with the status of a 'regular' student in regular classes.b) Students labeled with poor German language skills (level M 'mangelhaft') are required to attend a pull-out GLSCO for weekly six hours parallel to their regular classes and are assigned the status of 'irregular student'.c)Finally, students labeled with insufficient German language skills (level U 'unzureichend') are also given the status of a so-called 'irregular student' and must spend the majority of their class time in a separate GLSC (15 h per week in elementary schools and 20 h per week in secondary).This policymakes the organization of these classes very difficult for teachers as well as school leaders (Resch et al., 2023;Hassani & Schwab, submitted) because it requires that GLSC students are present in the mainstream classroom for classes e.g.in physical education, music, and arts and crafts (BMBWF, 2019).
In total, about 34,075 students (13,058 in GLSC, 21,017 in GLSCO) attended one of these measures in the school year 2020/21 and thus had a so-called 'irregular ' status, (see BMBWF, 2022), which corresponds to about 3% of all students in Austria.Interestingly, the majority of students were born in Austria (approx.60%), and one out of three students attended kindergarten in Austria (BMBWF, 2022).Around one third of the students who are put in the GLSC also hold the Austrian citizenship (Expert Council on Integration, 2022).Therefore, GLSC are attended by students born in Austria, newly arrived students with a solid school background, as well as newly arrived students who lack previous schooling in their country of origin.
Students attending a GLSC or a GLSCO undergo the MIKA-D screening at the end of each semester to assess if a transition to the other measure is necessary/possible or if a 'regular' status can be obtained.Students however attend this GLSC/GLSCO for a maximum of four semesters and transfer regardless of their MIKA-D scores to the regular class and thus to a so-called 'regular' status after the fourth semester at the latest.As long asstudents attend GLSC, they are generally not allowed to advance to higher grade levels and have to repeat the same grade in a new mainstream class.The MIKA-D itself is heavily criticized in Austria (see e.g.Hassani et al., submitted;by Spiel et al., 2021) for lack of evidence on reliability and validity and for not taking into account students' multilingual repertoires.In this context, international research stressed that tests for decision-making needto be unquestioned in their reliability and validity (e.g.Bachman & Purpura, 2008).Moreover, long-term disadvantages for multilingual students due to inappropriate language screenings (e.g.limited access tohigher education) (Nikolov & Timpe-Laughlin, 2021) and broader implications of testing on teaching and learning (Schissel, 2019) should be taken into account before screenings such as MIKA-D are implemented.
While international research stresses the role of school principals in implementing language support measures (e.g.Arar & Orucu, 2022;Arar et al., 2022;Bunar & Juvonen, 2022;Wilkinson & Kaukko, 2020), educational policy in Austria has drastically limited the official scope of action by introducing a mandatory segregated approach.The most recent amendment to the law explicitly stipulates that the language support (if there is any special possibility) needs to be outside of the classroom.Prior to this, schools could decide autonomously whether language support should take place inside or outside the mainstream class.Therefore,school principals had greater flexibility in implementingGerman language support.A quantitative study indicated that approximately one third of school principals do not or tend not to adhere to the ministerial guidelines for the implementation of German language support (Schwab et al., 2023).However, the study did not include information on how German language support was implemented, which is one of the central research aims of the study presented in this article.
To summarize, the current paper aims to provide in-depth insight into the current implementation practice of GLSC comparing language support policy and practice.Based on previous knowledge and researchin the field and working closely with schools, a variation in the implementation of the GLSC was expected.Nevertheless, the aim of the study here is to understand the variation of implementation practice in detail against the background of existing policy.This is particularly important because published studies related to GLSCs have limited interpretability unless knowledge of the gap between policy and current implementation practices is updated.

