Teaching and Learning of Mathematics through CLIL, CBI, or EMI—A Systematic Literature Review

Abstract

This study provides a systematic literature review of research in the field of teaching and learning mathematics through Content and Language Integrated Learning (CLIL), Content-Based Instruction (CBI), and English Medium Instruction (EMI). The review aims to examine the most relevant literature with a focus on mathematics and CLIL, CBI, or EMI in Scopus and Web of Science per the Preferred Reporting Items for Systematic Reviews (PRISMA) guidelines. Based on 151 sources, 52 papers were selected according to predefined selection criteria. The papers were analysed and coded according to the following categories: (1) geographical productivity, (2) diachronic productivity growth, (3) main objectives, (4) methodology, and research design, (5) variables and measurement instruments, (6) context and sample details, and (7) main findings. The results reveal that most of the research analysed has been carried out in the Asian continent, followed by Europe. The analysis of diachronic productivity shows that the study of the application of CLIL, CBI, and EMI programs in the teaching-learning process of mathematics has notably increased in recent years, especially in the last triennium (2020–2022). Regarding the objectives of the selected corpus, the majority aimed at teaching practices and learning processes, and outcomes in mathematics and language proficiency in CLIL, CBI, and/or EMI classrooms. The corpus analysed fits within one or more of the following categories: empirical, qualitative, descriptive, correlational, and cross-sectional. The samples utilized in different studies differ significantly, both in terms of quantity –ranging from one person to 700– and in the educational level being studied, which would be primary, secondary, or tertiary education. The main variables studied in the corpus focus on mathematical competence, language proficiency, teaching practices, teacher training, science competence, and teachers’ perceptions. The most widely used instruments have been objective tests, such as questionnaires, together with standardized tests to measure some aspects related to mathematical competence and language proficiency. They are followed by an analysis of documents (academic records, teaching materials, official documents...), participant or non-participant observation, interviews, and video and audio recordings. In summary, in the scientific literature analysed, a positive or neutral view predominates on the effects of the CLIL, CBI, and EMI approaches on the learning of mathematics and the L2. This can be due to methodological issues fundamentally related to the methodology, research design, sample, and measurement instruments. Thus, we must highlight that some of the results from the selected papers must be interpreted with caution. Taking this factor into consideration, further comparative studies on a wider scale are required to examine thoroughly the effects of CLIL, CBI, and EMI on the teaching and learning of mathematics in an L2. Besides, it is important to study in greater depth the different levels of language acquisition since the research analysed shows that these have not been sufficiently addressed in the mathematical field of knowledge.

1. Introduction

Nowadays, the vast majority of educational policies in Europe and beyond consider the promotion of the learning of an additional language to be fundamental. Similarly, learning the contents of a specific academic subject such as mathematics through a foreign language has become an increasingly accepted framework in the field of education, especially in higher and secondary education stages. Thus, as English has been recognized as the international language of sciences and technology, it has turned out to be the quintessence key of bilingual policies hinged mainly upon the CLIL (Content and Language Integrated Learning), CBI (Content-Based Instruction), or EMI (English as a Medium of Instruction) educational approaches.

Stoller [1] (p. 59) conceives CBI as an umbrella term for approaches that amalgamate language and content learning aims, even if there may exist differences in the degree of importance ascribed to language and content per the type of program applied in a certain context. Therefore, following Cenoz’s research [2], CBI and CLIL programs “share the same essential properties and are not pedagogically different from each other” (p. 8). The use of one term over the other is mainly due to contextual and accidental characteristics: CLIL is usually used in Europe, while CBI is the preferred term in American and Canadian contexts. Nonetheless, we must consider that this use may not always be linked to a specific context or educational stage as some countries have found difficulties when defining the type of program they are applying [3].

CLIL represents a pedagogical approach that pivots both on the learning of academic subject content and the learning of a foreign language, which is the medium of instruction for the content [4]. According to Carrió-Pastor and Bellés-Fortuño [5], even if CLIL and EMI programs share most of the common CBI pedagogical principles, there are some major differences between CLIL and EMI that we must highlight. Thus, the former is the approach used to teach both content and (foreign) language, while the latter focuses solely on the learning of contents through English, and hence the linguistic-related objective is peripheral. In addition, CLIL is most often put into effect in primary and secondary education, while EMI is mostly implemented in university and bilingual secondary schools since the level of English proficiency is often higher in these contexts. Another of the differences that these authors point out, referring to Dearden’s words [3], is that EMI does not have a specific context of origin. English is envisaged as the language of education (with the geopolitical and sociocultural repercussions that this may entail), while CLIL refers to a clearly defined context of origin (i.e., Europe).

Table 1. Commonalities and differences between CLIL, CBI, and EMI.

	CLIL	CBI	EMI
Term used in Europe	X		X
Term used in USA and Canada		X	X
Teaches both content and foreign language	X	X
Focuses only on learning content through English			X
Primary and secondary educational levels	X	X
University and bilingual secondary schools			X
Has no specific context of origin			X
Has a specific context of origin	X	X

Source: own elaboration.

lists the main commonalities and differences between CLIL, CBI, and EMI.

Therefore, we consider it convenient to carry out this study based on the premise that the CLIL and EMI approaches are very widespread in Europe and Spain, with the latter being one of the countries where more research on CLIL and EMI has been carried out [6].

Regarding the types of CLIL, we can distinguish [7] ‘hard CLIL’, consisting of a partial immersion process through which half of the curriculum is taught in another language; and ‘soft CLIL’ through which the teaching staff makes use of an integrated approach to languages, holistic, and often with an interdisciplinary extent for the teaching of the contents. Similarly, various forms of EMI can occur in the same country, school, and even in the same class whose characteristics can be classified based on the following criteria proposed by Richards and Pun [8]: purposes, assessment, curriculum models, the introduction of EMI, access to EMI, the English course and EMI, the EMI teacher, the English subject teacher, the EMI learner, and instructional materials. Therefore, as we have also observed in the literature, in some cases, the limits between the CLIL and EMI approaches are blurred and sometimes confusing. Thus, EMI is ‘CLIL-ised’ in practice, i.e., the L1 (mother tongue) is used at certain times both to attend to the learning needs of the content and linguistic students [9], as well as to respond to linguistic and cultural diversity in the face of phenomena such as multiculturalism [10,11,12].

