Language use in an English-medium instruction university in Lebanon: Implications for the validity of international and local English tests for admissions

English-medium instruction (EMI) universities in non-English dominant contexts often use internationally available tests of English (e.g., TOEFL) and/or locally developed tests to assess the academic English-language proficiency of applicants prior to admission. However, little research has been conducted to establish the extent to which either type of test reflects the real-world classroom discourse, and associated communicative demands, within these often multilingual educational contexts. In this paper, we report on one part of a larger study designed to address this research gap in a particular EMI environment – the American University of Beirut (AUB) in Lebanon. We collected and analysed samples of language use at AUB through non-participant ethnographic observations and video recordings of eight undergraduate classes across academic disciplines. Findings indicated that classes were varied in nature, shifting between registers and levels of interaction, involved a range of translanguaging and English as a Lingua Franca-oriented phenomena, and included meta-commentary about language use. We argue that speaking and listening constructs in English-language admissions tests used in EMI contexts ideally need to be localised or locally developed to best match the unique language use characteristics of these settings.


Introduction
As communicative contexts change, or as new insights are provided into language use in such contexts, it is necessary to provide corresponding evidence to support validity arguments establishing a case for test use.This is of special importance for high-stakes tests used in university admissions, which often function as gatekeeping instruments determining who can be admitted to study, with far-reaching consequences for test takers.In recent years, the rise of English-medium instruction (EMI) higher education has led to critiques of current admissions testing practices (e.g., Galloway et al., 2020) on the grounds that, although designed specifically to determine students' readiness to study in English, existing international English-language tests do not adequately capture the communicative demands of EMI domains, characterised by multilingual practices and English as a Lingua Franca (ELF) communication.The growing prevalence of EMI education therefore represents a fundamental challenge for English-language testing practices, as large-scale international tests may be implemented across different geographical and sociolinguistic contexts before additional validation evidence has been provided to support their use.
In this study, we investigate a particular EMI context -the American University of Beirut (AUB), in Lebanon -and consider the extent to which communicative practices observed in AUB classes would be likely to align with the task features of English-language tests (one international and one locally developed) currently used for admissions purposes.
This study does not aim to provide a full validation of the use of either test for admissions in this context.Rather, we seek to identify overlaps and possible gaps, leading to discussion of potential mismatches between the characteristics of oral and aural communication in the target language use domain and the corresponding demands of current tests used for admissions purposes.

