Which pupils have access to CLIL? Investigating the schools’ choice for English-based CLIL programmes in Belgium

Abstract

Despite its non-elitist goals, there is growing concern over the potentially elitist nature of content and language integrated learning (CLIL) programmes in English. Previous European studies have reported more advantaged pupils’ profiles in these programmes compared to regular programmes. It is unclear from the literature however which pupils have access to CLIL programmes based on the choices made at the school level. In order to remedy this gap in the current literature, the present quantitative study analysed all 948 secondary schools in the Flemish Community of Belgium based on three main categories of predictors: home language, socio-economic status (SES) and school size. Resulting of a stepwise logistic regression analysis, only SES and school size appear to be strong predictors of the presence of an English CLIL programme. In a region where low-SES pupils have an average of three-year delay in development in comparison to their high-SES peers (Franck and Nicaise 2018), the identification (and transformation) of these kinds of selective mechanisms is crucial to reaching educational equity. We relate these mechanisms to some specific characteristics of CLIL in the Flemish Community of Belgium, which may also apply, to some extent, to other educational contexts in Europe.

Keywords:

• CLIL
• EMI
• English
• schools
• selectivity
• SES

Introduction

Content and language integrated learning (CLIL) was introduced as the answer to the increased need for pluriliteracies when regular language education did not live up to this need. It has often been hailed as an egalitarian pedagogical approach with the potential to narrow down the achievement gap between students with high and low socio-economic status (SES) (e.g., Surmont et al. 2022). However, some studies have found that despite the egalitarian intent, it is mainly pupils with high SES who opt for CLIL (e.g., Van Mensel et al. 2020). This leads to the question if the full potential of CLIL is currently exploited. When investigating CLIL (implementation), we deem it important to distinguish two distinct levels of selectivity: the individual or ‘within-school’ level (i.e., which pupils opt for a CLIL trajectory?) on the one hand and the institutional or ‘between-school’ level (i.e., Which schools choose to offer a CLIL programme to their pupils?) on the other hand. From this distinction, it is evident that the institutional/between-school level greatly influences the individual pupils in their opportunities to even consider enrolling in a CLIL programme. So far, studies on a potential selection effect have focussed on individuals, but it is not clear whether other mechanisms at institutional levels exist. Therefore, the current article intends to investigate these institutional mechanisms by looking for the first time at which factors influence schools in their choice to offer an English CLIL programme. The study is conducted using data from all the secondary schools of the Flemish Community in Belgium, a region in Western Europe. Here, CLIL was only implemented in 2014, but it has known a rapid increase in CLIL programmes since then. In the following paragraphs, a literature review on the SES effect in CLIL will be presented first, after which the context of the study will be introduced in detail.

Literature review

CLIL or SES effect?

Over the past two decades, CLIL has seen a major rise in popularity in a large number of European regions, especially programmes with English as the target language. Being defined as ‘a generic term to describe all types of provision in which a second language (a foreign, regional or minority language and/or another official state language) is used to teach certain subjects in the curriculum other than languages lessons themselves’ (Eurydice 2006, p. 8), this educational programme, as well as its effectiveness, have been well-documented over time (see Graham et al. 2018 and Goris et al. 2019 for systematic reviews). Generally, CLIL has been shown to engender positive outcomes in second language (L2) proficiency for various language skills: listening (e.g., Bulté et al. 2022), speaking (e.g., Lorenzo et al. 2010) and writing (e.g., Van Mensel et al. 2020). In addition to this positive effect on L2 skills, content acquisition (Graham et al. 2018) and mother tongue (L1) abilities (Bulté et al. 2022) do not seem to be negatively affected by this approach in which content is being taught in a second language. Other (positive) effects that have been reported relate to motivation (Lasagabaster 2019; Bulté et al. 2022; Mearns et al. 2020) and cognition (Nicolay and Poncelet 2013; Woumans et al. 2016). Taken together, these findings suggest that CLIL enhances the linguistic capital of a pupil, and thus serves as an important leverage for achieving individual multilingualism through schooling.