Likewise, existing literature recognizes different types of bilingualism, which we highlight since they fit better with the objectives of our study, ‘balanced bilingualism’ in which individuals are capable of using two languages regularly; and ‘unbalanced bilingualism’, also known as ‘dominant bilingualism,’ in which individuals do not use the L2 at the same proficiency level as their L1 [13]. One fact that we must consider in this regard, according to the scholars, is that the latter is the one that prevails in most countries implementing CBI, CLIL, and/or EMI.

Among the benefits reported by the scientific literature about the bilingual teaching of mathematics is that it provides students with greater insight into language structures [14] which may have a positive impact on cognitive development and, more specifically, on metalinguistic awareness. The latter has been reported to be fundamental to the development of mathematical competence. Essentially, metalinguistic skills help bilingual students when solving mathematics word problems stimulating cognitive development and content knowledge compared to monolinguals [15,16]. Similarly, Bernardo and Calleja [17] contend that a good command of the students’ linguistic and communicative competence in the language of instruction, be it L1 or L2 (first foreign language), has a direct influence on the development of mathematical competence. This complies with Cummins’s critical threshold hypothesis [18] as he states that it is essential to attain a certain level of L1 and L2 before the positive effects of bilingualism are acknowledged. This requires a longer time and the right context in addition to quality instruction [19].

One of the big questions that have been raised about the implementation of the CLIL, CBI, and EMI programs is whether they promote language proficiency at the expense of learning academic content and, specifically, mathematics. In this sense, various studies [13,19,20] have criticized the premise from which CLIL, CBI, and EMI start, namely, that content learning and language learning are independent. Certainly, mathematics differs from other subjects in that it contains its own mathematical register, its own technical and academic language, which requires a certain degree of abstraction, and consequently, it embodies a challenge for students who must become familiar with it both in the L1 and the L2 to learn the content of the subject in addition to mastering the language in which it is taught. So, the type of language used in CLIL, CBI, and EMI generally differs from that brought into play in regular foreign language classrooms. This means that whereas foreign language lessons make informal interactions and face-to-face communication easier, CLIL, CBI, and EMI entail academic aspects of language proficiency. Following Cummins [18], language uttered in daily conversations (BICS—Basic Interpersonal Communication Skills) diverges from the language used in acquiring knowledge at school (CALP—Cognitive Academic Language Proficiency). CALP consists of the ability to produce complex meanings explicitly in a written or oral form employing language itself instead of intonation, gestures, and other non-verbal hints such as visual aids, and so on. Hence, CALP requires higher cognitive demands than BICS as it is more abstract, context-reduced, and develops relatively slower than BICS in foreign language learners, which may have a negative impact on their motivation.

After having mentioned some of the benefits and challenges of teaching mathematics through CLIL, CBI, and EMI, we agree with Bermejo et al. [21] on the necessity to study in greater depth the different levels of language acquisition, as well as “assessing to what extent the development of mathematical thinking can be constrained if it is taught in an additional language or a ‘non-dominant language’” [21] (p. 2).

In the same vein, some teaching practices have caught researchers’ attention in the last decade concerning teaching mathematics through an L2. Thus, interaction has become an essential element in learning mathematics in bilingual and non-bilingual teaching contexts through dialogical learning among peers [22,23]. One of the most recent approaches that also assigns an essential role to interaction and goes beyond code-switching is translanguaging. It consists of making use of a strategic discourse both in L1 and L2, or even in an L3 (second foreign language), to enhance the learning of the language and the content in a natural, flexible, and dynamic way, favouring the communicative dimension of the language through speaking, listening, reading, signing, and remembering. Its positive impact on the teaching of mathematics is being reported by a growing number of investigations [12,24]. Another approach that strengthens the implementation of CLIL, CBI, and EMI is Task-Based Learning (TBL), which is recommended by the Council of Europe [25] and is also receiving positive reviews [26]. These authors have created a taxonomy of activity types based on TBL for teaching mathematics, among other subjects, suitable for primary and secondary CLIL classrooms. Abedi and Lord [27] also propose a list of possible adaptations of the linguistic register of mathematical word problems to the students’ linguistic level that we have found of interest.

We find it useful to refer to some prior systematic reviews related to CLIL, CBI, and/or EMI. For instance, Graham et al.’s study [6] focuses on the language and content outcomes of CLIL and EMI although not in mathematics, but also other subjects such as natural and social sciences. However, the number of studies reviewed that specifically pertain to mathematics is limited to two. Macaro et al.’s study [28] exclusively examines EMI in higher education, with a focus on teachers’ and students’ beliefs, and explores whether teaching academic subjects in English as a foreign language benefits English proficiency without harming content learning. Finally, we have come across a review that delves into CLIL and EMI in the field of physical education [29]. To the best of our knowledge, no previous systematic review has solely focused on exploring mathematics learning through CLIL, CBI, and EMI, and that is the original contribution of this paper.

2. Materials and Methods

To respond to the study objective, i.e., to examine specifically the most relevant literature with a focus on mathematics and CLIL, CBI, or EMI, this work follows a quantitative and descriptive method, falling within the literature review studies, as it intends to explain the future of a line of inquiry in the international panorama of scientific research.

This research seeks to address the following questions:

-

RQ1: How is geographical productivity distributed?

-

RQ2: What is the diachronic productivity growth?

-

RQ3: What are the objectives of the existing research?

-

RQ4: What methodology and research design are used?

-

RQ5: What are the variables and measuring instruments?

-

RQ6: What type of population is selected?

-

RQ7: What are the results of existing research?

To achieve the objectives described above and answer the research questions, we conducted a systematic review which was done per the Preferred Reporting Items for Systematic Reviews (PRISMA) guidelines [30].