Background
We define EMI following Macaro, Curle, et al. (2018) 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" (p.37).This phenomenon of EMI has been growing across higher education contexts around the world (Dearden, 2015) with a concomitant expansion in research interest among language education scholars.Research has focused principally on the effectiveness of EMI education (e.g., Dafouz & Camacho-Miñano, 2016), the perceptions of students (e.g., Kuteeva, 2020), the preparedness of university teaching staff to instruct through the medium of English (e.g., Macaro et al., 2019), language use in the EMI setting (Macaro, Tian, & Chu, 2018), and policy issues in implementing EMI within institutions (Rose et al., 2020).The use of English-language admissions tests, however, has been less extensively researched (see Dimova, 2020), despite their crucial role as gatekeeping devices for entry to EMI institutions.
Existing studies of English-language admissions testing in EMI contexts have generally addressed two areas: the suitability or predictive validity of cut-scores, and the suitability of test constructs for EMI settings.The first topic has attracted considerably more attention, with several recent studies exploring the appropriateness of existing score-based admissions requirements for entry to EMI programs.Researchers have investigated, for example, the predictive validity of TOEFL iBT for EMI admissions in Turkey (O'Dwyer et al., 2018), suitable cut-scores for the use of IELTS in the United Arab Emirates (Schoepp, 2018), and perceptions of the suitability of various methods of demonstrating Englishlanguage proficiency at the University of Copenhagen (Dimova, 2020).
Research is only beginning to appear, however, regarding the suitability of existing test constructs for EMI admissions decisions.For example, Owen et al. (2021) focused on the suitability of TOEFL iBT reading tasks for two distinct EMI higher education contexts: Nepal and Sweden.They found that TOEFL iBT could provide useful evidence for drawing inferences about reading ability in both contexts.However, the authors noted that, specifically in Nepal, academic reading often involved "the use of translation, Nepali language and [first language]-based strategies, such as asking peers or tutors to translate" (p.22).While there is less empirical research on the suitability of speaking and listening test constructs, there have been notable critiques of existing test options for EMI admissions.For example, Galloway et al. (2020) argue: The continued use of tests such as IELTS, TOEFL and TOEIC, as both placement and proficiency tests in EMI programmes, are [sic] problematic given that they are based on "native" English norms and language-related challenges as well as how to define "proficiency" for EMI settings are in need of further investigation.(p.399) The notion that the nature of proficiency might differ in EMI contexts is underscored by recent research that views EMI contexts fundamentally as sites of ELF communication and translingual practice (see Dafouz & Smit, 2016;Kuteeva, 2020).Within the ELF paradigm, proficiency is conceptualised as a fluid and dynamic phenomenon, contextually bound and co-constructed in situ by interactants, foregrounding accommodation, negotiation of meaning and clarification strategies that are implemented to achieve understanding (see Canagarajah, 2006;Harding & McNamara, 2018;Jenkins & Leung, 2014).From the translingual perspective, EMI classroom discourse has been characterised by the prevalence of translingual/multilingual practices (Smit, 2019).Such practices are considered common features of classroom discourse in EMI settings, as they are "spontaneous" and closely connected to "local language practices" (Kuteeva, 2020, p. 298).In addition, translingual practices have been found to express a range of important functions in classroom discourse (Yuan & Yang, 2020) and to mark identity within classroom communities (Smit, 2019).
If evidence suggests that communicative practices in EMI higher education contexts differ from those of, say, North American or British institutions, it follows that the linguistic knowledge, skills, and abilities required to operate in EMI contexts are likely to differ as well (see Papageorgiou et al., 2021).This situation threatens what Chapelle et al. (2008) label the domain description inference, the claim -foundational to an assessment validity argumentthat test tasks represent the skills, knowledge, and abilities required to operate in the target language use domain.However, this critique is not entirely straightforward.EMI higher education contexts are considered highly idiosyncratic (Doiz et al., 2013), with a constellation of factors determining the ways in which (de facto) language policies are understood and enacted (see Dimova & Kling, 2020).This means that different EMI domains may operate closer to, or further away from, the norms embodied in large-scale international English-language proficiency tests, perhaps varying even within the same class.As Kuteeva (2020) explains, "each EMI context constructs its own language regime; the idea of what is acceptable is not static and can move along the standard-non-standard continuum" (p.298).
As such, it is difficult to generalise from existing research to determine the appropriateness of test instruments without scrutiny of the language use domain in a specific institution.

Research context
The site for our study is the American University of Beirut (AUB), a liberal arts university in Lebanon that follows the American higher education model.AUB was established by During AUB's first thirteen years, Arabic was used as the medium of instruction, but this was changed to English in 1879, first in the Literary Department, then the Medical Department (Anderson, 2011).Currently, teaching in most contexts at the university is in English (AUB, n.d.-a).AUB offers insight into a specific EMI context, whose characteristics have been shaped by the historical foundations and development of the institution.The university was an EMI institution long before the relatively recent spread of EMI in higher education institutions around the world and is oriented towards the educational values of the United States in ways that also might set it apart from more recently established EMI settings.
Nevertheless, the early change to EMI at AUB was arguably still motivated by the goal of internationalisation, but from a mid-19 th -century American and Christian perspective.
Unlike many universities that have recently introduced teaching in English (e.g., University of Amsterdam), AUB has not published an extensive language policy.This may be due to the taken-for-granted nature of EMI at the university, because the change was made over 140 years ago.The only explicit evidence of AUB's EMI status is a short statement on the university website: All teaching at AUB is carried out in English (with the exception of some languagespecific teaching, especially in the Department of Arabic and Near Eastern Languages), and undergraduate as well as graduate applicants must demonstrate that they have sufficient fluency in written and spoken English to cope with the language demands of their chosen field of study.Therefore, AUB requires all applicants, regardless of citizenship, to demonstrate English language proficiency or Readiness for University Study in English (RUSE).(AUB,para. 1) This situation allows us to inspect an "organic" locus of EMI situated in a multilingual context where English and French are used alongside Arabic 1 (Esseili, 2017;Shaaban, 2017).
AUB requires that enrolling students provide evidence of English proficiency (AUB, n.d.-c); no proficiency in Arabic is required.Proof of English proficiency can be evidenced by scores on international tests, including TOEFL, IELTS and, more recently, the Duolingo English Test (AUB, n.d.-c).Alternatively, students can take a locally developed English test, the AUB English Entrance Examination (AUB-EN), which was re-designed in 2015 (see Pill, 2016b) to represent the "general experience" at the university, and which aims "to assess a test taker's readiness, in terms of language and communication skills, to participate in academic, administrative and social activities" (Pill, 2016a, p. 4).The AUB-EN is designed to cover the four macro-skills: reading, writing, listening, and speaking (Pill, 2016a); however, the speaking test has not yet been implemented.The university requires a minimum overall score of 79 for TOEFL iBT and 40 for AUB-EN (AUB, n.d.-c).Our study considers these two tests, as they are currently the most frequently used by applicants to AUB.Table 1 presents features of the two tests salient to this study, with a focus on the speaking and listening components.Accents and speech patterns of the speakers in the recordings reflect the diversity of the AUB community; some speakers may be competent users of English as an additional language.