Given these positive effects, the question can be raised if all pupils might equally profit from this promising approach to language teaching. Many scholars have answered this question affirmatively because, from its inception onwards, CLIL has been heralded as a non-elitist approach: all pupils should be able to benefit from CLIL, no matter their background or achievement level (see, for example, Marsh 2002; Wolff 2002; Coyle et al. 2010). Indeed, Dalton-Puffer et al. (2009) report for example how pupils in vocational schools benefit from CLIL programmes. Yet even in the initial years of CLIL, there was an awareness that CLIL was likely to be available mostly for the academically brighter students (see for example Mehisto 2007 and Lasagabaster and Sierra 2010). As we know, a direct link between SES and a pupil’s general (see Sirin 2005 for a review) and linguistic (e.g., Butler and Le 2018) achievement in a school context has been reported. If CLIL indeed aims to be an egalitarian approach which helps promote educational equity (e.g., Lorenzo et al. 2021; Surmont et al. 2022), potential selectivity on the basis of factors like SES certainly needs to be considered.

Bruton (2011) pleads in favour of a reconsideration of previous research on the multiple advantages from which CLIL pupils tend to benefit. Because these programmes function on a voluntary basis, meaning that parents (or pupils) may freely choose whether their children enrol in a CLIL programme, pupils who follow a CLIL curriculum may have different background characteristics compared to other students, such as a possible higher level of motivation to learn foreign languages, and a more positive attitude towards the target language (TL), in most cases English (San Isidro 2010 in Bruton 2011). Moreover, some studies even report more TL-teaching support for pupils enrolled in CLIL, thereby raising the question if potential benefits should only be attributed to CLIL (see Lorenzo et al. 2010 in Bruton 2011). According to Hüttner and Smit (2014), these claims disregard the ‘high degree of social determinism’ (p. 161) and ‘struggles of multilingual learner groups within the traditionally “monolingual habitus” of European schools’ (p. 162) which already feature as discriminatory elements in non-CLIL education. However, the authors do not discredit the concern that a CLIL approach could replicate these inequalities if not implemented consciously of this risk. Issues concerning SES-related selectivity are certainly not restricted to European contexts either. In South America, CLIL is often seen as an exclusive approach reserved for the dominant classes (see Banegas 2022 for an overview and Landau et al. 2021 for the case of Brazil specifically). A similar observation is made in (East) Asia where Butler (2014) reports that most ‘immersion programs tend to be highly selective’ (p. 15). Even in Canada, where immersion programmes first emerged, ‘middle and upper SES parents are more likely to have the inclination (parentocratic habitus) and the resources (economic, social, and cultural capital) to realise the benefits of FI for their children’ (DeWiele and Edgerton 2021, p. 11). Overall, the critical considerations raised by Bruton (2011, 2013, 2015) certainly present the merit of having raised awareness about the importance of non-linguistic characteristics (especially SES) when comparing CLIL and non-CLIL pupil groups in future research (Perez Cañado 2016).

Several studies have indeed confirmed the validity of these concerns. Through a self­-reported questionnaire administered to over 200 12-year-old students in the Spanish region of Andalusia, Broca (2016) investigated what influences the pupils’ choice of CLIL, their views on the programme and their grades, especially for English. The results showed that pupils seem to recognize the selective character of CLIL, although not particularly regarding it as a discriminatory element. Furthermore, this study reveals a tendency of CLIL curricula to recruit students with higher general and language grades prior to the programme’s onset. This phenomenon reinforces Bruton’s (2011) claim that, in many cases, the differences found between CLIL and non-CLIL pupils after a certain amount of time are already present at the start of the programme, and cannot purely be attributed to the programme itself.