Empirical study
The empirical findings presented in this paper stem from the first phase of an ongoing study ('A multiperspective study on German language support classes') which aims to explore the academic, social, emotional, and linguistic development of students in GLSC, GLSCO as well as mainstream education in Austria.As outlined above, no study has so far investigated how GLSC is currentlyimplemented in Austria.The present study endeavors to address this research gap between policy and implementation practice by providing insights into various approaches to language support instruction and the underlying rationales of the responsible school principals.This will be helpful for the broader study and future data collection on GLSC to better interpret the data and understand the differences in students' academic, social, emotional, and linguistic development, which mayvary greatly across schools.It is important to knowthe differences in how language support measures are implemented in order toknow why schools are successful or not.
The current study is being conducted in the city of Vienna, which is the capital of Austria, where the highest numbers of multilingual students as well as GLSC can be found (Statistik Austria, 2022).At the beginning of the empirical work, school principals were contacted to personally introduce the study, and recruit the school to participate in the project.In this phase, the first interviews with school principals and GLSC teachers were conducted.By establishing rapport and a trusting relationship, it was possible for both school principals and teachers to openly discuss the way that the model was implemented in their schools with the research team.Soon it became apparent that despite the legal framework which demanded uniform implementation of the policy, there were various divergences in practice, as school principals needed to implement the support measures in consideration of their school environment as well as what they felt that language learners needed to be successful.This paper attempts to map the models of German language support that could be found at schools that participated in the initial phases of the study as a first stage in the development of typologies of implementation types.This mapping provides insight into how and why schools do (and do not) conform to the mandated policy and how this relates to school contexts, resources, the student body, and the school principals' understanding of language learning.

Ethical considerations and researcher positionality
The study was approved by the Ethical Committee of the University of Vienna (approval number: XXXXX00816).School principals and teachers were informed by a project member and questions were answered either face-to-face or via phone.All participants provided voluntarily informed consent, and schools had the autonomy of choosing their preferred research methodology for participation (interview with principal, interview with teacher, etc.).Schools were assured that they could withdraw their participation at any time.The collaborative atmosphere was ensured by project members by being approachable, holding workshops, and inviting schools to project presentations.
In the spirit of self-reflexivity, the research team acknowledges their standpoint as women with high socio-economic status (e.g.being educated at university level).While most of the authors involved are autochthonous Austrians ('White Europeans'), speaking German as first/family language, some of the team members have a migration biography.Overall, within the author team, experiences of language barriers (e.g.due to learning German while moving to Austria during school time), experiences of full submersion, experiences of racial discrimination and experiences of social discrimination because of a working-class background can be traced.

Methodology
The interviews with school principals and teachers as well as classroom observations mainly focused on implementation practices: How are GLSC implemented against the background of segregation policy requirements?Throughout the initial and continuing fieldwork, a circular, processual, reflexive, communicative and comparative approach in the sense of grounded theory was applied by the researchers (e.g.Charmaz, 2014;Corbin & Strauss, 1990), which made it possible to gradually map different models of implementation.By writing theoretical memos or taking audio memos in the sense of Grounded Theory (Charmaz, 2014;Corbin & Strauss, 1990), particularities, implementation strategies, school characteristics, etc. were recorded and continuously included in the interview guideline.This also facilitated the development of further documentation tools to record and classify generated information and toask more precise questions about the implementation of GLSC, thus substantiating or understanding them or adding further types.

Data collection and interview guideline
The qualitative interviews (Witzel & Reiter, 2012) selected for this paper (n = 9) were conducted in German either in person at the local schools (n = 7) or online via Zoom (n = 2).All interviews were audio recorded and transcribed verbatim.Any quotes given below have been neither edited nor 'polished', but were translated to English by the research team.On average, the interviews took 71 min, ranging from 32 min as the shortest interview to 105 min as the longest interview.
As a narrative-generating introductory question, the interviewees were asked to think back to the phase when German language support classes and courses were implemented in autumn 2018 for the first time and how they experienced these times.The teachers' answers structured the further thematic course of the interview and provided the interviewers with opportunities to address the following additional topics: teaching in GLSC (e.g.Please describe a typical day of teaching in GLSC from your experience.),beliefs and professionalization in second language learning (e.g.How do you perceive language support measures in your school?), development of students with a first language other than German (e.g.How would you describe the relationships of students attending GLSC and their peers in regular classes?), opinions about the mainstream model (e.g.What do you think about the mainstream model for German language support in Austria?).All questions had additional sub-questions, which were optional to the interviewers.

Sample
School principals and GLSC teachers who had expressed interest in taking part in the study were asked to take part in the initial interviews.In total, seven teachers and two school leaders (T_2, T_8) from six different schools took part in the study (T_6, T_7 as well as T_8 worked at the same school).All participants were female.In terms of school type, six participants were working in Austrian primary schools, and three participants in middle schools (T_1, T_3, T_5).Both school types were included to explore the range of implementation types at the primary and secondary levels.