Speaking topics
Topics that test takers are somewhat knowledgeable about, but also topics they know less well because they are just learning about them.
Tasks simulate those likely to be required of students in the university context.Specific academic or procedural knowledge is not expected.Note.Drawn from https://www.ets.org/toefl;Pill, 2016a; also Bejar et al., 2000; Butler et al., 2000.a https://www.ets.org/toefl/score-users/about/content

Aims and research questions
Despite a recent surge of research interest in the nature of discourse in EMI university settings (see Smit, 2019), there remain areas where our understanding is limited.In this article, we seek to build on existing theory by (a) exploring language use in a particular EMI context, and (b) using those findings to compare characteristics of the target domain with characteristics of language tests used for admissions purposes.On the first point, AUB functions as an instrumental case of an EMI institution in a wider multilingual social context where Arabic, English, and French are the main languages.As set out above, AUB's history and role within the landscape of higher education in Lebanon make it pertinent for study.It is a leading university nationally and regionally, and pedagogic and administrative practices at AUB are often emulated at other universities in Lebanon and across the Middle East.On the second point, while other researchers have noted the potential unsuitability of current English-language admissions tests used in EMI contexts, no one has conducted empirical domain analyses of language use in specific EMI institutions with a view to evaluating the strength of the domain description inference of specific English-language tests.
Against this background, we address a general research question in this paper: What is the nature of spoken language use in undergraduate classes across disciplines at AUB? Based on our analysis, we will then discuss the extent to which salient features of lecture/classroom language use could potentially be captured in TOEFL iBT and AUB-EN listening and speaking specifications and tasks.

Methods
Data for the current study are drawn from a larger project designed to explore the validity of international and local English-language proficiency tests used for admissions at AUB.Data in that project include observations/recordings of eight undergraduate classes across different disciplines, recordings of lecturers' office hours, recordings of transactional and service-related encounters on campus, and interviews with lecturing staff, students, and admissions personnel at AUB concerning perceptions of and attitudes towards current English-language admissions procedures and expectations.In this paper, to address the issue of language use in lectures/classes, we focus on the first set of data: observations and recordings of eight classes at AUB.