Similar observations have also been made elsewhere in Europe. In Germany, Dallinger et al. (2018) assembled a large pool of 1362 participants who were in their second year of secondary education in order to analyse a wide array of background characteristics that could have influenced their choice for a CLIL programme. In addition to variables related to prior achievement and motivation towards content and language learning, other variables, such as the pupils’ cognitive abilities (through matching figures and word pairs), and their socioeconomic background (through the level of parental education and current occupation as well as their immigration background), were included in this study. The results not only confirmed but also supplemented the findings from previous studies, showing that CLIL students were not only more motivated and had higher prior achievement levels than their peers in a traditional trajectory, but that they also outperformed their peers in terms of cognitive abilities and SES.

Building on these important observations about the effect of cognitive abilities and SES on CLIL enrolment, findings reported by Van Mensel et al. (2020) suggest that the latter might be a more decisive factor in explaining access to CLIL. Van Mensel et al. (2020) surveyed over 900 pupils in primary (mean age = 10,5 years old) and secondary (mean age = 16,5 years old) education in the French-speaking Community of Belgium. In addition to the testing of the pupil’s verbal (through receptive vocabulary knowledge in their L1) and non-verbal (through a visuospatial segment identification task) intelligence, the other factors of analysis included the pupils’ SES, household structure and school success. No significant differences were found in terms of verbal and non-verbal intelligence. However, the opposite is true for the other three factors, especially in terms of SES. Based on these findings, the authors conclude that CLIL in the French-speaking part of Belgium should be seen ‘as not simply a selective but even elitist education model, since it clearly attracts pupils from better-off families albeit not pupils that are necessarily “brighter”’ (Van Mensel et al. 2020, p. 14).

From this brief literature review, it becomes clear that, in the current debate on CLIL’s potential selective nature, the focus so far has been exclusively on individual factors that might lead to selectiveness in CLIL programmes, leaving more structural inequalities on higher levels of analysis (class, school, legislation, etc.) underinvestigated. However, a more profound investigation of selectiveness at these institutional levels might hold the key to a better understanding of the mechanisms behind the selectiveness of CLIL programmes at the individual level. In a study on the motivation of pupils to choose for CLIL, Bakken and Brevik (2022, p. 21) conclude by saying that ‘Elitism should not be perceived as an inherent quality of CLIL but as a possible consequence if the benefits of a CLIL education are only offered to students from a high SES background. It is a question of school policy, not pedagogy’. In the present study, we will, therefore, for the first time, explore potentially selective mechanisms inherent to CLIL at the between-school level by investigating how intrinsic school composition characteristics related to various indicators of SES, language background, and size, might have an impact on whether the school offers a CLIL curriculum in the first place. This study will thus shift the unit of analysis from individual students to schools, to investigate how school characteristics might constrain the options available to students with lower SES or another home language than the instruction language. We believe that individual characteristics are not the sole contributors to this selectiveness, but that the between-school level must be included to achieve a more complete picture of how CLIL selectiveness, if present, might operate.

CLIL in Belgium

The organization of the educational system(s) in Belgium is the responsibility the country’s three Communities. The Flemish Community is responsible for education in Flanders (6.5 million inhabitants), whereas the French Community takes on the identical responsibility in Wallonia (3.6 million inhabitants). Both communities share the authority over educational affairs in the Brussels-Capital Region (1.2 million inhabitants), although the two parallel school networks are strictly separated. The third national language is represented by the German Community (77 000 people), situated in the eastern part of Wallonia. In Flanders and Wallonia, the instructional language is Dutch and French respectively, with the only exception being a number of municipalities close to the language border, with facilities for linguistic minorities. In Brussels, schools may offer education in Dutch or French, depending if they belong to the Dutch- or French-speaking community. The freedom of education is well -protected in Belgian legislation: the Belgian constitution guarantees freedom of choice for parents, meaning that access to a school of their choice withing reasonable distance of their residence must be provided (for more information, see Eurydice 2023). Despite the rapid rise in Flemish CLIL schools since its introduction in 2014, policymakers in Flanders have shown a lukewarm attitude towards the growth of CLIL, as evidenced by four thresholds that restrict or even complicate the organisation of CLIL curricula in individual schools. The first threshold is related to the number of hours that CLIL might be taught in Flanders. In Flemish Community schools, courses taught in the CLIL language (i.e., English, French or German) are limited to 20% of the entire curriculum (excluding language courses) (Onderwijs Vlaanderen 2014).