2.1. Search Strategy

From January to September 2022, we developed the framework of this research. We began by defining the keywords to respond to the research objective, which were: “Mathematics”, “Content and Language Integrated Learning” (CLIL), “Content Based Instruction” (CBI) and “English Medium Instruction” (EMI). Once the keywords had been selected, nine combined searches were carried out by introducing the following search equations: “Mathematic*” AND “Content Language Integrated Learning”; “Mathematic*” AND CBI; and “Mathematic*” AND “English medium instruction” in all possible fields (title, abstract and keywords) in Scopus, ERIC (Educational Resources Information Center) and in the Core Collection of the Web of Science (Science Citation Index Expandex (SCI-Expanded); Social Sciences Citation Index (SSCI); Arts & Humanities Citation Index (AHCI); Conference Proceedings Citation Index–Science (CPCI-S); Conference Proceedings Citation Index–Social Sciences & Humanities (CPCI-SSH); Book Citation Index – Science (BKCI-S); Book Citation Index–Social Sciences & Humanities (BKCI-SSH); Emerging Sources Citation Index (ESCI); Current Chemical Reactions (CCR-EXPANDED); and Index Chemicus (IC)). We reviewed all potential outcomes available to date, as these databases currently encompass the majority of research sources are widely used by researchers and experts across various fields of knowledge [31]. They were not restrictions placed on geographic location, language, year of publication, document, type, or any other criteria.

2.2. Inclusion Criteria

To obtain the analysis units that made up the final sample, the PRISMA protocol for systematic reviews [30] was followed, according to the inclusion and exclusion criteria outlined in

Table 2. Inclusion/exclusion criteria.

Inclusion Criteria	Exclusion Criteria
(a) Papers (article and article review [including both bibliographical and systematic reviews]).	(a) Book chapters, proceedings, and other types of publications.
(b) Full text in English or Spanish.	(b) Restricted access (only abstract or incomplete text) or full text in other languages.
(c) Research on the teaching and learning of mathematics.	(c) Research not including the teaching and learning of mathematics.
(d) Implementation of CLIL, CBI, or EMI.	(d) Non-application of CLIL, CBI, or EMI (English for academic purposes or English for specific purposes).
(e) Research carried out in contexts where the use of CLIL, CBI, or EMI for the teaching of mathematics is through English as L2.	(e) Research carried out in contexts where the use of CLIL, CBI, or EMI for the teaching of mathematics is not English as L2.
(f) Research carried out in an instructional setting where the target population’s L1 is not English.	(f) Research carried out in an instructional setting where the target population’s L1 is English.
(g) Research that reports on applied empirical studies (quantitative, qualitative, mixed methods, and reviews).	(g) Non-empirical studies (i.e., those obtained through observation, perception, and interaction with our surroundings).

Source: own elaboration.

.

Until reaching the definitive corpus, we discarded, in the case of the sample rescued from Web of Science (n = 87), the references that were not papers (n = 30). Similarly, we performed this procedure in Scopus (n = 63) and ERIC (n = 51), excluding a total of 16 and 4 references respectively that did not meet the inclusion criteria regarding the type of publication. In this way, we analysed 151 references, 56 of which were excluded because they were papers found in both databases. The remaining 95 were evaluated for suitability based on their abstracts, and if there was any uncertainty, the complete text was examined. Finally, this figure was reduced to 52 (Figure 1) due to the following exclusion criteria: the full text was not in English or Spanish (exclusion criterion b) [32,33,34,35,36]; CLIL or EMI was applied in disciplines other than mathematics (exclusion criterion c) [32,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60]; the immersion or bilingual program was not CLIL, CBI and/or EMI (exclusion criterion d) [57,61,62]; research carried out in contexts where the use of CLIL, CBI or EMI for the teaching of mathematics is not English as L2 (exclusion criterion e) [13,17,62,63,64,65]; research carried out in an instructional setting where the target population’s L1 is English (exclusion criterion f) [56,66,67,68,69]; and non-empirical studies (exclusion criterion g) [70,71].

3. Results

Overall, 52 studies were included in the systematic review (see Appendix A): 48 applied empirical papers (92.3%) (quantitative, qualitative, and mixed methods), and four reviews of scientific literature (7.7%) [72,73,74,75]. All of them are written in English and published between 2005 and 2022.

3.1. Geographical Productivity

In terms of geographic productivity, most of the research analysed has been carried out on the Asian continent (n = 27; 51.9%) (Bangladesh 1.9%; China 13.4%; India 1.9%; Indonesia 7.6%; Iran 3.8%; Iraq 1.9%; Japan 1.9%; Malaysia 11.5%; Russia 3.8%; Thailand 1.9%), followed by the European continent (n = 24; 46.2%) (Austria 1.9%; Belgium 1.9%; Czech Republic 9.6%; Finland 1.9%; Germany 1.9%; Latvia 1.9%; Norway 1.9%; Poland 3.8%; Slovakia 1.9%; Spain 11.5%; Sweden 1.9%; Turkey 3.8%; and Ukraine 1.9%). We only found one result from the African continent (South Africa 1.9%), and none from the Americas or Oceania.

3.2. Diachronic Productivity

The analysis of diachronic productivity (Figure 2) shows that the study of the application of CLIL, CBI, or EMI programs in the teaching-learning process of mathematics has been increasing in recent years. If we analyse the total productivity, we can see that the largest number of publications is concentrated in the six years 2017–2022 (n = 35; 67.3%), with this period seeing exponential growth in research. The 2020–2022 triennium is especially noteworthy, in which a total of 21 studies are published (accounting for 40.4% of the total number of studies in this systematic review).

3.3. Aims

The aims of the papers included in this systematic review are diverse. On the one hand, most of the research (n = 28, 53.8%) focuses on CLIL effects on the teaching and/or learning of mathematics [11,19,21,26,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95]. On the other hand, a smaller number of studies (n = 18, 34.6%) zero in on EMI effects on the teaching and/or learning of mathematics [10,12,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110]. In addition, just three articles (n = 3, 5.7%) focus only on CBI without distinguishing between CLIL or EMI [111,112,113]. In addition, three of them (n = 3, 5.7%) address the interrelationship that is established in some contexts between the EMI and CLIL programs [9,114,115].

Similarly, we have observed that these publications spotlight the study of teaching and/or learning processes and outcomes fundamentally related to one or more of the following categories: mathematical competence (mathematical thinking, arithmetic, and problem-solving); linguistic development (linguistic skills, language proficiency, language switching, discourse competence, translanguaging, playful talk and metalinguistic awareness); cognitive development (executive functions); and digital competence. Besides, we have noticed that several investigations attempted to study mathematical competence and science competence separately without establishing connections among them [81,82,88,97,99,102,108,109,110]. It is worth noting that some cases fit within the framework of the STEM approach [9,76,98,101,114].