Data collection procedures
Ethical approval was obtained at AUB and Lancaster University prior to data collection.
Teachers of specific classes were recruited through purposive sampling, and appointments were made to observe and record the eight classes.Each observation/recording followed a similar protocol.Before the class started, two video cameras were placed in each teaching space to capture both the view of the front of the room and the reverse aspect (one camera was used for the first part of the Landscape Design class).Audio recorders were placed around the classroom to capture student conversations.Classes were typically set up with teachers (or presenting students) at the front of the class near a whiteboard or equivalent, and students seated in rows or on lecture theatre chairs facing them.In the Education class, the students sat in groups on the two sides of a U-shape structuring of tables, while in the English class the students, though still in rows, turned to form pairs and small groups for part of the session.The Landscape Design class began in a lab but then moved outdoors; outside the lab, only audio recordings were collected.
In addition to making recordings, the first author observed all classes and took detailed field notes following the "open observation" method (see Van Note Chism, 1999), except for the Chemistry class, where circumstances made taking field notes unfeasible.To gain further insight into the linguistic diversity within each class, students were asked to complete a language background questionnaire based on Smit (2010) before or after the class (depending on the teacher's preference).Questionnaire items asked respondents to indicate their nationalities, the language(s) they spoke at home, other languages they had studied, and if they had spent substantial time studying in other countries.Background information on teachers was collected through interviews conducted after the observed class had taken place.
(One teacher was interviewed in advance; one was not available to be interviewed.)This information was supplemented in some cases with reference to publicly available resources (e.g., CVs posted on institutional websites).
For the current study, this background information provided a useful insight into the range of linguistic resources among teachers and students in the various classes observed (see Table 3).Also noteworthy through observation, but not captured directly in the questionnaire or interview data, were the many varieties and accents of English used among students and teachers.

Analysis
The Ethnography of Communication, and the TOEFL framework papers for speaking and listening (Bejar et al., 2000;Butler et al., 2000).The observation grid is presented in Table 4.In the next step, five classes were selected for a more fine-grained, micro-level discourse analysis of salient language-related episodes.These were Biology, Chemistry, Civilization Studies, Education, and History.This selection allowed for a deeper insight into Humanities as well as Sciences, with a variety of language backgrounds for teachers and of class types Episodes and themes were discussed by the three researchers, with the analysis fine-tuned through a shared interpretation of the meaning and significance of the data.
Consideration of the macro-and micro-analyses, together, provides a fuller picture of potential alignments (and misalignments) between the target language use domain and characteristics of current admissions tests.For example, the macro-analysis provided the foundation for a high-level, broad descriptive account of situational and linguistic characteristics, and activities, across classes.The micro-analysis provided a detailed insight into salient features of spoken discourse.While some interactions may be similar to those in non-EMI contexts, they remain important to an inclusive domain analysis for the AUB context, providing valuable insight concerning the suitability of tests used to decide access to this context.

Macro-analysis
The macro-analysis based on the observation grid revealed similarities and differences across the eight classes.Table 5 provides a summary of key observations against each area of focus, with more detail provided below.