The second threshold relates to the allowed onset age of a CLIL trajectory. While pre-­primary and primary schools in the French- and German-speaking communities might offer CLIL activities to toddlers and young children, CLIL is still prohibited in Flemish primary education, and only available at the start of secondary education.

A third threshold relates to the application procedure to be allowed as a CLIL school. For a Flemish school to start with a CLIL programme, they need approval from the Flemish government before the programme’s onset. To get this approval, they need to prove, for example, that the future CLIL teachers hold a C1-level certification (in accordance with the CEFR) in order to be eligible (Onderwijs Vlaanderen, n.d.). Interestingly and particularly relevant to the present study on the topic of CLIL selectiveness, requirements also apply to the pupils, as they need approval ‘from the class council, which should at least consider the individual pupil’s sufficient knowledge and command of the [(non-CLIL)] language of instruction, namely Dutch’ (Onderwijs Vlaanderen 2014, appendix) in order to be admitted. This regulation seems to be at odds with the CLIL’s proclaimed egalitarian character, although sufficient language-related knowledge (e.g., in France) sometimes in combination with general knowledge (e.g., the Netherlands) are also known to be admission criteria in neighbouring countries (Eurydice 2006). However, in the French Community of Belgium, no such conditions for access to CLIL have been implemented. On the other hand, Flemish policymakers seem to subscribe to the egalitarian objective as they believe that ‘it is important that all students can opt for a CLIL trajectory’ (Onderwijs Vlaanderen 2014, appendix), leading to a paradox with the above-mentioned regulation that allows for the rejection of pupils on the basis of limited proficiency in the (non-CLIL) instructional language. This third threshold thus highly relates to the impact of the multilingual reality in the Flemish Community on the implementation of CLIL.

The final threshold relates to the necessity for all Flemish secondary schools that choose CLIL to offer a parallel non-CLIL trajectory to their pupils (Onderwijs Vlaanderen 2014). While such a regulation might be feasible for large schools with many classes in parallel, this is rarely reachable for small schools.

Research questions

With the aim to further investigate the potential selective nature of CLIL, but with the novelty that this investigation takes place on the level of the schools, our research questions are as follows:

1. Is the home language of the pupils a predictor for a school’s choice to offer an English CLIL programme?

2. Is the pupils’ SES a predictor for a school’s choice to offer an English CLIL programme?

3. Is school size a predictor for a school’s choice to offer an English CLIL programme?

There are several indications that lead us to believe all three of these factors will bring about divergences between CLIL schools and their non-CLIL counterparts. First, we know from previous research that a pupil’s home language can generally be considered a predictor for their educational achievement level in Flanders (Jacobs et al. 2007). Pupils with the school’s language of instruction (i.e., Dutch) as a home language hold an advantage over pupils with a different language within their household. If CLIL indeed attracts higher-achieving pupils, we expect schools to be influenced by the degree of multilingualism of their student population when it comes to their choice to offer CLIL or not. This hypothesis also rests on the admission criteria that the Flemish educational system provides based on proficiency in the language of instruction as a condition for CLIL access, which could be restrictive towards pupils with a multilingual background. Since home language and instruction language proficiency play a determining role in non-CLIL education, we believe these factors should be considered when discussing CLIL settings as well. Secondly, higher SES is even more heavily linked with better results in educational settings (Franck and Nicaise 2018). Since CLIL pupils more often originate from higher SES backgrounds (Dallinger et al. 2018; Van Mensel et al. 2020), we expect this difference to be reflected at the school level as well, meaning that we expect that schools with a higher percentage of students with a low SES are less inclined to offer a CLIL curriculum than schools with a more privileged population. Finally, we expect the school size to be a significant predictor as well. As mentioned previously, all CLIL schools within the Flemish Community have the obligation to provide a parallel non-CLIL trajectory, although they receive no additional financial support in doing so. This suggests to us that schools with a larger size are more likely to be able to afford a CLIL programme.

