A comparative study of Japanese and English medium instruction: the driving factors behind academic success

ABSTRACT

This paper compares direct measures of content knowledge for students studying Chemistry through English Medium Instruction (EMI) and Japanese Medium Instruction (JMI). It also examines a range of contextual differences between the two mediums of instruction. A total of 53 undergraduate students undertook pre–post course content tests at a university in Japan. These measurable test outcomes were triangulated with data from semi-structured interviews with eight EMI and five JMI students and two content teachers as well as field notes from ten EMI and ten JMI lecture observations. While the quantitative results revealed no statistically significant differences in the post-test scores between the two courses, the qualitative data illuminated a huge range of context-dependent differences which shed light on possible explanations of this non-significant result. Findings revealed that while there were unique challenges only experienced by EMI students, many of the students navigated their EMI studies successfully through a combination of effective institutional and academic language support, teacher support and – not least – collaborative peer networks. This comparative study of an EMI and non-EMI course contributes to the currently limited research evidence on medium of instruction differences on content learning outcomes.

KEYWORDS:

• English medium instruction
• higher education
• academic success
• language support
• Japan

Introduction

In a globalised era in which English has become the de facto lingua franca in higher education (Rose, Sahan, and Zhou 2022), universities around the world have observed an increasing popularity of EMI programmes in recent decades. While the perceived benefits of EMI have now been well documented (e.g. Macaro 2018), there are few studies that have investigated the potential drawback of EMI on the quality of disciplinary learning, particularly whether ‘learning content through L2 English leads to at least as good learning of academic content as learning content through the students’ first language’ (Macaro 2018, 154). In response to this research agenda, this study explores the impact of medium of instruction through a comparison of students’ content learning outcomes between an EMI and JMI course as a direct L1 comparison group. Consistent with Macaro’s (2018) observation of EMI research as ‘an evaluation of naturally occurring phenomenon’, this study acknowledges that EMI is ‘an educational practice that is already happening and needs evaluation for its effectiveness’ (155). Accordingly, this article adopted classroom-based research, ensuring its richness in ecological validity to capture the authenticity of an EMI course whereby students are self-assigned to two learning conditions. Macaro explains this as ‘voluntary enrolment’ serving as a crucial factor in ‘outcome studies’ – leading to numerous pre-existing differences (i.e. ‘messiness’). As outcome research is prone to various uncontrolled factors, such as these confronting researchers as methodological barriers, this study adopted both score gain measures and follow-up perception-based methods to capture the natural climate of instructional activities pertaining to the process of ‘lived through’ learning experiences (Curdt-Christiansen 2020, 346). This research thus harnessed the complexity surrounding the constellation of key factors driving success in EMI.

Background to the study

EMI outcome studies

There is now a growing body of research evidence on medium of instruction differences on academic outcomes in higher education (e.g. Lin and Lei 2021). Arroyo-Barrigüete et al. (2022) conducted one of the largest scale studies to date, revealing no significant differences in the academic performance of EMI students (n = 229) and Non-EMI students (n = 635) enrolled in the Business Administration programme in Spain. The study controlled for some learner variables, including gender, high school subjects, resident origin and pre-university performance. However, the authors themselves pointed to various other uncontrolled factors associated with a pre-selection bias of the EMI cohort. Similarly, Dafouz, Camacho, and Urquia (2014) explored the effect of EMI on undergraduates’ academic performance measured using the results of the coursework and final exam. The findings illustrated that both EMI and Spanish medium cohorts yielded a similar academic achievement, indicating that EMI did not compromise the students’ performance. However, comparing the final grades posed methodological issues, noted by the authors themselves that ‘differences as to the precise criteria applied in the final evaluation of students’ performance seem to be largely dependent on individual teacher decisions’ (229). Similar methods were observed in their subsequent study (see Dafouz and Camacho-Miñano 2016). Hence, whether the academic achievement was comparable between the two groups remained questionable. Without adopting a direct comparison group of an L1 medium instruction group, EL-Daou and Abdallah (2019) examined undergraduate students’ maths test performance in Lebanon (n = 21). The descriptive statistics and paired samples test of pre-test and post-test scores of a maths achievement test showed significant differences in mean test scores at the start and end of the ‘40 instructional periods’ (6), detecting significant disciplinary gains among the English medium students. However, no data were collected regarding how the course grades were comparable between the pre–post-test conditions.