Likewise, some investigations supplement the categories described above by adding the perceptions of in-service teachers in primary [88,96], secondary education [12,85,93,103,104,105,108,109,113,115,116], and tertiary education [94,101,110]. Concerning the latter, we have also detected research at the tertiary level whose objective is to examine STEM (Science, Technology, Engineering, and Mathematics), EMI lecturers’ perceptions about translanguaging [101], and the assessment of the four language skills of university students [110]. It is worth highlighting two studies focusing on students’ perceptions, such as the mixed-method research of Rethinasamy, Chuah & Hashim [102] that centres on elementary students’ perceptions about learning mathematics and science through EMI; and the quantitative research of Akbarov, Gönen & Aydogan [91] that focuses on university students who learned mathematics through a CLIL program.

On the other hand, a smaller number of investigations aim to inquire into STEM lecturers’ beliefs and identities concerning the English Language Teaching (ELT) gaze [9,114], as well as secondary teachers’ beliefs in mathematics following an EMI approach [96,108,109].

At the same time, we have found a few studies whose objectives revolve around the affective aspects in the teaching and learning of mathematics through CLIL, CBI, and/or EMI in all educational stages. It is noteworthy that the studies found focus solely on the teaching of mathematics through the CLIL approach. For instance, the study by Otwinowska and Foryś [83] pinpoints the relationship established between affectivity and cognitive development in CLIL mathematics in elementary classrooms. The investigation by Binterová and Šulista [78] analyses elementary school pupils’ attitudes towards learning mathematics in an L2 using CLIL and an ICT tool. The paper by Akabarov, Aydogan & Gonen [91] takes into consideration CLIL university students’ attitudes and preferences. Finally, the research work by Kosybayeva et al. [76] measures the psychological wellbeing of 80 university students enrolled in two mathematics and science educational CLIL courses.

Similarly, some studies concentrate on teacher training, specifically concerning: the analysis of teaching practices at secondary education [93,113] and at university [94]; the design of a proto-English syllabus using a CLIL approach for mathematics secondary teachers [87]; the implementation of a development program [80]; the restructuring of CBI mathematics university courses in Oman by using active learning through educational technology [111]; the description and discussion of a CLIL project for mathematics teaching based on TBLT [26,112]; a literature review about CLIL teaching practices in a mathematics classroom in India [75]; the implementation of a CLIL model for the learning of mathematics in English at elementary education [92]; and at university level [89,95] although from an EMI perspective [107]; and the design of a specific teaching material which is a phrasebook for English-mathematics teachers in elementary schools [100].

3.4. Methodology

The corpus analysed is primarily empirical and includes qualitative, descriptive, correlational, and cross-sectional. In terms of research design, 32.7% (n = 17) used mixed methods [11,76,78,79,83,85,87,89,92,94,95,99,100,102,107,108,111]; 30.7% (n = 16) were qualitative [9,12,26,77,80,82,90,93,96,98,103,104,105,110,112,116]; and 15.4% (n = 8) utilized a quantitative design [19,21,81,86,91,97,106,113,116]. Additionally, 13.5% (n = 7) were case studies [10,84,88,101,109,114,115], and 7.7% (n = 4) were scientific literature reviews [72,73,74,75].

Conversely, there were a significant number of studies with an experimental design (13.5%, n = 7) [11,19,21,78,81,92,106]. Regarding temporality, most of the studies showcased transversal results (71.2%; n = 37), while only 11.5% (n = 6) were designed longitudinally [76,78,81,106,108,111].

3.5. Main Variables and Measuring Instruments

The main variables studied in the corpus home focus on mathematical competence (100%; n = 52) and language proficiency (100%; n = 52). The presence of other variables is also significant, such as teaching practices (78.8%, n = 41) e.g., [65,72,78,92,93,94,95,96,106,107,108,109,111,112]; teacher training (55.8%; n = 29) e.g., [9,95,96,100,107,109,111,112]; teachers’ perceptions (30.8%; n = 16) e.g., [94,107]; students’ perceptions e.g., [91,108]; and students’ attitudes e.g., [91,92]. Another representative variable is science competence (28.8%; n = 15) e.g., [81,82,88,97,99,102,106]. Other variables are less well represented, such as students’ vocabulary, teachers’ identities, etc. (detailed information is provided in Appendix A).

To carry out the different studies, questionnaires/tests were used [11,19,21,63,76,78,79,83,85,86,87,88,89,91,94,95,97,98,99,100,102,107,108,111,113] (n = 25; 46.2%); analysis of documents (academic record, tests or exercises...) (n = 22; 42.3%, e.g., [95,107,111,112]); participant or non-participant observation (n = 20; 38.5%) [10,12,76,78,79,84,87,88,90,94,95,96,103,104,105,108,109,114,115,116]; interviews (n = 20; 38.5%) [9,12,78,87,88,90,94,96,99,101,103,104,105,107,108,109,110,114,115,116]; and video recordings (n = 9; 17.3%) [10,12,77,82,98,103,104,105,116]. To a lesser extent, audio recordings (n = 2; 3.8%) [11,84]; or focus groups (n = 1; 1.9%) [79] have been utilized.

3.6. Context/Sample Details

The samples utilized in different studies differ significantly, both in terms of quantity –ranging from one person to 700– and in the educational level being studied, which would be primary, secondary, or tertiary education. In this sense, most of them focus on secondary education (38.5%; n = 20) [11,12,77,81,82,84,85,86,87,93,103,104,105,106,108,109,112,113,115,116]. However, a relevant number of studies are carried out in the primary education stage (19.2%; n = 10) [10,19,21,65,78,83,88,92,100,102], as well as at the university level (26.9%; n = 13) [9,76,79,89,90,91,94,95,97,98,101,111,114]. Additionally, regarding the population under study, some studies zoom in exclusively on students (30.8%; n = 16) [11,19,21,77,81,83,84,86,91,92,95,97,102,106,111,113]; pre-service teachers (3.8%; n = 2) [79,89]; in-service teachers (21.2%; n = 11) [9,85,93,94,96,98,108,109,110,112,114], both students and in-service teachers (25%; n = 13) [10,12,78,82,87,90,100,101,103,104,105,115,116] and others which, together with the agents already mentioned, consider other members of the educational community, such as the families or tutors of the student body (1.9%; n = 1) [99]. Some, however, do not carry out any field research, and their sample is made up of scientific documents or results of educational projects or programs (13.4%; n = 7) [26,72,73,74,75,80,88].

