399 Studies in Second Language Learning and Teaching Department of English Studies, Faculty of Pedagogy and Fine Arts, Adam Mickiewicz University, Kalisz SSLLT 13 (2). 2023. 399-426 https://doi.org/10.14746/ssllt.38280 http://pressto.amu.edu.pl/index.php/ssllt Examining the role of English language proficiency, language learning anxiety, and self-regulation skills in EMI students’ academic success Dogan Yuksel Open University, Milton Keynes, United Kingdom https://orcid.org/0000-0001-9131-3907 dogan.yuksel@open.ac.uk Adem Soruç University of Samsun, Turkey https://orcid.org/0000-0003-4165-6260 adem.soruc@samsun.edu.tr Barıs Horzum Sakarya University, Turkey https://orcid.org/0000-0003-3567-0779 mhorzum@sakarya.edu.tr Jim McKinley University College London, UK https://orcid.org/0000-0002-9949-8368 j.mckinley@ucl.ac.uk Abstract This study focuses on the predictive power of linguistic (i.e., general English proficiency; identified simply as “proficiency” in this paper) and non-linguistic (i.e., language learning anxiety and self-regulation) factors on the academic success of English medium instruction (EMI) students studying in engineering and social sciences programs in a Turkish university setting. Data were col- lected from 705 conveniently sampled EMI students of four academic subjects mailto:adem.soruc@samsun.edu.tr Dogan Yuksel, Adem Soruç, Barıs Horzum, Jim McKinley 400 (international relations; N = 158; business administration; N = 184; mecha- tronics engineering; N = 181; mechanical engineering: N = 182) representing two disciplines (i.e., social sciences and engineering) from a public university. Pearson correlation and SEM analyses were run to determine the relation- ships among language learning anxiety, self-regulation, proficiency and EMI success. Findings revealed that anxiety and self-regulation skills do affect EMI students’ proficiency irrespective of academic disciplines. Both self-regulation and proficiency impacted EMI students’ academic success in engineering, while only proficiency predicted academic success in the social sciences. These results are discussed and pedagogical implications are given related to the impact of linguistic and non-linguistic factors in EMI contexts. Keywords: English medium instruction; individual differences; general English proficiency academic success; discipline-based differences; structural equa- tion modelling (SEM) 1. Introduction Scholars concur that the number of English medium instruction (EMI) programs have increased exponentially in higher education (HE) institutions, notably since 2012 (Rose, McKinley, et al., 2020), because of globalization and internationaliza- tion (Galloway et al., 2020). EMI is defined as “the use of the English language to teach academic subjects (other than English itself) in countries or jurisdictions where the first language of the majority of the population is not English” (Macaro, 2018, p. 19). As a global educational phenomenon, EMI has been implemented in various contexts (McKinley & Galloway, 2022), including Turkey (Altay et al., 2022; Sahan et al., 2021; Soruç & Griffiths, 2018), where our study was conducted. Different strands of EMI research can be identified in the literature. While earlier EMI research centered on specific language policies and their implemen- tation in various contexts (e.g., Evans, 2000; Kırkgöz, 2009), more recent research has focused on beliefs, perceptions, and attitudes of students and teachers to- wards EMI (Macaro et al., 2018). Linguistic factors in EMI academic achievement, specifically the effect of language proficiency on academic success in EMI, has also been investigated (Curle et al., 2020; Rose, Curle, et al., 2020; Xie & Curle, 2022). In these studies, academic success entails the mastery of disciplinary knowledge and is usually measured by calculating the semester-long test scores of the students (Curle et al., 2020; Rose, Curle, et al., 2020). Although the relationship between individual differences (or non-linguistic factors) and academic success has been investigated with respect to, for example, self-efficacy and self-concept (Thompson et al., 2019), ideal L2 self (Rose, Curle, et al., 2020; Xie & Curle, 2022), mindsets (Kaya et al., 2021) Examining the role of English language proficiency, language learning anxiety, and self-regulation . . . 401 and motivation (Lasagabaster, 2016), there is an urgent need for research includ- ing other individual differences such as anxiety and self-regulation (Thompson et al., 2019). The current correlational design study, therefore, aimed to investigate the role of linguistic factors, that is, language proficiency, and that of non-lin- guistic factors, that is, anxiety and self-regulation, on EMI academic success by gathering data from both social sciences and engineering programs. 2. Literature review 2.1. English language proficiency and EMI academic success The interplay between English language proficiency and academic attainment in EMI has been examined in a number of studies (e.g., Curle et al., 2020; Rose, Curle, et al., 2020; Thompson et al., 2019; Xie & Curle, 2022). In their study conducted with Year 2 students (N = 139) in a Japanese context, Thompson et al. (2019) found language pro- ficiency (i.e., TOEIC scores) and preparatory performance (i.e., English for specific pur- poses, or ESP, course test scores) to be significant predictors of EMI success in an in- ternational business course. Similarly, when Rose, Curle, et al. (2020) investigated the relationship between TOEIC language proficiency scores and EMI international busi- ness course scores of Year 2 Japanese students (N = 146), they found a statistically significant relationship between language proficiency (i.e., TOEIC) and EMI success (i.e., content course scores). Another study on this relationship was conducted with Year 2 students (N = 106) in the Chinese higher education context by Xie and Curle (2022). They used their participants’ Business English proficiency scores and content scores in an International Business course and found a statistically significant relation- ship between Business English proficiency and content scores (i.e., academic success). When Curle et al. (2020) examined this relationship in Turkey, with Year 4 students (N = 212) majoring in economics, they found contradictory results when compared to studies in the Japanese and Chinese contexts. In their study, general English profi- ciency, operationalized as scores on the Cambridge Preliminary English Test, was not a statistically significant predictor of EMI academic success. Of these four studies, three have addressed the relationship between lan- guage proficiency and EMI academic success in international business courses in Asian contexts (Rose, Curle, et al., 2020; Thompson et al., 2019; Xie & Curle, 2022), while the remaining one (Curle et al., 2020) was conducted with economics stu- dents in the Turkish context. However, studies are still needed in programs other than international business courses or economics in EMI. Our study, therefore, extends the scope of the aforementioned research related to the role of language proficiency in EMI success in recruiting EMI engineering students, which is a common EMI discipline area (see Sandström & Neghina, 2017), as well as EMI Dogan Yuksel, Adem Soruç, Barıs Horzum, Jim McKinley 402 social sciences students, which is a more linguistically demanding discipline area (see Bolton & Kuteeva, 2012; Kuteeva & Airey, 2014). In our study, we maintain separate investigations into these different disciplines as the variation of EMI between them is worthy of investigation. More to the point, it is hypothesized that the comparison of multiple EMI programs from the same setting is most likely to yield more generalizable results in terms of the impact of language pro- ficiency on academic success. 