While these studies indicate no adverse effects of EMI on content learning, Civan and Coşkun (2016) found that Turkish medium instruction (TMI) was more effective than EMI. The study compared academic outcomes of EMI and TMI students with regards to their semester point averages across nine academic disciplines at a Turkish university. Meanwhile, no qualitative data were collected regarding how the point averages were comparable between the two conditions. The similar methodological limitations considering the absence of qualitative data were also identified in Hernandez-Nanclares and Jimenez-Munoz’s (2017) study which compared 172 EMI and 482 Spanish medium students solely based on course grades obtained from their final examination results. Their findings revealed similar grades in both the cohorts but a notable gap in higher bands. From the few studies carried out thus far, findings are mixed. Thus, more research is warranted not only to quantitatively measure medium of instruction differences, but also to qualitatively describe how contextual differences are manifested between the two mediums of instruction.

Classroom-based EMI outcome research

In classroom-based EMI research, observation is a crucial means to shed light on contextual factors. Jiang, Zhang, and May (2019) adopted nine lecture observations, three post-observation interviews and a questionnaire survey in exploring undergraduate students’ language-related challenges in EMI. The findings suggested that whether continued language support was offered alongside their regular EMI courses was shown to reduce the degree of their challenges. In another observation study, Trang and Trinant (2017) adopted a questionnaire, semi-structured interviews and observations to identify patterns of classroom interactions, demonstrating that the absence of interaction impinged on their content learning outcomes. However, these observation studies included no comparable L1 learner groups. In a growing body of EMI outcome research, little to no research has compared medium of instruction differences via observation. In one of the few observation studies, Guo et al. (2018) compared medium of instruction differences, recording wider contextual factors, such as admission policies and instructional practices. The findings identified a range of contextual differences between the EMI and Chinese streams of the television art programme, revealing that L1 use in EMI was shown to facilitate the degree of EMI students’ disciplinary gains.

Observation thus captures the fidelity of the real-world teaching practices and speculates on potential contextual, linguistic, institutional and learner factors (Pun 2021). However, few studies have incorporated score gain measures and follow-up qualitative methods to compare the medium of instruction differences. To fill this methodological gap, the current study draws on rich sets of qualitative data to elicit detailed and illuminating contextual data required to achieve a deeper understanding of students’ experience and possible sources of their challenges.

Summary: methodological concerns for EMI outcomes studies

In the growing body of research on L2 medium of instruction, academic outcomes are often investigated via one single method, rather than multiple data collection methods. There is a paucity of mixed-methods research in particular that triangulates direct measures of test outcomes with perception-based measures, indicating that comprehensive contextual factors are rarely examined in EMI research (e.g. Macaro 2018). As quasi-experimental research introduces various contextual factors, the current study gathered follow-up qualitative data to capture the underlying factors that led to success in EMI. Accordingly, this study aims to account for a highly contextualised and multifaceted nature of direct measures and perceptions of content knowledge acquisition.

The study

The study addresses the following research questions:

1. To what extent does the achievement of direct measures of content knowledge of the EMI and JMI students differ?

2. What contextual differences between the EMI and JMI courses do the students ascribe to their achievement of content learning?

Research contexts and participants

The study was conducted at a single university in Japan to take into account the environmental and context-specific nature of the implementation of the Top Global University Project (TGUP) initiative. This institution was purposively chosen for the following reasons:

• The university has been annually funded as part of the TGUP by the government to promote internationalisation since 2014, thus providing uniform and consistent EMI provision (Rose and McKinley 2018).

• The university has proposed international strategic plans to increase the proportion of EMI programmes from 16.5% to 40.0% from 2013 to 2023, thus offering a number of EMI courses.

• The university values its strong tradition of bilingual education, providing a number of parallel English and Japanese programmes in a range of academic disciplines.