Data analysis
The analysis of the empirical data was strongly oriented to the method of type formation.We proceeded in an 'ideal-typical' manner following the steps of typology construction provided by Kluge (2000): In the first step, each interview transcript was coded by assigning relevant statements to the typological categories and subcategories.Based on these codes, we developed comparative dimensions that resulted in specific 'attribute spaces' (Kluge, 2000).In this study, the following dimensions emerged: physical placement of students and hours of support per week.In the second step, the individual cases were grouped according to their attributes and assigned to the respective 'attribute space', which paved the way for step 3the more in-depth analysis of the cases.In this step the technique of case contrasting played a central role: On the one hand, those cases that were assigned to a certain group ('attribute space') were compared with each other in order to ensure 'internal homogeneity' (Kluge, 2000) of the patterns formed.On the other hand, the formed patterns were contrasted to ensure 'external heterogeneity' (Kluge, 2000) at the level of typology.After this step, some cases were re-assigned to other groups to whom they were more similar.The final step (4) involved the precise characterization of the four types.

Results
The data analysis revealed the following four types of language support models in practice, in contrast to the established two segregative language support measures by the Austrian government: (1) Segregation model, (2) Mixed segregated and integrated model, (3) Individualized language support model, and (4) Inclusive model.The identified models reflect the diversity in the implementation of German language support found in schools in Vienna.In the following table, these classifications are presented along with the degree of compliance with the policy-mandated guidelines.As such, fully conform indicates that the model fully conforms to the policy whereas, partly conform indicates that there are aspects of the model that align with overall policy (Table 1).

Description
In the segregation model, students with beginner-level German language skills assigned to the 'irregular' status are taught in segregated pull-out classes.The legal basis of offering 15-20 h of segregated language support (15 h in primary and 20 h in secondary schools) is fully covered in this model, thus, it conforms to the mainstream model enforced by the law.

Empirical case
The segregationmodel was described by T_5, who works at a secondary school.Students with beginner-level German language skills and labeled with the 'irregular' status are supposed to attend the GLSC for 20 h per week.The aforementioned school fully corresponded to the legal model: Students received 20 h of GLSC in separate classes and 9 additional hours of teaching in different subjects within the mainstream class.'We are handling this like this, from 8 am to 12 am every day they are in the German language support class.Mondays from eight to one, on all other days from eight to twelve and Fridays from eight to eleven.' (T_5: 727-728) The teacher described the level of conformity to legal regulations like this: Well, we do really have a German language support class (…) and 20 h, so our way of implementation really is a textbook example.We could be monitored any time.(…) in individual cases like mine now, the Hungarian [student], I mean we just took him out [of the GLSC], that we dismiss students into mainstream classes earlier than … (…) sometimes I dismiss students to their new mainstream class maybe three months before the end of school and not (…) Well, we are playing in a borderland (T_5: 1177-1187) Model 2: mixed segregated and integrated model Description Within this type, students with beginner-level German language proficiency with a labeled 'irregular' status participate in both an integrated and separated setting.However, schools in this sample balance the degree of integration and segregationin different ways (e.g. can you give an example of a different way), thus different specifications of this model were identified: a balanced model mixing integration and segre-gation50:50 and an imbalanced model mixing segregationand integration 80:20.This seems to depend on how many students have a labeled 'irregular' status and the level of German language skills of the students in the class.Both models will be described below: Model 2a.Balanced model of mixing integration and segregation 50:50

Empirical case
Based on the interviews, the balanced model of German language support could be assigned to one primary school in the sample.The teachers (T_6 and T_7) explain that their students receive eight hours of German language support in a separated class and seven hours integrated in their mainstream class.In the morning, students participate in their mainstream classes.Later in the day, the German language support teacher pulls them out for separate language learning.T_7 stated that it is 'completely unrealistic' that every child receives 15 h of segregated language support: 'It [the model] is simply unconvertable to practice (…) It was impossible to offer 15 h (…) I don't know any school, which is able to pull this through.'(T_7: 834-843).
The students had various levels of German language proficiency and were in different grades (1-4).Hence, it was difficult for the GLSC teachers to meet the diverse range of needs within one class.'Well, in total, eight hours [separately].Then I teach one hour with those who repeat the class, exactly, three hours with newly arrived, and that means the rest of the three to four hours are integrative support hours in the first two grades' (T_7: 349-353).