Internationalisation and glocalisation
Mix of international and local focus Some degree of local knowledge assumed (especially in Humanities classes) In terms of Locale, classes spanned a range of physical settings, from smaller rooms (e.g., Mathematics) to lecture theatres (e.g., Chemistry).Instruments also varied widely, with more traditional set-ups (slideshows, U-shaped or rowed seating) observed, for example, in the Biology and Education classes, as well as discipline-specific arrangements, such as students seated around larger tables in labs and going outside "into the field" to draw (Landscape Design).
With respect to the Role of language, all classes were primarily in English.Arabic Often, these genres were mixed within the same class.For example, the Biology class was partly teacher-led, with the teacher transmitting content in a lecture-like manner.However, the teacher also engaged students through questions that placed them more explicitly at the centre of the class and led to seminar-like exchanges.Other classes leaned more towards one genre: for example, the Mathematics class was mainly lecture-style, while the observed Education class was mostly student-led.
In line with these variations in genre, the spoken register varied from more informational to more involved, within and across classes.Classroom management was often accomplished in an informal register, and it is here that we also observed an important role for Arabic (see below).Norms of the exchange in each class varied, with some classes maintaining a more stable use of English throughout (e.g., Civilization), and others moving rapidly between languages, registers, and levels of linguistic complexity.In the Education class, for example, a broad discussion about questions and critical thinking included an informal digression when students leading the class noted how humans are different from animals in their ability to formulate questions, prompting the teacher to relay a story about how her own pet asks questions.In the Mathematics class, in contrast, the focus was primarily on conveying specific content, and a formal, academic register was maintained.
Agents were the teachers and students.Others occasionally took up roles in the classes -for example, the History class was visited by a teaching assistant who discussed the assignment, while Landscape Design students interacted with others outside the class during their fieldwork.Teachers typically managed the class, but the floor was given to students on occasion to present, speak, and even lead the session (under the teacher's guidance).Most of the classes observed involved substantial interaction: students either spoke out unprompted, indicated their desire to speak then spoke, or were called on to speak.Students also spoke amongst themselves occasionally.These patterns were also reflected in the observation of Practices and purposes for language use.These ranged from lectures, presentations, explanations, and instruction-giving, to greetings, classroom management, group-work, pairwork, and incidental conversation.
Internationalisation and glocalisation were re-interpreted for this study considering the sociohistorical context of AUB.These concepts are already signaled in the name of the institution: "American University" shows internationalisation (from the United States to Lebanon and an international outlook from Lebanon) and "of Beirut" indicates glocalisation (a local version of an American model) (see also Anderson, 2011;Pill, 2016b).Our analysis looked for evidence in the data of relationships with a wider world and of references of local relevance.Some classes were obviously international in focus -the Civilization class involved a discussion of the play Hamlet, while the Biology class contained discussion about international medical examinations the students might take.In other classes, explicit connections were made with the local context (e.g., in the History class, students were asked to interview local community members about the Lebanese Civil War).At other points, explicit and implicit references were made to local knowledge.For example, in the Education classroom, the teacher mentioned West Beirut and East Beirut as terms used during the Civil War but did not offer an extended definition, although the terms hold strong connotations regarding the division of the city.As a general observation, Humanities classes were more oriented towards local concerns than Science classes.
Overall, the macro-analysis suggested that language use in AUB classes was complex, though this complexity cannot necessarily simply be attributed to the "EMI-ness" of the setting.For example, the observation that classes spanned a range of locales, included a variety of instruments, and represented mixed genre and register patterns, aligns AUB with the characteristics of many non-EMI university settings (see, for example, Lee, 2009).
Further, it was clear that in many classes, teachers and students were oriented towards disciplinary conventions in terms of topics, activities, and specific vocabulary.Nevertheless, there were clear points of interest in the observations around language, and specifically the use of Arabic, the flexible way in which English was used among the diverse staff and student population, and the explicit management of language-related issues in the classroom discourse.It is on these elements that we focused the micro-analysis.

Micro-analysis
The micro-analysis of critical incidents in the discourse of five classes revealed three salient (inaudible) S1: forty-eight S2: forty-five (inaudible) T: forty-five forty-eight teb khalas <A_so enough> okay ma 'ento <A_but you> you arrived late so I should have kept you five minutes extra but mesheh el hal <A_it is okay> In this example, the use of Arabic discourse markers and conventional phrases (lahza di'a shway; teb khalas; ma 'ento; mesheh el hal) accompanies a shift to a less formal register for classroom management.
Moving fluidly between English and Arabic was also commonly observed when students discussed classroom management issues.Earlier in the same Biology class, a student interrupted the teacher discussing a medical student's project to ask if there is an attendance sheet (Excerpt 2).
[2] Biology (2-00:43) T: ... is a med one student going to to <<student waves raised hand>> yes S1: eh fi <A_is there> attendance sheet?T: na`am? <A_pardon?>S1: fi <A_is there> attendance sheet?T: eh yes SS: @ T: I don't know where it's going S2: la' <A_no> you never gave out In this excerpt, it is notable that the teacher immediately switches to Arabic (na`am) to manage the request, indicating that moving between English and Arabic was an unremarkable phenomenon in this classroom.
A second site of translingual practice concerned instances of rapport-building between teachers and students.As an example, Excerpt 3 shows an extract of talk from the Civilization class on Shakespeare's Hamlet.The teacher makes a play of words on the name of Hamlet's castle, Elsinore.
[3] Civilization (12:02) The humour in this quip is derived from understanding that el is a definite article in Arabic and can be translated as the in English; thus, Elsinore could be translated as the sinore.Tai and Li (2020) have shown this "playfulness" with language to be a salient resource for building rapport in the EMI classroom.It is also noteworthy how, in making this joke, the teacher anticipates that the students will understand the linguistic connection that makes the play on words work.
A third area where translingual practices appeared extensively was the authentic rendering of quotations or names.In Excerpt 4, from the History class, during a discussion of whether Lebanese political parties are secular; a student switches to Arabic to quote a politician -Saad El Hariri, a former Prime Minister of Lebanon -verbatim.
[4] History (47:22) T: … they don't talk they don't talk about they all promise us a secular state S1: no they don't eh Saad El Hariri for example says 'ana 'ana bay el Sunna <A_I I am the father of the Sunnis> in public In the same class, the teacher introduces the name of a political group in Arabic terms (Excerpt 5).
[5] History (31:44) Here, it is notable that the teacher builds meaning using the multilingual repertoire available to the class, providing the name that will be most authentic and familiar to the Lebanese students in the class, El Helef El Thoulathy, before extending this in an English gloss.
ELF-oriented practices.As well as translingual practices, several ELF-oriented practices were identified in the micro-analysis.One salient practice, observed across several classes, was creativity (see Kecskes, 2019;Pitzl, 2018).In the Education class, for example, a student referred to deep argumentising (02:31).Another student, when discussing questions posed in school classrooms and how they are handled by teachers, asked what handles or mishandles happened in your classroom?(28:41).In Excerpt 6, the same student goes on to use wow handling to mean effective or impressive handling of a question.
[6] Education (23:00) ST1: … okay so eh can we take please five minutes to discuss any mistakes you've seen in your practice teaching experience that is mishandling a question and not ya`ne ya <A_I mean either> mishandling ya <A_or> wow handling of a question SS: @ ST1: okay?
In this exchange, ELF creativity co-occurs with translanguaging phenomena, adding to a picture of the rich and complex linguistic repertoires being drawn on as a natural feature of classroom communication.
Negotiation of meaning also occurs in the classroom interactions, either between students or between teacher and students.In Excerpt 7, a student and teacher are discussing an assigned reading on the 1970 elections in Lebanon, and jointly work to find an appropriate (discipline-relevant) term to refer to political groupings.They decide on blocs.