2.2. Language learning anxiety Language learning anxiety refers to tendencies towards nervousness and hesita- tion when communicating or learning in second language contexts. At various lev- els, almost all language learners experience language-learning anxiety at some point in their language-learning process (Thompson & Lee, 2014) and there might be various anxiety levels for different skills (Cheng, 2017). Dörnyei and Ryan (2015) argue that without any doubt “anxiety affects L2 performance” (p. 176). Previous research that has investigated the interplay between language learning anxiety and various issues related to language attainment confirmed almost unequivocally that higher levels of language anxiety were closely connected with lower levels of language success or attainment (e.g., Zheng & Cheng, 2018). Hence, students’ self-regulation skills (see below) can be strengthened to moderate the negative effects of anxiety by helping them structure their learning environment (Zimmer- man, 2002) and transmit their efforts to self-direct the learning process (Wang & Zhan, 2020). Anxiety caused by studying in English has been reported in numerous studies, which have generally examined the perceptions and responses of EMI students (e.g., Chun et al., 2017). Also, several studies have investigated the pres- ence and/or types of speaking anxiety in EMI contexts (e.g., Suzuki et al., 2017). However, studies examining the relationship between language learning anxiety and academic success in EMI settings are rare (Macaro et al., 2018). 2.3. Self-regulation Self-regulation refers to the ability of students to control, direct, and modify their learning experiences to become active agents of their own learning process (Zim- merman, 2002). According to Iwaniec (2014), self-regulation allows students to become more proactive in their learning and scaffolds them to reach their self- established goals. Self-regulation can be observed through dedication, commit- ment, and resilience towards a specific field of study or topic (Henry, 2019). The concepts of autonomy and self-regulation have also been used interchangeably in the field of language learning (Griffiths & Soruç, 2020). Self-regulation is therefore Examining the role of English language proficiency, language learning anxiety, and self-regulation . . . 403 accepted as a multi-faceted construct which contains “cognitive, metacognitive, motivational, behavioral, and environmental processes that learners can apply to enhance academic achievement” (Dörnyei, 2005, p. 191). Previous research has provided strong evidence demonstrating the effects of the level of self-regulation on students’ learning outcomes (e.g., Bai & Wang, 2023; Zimmerman, 2008). In the field of language learning, for instance, self-reg- ulation is considered a key aspect that leads to improved language competence (Bai, 2018). However, few studies have explored self-regulation in EMI contexts. In a study conducted in Hong Kong, Hu and Gao (2018) examined the processes of the utilization of resources in high-achieving and underachieving secondary students’ self-regulated strategic writing in their EMI academic studies. Their find- ings indicated that high achievers used a wider range of learning strategies com- pared to their low-achieving counterparts. In a recent study in the Turkish context, Soruç et al. (2022) found that self-regulation was a significant contributor to EMI students’ academic success both in engineering and social sciences disciplines. In this study, the self-efficacy of the students also contributed to the academic suc- cess of the students. Recent research in non-EMI contexts has examined the rela- tionship between self-regulation and language learning anxiety (Guo et al., 2018; Wang & Zhan, 2020). To the best of our knowledge, no prior study has examined the co-impact of self-regulation and language learning anxiety in EMI contexts. 2.4. Non-linguistic factors as predictors of academic success in EMI There is a strand of studies that examines the relationship between non-linguistic factors including motivation and EMI academic success. In one of the first studies, Lasagabaster (2016) investigated the interplay among motivation, gender, L1 and possible selves in English medium instruction (except for success). His findings demonstrated that ideal L2 self and integrativeness were positively correlated, in- dicating that the two factors tapped into the same domain and that the students had a positive attitude toward the second language. According to the proposed model, ideal L2 self, attitudes towards EMI, family influence and instrumental pro- motion accounted for 47.6% of the variance of the criterion measures. In their study, Thompson et al. (2019) explored the interplay among language proficiency, self-ef- ficacy beliefs, self-concept, and EMI success, finding that L2 self-concept were not predictors of EMI academic attainment, while L2 self-efficacy was a significant pre- dictor. The inconsistency of the relationship between non-linguistic factors and EMI academic success led Thompson et al. (2019) to call for further research in different academic subjects, other than internationally mediated EMI business course, as well as developing multi-item measures and different models based on more ad- vanced statistical analyses such as structural equation modelling (SEM). Dogan Yuksel, Adem Soruç, Barıs Horzum, Jim McKinley 404 When Rose, Curle, et al. (2020) examined the relationship between moti- vation (i.e., ideal L2 self) and academic success, they found that motivation did not correlate with higher EMI success grades. Therefore, Rose, Curle, et al. (2020) argued that, because “EMI motivation warrants further investigation in future re- search” (p. 2158), alternative motivational constructs should be explored. Simi- larly, Xie and Curle (2022) investigated the interplay between motivation (i.e., ideal L2 self) and academic attainment. Their results demonstrated likewise that motivation did not predict EMI success. This result led Xie and Curle to “highlight the need for further research in this domain” (p. 594). Table 1 Overview of the studies on non-linguistic factors and success in EMI Studies on non- linguistic factors Set- ting EMI Course Year of study Number of participants Non-linguistic factors Key findings of the study 1 Lasagabaster, (2016) Spain Different EMI courses from four different faculties various years of study 189 Ten factors including ideal L2 self Ideal LE self and students’ attitudes towards EMI signifi- cantly predicted their intended learning effort. 2 Thompson et al. (2019) Japan International business Year 2 139 Self-concept and self-effi- cacy L2 self-concept did not predict EMI ac- ademic success, while L2 self-effi- cacy was a signifi- cant predictor. 3 Rose, Curle, et al. (2020) Japan International business Year 2 146 Motivation (i.e., ideal L2 self) Motivation did not correlate with higher EMI success grades. 4 Xie & Curle (2022) China International business Year 2 106 Motivation (i.e., ideal L2 self) Motivation did not correlate with higher EMI success grades. We argue that more specific research is needed in various EMI contexts or EMI academic disciplines because the studies in Table 1 are limited to one EMI subject: English-mediated international business. We still do not know, for instance, whether the same findings would be obtained if students from an- other EMI subject were surveyed, in particular such where English does not play an international role (Thompson et al., 2019). Besides, three studies out of four were conducted in the East Asian contexts of Japan and China, potentially limit- ing transferability to other EMI contexts. Based on the review of the literature above, the following specific gaps have been identified: • Studies investigating the interplay between English language proficiency and EMI academic attainment did not yield conclusive results. Examining the role of English language proficiency, language learning anxiety, and self-regulation . . . 