Materials and methods

Dataset & variables

Using publicly available data published by an entity within the Flemish department of education (AGODI), we established a dataset containing information from all 948 registered secondary public schools of the Flemish Community as of February 2019. In addition to generic information (i.e., the schools’ names, addresses and institution numbers), the dataset also contains the necessary data to assess the independent variables (home language, SES, and school size) we put forward in our research questions. For the dependent variable (English CLIL trajectory, see below), we relied on the files that the schools had to complete upon submission of their initial CLIL programme to the Flemish government, and on the outcome of the application procedure (positive or negative). All five independent variables and this single dependent variable are described in detail below. The definitions we use coincide with those provided in the Flemish legal framework (AGODI n.d.).

English CLIL trajectory

The first binary categorical variable pertains to the presence (1) or absence (0) of an English CLIL trajectory in all 948 schools included in the dataset. Only schools that received official approval from the Flemish government to start with such a trajectory after submission and positive evaluation of the original application file, and were still offering CLIL at the time that this study was conducted, were included.

Home language

The second variable is linked to research question (a). For every school, the total number of pupils who speak a language other than the school’s main language of instruction (i.e., Dutch) at home was reviewed. Following the guidelines of the Flemish legal framework, a distinction was made between families with one child and families with multiple children. In the case of a one-child family, the child was considered to have another home language than Dutch if they speak that other language with both parents. In case of a family with multiple children, the child was considered multilingual, if they speak another language than Dutch with at least two members of the family, whereby siblings (if multiple) are counted as one family member (AGODI n.d.). So, in case a child speaks Dutch with their siblings but another language with both parents, the child would be included in this parameter. As an illustration, a pupil speaking Italian with both parents, but Dutch with their three brothers will not be counted as a Dutch speaker. On the other hand, a pupil speaking Arabic with their mother, but Dutch with both their father and their two sisters will be considered as a Dutch speaker.

SES

For research question (b), SES was divided by AGODI into three distinct variables, corresponding to multiple facets of this construct, namely parental education, parental income through school allowance, and neighbourhood of living.

• maternal education: a child scored on this parameter if the mother of the child had not completed secondary education (AGODI 2019). The level of education of the father was not taken into account. The database guidelines do not mention any specificities with regard to cases in which no mother is part of the family context.

• School allowance: a child scored on this parameter if it benefited from financial aid during its educational career as a result of low family income (AGODI 2019).

• Neighbourhood: a child scored on this parameter if it belonged to the 25% of pupils who live in neighbourhoods with the highest percentage of school delay (2 years minimum) (AGODI 2019).

School size

Finally, for the last research question (c), we used the total number of pupils enrolled in order to determine the size of every school.

Data analysis

Our data analysis thus involves a total of six variables divided into five independent variables and one dependent variable. The independent variables were categorised into three potential predictors of divergence between CLIL and non-CLIL schools, each corresponding to one of the three research questions (i.e., home language, SES of the pupils and the school’s size). The analytical plan for this study was composed of three components. First, means, medians and standard deviations were calculated for each of the five independent variables, with a further division according to the dependent variable (English CLIL school or not). Second, a set of Welch’s independent samples t-tests were conducted to test the research questions for each independent variable separately. Third, a stepwise regression analysis was executed, where each additional predictor was added incrementally to the model, leading up to a full model where all independent variables are jointly present in order to explain the dependent variable. The outcomes of these descriptive and inferential analyses, which were all executed using R, can be found in the results section below.

Results

Descriptive statistics

Means, medians and standard deviations were calculated for all independent variables based on their ratios compared to the school’s size. This also applies for the t-tests below. Table 1 provides an overview of the descriptive statistics for the schools with an English CLIL programme, as well as schools not offering such a trajectory.

Table 1. Descriptive statistics for schools with and without an English CLIL programme.