This study was conducted in an introductory Chemistry course, which was available in either of two languages of instruction. This course offers 70 min of lectures twice a week for 12 weeks; 21 lectures in total. The Division of Natural Sciences offers this bilingual parallel course, offering students options to study in their preferred medium of instruction. Accordingly, the course syllabus, assessments, materials, objectives and textbooks were the same between the two streams.

All students enrolled in the EMI (n = 27) and JMI (n = 31) courses were invited to participate in the study. However, five JMI students withdrew from the study. Thus, responses from a total of 27 EMI and 26 JMI students were ultimately analysed in the quantitative analyses, which represented 100% and 84% participation rates, respectively.

Adopting a maximal variation sampling strategy, interview participants were purposively selected as individual cases to explore a range of variables shown to influence success in content learning. Table 1 shows the selected cases of eight EMI and five JMI students representing a range of content knowledge and L2 proficiency levels, and the two mediums of instruction.

Table 1. Purposive sampling of students (n = 8 EMI; n = 5 JMI).

Content learning (Post-test scores)		English Proficiency (IELTS Scores)
		Pass (1–6) (n = 5)	Merit (7–12) (n = 4)	Distinction (13–17) (n = 4)
EMI (n = 8)	IELTS 6.5 (n = 2)	Student A (Post-test 5)	Student B (Post-test 7.5)
	IELTS 7.0 (n = 3)	Student C (Post-test 4) Student E (Post-test 1)	Student D (Post-test 8.5)
	IELTS 7.5 (n = 1)			Student F (Post-test 2.5)
	IELTS 8.0 (n = 2)		Student G (Post-test 9.5)	Student H (Post-test 15)
JMI (n = 5)	N.A. (n = 5)	Student I (Post-test 5) Student L (Post-test 3)	Student J (Post-test 8)	Student K (Post-test 15) Student M (Post-test 16)

The classification of Pass (0–40%), Merit (40–70%) and Distinction (70–100%) was adopted from Assessment Criteria of the university.

In addition to students, two content teachers, one EMI and one JMI, were also interviewed. Table 2 shows information about the content teachers who participated in the study.

Table 2. Information of content teachers (n = 1 EMI; n = 1 JMI).

MOI	Position	Gender	L1	Teaching history
EMI	Professor	Male	L1 English	Teaching in Japan over 15 years
JMI	Professor	Male	L1 Japanese	Teaching in Japan over 20 years

Data collection instruments

The present study draws upon the following research instruments and measures:

• Pre–post-test scores on a content knowledge test of introductory Chemistry completed by EMI and JMI students at the start and end of the semester:

The mastery of content knowledge was operationalised in terms of students’ test scores of a Chemistry test (see Appendix 1). The content teachers devised 13 open ended questions, ensuring the high-level representation of the final course exam. To devise the tests, the teachers adopted the same learning objectives and marking rubric as those used in the Chemistry course and ensured that the test items were equivalent between the two courses. The pre-test and post-test were administered under supervision of the content teachers during the first and last lecture of the semester. All test items were presented to the EMI and JMI students in English and Japanese, respectively. The participants were offered 25 min to complete the tests.

• Semi-structured interviews with eight EMI and five JMI students (see Table 1):

Student interviews were conducted to draw out perceived learning experience while identifying underlying factors affecting their content learning. The average length of interviews was 26.8 min (Mean 26.8; SD 6.4; Min 14; Max 36) (Appendix 4 for interview schedule).

• Semi-structured interviews with an EMI and JMI teacher:

Teacher interviews were conducted to understand how teaching was operationalised in practice in terms of language use, classroom discussions, pedagogical styles, pop quizzes, student expectations, faculty meetings and teaching practice. Although the EMI and JMI courses shared the same learning objectives, it was crucial to gain their situated knowledge and views on their teaching practices (Appendix 5 for interview schedule).

• Observation fieldnotes (n = 10 EMI; n = 10 JMI):

Unstructured observations were adopted as a medium to elucidate as many contextual differences as possible between the two courses (e.g. educational culture, interaction, classroom activities). Based on an observation summary guide adopted from Huberman and Miles’s (2002) observation questions from their contact summary form (see Appendix 2), the researcher himself acted as a non-participant observer and recorded detailed field notes of ten lectures each for the JMI and EMI courses.