Description
Compared to the 'balanced model', the ratio of integration and segregationis not balanced, but segregation is the predominant, at least in the cases of this sample.

Empirical case
The imbalanced model at this secondary school largely meets the policy requirements, with slight differences.The teacher interviewed described that students in this school spend 16 h per week in separate classes and the remaining 4 h in mainstream classes.They are pulled out of the mainstream classes on four days a week, but 'one day a week, every Tuesday, they stay in their classes, they attend mainstream classes in a normal way, yes.' (T_3, 26-27).

Description
This individualized model of language support describes a comprehensive form of support for the whole school, rather than for single students with elementary or beginner-level German language proficiency labeled 'irregular'.All resources for German language support are pooled together from pre-school, GLSC, and GLSCO so that all students can benefit from all forms of support.Teachers in this model work collaboratively and identify and utilize synergies between the existing support models.

Empirical case
T_4 described a mixed model of implementation: Students within this model receive segregated language support.However, not all students are pulled out for the full amount of 15 h a week according to the legal model.In difference, they are pulled out according to their needs and then receive language support in small groups.Thus, it is a flexible and competence-oriented approach, which aims to keep students in mainstream classes as much as possible.This model facilitates fluid transitions across different support measures.Almost every newly enrolled student at this school has a migration background, and according to the interviewed teacher 'the German support classes are counterproductive on the whole; they need to be merged with pre-school classes.' (T_4) The responsible teacher has a comprehensive view on the available support models and promotes synergies across them.T_4 aims a holistic plan for the whole school with the possibility to mix students with elementary and beginner-level German language skills across ages, grades and classes.
According to T_4, teachers in this school act collaboratively as a teacher tandem and try to build a unit of German language support no matter which 'model' the children are officially assigned to.'It's wrong that there are courses and German language support classes.There should be only one German support.(…) everything could be a unit, pre-school and German language support class as it is called now, a unit and that children of all classes who need German support, can be pulled out.' (T_4, 422-435).According to her, it is impossible to implement the legal model and that one form of support is not productive, especially where kindergarten and pre-school are not taken into account.Alternatively, all support measures should be valued as a collective system of support with overlaps and interactions.

Description
In the inclusive model, all students with a first language other than German and with 'irregular' student status are taught in mainstream classes.T2's school does not offer any separate GLSCs, which might be labeled as such, but has an inclusive understanding of being a multilingual school perceiving (foreign) languages as a resource.Students with German as an additional language receive language support in the mainstream class in so-called 'Abteilungsunterricht'.

Empirical case
T_2 described the model of implementation as a rather inclusive model: students with German as an additional language are taught in the mainstream classes and are treated equally to all others.Two teachers in the school have additional resources for language support (German as a second language) and they support these students who need support within the mainstream class.This model does not apply to the official model supported by the Ministry and therefore does not conform to the segregation model; however, the school leader underlines that there is a legal base for inclusive schooling in a 'special contract' of the school with the Ministry.Hence, the segregation model is rejected for both ethical and pedagogical-didactical reasons (e.g.peer learning).The exception is officially made because of the lack of spatial resources.
'So I and a few other principals, we opposed the introduction of the law with the German classes, yes, and then for us and I think 40 other schools a special arrangement was found, which looks like this, that we actually cannot do the German classes in the very exclusive sense, because we have no space for it (…) our school is so crowded, we cannot do it and of course we are not allowed to do it integrative (…) but we do it in the departmental teaching lessons, yes, that means the language support resources are in the classes '. (98-105) Moreover, the student's first language and German should be learned as thoroughly as possible.
'(…) there are so many languages on offer in Vienna that I think it's completely insane to let them go to waste in school ' (55-58).
All children in the school, no matter if they have a migration background or not (over 50% do so), have the opportunity to learn another language for six months within the regular school day (e.g.Bosnian).Everyday schooling is multilingual, and all classes are multilingual.