S1:
… mostly the way he does it is that he explains the the different eh factions the different groups that were involved in the or the different eh not groups the different eh ya`ne <A_ so to speak > T: they were groups the factions S1: d-different blocs T: blocs S1: the different blocs though is that T: this is we call them blocs because they were voting in parliament right … The need for negotiation does not appear to be due to a lack of English-language proficiency, especially since it is the student who eventually identifies the word and since the teacher is able to explain why that word is correct.The student signals a word-search strategy (eh not groups) and, when they do not manage to find the term immediately, the teacher assists (they were … the factions).The student then identifies the term, and the teacher uses this negotiation of meaning to help explain to the class why blocs is a more accurate label.This kind of co-construction of meaning was a feature of the more discursive, seminar-style teaching observed across several classes.
Meta-commentary around language use.In addition to translingual and ELForiented strategies, a third noteworthy observation that emerged in the micro-analysis was meta-commentary and negotiation around the use of language itself.Teachers and students sometimes specifically oriented towards language regulation in their exchanges, pointing to the fact that the language policy "on-the-ground" in each classroom was in constant flux and subject to the agreement of both teacher and students.Excerpt 8 is an example from the History class.In the discussion, the teacher, a teaching assistant, and a student are setting up a class assignment that involves interviewing someone from their local area about their experiences of the Civil War.The student asks the teacher is Arabic fine?, checking that the video assignment can be presented in class in the language in which the interview will be conducted.The teacher agrees, noting that subtitles would be ideal.
[8] History (24:58) T: … everyone is supposed to interview one person which is actually it will only take maybe an afternoon of your time to both pick pick the pick the to actually listen to the interview again do it listen to it again and pick the stuff you want to use (.) yes S1: eh is is Arabic fine?T: yes of course TA: yeah T: of course S1: no translation needed right?T: if you have the skill to insert eh subtitles that would be perfect but Arabic is fine The extract is notable for two reasons.First, both the student and the teacher recognise the importance of Arabic in engaging with the community directly outside the class.The classroom is positioned, implicitly, as a particular kind of space where dominant language practices in the wider social context may (or may not) require mediation.Interactions of this kind reveal the commonplace policy issues that are frequently (and, in this case straightforwardly) negotiated in the EMI classroom.Second, the teacher's quick decision not to require an English translation of the interview signals, again, that Arabic is understood to be a shared code within the classroom environment, forming part of the repertoire of language practices that teachers and students regularly engage in.