405 • The relationship between some individual differences (i.e., self-regula- tion and anxiety) has not yet been explored in EMI settings. • The current research on academic success in EMI has generally sampled international business students, ignoring other disciplines. • The relationship among linguistic and non-linguistic factors and EMI ac- ademic attainment has not been examined considering the EMI aca- demic division or from the perspective of potential discipline-based dif- ferences (i.e., social sciences and engineering). • Multi-item scales and different models of analysis, such as structural equation modelling (SEM), have not been used to explore the relation- ship between linguistic and non-linguistic factors and EMI success. 2.5. Proposing a hypothesized structural model In this study, we hypothesized that language learning anxiety, self-regulation, and language proficiency would directly or indirectly predict EMI success. We also proposed that these three variables would have a direct or indirect impact on each other. Previous research in the field of second and foreign language learning settings identified a negative relationship between language learning anxiety and second language achievement (e.g., Oteir & Al-Otaibi, 2019; Tei- mouri et al., 2019). More specifically, a review of around 60 years of research in the field of language learning anxiety showed that anxiety was one of the major causes of lower language proficiency scores (Oteir & Al-Otaibi, 2019). In the field of second and foreign language learning, studies have revealed the positive im- pact of self-regulation skills on improving language proficiency (e.g., Andrade & Evans, 2013). Previous studies confirmed that self-regulation skills contribute positively to success in language learning (Oxford, 2011). In the field of EMI, a recent study by Soruç et al. (2022) confirmed the relationship between self-reg- ulation and EMI success. The interplay between language proficiency and EMI academic attainment has been proposed in recent research (e.g., Altay et al., 2022; Soruç et al., 2022), although there are mixed results regarding the effects of language proficiency and EMI academic attainment (e.g., Curle et al., 2020) and implications for discipline- based differences (e.g., Soruç et al., 2022). Our study aims to examine the rela- tionship among these four constructs utilizing SEM (see Figure 1). Dogan Yuksel, Adem Soruç, Barıs Horzum, Jim McKinley 406 Language learning anxiety Language proficiency EMI successSelf-regulation Figure 1 The rationale underlying the structural model hypothesized in this study Based on our hypothesized structural model and responding to calls in previous studies (e.g., Thompson et al., 2019; Xie & Curle, 2022), the present study aimed to address the following research questions: RQ1a: To what extent do language learning anxiety, self-regulation, and language proficiency predict the success of EMI students in engineering programs? RQ1b: To what extent do language learning anxiety, self-regulation, and lan- guage proficiency predict the success of EMI students in social sciences? RQ2: To what extent do the SEM models produced for EMI programs from engi- neering and social sciences programs explain the relationship among learn- ing anxiety, self-regulation, language proficiency and academic success? 3. Methodology 3.1. Research strategy Correlational research design, a quantitative research method, was used in this study. Correlational studies explore the interplay among various variables measured at once without any intervention. The current study adopted a correlational research design because the interplay among university students’ language learning anxiety, self-regulation skills, language proficiency and success in EMI courses, operational- ized as grade point average (GPA) scores, were measured at one time. 3.2. Context of the study Turkish higher education institutions offer two types of EMI programs: partial and full (Curle et al., 2020). The data for this study came from a partial EMI program, which is also described as a “multilingual model” (Macaro, 2018), where students are required to register for a minimum of two courses taught in English per se- mester. Data were collected from four partial EMI programs (i.e., two programs Examining the role of English language proficiency, language learning anxiety, and self-regulation . . . 407 from social sciences, two from engineering) of a major public university in Turkey with a population of more than 50,000 students, enabling the researchers to sample a higher number of EMI students from various EMI programs to examine any discipline-specific differences in terms of the relationship among linguistic and non-linguistic factors and EMI success. Selecting four EMI programs from the same setting further helped us to tackle potential subject-related confounding variables (e.g., Margić & Vodopija-Krstanović, 2018). 3.3. Participants Data were collected from students who were conveniently sampled from EMI academic programs to represent two disciplines from a public university in the western part of Turkey (see Table 2 for details). Table 2 Distribution of the participants according to the academic programs Academic programs F (%) M (%) Total (%) Age (M) (SD) Overall ratio of the whole sample So ci a l sc ie n ce s International relations 77 (49) 81 (51) 158 (100) 21.5 (3.1) 22.4% Business administration 86 (47) 98 (53) 184 (100) 23.8 (4.7) 26.1% E n gi n e e ri n g Mechatronics engineering 52 (29) 129 (71) 181 (100) 24.3 (4.1) 25.7% Mechanical engineering 49 (27) 133 (73) 182 (100) 25.1 (5.2) 25.8% In total, there were 705 participants from four EMI programs: 215 of the participants were second-year students (30.5%), 194 were from the third year (27.5%), and 296 were fourth-year students (42.0%). First-year students were not included because they did not have EMI success scores. A large proportion of participants were intentionally recruited from the four academic programs to run more reliable and valid analyses; for instance, structural equation modelling requires 150 or more participants as the ideal sampling size (Kline, 2011). Fur- ther details about the selection of the participants are the following: • All participants had completed at least two years of study (data collection took place at the end of the academic year) in one of the four EMI programs. • Turkish was the main language of all participants and they used English as a foreign language only. They all had comparable formal English lan- guage learning experiences within the Turkish education system includ- ing the higher education institution. • No participant reported any high school EMI experience. Dogan Yuksel, Adem Soruç, Barıs Horzum, Jim McKinley 408 • All participants had taken the same assessments and exam procedures in their courses mitigating any testing effect on the scores used in this analysis. • All participants had completed a minimum of eight EMI courses and 20 Turkish-medium instruction courses during their studies in their programs. 3.4. Data collection instruments and procedure With the ethical clearance obtained, students were invited to participate in the study at the end of the academic year. The data related to the students’ level of anxiety and self-regulation were collected via an online five-point Likert scale ques- tionnaire (Iwaniec, 2014), while EMI course scores (GPAs) and language proficiency scores were obtained from the University Registrar’s office. The whole process of data collection was conducted anonymously, and any personal identification infor- mation about participants was removed after determining the demographics. The following measures were used for language proficiency, success in EMI courses, anx- iety, and self-regulation. 