Data analysis

Before conducting statistical tests, data were tested for normality of distribution. A z-score of skewness and kurtosis was computed, demonstrating approximately normal distributions of the post-test scores. The pre-test scores, on the contrary, appeared to be positively skewed, indicating that a large proportion of scores clustered at the lower end of the distribution. As a result, a Wilcoxon Signed Rank test was used to contrast the pre–post-test scores within the groups (with one within-subjects variable). A one-way ANCOVA was then performed for the subsequent analysis that explored the differences in post-test scores between the groups while controlling for pre-test scores (between groups comparison of the post-test scores).

To code the volume of qualitative data, all interviews were transcribed and analysed in NVivo. Adopting an abductive approach, main themes were initially developed deductively based on EMI literature and inductively from raw data. Initial themes emerged from transcripts which represented an array of contextual differences between the two mediums of instruction (e.g. educational culture). These themes were considered across all of the data sources, whereby each of them was further collapsed into subthemes (e.g. teacher centred pedagogy). Data from the fieldnotes were then consulted to add a further dimension and ascertain any contextual similarities and differences between themes identified in the interview data related to related to context-related differences in the areas of language use, peer learning, support schemes and educational culture.

Findings

Content learning differences according to the medium of instruction

Research question one examined the extent to which the achievement of content learning differed between the EMI and JMI students. Table 3 shows the descriptive statistics for the pre-test and post-test scores for the whole sample and for the EMI and JMI samples separately. First, the results of a Mann–Whitney U Test found no differences in the pre-test scores of the EMI (Md = 0, n = 27) and JMI students (Md = 2, n = 26), U = 259, z = −1.69, p = .09, r = .23, suggesting that the two groups were not significantly different in their prior knowledge. Subsequently, the results of a Wilcoxon Signed Rank test suggested a significant score gains for both the EMI group, z = −4.545, p < .001 with a large effect size (r = .62) and for the JMI group, z = −4.463, p < .001 with a large effect size (r = .62). The median for the EMI and JMI groups on the pre-test scores increased from (Md = 0) and (Md = 2.0) to the post-test scores (Md = 12.0) and (Md = 9.75), respectively. Thus, both groups improved their scores between the start and end of the semester (see Appendix 3 for the distributions of pre-test and post-test scores).

Table 3. Descriptive statistics – test scores for the EMI (n = 27), JMI (n = 26) and total sample (n = 53).

	MOI	N	Min	Max	Mean	Median	SD
Pre-test scores	EMI	27	0	13.5	1.80	0.5	3.21
	JMI	26	0	7	2.54	2.0	2.45
	Total	53	0	13.5	2.16	1.0	2.87
Post-test scores	EMI	27	1	15	10.40	12.00	3.15
	JMI	26	2	17	9.64	9.750	5.04
	Total	53	1	17	10.01	9.500	4.60

A one-way ANCOVA was performed to explore the medium of instruction differences on content knowledge. Preliminary checks were carried out to ensure no violation of homogeneity of variances and reliable measurement of the covariate. No statistically significant interaction was observed between the dependent variable and covariate at an alpha level of .111, and thus no violation of the homogeneity of regression slopes. When adjusting for the pre-test scores, no significant differences were found between the two groups on the post-test scores, F(1, 50) =  1.415, p = .24, partial η2 = .027. Thus, the two groups were not significantly different in their post-test scores while accounting for the pre-existing differences in their prior content knowledge.

Contextual differences according to the medium of instruction

In light of this non-statistically significant result, research question two considered the extent to which the students attributed contextual factors to their achievement of content learning. Interview data revealed that success in EMI was underpinned by several contextual conditions.

1. Strong presence of L1 medium resources among EMI students

The use of L1 medium resources was the first contextual difference between the EMI and JMI courses. The EMI students regarded the use of an L1 as a pedagogical tool associated with better outcomes in content mastery. Student D (IELTS 7.0, post-test 8.5) reported using textbooks and notebooks from high school, noting that ‘it is helpful to revise basic academic concepts in Japanese’. Student B (IELTS 6.5, post-test 7.5) also referred to online resources to gain a deeper understanding of academic concepts in Japanese. She noted that ‘I like watching tutorial videos in Japanese on YouTube’. Although conducting lectures entirely in English, the EMI teacher also supported the use of L1 medium resources. He observed that students who had previously studied Chemistry in Japanese used both the English and Japanese versions of the same textbook, noting that ‘it is important that they can use their first language to build a strong foundation in the knowledge and understanding of Chemistry’.