Summary of findings
The macro-analysis based on field-note observations and video recordings revealed that the nature of communication in classes at AUB was varied and flexible.The observed classes share some similarities with generic descriptions of listening and speaking in non-EMI academic contexts: they include discussions of discipline-specific content material, instructions, group-work, presentations, and various other speech activities conducted through the medium of academic oral English.The U.S.-oriented provenance and traditions of AUB possibly lead to a stronger reflection of these practices at this institution than at universities implementing EMI more recently.At the same time, the classes we observed also departed, occasionally, from the conventions of academic English lecture-style discourse, that is, from predominantly monologic, teacher-led, formal academic language.Observed classes shifted between languages, registers, and levels of interaction.English was used predominantly, but use of Arabic was salient and, in at least one class (Education), extensive.
Topics ranged from those with a more international focus to those with a specific, local focus requiring a shared cultural understanding (particularly in Humanities The micro-analysis provided greater depth to these observations.Translingual practices were found to be salient features of classroom discourse, used for classroom management, for establishing rapport, and for rendering quotes and names authentically. ELF-oriented practices were also common.There were examples of creativity and frequent sites where negotiation of meaning underpinned classroom discourse and pedagogical effectiveness.Finally, teachers and students oriented towards language use itself, demonstrating a meta-awareness of the need to negotiate and update understandings of which language practices would be permitted in which contexts.These practices were fluid and dynamic, as has been noted by other researchers exploring EMI contexts (e.g., Kuteeva, 2020;Smit, 2019).

Comparison with existing admissions tests
Turning to a discussion of the applicability of these findings to current English-language tests used for admissions, we observe that there are both overlaps and areas where there may be a mismatch between the domain and the potential for each test to capture the "theoretical construct" (see Knoch & Macqueen, 2020).For TOEFL iBT, there are overlaps in terms of the range of contexts for academic listening and speaking.The listening section, for example, includes both conversations and lectures, including classroom discussion, while integrated speaking tasks involve re-presenting speech content previously read and/or heard (see Table 1).These genres and activities match those observed across the classes.However, there are also instances of mismatch between task characteristics in the current TOEFL iBT and the test's potential to capture characteristics of the AUB domain revealed in our study.The use of monologic speaking tasks in TOEFL iBT does not permit negotiation or co-construction of meaning, which was observed in our data.Similarly, the listening section of TOEFL iBT features a limited range of speaker accents (see Table 1) that overlaps only partially with those present in classes observed at AUB. Translanguaging and meta-commentary around language use were also salient in the data.While it is difficult to determine from materials available, we believe it is unlikely that TOEFL iBT listening materials currently include translanguaging phenomena -this is unsurprising as the test was originally developed for use in a North American higher education context assumed to have a de facto monolingual language policy.Moreover, referring to publicly available speaking descriptors (Educational Testing Service, 2019), no indication is given of how a TOEFL test taker would be scored if they demonstrated translingual competences in the TOEFL Speaking test.Thus, it is likely that while TOEFL iBT -currently used extensively for admissions processes at AUBcaptures core elements of general academic proficiency, it is limited in its ability to tap into the complex interactional and translingual skills and the heterogeneous varieties of English that are characteristic of this EMI setting.
For the AUB-EN, which explicitly "aims to reflect the diversity of the AUB as a community of people having different languages and cultures" (Pill, 2016a, p. 4), there is arguably slightly more overlap with the nature of the domain that has emerged through our analysis.For example, the AUB-EN Speaking test has the potential to elicit some of the interactional ELF-oriented skills through its group oral task, but, as with the TOEFL iBT speaking descriptors, neither the test information nor the scoring descriptors explicitly value this aspect of performance.Instead, test takers are "advised to use the time during the test to demonstrate their spoken communication skills in English" (Pill, 2016a, p. 21).Similarly, the AUB-EN Listening test introduces local accents of English specifically for the purpose of more closely modelling the target language use domain (see Table 1).Nevertheless, the locally developed test is currently no more sensitive to the translingual practices we observed in AUB classrooms than is the international test considered in this study.It seems that the AUB-EN test specifications seek to reflect the local context in terms of the English-language and communication challenges for many incoming students as they enter this educational setting, but they still do not take into account some aspects of language use at AUB, particularly translanguaging.The AUB-EN deals with all interaction as if it were monolingually English.This study, thus, requires us to question some of the current language testing practices for such contexts.
In summary, the specifications of TOEFL iBT and the AUB-EN suggest a good level of overlap with the target language use domain of undergraduate classes at AUB at the broad level, but neither represents the range of languages and the full variety of language use we observed in this setting.We acknowledge that the current study is limited in that it did not conduct a detailed analysis of sample tasks/items or performance samples from the tests under focus.Nevertheless, the analysis of specifications and task features alone provides useful insights into areas where mismatches are evident between characteristics of the observed domain and the broad features of these two admissions tests used to determine access to the same domain.