3.4.1. General English proficiency (“proficiency”) To determine participants’ proficiency levels, a version of the Cambridge Prelim- inary English Test (PET) at a B1 difficulty level (Cambridge ESOL, 2014) was used. This proficiency test included questions on the four language skills: reading, writing, listening, and speaking. Since the university provides partial or full EMI programs, students are required to take this language test to measure their pro- ficiency and when necessary to place the students in the appropriate level of language support classes. Even though the students took the tests in different years before starting their EMI programs, a very similar version of the Cambridge PET test was administered each year by the university administration. 3.4.2. EMI success To measure EMI academic success, the sum of the final course scores for all con- tent courses taken in English was divided by the number of English courses each student took. A minimum of 8 courses was determined as a unit threshold to be included in this study so that we could have robust evidence regarding our par- ticipants’ EMI academic attainment. The fact that Year 1 students were not sam- pled for the present study can be considered from this point of view since an inadequate number of EMI courses was most likely to reduce the generalizability of the findings. Following earlier EMI success research (e.g., Rose, Curle, et al., Examining the role of English language proficiency, language learning anxiety, and self-regulation . . . 409 2020; Thompson et al., 2019; Xie & Curle, 2022), we took the grade point average (GPA) scores of our participants as our measure of academic success. 3.4.3. Language Learning Anxiety Scale (LLAS) A 5-point Likert scale was used to determine the extent to which the participants felt anxious when studying in English-medium instruction contexts (see Appendix A). We used the anxiety questionnaire developed by Iwaniec (2014), who modelled the items on the scale using Horwitz, Horwitz, and Cope’s (1986) Foreign Language Classroom Anxiety Scale. The original scale included eight items, but this was re- duced to five after the factor analyses (see details in the section on data analysis). 3.4.4. Self-Regulation Scale (SRS) To collect the data about students’ level of self-regulation in the EMI context, we used the scale with 12 items developed by Iwaniec (2014) (see Appendix B); how- ever, after the factor analysis, just eight items were used in the analysis (see de- tails in the section on data analysis). 3.5. Data analysis Pearson correlation and SEM analyses were utilized to ascertain the relation- ships among language learning anxiety, self-regulation, language proficiency and EMI success. Maximum likelihood estimations were used in SEM. In the analyses, descriptive analysis and Pearson correlation coefficient analysis were performed using SPSS 21.0 and SEM was conducted using LISREL 8.54 programs. 3.5.1. Language Learning Anxiety Scale (LLAS) A series of analyses were conducted to examine the validity and reliability of the original scale in the Turkish context. Firstly, all the assumptions for the factor analyses were checked, confirming the factorability of the correlation matrix. Then, an exploratory factor analysis (EFA) was conducted with the data from business administration students (N = 184) to explore the factor structure of the data. According to the analysis, items 1, 4, and 5 were removed, so the scale was formed by five items with one factor. A confirmatory factor analysis (CFA) was run, using the data from mechanical engineering students (N = 182) to test if the factor structure fits the data. The analysis showed that five items with one factor all pertained to the LLAS, measuring the same construct (i.e., anxiety) accurately and consistently. After this, Cronbach’s alpha for internal reliability was examined Dogan Yuksel, Adem Soruç, Barıs Horzum, Jim McKinley 410 and the test returned an excellent value (r = .91). After all these analyses, the scale used in this study included five items (with one factor), measuring the level of anxiety of the students in the EMI context. According to the results of CFA and the values of Cronbach’s alpha, LLAS is a valid and reliable instrument. De- tailed results are presented in Appendix C. 3.5.2. Self-Regulation Scale (SRS) To examine the validity and reliability of the SRS in the Turkish context, as in the anxiety scale, we carried out several analyses. After confirming the factorability of the correlation matrix, EFA was, firstly, run using the data of the participants from international relations (N = 158), according to which the items 2, 3, 9, and 10 were removed from the scale, and remaining was an eight-item scale with one factor (i.e., self-regulation). Immediately after this, CFA was run, using the data from mechatronics engineering students (N = 181). The analysis revealed that the eight items in the scale comprised a unified construct (i.e., self-regulation). Finally, when Cronbach’s alpha for internal reliability was measured, the test returned an excellent value (r = .96). According to these analyses, we used an eight-item SRS with one factor. According to CFA and the values of Cronbach’s alpha, SRS is a valid and reliable instrument. Detailed results are presented in Appendix A. 4. Findings The data were analyzed separately for the engineering and the social sciences (RQ1a and RQ1b, respectively). 4.1. RQ1a: To what extent do language learning anxiety, self-regulation, and language proficiency predict the success of EMI students in engineering programs? The data from 363 students (mechatronics; N = 181; mechanical engineering; N = 182) were analyzed, producing a model based on the variables proficiency (i.e., general English proficiency), language learning anxiety, self-regulation, and overall EMI course scores. Standard deviations and correlation coefficients are presented in Table 3. When the data were closely examined, all the means proved to be above average. When the relationship between the variables was analyzed, language learning anxiety was found to be negatively correlated with self-regulation (r = -.41). Also, language learning anxiety correlated negatively with proficiency (r = -.32) and EMI success (r = -.11). Furthermore, self-regula- tion had a positive correlation with proficiency (r = .48) and EMI success (r = .24). Examining the role of English language proficiency, language learning anxiety, and self-regulation . . . 411 Additionally, language proficiency was found to be positively correlated with the EMI success (r = .31). Table 3 Mean, standard deviation and correlations among the variables Variables LLA SR Proficiency EMI success LLA - SR -.41** - Proficiency -.32** .48** - EMI success -.11* .24** .31** - Mean 16.74 30.29 73.93 60 95 Standard deviation 4.62 6.36 7.35 14.22 Note. **p < .01; *p < .05; LLA = language learning anxiety; SR = self-regulation; proficiency = general English proficiency; EMI success = GPA To examine these relationships in more detail, in the second stage, SEM analysis was carried out. In this model, we found that language learning anxiety and self-regulation were predictors of proficiency, while self-regulation and lan- guage proficiency significantly predicted EMI academic attainment. The fit crite- ria used in the methodological literature (e.g., Kline, 2011) were followed when reporting the findings of the present study. According to the model, fit indices were χ2/df = 0.15, RMSEA = 0.001, SRMR = 0.054, AGFI = 0.99, GFI = 1.00, NFI = 1.00, CFI = 1.00. When these fit values obtained for the model were compared with the critical values determined for SEM by Schermelleh-Engel et al. (2003), the fit indices were in the range determined for a good fit, thus leading us to maintain that it was the tested structural model which provided the model-data fit. The path diagram containing the standardized path coefficients for the model is provided in Figure 2. Language