	CLIL in English (n = 201)			No CLIL in English (n = 747)
	M	Mdn	SD	M	Mdn	SD
Home language	14 %	10 %	14 %	19 %	12 %	19 %
SES 1 (maternal education)	18 %	15 %	13 %	26 %	22 %	17 %
SES 2 (school allowance)	26 %	23 %	13 %	32 %	30 %	17 %
SES 3 (neighbourhood)	19 %	15 %	18 %	26 %	18 %	24 %
Size	527	516	307	448	389	271

Welch’s independent samples t-tests

The results from the Welch’s independent samples t-tests display similar tendencies. We found a statistically significant difference between schools with and without an English CLIL programme on the variables of school size [t(264) = −3.5, p < .001], pupils’ maternal education, [t(378) = 6.24, p < .001], school allowance [t(374) = 5.76, p < .001] neighbourhood [t(394) = 3.87, p < .001] and pupils’ home language, t(390) = 4.41, p =.01, with a higher overall number of pupils, a lower number of pupils with low SES, and a lower number of pupils who speak another language at home than the instruction language for schools that offer a CLIL programme in English as compared to schools who do not provide such a programme. The Cohen’s d values are .32 for school size, .43 for level of maternal education, .4 for school allowance, .26 for neighbourhood and .30 for home language respectively. These values only indicate a rather small effect size for the first two SES variables according to the benchmarks suggested by Plonsky and Oswald (2014).

Regression models

Our second inferential analysis consists of a stepwise regression including three different models. The first one features the three selected SES variables (i.e., maternal education, school allowance and neighbourhood). We subsequentially added the variable of home language to the second model and the size of the schools to the third one respectively. Since the school size itself already features as an independent variable in the analysis, the raw numbers were used for the home language and SES characteristics The results for the three models can be found in Table 2. It is important to note that, based on the outcome of Shapiro-Wilk normality tests, we established that the data were not normally distributed. We resolved that issue by using the ‘log1p(x)’ function in R, so by executing of a logarithmic transformation of the data of all independent variables.

In the first model featuring all three SES variables, maternal education and school allowance appeared to be extremely significant, whereas neighbourhood was not significant. In the second model, the degree of significance remains the same for the SES variables. The newly added variable of home language is not significant. Finally, in the third model, the variable of size was added. It was extremely significant and caused the variables of maternal education to drop in significance. The variable of school allowance became not significant and the other two variables introduced in the previous steps remained not significant. The odds ratios controlling for the other variables confirm the negative relationship between low maternal education and the presence of a CLIL trajectory at school: model 3 shows that a one-unit increase in the variable of maternal education is associated with a 55% lower likelihood (or odds) of a CLIL trajectory. In addition, the positive relationship between school size and CLIL is expressed by the association between a one-unit increase in that variable with a 121% increase in the likelihood of CLIL.

Table 2. Results of the regression analyses.

	Model I				Model II				Model III
	b (SE)	Or	L	U	b (SE)	Or	L	U	b (SE)	Or	L	U
SES (maternal education)	−1.36***	0.26	0.15	0.42	−1.34***	0.26	0.16	0.44	−0.8**	0.45	0.25	0.78
	(0.26)				(0.26)				(0.29)
SES (school allowance)	1.51***	4.53	2.32	9.08	1.51***	4.55	2.33	9.13	0.46	1.59	0.7	3.74
	(0.35)				(0.35)				(0.43)
SES (neighbourhood)	<- 0.01	1	0.83	1.21	0.05	1.06	0.83	1.35	0.08	1.09	0.86	1.39
	(0.09)				(0.12)				(0.12)
Home language					−0.1	0.91	0.69	1.19	−0.17	0.85	0.63	1.11
					(0.14)				(0.14)
Size									0.79***	2.21	1.46	3.35
									(0.21)