Conversely, the JMI students showed little to no interest in using English medium materials. Student L (post-test 3) exclusively relied on Japanese medium resources, noting that ‘I do not see the point of reading materials in English. The class is conducted in Japanese’. Despite student L's negative stance, the JMI content teacher was supportive of using English:

JMI teacher: I would like my students to learn technical terms in English. Being able to read research articles in English has become important in many of the science disciplines.

He emphasised English as an academic lingua franca in STEM. He also allowed students to submit problem-solving worksheets in English. To support this view, the fieldnotes also revealed that the majority of JMI lecture slides were presented bilingually. A list of suggested reading and resources was also available bilingually. Given that the JMI students rarely utilised English medium resources, there was a gap between how the language of instruction was envisaged and how it was being put in practice. Thus, the current learning environment where the EMI students utilised L1 was one of the conditions which may have contributed to successful learning. That is, the students utilised their L1 to support their EMI studies.

2	Peer learning as a pedagogical tool to ensure content mastery in EM

The EMI students valued peer learning as a pedagogical tool to ensure content mastery. For example, a high performing learner, student H (IELTS 8.0, post-test 15) noted that ‘I often sit next to my course mate in lectures and we study together afterwards’. While playing a more dominant role in aiding in her peers’ academic development, student H was still able to enrich and expand her own understanding through explaining difficult concepts with clarity and ease. Student D (IELTS 7.0, post-test 8.5) also shared her positive experience of working closely with her peers who ultimately became close friends. She attributed supportive peer networks to her long-term academic success.

Student D (EMI): I always do my homework with my classmate. Having close friends from the course has helped me make it through some of my difficult challenges at university.

Thus, the EMI students turned to their peers for help even when they did not necessarily share the same level of content knowledge. To mirror this positive view, the EMI teacher also pointed to clear long-term benefits associated with peer collaboration particularly for science degree students.

EMI teacher: It is important for Chemistry major students to build good relationships with their course members in their first year. They need to work collaboratively when they advance to junior and senior year courses where they work on laboratory based projects.

However, none of the JMI students were able to establish supportive peer networks. Student K (post-test 15) noted that ‘I understand almost all concepts in this introductory course so I rarely need help from my course members’. Student J (post-test 8) also supported the absence of peer learning that ‘I have not made any friends in the course because the lecture room is too big and there are too many students’. One possible explanation for the perceived absence of peer learning in JMI was the size and space of the classroom. The JMI content teacher noted below that a combination of the large class size and teacher-centred teaching was barely conducive to collaborative learning.

JMI teacher: In the introductory level courses, it is often difficult for students to interact with each other because we usually give formal lectures. This works well when we teach a large group of students in a spacious lecture hall.

Thus, the large class size coupled with the tightly constructed lecture style both contributed to the absence of peer learning. Unlike the JMI students, the EMI students enjoyed a range of benefits associated with peer learning (e.g. improving content knowledge, establishing close friendships and creating long term support networks). These benefits significantly improved the speed and ease of the transition into EMI education. It also became clear that a mixed ability learning environment offered opportunities to learn to communicate and breakdown complex concepts in a way which was easier for other students to understand. Overall, the positive evidence found in this study appeared to be attributed to a context where the EMI students established collaborative peer networks to combat their academic challenges.

3	Support provision was put in place differently in EMI and JMI

The EMI students utilised frequent learning support called the Comment Sheet, where they could ask questions to their lecturer. Student C (IELTS 7.0, post-test 4) could express her concerns and issues in writing, stating that ‘I am often too shy to ask questions in front of my classmates. But I can use the Comment Sheet to share my questions with my professor’. Student A (IELTS 6.5, post-test 5) also illustrated how this support scheme enhanced his overall learning experience.

Student A (EMI): I appreciate the teacher taking the time to address our questions using the Comment Sheet. It helps me to stay motivated and enjoy learning Chemistry.