learning anxiety Language proficiency EMI success -0.16 Self-regulation 0.39 0.76 0.89 1.00 0.26 1.00 0.11 Figure 2 Standardized path coefficients calculated for the structural model When the path diagram was examined, the path coefficients of some var- iables were low, and some were medium. In this respect, the standardized path coefficients obtained for the structural model and the t values and significance levels for these path coefficients are presented in Table 4. Dogan Yuksel, Adem Soruç, Barıs Horzum, Jim McKinley 412 Table 4 Standardized path coefficients, t values and significance levels regarding EMI courses GPA model Direct effect Standardized β value t LLA→ proficiency -0.16* -3.20 SR → proficiency 0.39* 7.59 SR → EMI success 0.11* 2.02 proficiency → EMI success 0.26* 4.58 Note. * p < .05 LLA = Language learning anxiety; SR = Self-regulation; proficiency = general English proficiency; EMI success = GPA When the data in Table 4 were closely examined, language learning anxiety (β = -0.16, t = -3.20) and self-regulation (β = 0.39, t = 7.59) were significant predic- tors of proficiency (p < .05). Furthermore, self-regulation (β = 0.11, t = 2.02) and proficiency (β = 0.26, t = 4.58) were significant predictors of EMI success (p < .05). These data showed a negative relationship between language learning anxiety and proficiency; therefore, when students in EMI contexts had low anx- iety levels, they developed higher language proficiency. Also, the self-regulation skills of the students were significant predictors of language proficiency, mean- ing that when students had a higher sense of self-regulation, they improved their proficiency. Both language learning anxiety and self-regulation skills ac- counted for 24% of proficiency. Furthermore, in this EMI context, the self-regulation skills of the students were found to be significant predictors of EMI success, indicating that as stu- dents develop improved self-regulation skills, they become more successful in their EMI courses. Lastly, proficiency was also found to predict the students’ EMI success significantly, which has revealed that as students become more profi- cient in English, they become more successful in EMI courses. The students’ skills of self-regulation and their proficiency explained 11% of their EMI success, while the model also demonstrated that language learning anxiety impacted EMI success indirectly, explaining 5.6% of success in EMI. 4.2. RQ1b: To what extent do language learning anxiety, self-regulation, and language proficiency predict the success of EMI students in social sciences? Following the same procedures as those outlined above, the data from 342 social sciences EMI students (international relations; N = 158; business administration; N = 184) were analyzed and then a model was produced based on the same vari- ables as the engineering students. Standard deviations and correlation coefficient values are given in Table 5. Overall, the data showed that the variables had scores above the average, and we found that language learning anxiety was neg- atively correlated with self-regulation (r = -.38). Language learning anxiety was Examining the role of English language proficiency, language learning anxiety, and self-regulation . . . 413 likewise correlated negatively with proficiency (r = -.34) and EMI success (r = -.12). Conversely, self-regulation was correlated positively with proficiency (r = .50) and EMI success (r = .21). Additionally, proficiency was positively corre- lated with the EMI success (r = .33). Table 5 Means, standard deviations and correlations among the variables Variables LLA SR Proficiency EMI success LLA - SR -.38** - Proficiency -.34** .50** - EMI success -.12* .21** .33** - M 17.89 28.52 73.29 66.39 SD 3.40 6.23 7.10 11.76 Note. ** p < .01; * p < .05; LLA= Language learning anxiety; SR= Self-regulation; Proficiency = general English proficiency; EMI success = GPA In the second stage, the SEM analysis yielded the same results as in the case of the engineering students. The same fit criteria used for the engineering student data were followed when reporting the findings of the social sciences student data. In this model, fit indices were χ2/df = 0.25, RMSEA = 0.001, SRMR = 0.093, AGFI = 1.00, GFI = 1.00, NFI = 1.00, CFI = 1.00. When these fit values obtained for the model were compared with the critical values determined for SEM, the fit in- dices were in the range determined for a good fit, indicating that the structural model tested here provided the model-data fit. The path diagram containing the standardized path coefficients for the model is presented in Figure 3. Language learning anxiety Language proficiency EMI success -0.16 Self-regulation 0.42 0.75 0.89 1.00 0.33 1.00 Figure 3 Standardized path coefficients calculated for the structural model Analysis of the path diagram for the students of the social sciences, like the engineering students, showed that the path coefficients of some variables were low, and some were medium. In this respect, standardized path coeffi- cients obtained for the structural model and the t values and significance levels for these path coefficients are presented in Table 6. Dogan Yuksel, Adem Soruç, Barıs Horzum, Jim McKinley 414 Table 6 Standardized path coefficients, t values and significance levels regarding EMI courses GPA model Direct effect Standardized β value t LLA→ proficiency -0.16* -3.33 SR → proficiency 0.42* 8.56 proficiency → EMI success 0.33* 6.39 Note. * Significant at level p <.05; LLA = language learning anxiety; SR = self-regulation; proficiency = general English proficiency; EMI success = GPA According to the data in Table 6, language learning anxiety (β = 0.16, t = - 3.33) and self-regulation (β = 0.42, t = 8.56) were significant predictors of profi- ciency (p < .05). However, language proficiency (β = 0.33, t = 6.39) was the only significant predictor of EMI success (p < .05). Such results indicated that when social sciences EMI students feel less anxious (as engineering EMI students do), they develop higher language proficiency in EMI courses. Also, the data showed that self-regulation was a significant predictor of proficiency. Similar to engi- neering EMI students, students studying in social sciences EMI programs with higher self-regulation skills had improved language proficiency. Both anxiety and self-regulation account for 25% of improvement in proficiency. Conversely, the non-linguistic factors of anxiety and self-regulation did not predict success in EMI programs in social sciences, whereas proficiency did. The model showed that as the students of the social sciences increased their lan- guage proficiency, they became more successful in EMI courses like interna- tional relations and business administration. Language proficiency explained 11% of EMI success. Although the impact was not found directly, anxiety and self- regulation both accounted for the EMI success, only 2.7%. 4.3. RQ2: To what extent do the SEM models produced for EMI programs from engineering and social sciences programs explain the interplay between learning anxiety, self-regulation, language proficiency and academic success? The data showed that the language learning anxiety levels of engineering students were lower than the students from social sciences, while their self-regulation skills were higher. Related to their EMI success, however, the students in social sciences had higher GPA scores, but they were more anxious and their self-regulation skills were much lower. The fact that these two non-linguistic variables influenced each other negatively should be closely examined in future studies. When the two SEM models were compared using Akaike’s Information Cri- teria (AIC), Consistent AIC (CAIC) and Bayes Information Criteria (BIC), we found that the value of AIC for the model of Engineering programs was 18.15, while the value of CAIC was 62.20. The value of AIC for the programs in social sciences was Examining the role of English language proficiency, language learning anxiety, and self-regulation . . . 