*** < 0.001 < ** < 0.01 < * < 0.05 < n.s. SE = Standard error, OR = Odds ratios, L = Lower, U = Upper.

Discussion

The goal of the present study was to investigate whether the reported selectivity of CLIL programmes on the within-school level (Bruton 2011; Broca 2016; Dallinger et al. 2018; Van Mensel et al. 2020) could also be observed at the between-school level. More specifically, we explored which factors influence the schools in their choice to offer an English CLIL programme or not. The factors we have taken into account were the size of the schools (i.e., the number of pupils), the home language(s) and SES (i.e., level of maternal education, school allowance and school delay in the neighbourhood) of their student population. We consulted a database compiling these characteristics for all of the 948 public secondary schools in the Flemish Community. These variables were subsequentially analysed using independent-sample t-tests and a stepwise regression analysis. Below, we provide an answer to each one of our research questions by interpreting and elaborating on the results presented above.

Home language

For this first variable, the results of the t-test initially indicated that the pupil populations in schools with an English CLIL programme contain fewer pupils with a language other than the school’s language of instruction (i.e., Dutch) being spoken at home compared to the average of all Flemish schools. However, our regression analysis shows that the home language of the pupils is actually not a significant predictor for a Flemish secondary school to offer this type of education, when all other variables are kept constant, in school size and SES. These results struck us as somewhat surprising considering the emphasis that educational policy-makers put on Dutch proficiency when it comes to CLIL implementation. On the one hand, the general idea is that every pupil should be able to benefit from CLIL, but on the other hand, rejection from enrolling in a CLIL programme can be motivated on the basis of school language proficiency. Our data suggest that even though schools are to some extent allowed to restrict CLIL programmes to native speakers of the instruction language only, the home language composition of their student population is not decisive in their choice for offering a CLIL trajectory in English. This finding, however, does not provide any insights in terms of how these pupils’ level of achievement evolves within a CLIL programme compared to their monolingual peers.

SES

One of the main aims of this study was to examine if the SES effect that was previously reported at the within-school level, also manifested itself at the between-school level, meaning that specifically schools with high-SES pupils opt for CLIL. The current study used a threefold operationalisation of the SES construct: maternal education, parental income, and neighbourhood. From these three variables, only maternal education turned out to be a solid predictor of whether a school offered a CLIL trajectory in English. The effect of the number of students with low maternal education was highly significant in both our t-test and regression analyses even when factoring in school size. In other words, the level of maternal education differed in favour of pupils in CLIL schools with a trajectory in English when comparing them to the average student population across all secondary schools. These findings are in line with our hypothesis as well as the results reported in previous studies at the within-school level (Dallinger et al. 2018; Van Mensel et al. 2020). Our results suggest that the SES effect previously detected at the within-school level might be the result of an SES effect at a higher level, i.e., the between-school level. It might be that low-SES parents and children do not freely choose not to opt for CLIL in English, but that their options are heavily impacted by the decision that the school has made not to offer a CLIL trajectory. Previous studies have mainly focused on which pupils opt for or can be found in a CLIL trajectory based on individual factors, whereas we have investigated which pupils have access to CLIL based on the schools’ choice to offer such a programme to its pupils or not. The current study does not allow for a precise analysis of the reasoning that led up to the decision of a school (not) to offer a CLIL trajectory in English, but we suggest two explanations for this SES effect at the school level. First, it might be that schools do not believe that low-SES students are capable of following courses in another language than the instruction language and that these lower expectations guide their decision not to offer such a trajectory in schools where the proportion with low-SES students is substantial. Second, it might be that a high number of low-SES students creates additional challenges for schools, in terms of pedagogy, staff as well as financial resources and that schools feel that offering a CLIL trajectory in English is an additional burden that the school cannot cope with, especially in times of teacher shortages, and financial perils. We would suggest future (qualitative) research to focus on the accuracy of these possible explanations.