Discussion
The current article aimed to describe how German language support is implemented in a sample of schools in Austria, considering the currently valid legal educational policy situation.Results of the interviews identified four different language support models, which vary in the extent to which they conform to and deviate from legal regulations for German language support.This is especially true for the inclusive learning in mainstream classes versus being educated in a spatially separated context that varies widely between the models.Within the first model (segregation model), a high range of conformity of the implementation to legal regulations was identified.Interestingly, the GLSC teacher referred to the possibility that the local school authority could monitor the implementation anytime.This might imply that schools are aware of the fact that the government implemented a policy that contradicts expert opinion and in the face of much criticism (for an overview see e.g.Erling et al., 2022;Flubacher, 2021).There were schools that were inspected when there were concerns that they were not implementing the policy.Therefore, the possibility of being monitored might be a reason for implementing GLSC strictly according to the legal framework, even though there are strong concerns about this model.In other words, schools and teachers face institutional risks if theydeviate from the policy despite their potential concerns.However, even the school that was identified in this model, circumvents the legal framework for individual exceptions.For example, the school allows some students (by the end of the school year) to attend their future mainstream class, which indicates that inclusion is of high importance to themas they interpret the related parts of the legal framework as a gray area that allows them to implement more inclusive language support.A key goal of some interviewed teachers is to enable inclusion as soon as possibleby passing MIKA-D (see e.g.Hassani et al. submitted).The fact that students' results on the MIKA-D might hinder the inclusive approach is in line with outcomes from Hassani et al. (submitted) indicating that teachers sometimes may fake good MIKA-D results.The transition from GLSC to mainstream classes was addressed as an obstacle.Bunar and Juvonen (2022) stated that it is somehow ironic, that schools first excluded the students in separate classes and then struggled with the inclusion in mainstream classes.A serious problem addressed in the context of a lack of inclusion in mainstream classes during the assignment to GSLC is the lack of a sense of responsibility mainstream teachers may feel for GLSC students when realizing that they will probably not teach these students in the future.Generally, it might be easier to implement the segregation model in full-day schools or secondary schools, as the school day is longer in these contexts.
In contrast to the first model, in the mixed segregated and integrated model,the number of hours in mainstream education for students with irregular status (not passed MIKA-D) is higher.Two different implementations have been found within our data using 50% or 80% of the total GLSC time allocated in the legal framework.Of course, other forms of mixing segregation and integration of GLSC students are theoretically and practically possible but were not included in the presented data.One of the main problems of this model is that students do not receive the full amount of support hours, which is officially granted to them.However, this is partly rooted in teachers' beliefs about the advantages of inclusive education and the need to be in contact with language role models.Additionally, organizational barriers were mentioned as reasons for the GLSC implementation against law.Further, representatives of these models argue that students need to be divided according to be taught in homogenous groups regarding their language levels and topics of interest.However, the division by language levels and interests is not sufficient.On the one hand, neglecting students' various levels of school experience and potentially limited previous experiences of formal schooling when preparing educational support can be seen critically (Brännström, 2021; see also Brännström & Ottemo, 2022).On the other hand, it might be problematic when social and emotional needs are not considered in a diverse group of GLSC and GLSCO students.For instance, if newly arrived students experienced trauma (e.g. in their country of origin, during forced migration), their individual needs should be considered by schools rather than seeing them as a homogenous group(see e.g.Umansky et al., 2018; see also Cerna, 2019).Concerning the social and emotional needs of students, one participating school did not implement GLSC in the first school week given that they believed it was aa sensitive period for students and instead focused on students' social participation.This might be a single case and might be linked to the fact that MIKA-D screening is often conducted by GLSC teachers during their allocated language support hours (see also Hassani et al. submitted).Hence, a systematic lack of resources for conducting time-consuming mandatory language screenings is implied.Further, a case was demonstrated wherethe mixed model of segregationand integration is based on the refusal of mainstream class teachers to let GLSC students leave class during the day.
The 'Individualized language support' is a model where all resources (e.g. from GLSC, GLSCO, pre-school classes, mainstream classes) are flexibly combined and used for the benefit of all students.Similar to the second model, students' individual needs (e.g.language level, need for support) are taken into account for decisions about inclusive or separated education.Teacher collaboration can be identified as a significant prerequisite of this model taking advantage of synergies.This model is justified by the school staff's collaborative beliefs.However, this might also occur from the fact that in this school the majority of students have a migration background.
In the 'inclusive support model' students who did not pass MIKA-D are educated only in mainstream classes and GLSC is not offered at all.Multilingualism is seen as a resource and all students have the possibility to learn further languages (e.g.Bosnian).German language support takes part in the so-called 'Abteilungsunterricht' which means departmental teaching, which is a specific exemption.Due to the inclusive support, a transition to mainstream classes does not take place in this model.Consequently, the associated difficulties, e.g.social integration, do not occur.