3.7. Main Results

As we have observed and analysed, most of the studies display positive results for the teaching and learning of mathematics and the L2 through CLIL, CBI, or EMI (n = 28; 53.8%) [11,12,26,73,75,78,80,81,84,85,87,90,91,92,94,95,99,100,102,103,104,105,106,110,111,113,115,116]. Moreover, papers that offer neutral results stand out (n = 25; 48%) [9,10,19,21,72,74,80,82,84,86,88,89,90,91,93,96,97,101,107,108,109,110,112,113,114]. However, there are very few studies that present negative results (n = 9; 17.3%) [74,83,87,93,98,107,108,109,112].

The positive results correspond mainly to an improvement in mathematical competence, cognitive development, socio-cultural competence and awareness, motivation, engagement, and language proficiency in the target language (English) of the sample studied. These are supported by: teaching approaches such as task-based [26,95] and multilingual approaches [75,91]; teaching methods based predominantly on active learning [92]; language teaching strategies such as code-switching [85,115], scaffolding [90], and translanguaging [12,103,105,115,116]; and new and restructured teaching materials [26,80,87,100,106,111] including the use of ICT tools [73,92,104]. Similarly, in some cases, these results are validated/supported by the positive perceptions of learners, teachers, stakeholders, and even families about the teaching and learning of mathematics through CLIL, CBI, or EMI [12,92,94,99,102,103,104,105,110,113,116]. It is also important to refer to some studies with an experimental design that show positive effects of teaching and learning mathematics through CLIL, CBI, and EMI on students’ attention, motivation, and teaching efficacy [78,92]; the highest improvements in English proficiency [11,106] and mathematical competence [11]; and faster cognitive development compared to the control group.

Neutral and negative results are especially related to:

-

Age. Results from two experimental studies [19,21] demonstrate that younger students display lower effectiveness in solving verbal problem tasks in mathematics when they do it in the target language (L2) compared to the control group that learned mathematics in their mother tongue. However, these differences decreased as time progressed and exposure to the L2 increased.

-

Students’ low level of language proficiency [96] as well as the diversity of linguistic competence that can be present in a classroom [90]. In this sense, a study points out that previous learning of mathematical content in an L1 predicts success in EMI courses at the university level [97].

-

Lack of teacher training in CLIL or CBI approaches [72,74,80,88,89,91,93,96,98,108,110,112], language proficiency [87,96,112], and ELT methods [74,82,93,96,98,108,110,112] which implies greater content teachers’ language awareness [93], explicit teaching of intercultural features [84], and teacher training in classroom discourse competence [98]. A study also underlines teacher training gaps concerning teachers’ skills to overcome cognitive and affective barriers that have negative effects on CLIL mathematics classrooms [72].

-

Teachers’ perceptions and beliefs concerning their role as merely content teachers or including that of language teachers [9,109,114], especially in EMI classrooms that have become CLIL-ised [9,114]. One of these studies emphasizes a shortage of collaboration between content and language teachers [109].

-

Criticism has been directed towards top-down policies promoting CLIL, CBI, or EMI for mathematics learning, questioning their sustainability and potential for creating social disadvantages [10,101,107]

4. Discussion

The analysis of the 52 articles retrieved from Scopus, ERIC, and Web of Science has fulfilled the inclusion criteria delimited in the PRISMA review process, which has led us to the following inferences. Firstly, concerning geographic productivity, the vast majority of research on teaching and learning mathematics through CLIL, CBI, or EMI has been published in Asia and Europe, except for one study from South Africa. In the Asian continent, China (n = 7) is the country with a greater number of scientific production, followed by Malaysia (n = 5), and Indonesia (n = 4). Spain stands out as the European country with the highest number of scientific productions in this field of knowledge (n = 6), followed by the Czech Republic (n = 5). These results coincide with those produced by Macaro et al.’s [28] systematic review published in 2018 which focused on EMI in Higher Education including all disciplines (not only mathematics). Therefore, in the last 4–5 years, there has been no substantial change in terms of geographic productivity on this research focus. Other continents are very little represented such as Africa (n = 1) or are not represented at all such as America or Oceania.

Secondly, as far as diachronic productivity is concerned, we have verified that the number of studies tackling the teaching and learning of mathematics through CLIL, CBI, or EMI has experienced a dramatic increase during the last decade, in agreement with the study by Macaro et al. [28]. Since then, according to our research data, the number of papers has grown exponentially, with the greatest number of publications being found in the last six years (2017–2022), accounting for 67.3% of the research analysed in this study. More specifically, during the last triennium 2020–2022, there were 21 publications (40.4%). In this way, the foundations of this line of research have begun to be laid, providing increasingly solid knowledge in this regard. Notwithstanding this fact, to date, there is a dearth of studies in this research field aimed at comparing CLIL, CBI, or EMI effects on the mathematics teaching and learning processes and outcomes on a global scale.

Thirdly, regarding the objectives of the selected corpus, the majority of studies evaluated teaching practices and learning processes and outcomes in mathematics and language proficiency in CLIL and/or, CBI and/or EMI classrooms. It is also worth pointing out that not all the papers had the sole objective of studying mathematical competence and language proficiency in CLIL/CBI/EMI (n = 11; 27.5%). Several studies were conducted jointly with other disciplines, such as science [81,82,88,93,94,97,99,102,106,107,108,109,110,112,113], or within the framework of the STEM approach [9,76,98,101,114], which has become a new trend in the field of education. This suggests that these papers may not provide a very thorough analysis of the teaching and learning of mathematics since the research goals are split and shared with other disciplines such as science or technology. We believe that a greater number of studies focusing exclusively on the teaching and learning of mathematics is necessary to gain better and more quality insights into this topic. Likewise, we should consider the fact that CLIL outnumbers EMI studies. This may be because the CLIL approach has been implemented for longer and is more widely investigated than the EMI approach [5]. However, the number of publications on both programs has recently experienced a similar increase in the field of mathematics. In this sense, we cannot neglect, as Richards and Puns [8] point out, the different forms of EMI that can come together even in the same class, making it difficult to establish clear-cut boundaries between CLIL, CBI, and EMI. This has also been argued by other authors [9,114,115]. Therefore, we have been able to observe that, in addition to the aforementioned research, the objective of EMI-related studies is not solely the study of the teaching of mathematical content but that there is a clear linguistic aim under an ELT gaze [10,12,96,97,98,100,101,102,103,104,105,116].