415 16.49, while CAIC was 55.17. Given that the better model was the one having the lower index scores, we can suggest that the SEM model produced for social sci- ence programs (i.e., international relations and business administration) showed more effectiveness than the other model produced for engineering programs. The main reason for this can be because the self-regulation skills of the engineering students predicted EMI success and that the model was more complicated. When the two SEM models are compared, it can be maintained that EMI students’ individual differences like anxiety and self-regulation predicted profi- ciency significantly, which explained one-fourth of the improvement in proficiency in both engineering and social sciences programs. Regarding academic success in EMI in both disciplines, proficiency was likewise found to be a significant predictor of EMI success. The only difference observed in both SEM models was about the self-regulation skills of the students. While this non-linguistic factor, or individual difference, was a significant contributor to the success of engineering EMI pro- grams, it did not directly predict the success of students studying in the social sci- ences. Therefore, we suggest that not only proficiency but also the self-regulation skills of engineering students should be supported. This is not the same case in social sciences, because it was not self-regulation, but proficiency which contrib- uted directly and significantly to the increase in students’ success. 5. Discussion Utilizing SEM, this study is among the first few studies to examine the relation- ship between linguistic (i.e., general English proficiency) and non-linguistic (i.e., language learning anxiety and self-regulation) factors and EMI students’ aca- demic success comparing engineering and social sciences students. The first re- search question asked whether EMI students’ general English proficiency, sense of anxiety and self-regulation skills predicted their success in engineering EMI courses (RQ1a) and social science EMI courses (RQ1b). Our findings revealed that in both discipline areas, students’ individual differences such as their level of language learning anxiety and sense of self-regulation skills predicted their level of language proficiency. For engineering students, language learning anxi- ety and self-regulation skills accounted for 24% of proficiency. Moreover, self- regulation and proficiency predicted EMI students’ academic success explaining 11% of their EMI success. This finding is important for EMI practitioners or lec- turers when teaching engineering students through EMI because, according to this finding, when these students are trained to be more self-regulatory and when they are more proficient in the language, it is more likely they will achieve success. Also, learning anxiety explained 5.6% of EMI success via proficiency in- directly. However, there was a negative relationship between language learning Dogan Yuksel, Adem Soruç, Barıs Horzum, Jim McKinley 416 anxiety and proficiency, indicating that when the students in engineering EMI programs had low anxiety levels, they developed higher language proficiency. This finding is vital to the EMI literature because no prior study has examined the predictive power of language proficiency, language learning anxiety and self- regulation skills on engineering students’ academic success. In the social sciences division, our results similarly demonstrated that lan- guage learning anxiety and self-regulation were significant predictors of students’ proficiency. Anxiety and self-regulation accounted for 25% of improvement in pro- ficiency. However, only proficiency significantly predicted EMI success accounting for 11% of students’ level of success, while non-linguistic factors such as anxiety and self-regulation did not predict success in social sciences EMI programs. Alt- hough this finding suggests that anxiety and self-regulation skills are less important for social sciences students when compared to language proficiency level (see Kuteeva & Airey, 2014), we note that these non-linguistic factors still have an in- direct effect. Regarding the relationship between language proficiency and EMI academic success, Thompson et al. (2019), Rose, Curle, et al. (2020), as well as Xie and Curle (2022) found language proficiency as a significant predictor of EMI suc- cess in social sciences programs. Our results support the findings of these studies, as we also found that English language proficiency, among other factors, predicted EMI students’ academic success in the programs of social science. Given the fact that our participants were sampled from both international relations and business administration programs, including a higher number of students, these findings are transferable to other similar EMI settings. In non-EMI contexts, previous studies reported a strong link between stu- dents’ levels of self-regulation and their learning outcomes (e.g., Bai & Wang, 2023; Zimmerman, 2008). Similarly, for engineering students, a statistically sig- nificant relationship was found between self-regulation skills and academic suc- cess. However, we also found that self-regulation skills did not predict academic success in social sciences programs. These mixed findings should provide an im- pulse for further research because they emphasize the significance of discipline- based differences in EMI settings, as could be observed in some other studies (e.g., Altay et al., 2022). While few studies have explored the relationship between language learn- ing anxiety and academic success in EMI settings, in non-EMI settings, previous research demonstrated that higher levels of language anxiety were closely con- nected with lower levels of language success (e.g., Zheng & Cheng, 2018). Similarly, in our study, for both discipline areas, language anxiety exhibited significant cor- relations with English language proficiency, and in the engineering programs, it indirectly predicted EMI academic success via proficiency accounting for 5.6% Examining the role of English language proficiency, language learning anxiety, and self-regulation . . . 417 of EMI success. For social sciences students, conversely, language learning anxi- ety did not predict EMI academic success. The second research question concerned the extent to which the SEM models produced for the EMI programs from engineering and social sciences explain the link between anxiety, self-regulation, proficiency and academic suc- cess. When comparing the SEM models produced for the two groups, we found that language learning anxiety levels of engineering students were lower than those of students from social sciences, while their self-regulation skills were higher. Lower language learning anxiety may be related to the varying role of the English language in engineering programs. Dearden and Macaro (2016) pro- vided evidence regarding the varying role of language in EMI contexts, suggest- ing that EMI students as well as lecturers deal more with formulae than words. Similarly, Wellington and Osbourne (2001) claimed that the intensity of the lan- guage used in engineering programs might be low because students mostly de- pend on a combination of words, pictures, diagrams, images, animations, graphs, equations, tables and charts in these programs. However, in social sciences, stu- dents rely mostly on their English proficiency because the courses they take en- courage them to use language more flexibly and creatively (Kuteeva & Airey, 2014). Dearden and Macaro (2016) also found that social science lecturers stated that they focus more on language issues because, according to the lec- turers, teaching social sciences subjects requires students to be engaged in more communicative and small group activities, thus leading to more frequent dependence on language (Bolton & Kuteeva, 2012). These differences concern- ing language in different disciplines may help explain why social sciences stu- dents manifest higher levels of language learning anxiety. 