Implications for practice, policy, and research
Based on the current study, one of the most important implications for practice is the need to significantly increase the autonomy of school leaders.A too strong legal framework and policies that aim for a universal approach to meeting students' language needs can lead to overlooking the individual and specific needs and strengths of students' and target groups.For example, resources (e.g.spatial resources, quantity, and quality of teachers trained to teach multilingual students) vary widely across schools.
Rather than developing a strengths-based perspective for educating multilingual students, politically enforced segregated language models could reinforce monolingual educational practices and limit the responsibility of regular education teachers to consider the full multilingual repertoire of students.
For future research, the results of this study can provide guidance, especially for the interpretation of the results.Without knowing exactly how schools implement language support measures, it does not seem possible to evaluate their impact at all.For example, it may be difficult to evaluate whether a smaller gap between students with German as a first and second language would be solely due to the implementation of the GLSC or if other factors, such as non-compliance with the legal framework, could also play a role.Future results of large-scale tests like PISA and PIRLS may indicate a change in this gap, but it is important to consider all possible factors that could impact the results.If study results show that students who participate in GLSC feel socially integrated, is this because there are no negative social consequences for student social participation due to pull-out classes or is it because students were not ostracized by their peers because GLSC was not implemented at all or for very few hours.Overall, there is a need for more comprehensive information about what is happening in schools in Austria, and it is not certain that the official statutory implementation models are really implemented in (all) schools.
Future studies should distinguish not only between segregated and inclusive models, but also between different mixed segregated and integrated models, and should pay more attention to the number of hours of pull-out support.
Finally, since (according to the Ministry of Education inAustria) GLSC were originally implemented to create equal opportunities, it corresponds to a conceptual understanding that advocates for unequal treatment based on different starting conditions in the sense of equity of need (Brake & Büchner, 2012).Following Bourdieu (2001), however, such an understanding of equal opportunity means precisely not viewing access to education as a problem of fit between educational prerequisites and educational opportunities, but rather recognizing disadvantaged learners as fully capable of development and taking unequal social and linguistic starting conditions into account when considering specific, high-quality support (Rothe, 2015;Stojanov, 2013).Against this background, and given the empirical evidence on GLSC, their ethical viability must be fundamentally questioned.

Limitations
Within the current data sample, four different implementation models of German language support were identified.However, there might be more variations within these models or even additional models.Additionally, the current sample was limited to a metropolitan city and results from small cities and rural areas might differ.
When analyzing implementation models of German language support in Austrian schools, it would be of high importance to take the school environment and contextual factors of individual schools (e.g.school size, students' diversity) into account.For instance, the amount of GLSC students varies widely in Austrian schools.The results of this study underpinned the importance of certain school characteristics in the implementation of GLSC.Though, in this study it was decided not to describe schools and their environment in detail to ensure the anonymity of the sample.

Conclusion
In conclusion, a range of implementation models of GLSC is identified in the current sample.Hence, the evaluation of its effects might be challenging.The findings indicate that schools attempt to provide the best support for students within a tightly controlled education system.Some models opposed the legal framework, pointing to the autonomy of the school and that its decisions actually meet the needs of the students.Moreover, findings suggest that school principals might have a comprehensive knowledge of the resources available and a school's environmental and contextual framework.Following this, school leaders should have the autonomy to decide about the support for their students.Furthermore, it should then be proven whether the students receive the support they are entitled to in the regular classes, especially if separate German language support does not take place or takes place to a lesser extent than required by law.The strict legal regulations and limited school autonomy on German language support, alongside the absence of effective metalinguistic teaching approaches, can significantly impede the educational progress of multilingual students.This hasthe potential to perpetuate educational inequity.

Table 1 .
Overview of language support models.