Fourthly, the selected studies follow a predominantly empirical, mixed-methods, descriptive, correlational, and cross-sectional methodology. Bearing in mind the latter, the absence of a greater number of longitudinal studies that can yield more conclusive results regarding the impact of CLIL, CBI, and EMI on mathematical competence and language proficiency learning processes and outcomes, over time, is striking. Regarding the research method of the studies included, mixed-methods design (n = 17; 32.7%) prevails, followed by qualitative (n = 16; 30.7%), the latter doubling the number of purely quantitative designs (n = 8; 15.4%), case studies (n = 7; 13.4%), or literature reviews (n = 4; 7.7%). We, therefore, consider that the large body of literature is opening its scope to try to understand profoundly data from a specific context. However, it can still be difficult to determine accurately the impact of CLIL, CBI, or EMI on mathematical competence and language proficiency on a wider scale, as more quantitative studies are required in this sense. Indeed, the total number of studies including a quantitative measurement (fully or partially) does not exceed half of the selected papers (n = 25; 48%). Furthermore, we must bring to the fore the high number of experimental studies [16,19,21,63,78,81] that present methodological issues as they did not conduct a pre-test to confirm that the two groups had similar mathematical competence (content knowledge) before the intervention. In parallel, we have detected other methodological issues in the selected papers that must be addressed [6]. On the one hand, most of the CLIL, CBI, and EMI programs reviewed were electives, so they tended to attract students with greater motivation for learning a foreign language and/or higher aptitudes or previous learning outcomes. In addition, these programs offered extra instruction time, so the demonstrated benefits may simply be the result of more instruction. Furthermore, as we have been able to observe in our corpus, the participating teachers often decided to collaborate in the studies voluntarily and had years of previous experience in CLIL, CBI, or EMI. From our point of view, this factor must be considered, since it is likely that the level of quality instruction and the learning results of the students participating in the research are higher. In this respect, some important experts on CLIL and CBI [4] call attention to top-quality teaching as vital to the success of this approach.

In relation to the variables, mathematical competence (n = 52; 100%) and language proficiency (n = 52, 100%) are shared by all the studies, followed by teaching practices (n = 28; 70%), teacher training (n = 25; 62.5%), science competence (n = 15, 28.8%), and teachers’ perceptions (n = 13; 32.5%). Therefore, it is remarkable that researchers are interested in determining both the benefits and the challenges of learning mathematics and, in some cases, sciences too, as well as the acquisition of the L2. They attempt to carry out successful practices that inspire training plans and models in this area. It is also key to consider their perceptions to carry out adjustments that grant validity while improving the already-mentioned practices. As for measurement, the most common instruments have been objective tests, such as questionnaires, together with standardized tests (n = 24; 46.2%) to gain insight into some aspects related to mathematical competence (mathematical thinking, arithmetic, and problem-solving), and language proficiency (linguistic skills, discourse competence, etc.). They are followed by an analysis of documents (academic record, teaching materials, official documents...) (n = 22; 42.3%), participant or non-participant observation (n = 20; 38.5%), interviews (n = 20; 38.5%), video recordings (n = 9; 22.5%) and audio (n = 2; 3.8%), and one focus group (n = 1; 1.95%). In general terms, according to one of the results of the study by Graham et al. [6], we have also confirmed that the research instruments are fundamentally focused on the measurement of basic interpersonal communicative skills (BICS). In this sense, we have not detected studies that opt for any specific instrument to measure cognitive academic language proficiency (CALP). However, we have noticed a growing interest in linking these two types of language through scaffolding techniques such as code-switching, the use of visual aids, multimodality, translanguaging, etc., to reduce the difficulty that comes from both the discipline itself as well as the L2 proficiency (e.g., [12,72,82,85,90,96,100,101,102,103,104,105,115,116]).

When it comes to the context/sample details, the studies analysed have been carried out mainly in the educational stages of secondary education (n = 20; 38.5%), university (n = 14; 26.9%), and primary education (n = 10, 19.2%). The target population under study is students (n = 16; 30.8%), both students and in-service teachers (n = 13; 25%), or in-service teachers (n = 11; 21.1%), while research focused exclusively on teacher trainees (n = 2; 3.8%) is rather scarce. One piece of information that must be taken into consideration related to the context is that not all studies specify the type of school (public, private, or others). Many also have international training plans mostly on CLIL/EMI. In addition, in general, the groups and schools in which the studies were carried out are made up of students who had previously shown a positive attitude towards learning content in another language since the latter is often part of the idiosyncrasy of the school to which both students and their families adhere to in general terms. Following Macaro et al.’s [28] results, we agree that there seems to be a greater interest in comparing the teaching-learning results of mathematics in L2 with those of L1 in the stages of primary and secondary education. Nevertheless, since this paper [28] was released in 2018, a relevant number of studies addressing higher education have been published. However, there are not many studies that compare learning in different degrees/courses [19,21,63,73,78,81,87], nor do any of them focus on the transition between the different educational stages. Consequently, we have shown that there is a lack of comparative studies between institutions and countries, especially about the education methodology for learning mathematics in an L2 through CLIL, CBI, and EMI. The single most striking observation to emerge from the data comparison was the lack of studies from other European countries where CLIL, CBI, and EMI are widely implemented (for instance, France, Italy, Switzerland, or Portugal).

Finally, regarding the main results, there are several challenges for mathematics CLIL, CBI and EMI teachers, other stakeholders, and researchers need to face according to the most relevant neutral and negative studies results which focus on:

1.

All educational stages:

• The insufficient English proficiency of mathematics content teachers and their students highlights the need to improve the CLIL CBI, and EMI policies in theory and practice, with a particular emphasis on enhancing CALP and linguistic and pedagogical abilities [87,96,97].
• The varying levels of proficiency among students can make it challenging for mathematics and CLIL, CBI, and EMI teachers to contribute to their content and language development as highlighted in studies [74,113]. Therefore, getting support from stakeholders and even families become essential in addressing this issue.
• The challenge of finding a balance between content and language learning as well as a more flexible lesson plan and the use of time-management strategies in CLIL [74].
• The existing discrepancy between macrolevel English-only monolingual language policy adoption is encouraged by contemporary ideologies closely related to English, and micro-level stakeholders’ ideologies of translanguaging and STEM pedagogy [10,101]. Regarding this, the study by Heng & Tan published in 2006 [107] warns about the kind of top-down policy that the EMI approach represents. One of the main negative impacts of this approach is the anxiety and unease it can cause in multicultural and multilingual communities, where language-related matters can trigger feelings and sensitivities associated with identity, language allegiance, power dynamics, economic benefits, ideology, and domination.