Size

The size of the schools, expressed in terms of the total number of enrolled pupils, turned out to be a highly significant difference between CLIL schools and the total average as well as the greatest predictor of whether a school offers an English-based CLIL programme or not. It thereby surpassed all other variables that were considered, including the pupils’ SES. This observation is most probably the consequence of the aforementioned obligation for all CLIL schools to offer an equivalent non-CLIL trajectory too. In theory, on the condition that they follow the prescribed implementation procedure rigorously, all schools of the Flemish Community could opt for CLIL education. However, they cannot count on any form of additional funding in order to do so from the Flemish government. As a result, smaller schools most often lack the financial and/or infrastructural capacities to set up a CLIL path for their pupils. Although this justification is rather context-bound, it could be interesting to investigate this factor in other regions, especially given the fact that the size of the schools has repeatedly been overlooked in previous studies. With this in mind, we are of the opinion that, in addition to the previously reported pupil characteristics, the context of implementation certainly needs to be considered when discussing the selectivity mechanisms of CLIL.

Limitations & further research

Although the present study revealed a series of interesting findings, it has its limitations with the main one being that it solely relied on quantitative data. Using this type of data, some selectivity mechanisms were uncovered in terms of which schools grant access to CLIL on the basis of their size as well as the SES of the pupils. However, these results do not provide insights into how these mechanisms influence the concrete reality of CLIL’s implementation in Flanders. Are smaller schools simply not interested in the possibility to offer an English CLIL programme or do they believe they will not be able to get themselves through the prescribed procedure? Are there other factors at stake during their ­decision-making process? Further research including qualitative data could provide an answer to these questions. Additionally, although the nature of our analyses has indeed shed a light on which factors are predictors of a school’s choice to offer CLIL in English, the potential correlations between these factors are not investigated (e.g., between the pupil’s SES and the school’s size). Secondly, it is important to keep in mind that there are no exact data about how many pupils actually follow a CLIL trajectory in English in the Flemish Community, and what their profiles are in terms of SES, and home language. As a result, completing the present findings by examining selectivity mechanisms at the individual level of the pupils is rather difficult in our context of research. Moreover, an analysis of the materials used at the classroom level (i.e., which curricula are implemented) could prove to represent a significant addition to the current research field. However, seen as we laid our focus on the between-school level, this curriculum theory approach falls finds itself outside of the present study’s scope. Finally, we believe it is important to reiterate the value of taking context-bound variables, like the size of the schools, into account in future CLIL studies.

Conclusion

Using a quantitative approach, the goal of the present study was to contribute to the ongoing debate about CLIL’s (potential) selectivity mechanisms by investigating these mechanisms at the between-school level, instead of the individual or within-school level. Overall, we found that Flemish Community schools with a greater number of pupils as well as with a pupil population featuring a higher SES (in terms of the level of maternal education) are more likely to offer an English-based CLIL programme. Other potential factors, including the pupils’ home language (i.e., a school’s degree of multilingualism), were not significant predictors. In other words, there is a (possibly unconscious) selection effect which makes it more likely for large middle-class schools to opt for CLIL in English than other types of schools. As a by-effect, multilingual students, who are underrepresented in this school type, might have more limited access to CLIL with English as a medium of instruction, than other students. This mechanism is likely fuelled by the legislative framework in which schools have to operate. If CLIL education indeed aspires to bridge the current educational inequity by providing additional linguistic capital (in English) to all of its pupils, including those who come from lower-SES and/or multilingual backgrounds, current policy and practices in CLIL schools with English as a medium of instruction must be rethought by having these processes of selectivity addressed by future research and policymakers at various ­decision-making levels.

Disclosure statement

This publication was made possible by the funding of research mandate OZR4009 by the Vrije Universiteit Brussel (VUB).

Data availability statement

The school data that support the findings of this study are available at https://data-onderwijs.vlaanderen.be/documenten/bestanden/Publicaties_Leerlingenkenmerken_Overzicht_2019-2020_sec_0.pdf. An overview of the CLIL schools in the Flemish Community is available at https://onderwijs.vlaanderen.be/nl/directies-en-administraties/onderwijsinhoud-en-leerlingenbegeleiding/secundair-onderwijs/clil-onderwijs-in-een-vreemde-taal/clil-content-and-language-integrated-learning

Additional information

Funding

Vrije Universiteit Brussel (VUB).