In support of this positive view, the EMI teacher explained that ‘the Comment Sheet is an effective way to establish a rapport with students, especially with those who are quiet the class’. The fieldnotes also recorded that the EMI content teacher addressed students’ comments and questions at the start and end of every lecture to build a communicative relationship with students.

Conversely, the JMI teacher was not necessarily able to adopt the Comment Sheet in every lecture. He admitted that ‘I do not address students’ comments at the start of every lecture because it can take up too much of our precious lecture time. There are many topics we need to cover in an introductory course’. Overall, the use of the Comment Sheet was left open for each teacher, and was not consistent across the two courses. Thus, the differences in how this support was introduced could be another factor that contributed to the successful development of content learning.

In regard to another form of support, the EMI students attributed the academic English programme to their improvement of academic skills required to undertake EMI courses. However, none of the JMI students made reference to the benefit of such language support. The university offered first year students a range of in-sessional language courses organised concurrently alongside the mainstream courses. Student D (IELTS 7.0, post-test 8.5) believed that this support ensured her readiness for EMI education, noting that ‘the language programme is helping me to read the Chemistry textbook quickly’.

Similarly, student A (IELTS 6.5, post-test 5) asserted that this support scheme instilled greater confidence in using English in academic contexts and facilitated a smooth transition between language and content courses.

Student A (EMI): The English language program helps me feel confident in taking English medium courses because I get to read lots of authentic books in English.

Overall, the benefits of the academic English support were only reported by EMI students, and the two courses unveiled various contextual differences in how the support provision was integrated, and these contextual differences may have contributed to the progress of the content learning.

4	Interactive educational culture in EMI

EMI lectures typically involved a range of student-teacher interaction, which promoted an exchange of ideas and discussion of key academic concepts. Although the EMI teacher delivered a pre-planned, tightly formatted presentation, he frequently paused between class activities to encourage students to share ideas and engage in group discussions. The EMI students also regarded this interactive educational culture as a way to enliven and clarify difficult academic concepts. For example, student C (IELTS 7.0, post-test 4) commented as follows:

Student C (EMI): Our professor remembers our names and nominates us to share our ideas. I have many opportunities to interact with my professor, so I feel confident in visiting him during office hours.

Similarly, student E (IELTS 7.0, post-test 1) revealed that ‘it is not what I am used to at high school but I find it useful to regularly interact with my professor and classmates to exchange our ideas’.

To support these positive views, the fieldnotes also recorded numerous instances where the EMI teacher nominated students to come up to the front of the lecture theatre and explain key concepts to the rest of the class using a blackboard and chalk. The rest of the class listened to the explanation and pose follow-up questions, which then generated further student-centred discussions.

The EMI content teacher also suggested a range of techniques to promote greater student-teacher interaction, such as learning about his students, memorising their names, recognising those with low English proficiency, learning where they sat in the lecture hall, offering them individually tailored feedback, and encouraging them to share ideas with other course members.

EMI teacher: I try and remember their names and where they sit in the classroom. I also try and work out who is struggling and who is doing well, and who needs extra support.

Thus, the interactive educational culture enabled the EMI students to share their ideas and receive individually tailored feedback from their teacher.

In contrast, the JMI students perceived that their educational culture closely resembled that of their high school. Student J (post-test 8) described her lectures as ‘teacher-centred’ and ‘passive’. Student K (post-test 15) also revealed that ‘none of my course members did not answer any of the questions posed by our lecturer in today's lecture which lapsed into uncomfortable and awkward silences’. Both these students pointed to the strong reliance on the more traditional and formal teaching method. The fieldnotes also recorded no instances where interaction occurred between the teacher and students. In the excerpt below, the JMI content teacher offered two explanations for the absence of interaction. Firstly, the course involved delivering lectures to a group of up to 31 students in a large lecture hall. Secondly, the introductory course aimed at the mass transmission of a large volume of disciplinary knowledge from teachers to students through the medium of tightly constructed, well-organised lectures. The JMI teacher admitted that he typically offered a ‘chalk and talk’ lecture in which he mainly lectured while students quietly generated notes:

JMI teacher: I usually deliver lectures in a traditional way, using a blackboard and chalk because I have to teach a large group of students. There is also a lot of content to cover in these foundation level courses.