2.

At primary education stage:

• The difficulty of teaching mathematics through CLIL, CBI, and EMI in the first year of primary education affects the learning of the content such as mathematical thinking, arithmetic, and problem-solving. However, this difficulty is reduced as later courses progress [19,21,86]. This coincides with some research that suggests language proficiency is a strong predictor of mathematical academic achievement [96,106].
• Young CLIL primary school learners may exhibit symptoms of intellectual helplessness (IH) and negative affectivity [83].
• Traditional teaching methods for mathematics assessment in an L2 do not meet the needs of elementary school students [88].

3.

At secondary education stage:

• A lack of L2 language proficiency of CLIL [10] and EMI [112] in-service secondary teachers. In the same vein, the study by Hu & Gao [93] revealed a lack of language awareness, including the perceived role of language among four EMI secondary teachers whose pedagogical practices lacked work on English language forms and English language learning strategies.
• The importance of devoting more instructional time to reading and writing in L2 [82], as well as teaching L2 intercultural features that may hinder the learning of some mathematical learning concepts [84].
• Certain educational methods, such as translating and prioritizing keywords, may aid in the academic and linguistic development of students who are already proficient but can have adverse effects on the content and language acquisition of weaker students. This may also limit the development of their speaking and writing abilities [108].
• The absence of collaboration between content and language teachers delays sustainable professional development. In this sense, the convictions of CBI secondary teachers’ regarding their roles, as either solely content educators or solely language instructors can impede students’ chances of acquiring language skills [109].

4.

At the higher education stage:

• Finding a balance between the four language skills in content courses can be challenging at the university level. Reading comprehension is often prioritised over listening, speaking, and writing [110].
• Three EMI mathematics lecturers acknowledged a limited command of English and a lack of mathematical knowledge of their students [96].
• Cognitive costs on university students caused by language switching (from L1 to L2) both in problems requiring simple fact retrieval and problems requiring knowledge application in a new context, which increases students’ inhibition [96].
• The extra support and scaffolding that university students need to overcome the difficulty that learning mathematics entails concerning other subjects, even above L2 proficiency [97].
• A need for training of EMI mathematical teachers in discourse competence to promote students’ motivation and classroom interaction [9,98,114].

In parallel, to some extent, we can synthesise the study’s positive results on the following practical implications:

1.

All educational stages:

• Professional development programs can act as initial steps towards investigating the role of teachers in cultivating students, and mathematics-language registers within mathematics classrooms. The preparation of such professional development programs must be scrutinized in light of the existing knowledge regarding teachers’ skills, especially within a curriculum that aims to merge students’ language and mathematics education [80].
• CLIL should be an essential part of mathematics and English language teacher education programs and academics should be prepared to implement CLIL in higher/tertiary education. New materials must be created for implementing CLIL at all levels of education in collaboration with mathematics and English language teachers [79].
• In Finland, CLIL state-run schools have generally proven to provide advantageous, circumstances for mathematics and science content learning and cognitive development. In most instances, the cognitive advancement in the CLIL environments resembled that of teaching through the L1. In certain cases, the cognitive development of the experimental group appeared to be even more rapid than that of the control group [81].
• Positive results have been obtained regarding CLIL teaching techniques and support, such as lesson planning, the use of verbal-visual-metacognitive support, and language as the means of communication [72]. The creation of specific materials adapted to a particular context and students’ profiles paves the way toward successful outcomes in students’ mathematics learning [100].
• Educators ought to develop and adjust their own educational resources. However, there remains the necessity for pedagogical models that can assist them in the process. TBL is particularly compatible with a CLIL methodology and outlines a three-tier model for task creation in CLIL [26].
• The preliminary stages for executing CLIL for prospective mathematics teachers comprise the following steps: (1) identifying the mathematics course as a foundation for implementing the combination of mathematics and English language learning; (2) selecting specific themes or units from the course; (3) outlining the expecting outcomes of the educational process; (4) creating the organizational aspects of the educational process using the CLIL methodology; and (5) developing appropriate instructional and methodological support for the integrated learning of mathematics and English, based on CLIL principles. Also, the educational process should include a special system of multilevel bilingual assignments [89].
• The use of translanguaging as a strategy for teaching language and content is positive at all educational stages [101].
• The use of active learning and ICT tools may contribute to quality teaching in CLIL lessons [73,111].

2.

At primary education stage:

• The age of students affects how they retrieve information when the language of instruction is different from their mother tongue in CLIL lessons [21]. Variations in mathematical performance tend to decrease over time. For example, first-grade students demonstrate lower effectiveness in solving verbal math problems in an L2, compared to second-grade students [19].
• Incorporating modern interactive ICT tools and active teaching methods in mathematics lessons creates a better educational climate than teaching the same mathematical subject matter in the L1 without the use of these tools. They have a great impact on students’ motivation, activity, communication, and learning. For instance, the use of GeoGebra software has been shown to improve the effectiveness of students’ learning [78,92].

3.