6. Limitations Although our study offers important contributions to the field of EMI, it also has some limitations to consider when interpreting the results and transferring the findings to other contexts. First, the findings of this study came from a partial (i.e., multilingual) EMI context in Turkey. While this is generalizable to other par- tial EMI contexts, studies could be carried out in full (or fuller) EMI contexts to increase the generalizability of the conclusions more broadly. Moreover, the par- ticipants and the academic divisions were sampled conveniently. Random selec- tion of the participants could extend the validity of the study and it could pro- duce different results. Also, in this study, we examined only two non-linguistic factors, namely, anxiety and self-regulation. Therefore, future studies could inves- tigate other individual variables including instrumental and intrinsic motivation, self-efficacy, motivated behaviors, boredom, and so on. Due to the intensity of the Dogan Yuksel, Adem Soruç, Barıs Horzum, Jim McKinley 418 analysis, we reported only quantitative results without interviewing the partici- pants. Future studies should also use qualitative data to support quantitative data. Finally, since the data were collected from 2nd, 3rd and 4th-year students, differ- ent anxiety or language proficiency may have occurred at different learning levels. Future research can be carried out with a single education level. 7. Implications for the teaching/learning The main findings of our study revealed that anxiety and self-regulation skills do affect EMI students’ English language proficiency in both academic disciplines. Moreover, we observed that both self-regulation and language proficiency im- pacted EMI students’ academic success in engineering programs, whereas only English language proficiency impacted academic success in the social sciences discipline area. These two main findings indicate critical implications. • First, English language proficiency contributes to the student’s academic success in both academic areas (see also Yuksel et al., 2021); therefore, uninterrupted language support should be given to EMI students throughout their studies in order to boost their academic success. This language support can be in the form of in-sessional courses or English programs such as English for specific purposes and/or vocational English. • Second, strategy instruction should be given to the learners related to their challenges (e.g., Soruç et al., 2018, 2021) considering their specific language proficiency level and the academic discipline. However, we note that continuous language instruction or systematic strategy train- ing is not the only factor that paves the better road for students’ aca- demic success; students’ sense of anxiety should also be reduced, and self-regulation skills should be equally supported. • Finally, the self-regulation skills of engineering EMI students should be closely backed up through specific instructional, awareness-raising ac- tivities because it should be remembered that their self-regulation skills do affect their academic success as much as their language proficiency, while this may not be so significant in social sciences. 8. Conclusions This study is the first to find that while both self-regulation and general English lan- guage proficiency were significant predictors of academic success in engineering programs, it was only language proficiency in social sciences programs found as the predictor of academic success. These findings have pedagogical implications Examining the role of English language proficiency, language learning anxiety, and self-regulation . . . 419 for EMI practitioners, lecturers, and higher education institutions because it ex- plicitly highlights the importance of students’ individual characteristics such as the role of self-regulation skills on EMI students’ academic success, especially in engineering programs as well as the potential but unnoticed value of discipline- based differences in EMI settings. More meaningful and direct pedagogical in- terventions should therefore be implemented because EMI students in engi- neering programs might display different attitudes and have different needs from those studying in social sciences programs. Dogan Yuksel, Adem Soruç, Barıs Horzum, Jim McKinley 420 References Altay, M., Curle, S., Yuksel, D., & Soruç, A. (2022). Investigating academic achieve- ment of English medium instruction courses in Turkey. Studies in Second Language Learning and Teaching, 12(1), 117-141. Andrade, M. S., & Evans, N. W. (2013). Principles and practices for response in sec- ond language writing: Developing self-regulated learners. Taylor and Francis. Bai, B. (2018). Understanding primary school students’ use of self-regulated writ- ing strategies through think-aloud protocols. System, 78, 15-26. Bai, B., & Wang, J. (2023). The role of growth mindset, self-efficacy and intrinsic value in self-regulated learning and English language learning achieve- ments. Language Teaching Research, 27(1), 207-228. Bolton, K., & Kuteeva, M. (2012). English as an academic language at a Swedish university: Parallel language use and the “threat” of English. Journal of Multilingual and Multicultural Development, 33(5), 429-447. Cambridge, ESOL. (2014). Cambridge proficiency English test. Cambridge University Press. Cheng, Y. S. (2017). Development and preliminary validation of four brief measures of L2 language-skill-specific anxiety. System, 68, 15-25. Chun, S., Kim, H., Park, C. K., McDonald, K., Sun Ha, O., Kim, D. L., & Lee, S. M. (2017). South Korean students’ responses to English-medium instruction courses. Social Behavior and Personality, 45(6), 951-965. Curle, S., Yuksel, D., Soruç, A., & Altay, M. (2020). Predictors of English medium instruction academic success: English proficiency versus first language me- dium. System, 95, 102378. Dearden, J., & Macaro, E. (2016). Higher education teachers’ attitudes towards English medium instruction: A three-country comparison. Studies in Sec- ond Language Learning and Teaching, 6(3), 455-486. Dörnyei, Z. (2005). The psychology of the language learner: Individual factors in second language acquisition. Lawrence Erlbaum. Dörnyei, Z., & Ryan, S. (2015). The psychology of the language learner revisited. Routledge. Evans, S. (2000). Classroom language use in Hong Kong’s English-medium sec- ondary schools. Educational Research Journal, 15(1), 19-43. Galloway, N., Numajiri, T., & Rees, N. (2020). The “internationalization,” or “Englishiza- tion,” of higher education in East Asia. Higher Education, 80(3), 395-414. Griffiths, C., & Soruç, A. (2020). Individual differences in language learning. Pal- grave Macmillan. Guo, Y., Xu, J., & Liu, X. (2018). English language learners’ use of self-regulatory strategies for foreign language anxiety in China. System, 76, 49-61. Henry, A. (2019). Directed motivational currents: Extending the theory of L2 vision. In M. Lamb, K. Csizér, A. Henry, & S. Ryan (Eds.), The Palgrave handbook of motivation for language learning (pp. 139-161). Palgrave Macmillan. Examining the role of English language proficiency, language learning anxiety, and self-regulation . . . 