Thus, considering the myriad of additional benefits associated with interactive pedagogy, the differences in educational culture could be another key condition that led to success in EMI.

Discussion

Institutional language support

The results of the quantitative evidence unveiled no measurable differences in the development of learners’ understanding of Chemistry between the EMI and JMI students when accounting for pre-test scores. Such results are consistent with previous research on EMI content learning outcomes (e.g. Lin and Lei 2021), suggesting that EMI did not impinge on the participants’ academic success compared to that of their L1 medium counterparts. In response to this non-significant result, this study identified a range of key contextual factors that lead to academic success in EMI. The findings revealed that EMI students benefited from a range of support provisions, including English language programmes. Such findings concur with previous EMI research on academic skills support (e.g. Galloway and Rose 2021). Galloway and Rose (2022) in their recent conceptual work propose various measures to forge the relationship between the two fields of EAP and EMI research to assist EMI students in their discipline-specific work. This scholarly interests in academic skills support also corroborate other research evidence that major sources of academic challenges in EMI are associated with the insufficient English proficiency among students (e.g. Aizawa and Rose 2020; Pun and Jin 2021), and sustained and concurrent language support should be offered alongside their EMI courses, rather than short term, pre-sessional language support (Macaro 2018).

Bilingual models of EMI

The current study also draws upon rich sets of qualitative data to shed light on a broader range of language support measures (e.g. L1 medium teaching materials and bilingual course textbooks). It is possible that this broader language support, especially the use of L1, may have accounted for an even larger proportion of positive factors contributing to EMI students’ successful performance than the academic language support provision alone. This finding does concur with other research findings that emerged in other contexts, including Europe (e.g. Muguruza, Cenoz, and Gorter 2020), South Asia (e.g. Karim et al. 2021) and Middle East (e.g. Sahan, Rose, and Macaro 2021) that EMI students’ successful content learning experience was largely attributed to their bilingual medium instruction contexts where students are offered access to L1 medium instruction alongside their EMI courses. Rose et al. (2020) also found that institutional policy at two of their language-specialist universities in China mandated the use of Chinese and English in their teaching materials and assessments, indicating the prevalence of bilingual education policy in various EMI settings. Such findings indicate a strong need for a bilingual model of language support. In the current case university, students were offered options to select either of their preferred language of instruction, and the majority of students and content teachers valued the use of L1 support (e.g. L1 medium teaching materials) as useful pedagogical resources.

Interactions in learner-centred EMI classrooms

The current study also highlighted a myriad of other classroom-based factors, such as class size, interaction, peer feedback, group discussion, mixed ability learning and interactive educational culture. There is a growing body of research evidence in the area of classroom interaction in EMI (e.g. Sahan 2020 in Turkey; Tai and Wei 2021 in Hong Kong; Macaro et al. 2019 in Italy), many of which suggest that interaction in EMI classrooms is likely to be dominated by teachers, leading to low degree of student participation. Unlike this previous evidence that EMI lectures are unidirectional, the EMI class in this case study was highly interactive. This study yielded some evidence that small class size was a key driver behind the high degree of student engagement in the EMI class. In this study, the EMI teacher also adopted various pedagogical techniques (e.g. the Comment Sheet) to achieve interactive teaching. As the pedagogical benefits of interaction in EMI have been theoretically justified both in terms of content learning (An and Macaro 2022) and language learning (Gass 2018), this evidence on frequent student-teacher interaction may have, therefore, contributed to the positive conditions that underpinned successful content learning in EMI.

These classroom factors are highly context dependent but also shown to be paramount to students’ academic success in previous education research (see Spyros and Wei 2012, for a review of class size and educational outcomes). Students in this study navigated their EMI studies successfully through a combination of institutional provision, teacher support and – not least – peer networks. These results also underpin other findings on EMI pedagogy. Ismailov (2022) found that regardless of students’ English proficiency (ranging from B1 to C2), classroom interaction and teacher question types were vital to their perceived academic success when subjects were taught through L2 English.