At secondary education stage

• The interaction between language difficulty and mathematical complexity remains at central issue in word problem-solving [80]. On the one hand, individuals who learn through the CLIL approach are more inclined to use a given text more thoroughly to systematically deduce a mathematical model compared to monolingual individuals. On the other hand, mathematical activities can result in more rigorous language usage. Proficiency in mathematics is contingent upon effective text comprehension, and problem-solving in an additional language presents additional prospects for language-based and conceptual introspection [77,84].
• Research suggests that utilizing a CLIL approach is more effective than conventional teaching methodologies in facilitating learners’ achievement of advanced levels of proficiency in the target language, as well as promoting high levels of competency in mathematics [11].
• Code-switching can be a valuable strategy when asking for clarification, or explanation, looking for equivalence between L1 and L2 or discussing something private in CLIL lessons. It can facilitate the explanation of concepts, reduce learners’ stress, and establish a positive classroom climate [85]. Other code-switching methods that teachers can use include noticing, syllabification, morphological cues, think-pair-share, vocabulary-building strategies, questioning techniques, and immediate correction [115].
• Translanguaging can serve as a valuable tool for incorporating valuable out-of-school knowledge into the classroom, thereby aiding knowledge construction and content learning. It allows the teacher to switch between everyday language and academic register, enabling meaning-making processes to make the academic knowledge more relatable to students [103]. It also fosters an inclusive learning environment that promotes equity in knowledge construction, while also challenging the traditional hierarchical relationship between the teacher and the learner [105]. Furthermore, inclusive practices can be facilitated by translanguaging, which entails EMI teachers utilizing a range of available multilingual and semiotic resources to draw on students’ collective knowledge, thereby transcending cultural boundaries in L1, L2, and even L3 contexts [12].
• The use of the playful talk strategy can turn the classroom into a space where translanguaging is possible. This allows the teacher and students to engage in various creative acts and try out different voices to facilitate the process of making meaning and constructing knowledge [116].
• Using ICT tools such as tablets can provide teachers with opportunities to utilize visual and spatial elements to create a technology-enhanced learning environment. This can lead to a more immersive and interactive classroom experience for students [104].
• Learners need sizable vocabularies to cope with authentic school textbooks, especially in mathematics. Mathematics texts might be adapted by identifying and replacing low-frequency vocabulary items with high-frequency items. Besides, learning materials provided by teachers present a higher proportion of the most common vocabulary than textbooks, so they can act as a sort of lexical bridge between the classroom and the textbook [106].

4.

At the higher education stage

• Implementing a translanguaging perspective to understand university mathematics can promote a greater understanding of mathematics discourse, context, and culture [96].
• Some methods and techniques that teachers can use to accommodate different levels of students’ language proficiency include: using language scaffolding such as semantization (selecting synonyms, periphrasis, and reiteration); reducing the complexity of the instructional language; assessing students’ academic achievements individually during seminars; designing their own instructional material; and selecting adapted or authentic instructional materials [90].

To sum up, in order to respond to Bermejo et al.’s [21] statement on the necessity to study in greater depth the different levels of language acquisition, the research analysed shows that these have not been sufficiently addressed in the mathematical field of knowledge (only Yakaeva et al.’s [90] study takes this into account). After analysing the selected papers, we can also affirm that most of the studies have tried to provide an answer to whether mathematical competence is hampered if taught in an L1 or an L2. As we have observed, a positive or neutral view predominates on the effects of the CLIL, CBI, and EMI approaches on the learning of mathematics and L2. This result partially converges with that offered by Graham et al. [6] who revealed that their research findings had a positive or neutral impact of CLIL and EMI on student outcomes compared to monolingual classrooms. There is also a general tendency in scientific research to offer positive results over time, but these need still to be confirmed in comparative studies on a wider scale. To date, most concerns are still teachers’ and students’ language proficiency, and content teachers’ language awareness. Regarding this, intercultural features related to mathematics learning must be addressed. Nonetheless, teachers should pay more attention to cognitive and affective barriers. From this perspective, we agree with Sedek and McIntosh [117] in the sense that negative affectivity hinders complex cognitive processing that is crucial for problem-solving. These authors proved that experiences of failed intellectual effort may produce impaired information processing called intellectual helplessness (IH). Recent studies [83,118] suggest that this may be mainly provoked by ill-managed classes and linguistically inadequate materials too focused on CALP, which usually implies high cognitive demands reducing students’ affectivity and motivation [118]. Furthermore, methodological language adaptations like using high-frequency vocabulary to the learning of mathematics are important, as well as acknowledging students’ mother tongue and their L1 culture and using them to promote a positive learning climate to avoid intellectual helplessness.

5. Conclusions

After answering and discussing the questions posed at the beginning of this research, we can conclude that the study of the interrelationship established between content and L2 stands out as one of the main current questions raised by the scientific literature in this field of knowledge [RQ7]. Despite the methodological issues described above, most of the analysed research has tried to answer this question through the analysis of teaching practices that can serve as support for students to overcome the difficulty of learning both mathematics and a foreign language. These include the study of discourse competence and the subsequent use of language scaffolding, language switching, translanguaging, and playful talk, as well as the use of other visual supports and digital media that can effectively support communication, interaction, and content transmission [RQ7].

In the same vein, we consider it important to highlight the studies that relate mathematical competence and linguistic development together with cognitive development [19,21,76,77,81,83,86], describing the role that executive functions play in the learning of mathematics in L2. Among these functions, working memory, attentional control, and cognitive inhibition have been pinpointed by the research as those that are more involved in the learning processes of mathematical content in another language.

At the same time, even though great strides have been made to date, it is timely to draw attention to the scarce number of studies found in the analysed scientific literature whose objectives contemplate affective aspects related to the learning of mathematics in L2 [76,78,83,91,92]. This is essential for any type of learning [RQ3].

Likewise, we should note that positive results may have been a natural consequence of good practices and more instruction time [RQ3, RQ7]. So, we need to bear in mind some methodological issues fundamentally related to the methodology, research design, sample, and measurement instruments [RQ4, RQ5, RQ6]. Consequently, it is advisable to interpret this study’s results with caution.

One of the limitations of systematic reviews is that information may have been lost due to not having used correct descriptors or due to the low feasibility of covering all the existing databases at present. We avoided too the use of the term “maths” in the search equation as it is not listed in UNESCO’s thesaurus. We also considered this term as belonging to a less formal and non-academic language register. Nevertheless, we acknowledge that this might have shaped the results of the search. Another limitation that may have influenced the findings of this review is that all the papers are written in English, and we may have excluded papers in other languages that might have had important implications for the findings reported here. It is worth mentioning those references to which it has not been possible to have access and that would have enriched the results obtained in our research.

To conclude, we believe that further research is needed to investigate the effects of CLIL, CBI, or EMI on the learning of mathematics in languages other than English. Some experts have suggested that the learning of mathematics may differ depending on the intrinsic characteristics of the language in which it is taught [119,120]. Therefore, we consider that the current number of papers focused on L2s other than English is not noteworthy. In short, there is a need for more research in this field, both on a small and large scale, to determine more precisely the effects of CLIL, CBI, and EMI on the teaching and learning of mathematics in an L2 [RQ1, RQ2, RQ3, RQ4, RQ5, RQ6, RQ7].