421 Horwitz, E. K., Horwitz, M. B., & Cope, J. (1986). Foreign language classroom anxiety. Modern Language Journal, 70(2), 125-132. Hu, J., & Gao, X. (2018). Self-regulated strategic writing for academic studies in an English-medium-instruction context. Language and Education, 32(1), 1-20. Iwaniec, J. (2014). Motivation of pupils from southern Poland to learn English. System, 45, 67-78. Kaya, S., Yuksel, D., & Curle, S. (2021). The effects of language learning and maths mindsets on academic success in an engineering program in Turkey. Jour- nal of Engineering Education, 3(1), 50-62. Kırkgöz, Y. (2009). Students’ and lecturers’ perceptions of the effectiveness of foreign language instruction in an English-medium university in Turkey. Teaching in Higher Education, 14(1), 81-93. Kline, R. B. (2011). Principles and practice of structural equation modeling. Guil- ford Press. Kuteeva, M., & Airey, J. (2014). Disciplinary differences in the use of English in higher education: Reflections on recent language policy developments. Higher Education, 67(5), 533-549. Lasagabaster, D. (2016). The relationship between motivation, gender, L1 and possible selves in English-medium instruction. International Journal of Multilingualism, 13(3), 315-332. Macaro, E. (2018). English medium instruction. Oxford University Press. Macaro, E., Curle, S., Pun, J., An, J., & Dearden, J. (2018). A systematic review of English medium instruction in higher education. Language Teaching, 51(1), 36-76. Margić, B. D., & Vodopija-Krstanović, I. (2018). Language development for Eng- lish-medium instruction: Teachers’ perceptions, reflections, and learning. Journal of English for Academic Purposes, 35, 31-41. McKinley, J. & Galloway, N. (Eds.) (2022). English medium instruction practices in higher education: International perspectives. Bloomsbury. Oxford, R. L. (2011). Teaching and researching language learning strategies. Pear- son Longman. Oteir, N. I., & Al-Otaibi, N. A. (2019). Foreign language anxiety: A systematic re- view. Arab World English Journal, 10(3), 309-317. Rose, H., McKinley, J., Xu, X., & Zhou, S. (2020). Investigating policy and imple- mentation of English medium instruction in higher education institutions in China. London: British Council. Retrieved June 11, 2021. Rose, H., Curle, S., Aizawa, I., & Thompson, G. (2020). What drives success in English medium taught courses? The interplay between language proficiency, aca- demic skills, and motivation. Studies in Higher Education, 45(11), 2149-2161. Sahan, K., Rose, H., & Macaro, E. (2021). Models of EMI pedagogies: At the in- terface of language use and interaction. System, 101, 102616. Dogan Yuksel, Adem Soruç, Barıs Horzum, Jim McKinley 422 Sandström, A., & Neghina, C. (2017). English-taught bachelor’s programs: Inter- nationalizing European higher education. European Association for Inter- national Education. Schermelleh-Engel, K., Moosbrugger, H., & Müller, H. (2003). Evaluating the fit of structural equation models: Tests of significance and descriptive goodness- of-fit measures. Methods of Psychological Research Online, 8(2), 23-74. Soruç, A., & Griffiths, C. (2018). English as a medium of instruction: Students’ strategies. ELT Journal, 72(1), 38-48. Soruç, A., Dinler, A., & Griffiths, C. (2018). Listening comprehension strategies of EMI students in Turkey. In Y. Kırkgoz & K. Dikilitas (Eds.), Key issues in Eng- lish for specific purposes in higher education (pp. 265-287). Springer. Soruç, A., Altay, M., Curle, S., & Yuksel, D. (2021). Students’ academic language- related challenges in English Medium Instruction: The role of English pro- ficiency and language gain. System, 103, 102651. Soruç, A., Pawlak, M., Yuksel, D., & Horzum, B. (2022). Investigating the impact of linguistic and non-linguistic factors on EMI academic success. System, 107, 102794. Suzuki, S., Harada, T., Eguchi, M., Kudo, S., & Moriya, R. (2017). Investigating the relationship between students’ attitudes toward English-medium instruction and L2 Speaking. Essays on English Language and Literature, 46(1), 25-41. Thompson, A. S., & Lee, J. (2014). The impact of experience abroad and language pro- ficiency on language learning anxiety. TESOL Quarterly, 48(2), 252-274. Thompson, G., Aizawa, I., Curle, S., & Rose, H. (2019). Exploring the role of self- efficacy beliefs and learner success in English medium instruction. Inter- national Journal of Bilingual Education and Bilingualism. Wang, W., & Zhan, J. (2020). The relationship between English language learner characteristics and online self-regulation: A structural equation modeling approach. Sustainability, 12(7), 3009. Wellington, J., & Osborne, J. (2001). Language and literacy in science education. Open University Press. Xie, W., & Curle, S. (2022). Success in English medium instruction in China: Sig- nificant indicators and implications. International Journal of Bilingual Ed- ucation and Bilingualism, 25(2), 585-597. Zheng, Y., & Cheng, L. (2018). How does anxiety influence language perfor- mance? From the perspectives of foreign language classroom anxiety and cognitive test anxiety. Language Testing in Asia, 8(1), 1-19. Zimmerman, B. J. (2002). Becoming a self-regulated learner: An overview. The- ory into Practice, 41(2), 64-70. Examining the role of English language proficiency, language learning anxiety, and self-regulation . . . 423 Zimmerman, B. J. (2008). Investigating self-regulation and motivation: Historical background, methodological developments, and future prospects. Amer- ican Educational Research Journal, 45(1), 166-183. Yuksel, D., Soruç, A., Altay, M., & Curle, S. (2021). A longitudinal study at an Eng- lish medium instruction university in Turkey: The interplay between Eng- lish language improvement and academic success. Applied Linguistics Re- view. https://doi.org/10.1515/applirev-2020-0097 Dogan Yuksel, Adem Soruç, Barıs Horzum, Jim McKinley 424 APPENDIX A Language Learning Anxiety Scale (LLAS) 1. *I tremble when I know I’m going to be called on in language class. 2. Even if I’m well prepared for the class, I feel anxious about it. 3. In language class, I can get so nervous that I forget things I know. 4. *I worry about the consequences of failing tests, assignments and exams in English. 5. *It embarrasses me to volunteer answers in my language class. 6. I feel more tense and nervous in my language class than in my other classes. 7. I’m afraid other students will laugh at me when I speak English. 8. I start to panic when I have to speak without preparation in language class. * Items removed after the factor analyses. Examining the role of English language proficiency, language learning anxiety, and self-regulation . . . 425 APPENDIX B Self-Regulation Scale (SRS) 1. If there is something I don’t understand in English, I do my best to find the answer in a variety of resources (course books, dictionaries, online resources). 2. *If I can’t understand something in the English class, I ask others (my English teacher, friends etc.) for help. 3. *I try to find opportunities to practise my English. 4. I try to prepare for every English lesson, even if I know that I won’t be tested. 5. I try to learn English by watching films in English and listening to music in English. 6. I have my own ways of studying English vocabulary. 7. I plan my preparation and reviews before the test. 8. I have my own special techniques to make even the most boring activities more interesting. 9. *When studying English, I arrange my environment to make learning more efficient. 10. *I use my own techniques to keep me focused on studying English. 11. When studying English, I arrange my environment so that to avoid possible distrac- tions (TV, the Internet, mobile phone). 12. I study English as long as it takes me to achieve my own goals. * Items removed after the factor analyses. Dogan Yuksel, Adem Soruç, Barıs Horzum, Jim McKinley 426 APPENDIX C Reliability coefficients and goodness of fit indexes of confirmatory factor analysis of scales Scale Total items Factor loading EV Reliability Standard lolution Fit measure Model value Language learning anxiety 5 .539-.880 61.73 0.91 .69-.96 χ2/df 1.697 RMSEA 0.062 SRMR 0.016 CFI 0.99 NFI 0.99 NNFI 0.99 GFI 0.99 AGFI 0.94 Self-regulation 8 .719-.979 85.49 0.96 .53-.99 χ2/df 1.879 RMSEA 0.069 SRMR 0.033 CFI 0.99 NFI 0.99 NNFI 0.99 GFI 0.96 AGFI 0.91