In the case of the current study, students with high L2 proficiency and content knowledge still benefited greatly from working in collaboration with other course members. This evidence highlights the value of a mixed ability learning environment. Another study by Rivero-Menéndez et al. (2018) also found that students learning via EMI demonstrate greater mature learning strategies and motivation than their counterparts. The implications for this finding may require EMI content teachers to promote interactive educational culture, considering its effects on students’ positive learning behaviours (e.g. motivation, collaborative learning, learning strategies). This will ensure that EMI policymakers and course designers take into account classroom factors when planning and designing EMI courses (e.g. class size).

Implications, limitations and future research

This study embraced the ‘messiness’ of EMI implementation as part of its ecological validity – that is a myriad of contextual factors were considered part of the real-world representation of EMI programmes. It is worth emphasising that the findings of this single-site examination especially with small sample sizes such as this study may be limited to this educational setting where EMI students benefited from a range of institutional support schemes that greatly facilitated their success. Another factor that contributed to success in EMI was the academic subject explored in this study. Compared to conceptually oriented subjects (e.g. humanities), Chemistry involves single-word questions, multiple choice questions and problem sets regarding formula and calculation. As such, the current study might not have captured the full extent of how the language of instruction could affect students’ academic performance (see Yang and Farley 2019 for the impact of EMI on different academic subjects). As observed by Kim and Thompson (2022), EMI research should take a contextually sensitive approach accounting for individually unique research contexts, including the interaction between demographic and contextual factors (e.g. institutions, countries, course levels, student and teacher profiles, pedagogy and academic disciplines). More research is thus needed to examine factors that are held accountable for success across different research settings (including a range of soft and hard academic subjects), whether it be language support, collaborative learning, or frequent teacher-student interaction.

Regarding recommendations for teacher practice in university settings, as shown in this study (and others such as Sahan, Rose, and Macaro 2021), L1 use appears to contribute to EMI success despite it being relegated to outside of the classroom. Consistent with bilingual models of EMI education observed in other research contexts (e.g. Karim et al. 2021), this finding also highlights the need for EMI practitioners to embrace the use of L1 medium resources. With regards to methodological recommendations, this study adopted direct measures of content knowledge to investigate the impact of medium of instruction. It also recruited the JMI group as a baseline group which was comparable to their EMI counterparts to observe the impacts of EMI while accounting for the effects of prior content knowledge. These measurable test outcomes were also triangulated with semi-structured interviews and observations to explore an array of positive factors that contributed to the speed and ease of content learning in EMI. The triangulation of data sources thus achieved nuanced contextual data to improve the transparency of the inevitably messy classroom-based research where a range of differences existed between the two mediums of instruction. Future EMI outcome research should adopt not only direct measures but also qualitative measures to capture a contextually rich and detailed picture of students’ learning experience.

Conclusion

Although EMI is expanding at a rapid pace across the globe, research has identified a range of variations of EMI implementation with respect to language use (e.g. Sahan, Rose, and Macaro 2021). While policymakers insist on implementing English-only EMI policy, many institutions around the world embrace a bilingual and multilingual model of EMI pedagogies as normal practice, providing learners with options to study their chosen academic subjects in their preferred language of instruction (Aizawa and Rose 2019). When institutions offer an EMI and non-EMI parallel course, assessments, learning objectives, materials and teachers are often the same between the two mediums of instruction. However, when instructional language changes, various other aspects of course implementation should also reflect this policy change. While offering the same course in two instructional languages may be attractive to policymakers and university leaders, L2 students should be offered greater levels of support, and more opportunities for classroom interaction should be created to help them ameliorate their challenges which may be exclusive to EMI studies.

This study is one of the first outcome studies to compare medium of instruction differences while achieving a direct comparison group of JMI as non-EMI group. Data were collected in real-world classroom environments, rather than sanitised research conditions where contextual factors are sidelined. As a result, this study captured a range of context sensitive factors that facilitated EMI outcomes of disciplinary gains. As such, the positive evidence found in this case study was largely attributed to the context where the learners were offered a wide variety of institutional and pedagogical support. While acknowledging the messiness of such classroom-based research, this study successfully identified the crucial conditions which accounted for the successful development of content learning in EMI.

Disclosure statement

No potential conflict of interest was reported by the author(s).