Implications
The mismatches noted raise the question of how to address such issues in practice.The misalignments we have observed connect with more fundamental issues that have been discussed in different ways over recent years concerning, on the one hand, the constraints on large-scale international examinations to provide scores with validity in localised contexts, and, on the other, the challenges for locally designed tests to address the complexity of their domain with typically only thin resources (see Chalhoub-Deville & O'Sullivan, 2020;Dimova et al., 2020).Nevertheless, we suggest two potential ways in which these misalignments might be addressed.
First, providers of large-scale tests such as TOEFL iBT (and similar international exams) might consider a greater degree of localisation to allow for better representation of features of EMI contexts in their test constructs.It is likely that EMI contexts will vary, meaning that existing international formats would translate more easily to some EMI contexts (such as Sweden, as seen in Owen et al., 2021), but not so easily to others.Localisation could involve the adoption of, for example, a wider range of speaker accents and particularly local varieties (e.g., Motteram, 2020), the incorporation of translanguaging phenomena in listening assessment (see Baker & Hope, 2019), and the inclusion of interactive speaking tasks (e.g., through paired-or group-oral speaking exams), which allow for negotiation of meaning and meta-commentary around language use, and which value translingual practice in test taker performance in their grading schemes.
Addressing translanguaging in localised test constructs brings challenges, given that tests may be taken by candidates with varying levels of proficiency in languages observed in the domain.However, such challenges are surmountable.Using the case of AUB as an example, the inclusion of Arabic in listening texts could take the form of common interjections and asides, modelling conditions in which listeners must navigate the fluid linguistic repertoires of participants (students and teachers) but avoiding test takers being penalised for not understanding key content in Arabic.Similarly, scoring rubrics for speaking assessment could explicitly acknowledge that use of Arabic is common in spoken interaction so as not to penalise it through a lens of deficit.Notwithstanding these suggestions, the specific degree to which translanguaging should be accommodated in a test, and through what means, must be the topic of a different paper, as no tasks were trialed in this study.`ayn /ʕ/ (Arabic)

Endnote
1 As the diglossic situation of Arabic and the existence of several varieties is beyond the scope of this discussion, Arabic is used as an umbrella term for any variety.Language use in the context is yet more complex, with the further presence of minority languages (Esseili, 2017).
Protestant missionaries in 1866 (American University of Beirut, n.d.-d) and is accredited by the Middle States Commission on Higher Education in the United States (AUB, n.d.-b).
Eight undergraduate classes were selected for observation and video recording based on their representation of a range of disciplines and the lecturers' willingness to join the study.The sample comprised classes within the departments of Biology, Chemistry, Civilization Studies, Education, English (Enrichment Courses -developing communication and composition skills), History and Archeology, Landscape Design and Ecosystem Management, and Mathematics.The Chemistry, Civilization Studies, English, and History classes were part of the General Education Program (for all students, regardless of major, to meet the breadth requirement of a liberal arts education, but also taken by students majoring in those disciplines), while the Biology, Education, Landscape Design, and Mathematics classes were specifically for students majoring in those disciplines.Consequently, classroom interactions were likely to represent a range of more general and more specialised language use.Sampled classes varied in size, length, and location (see Table2).One session from each of the eight classes was observed (approximately 12 hours of class time in total).All classes were video recorded, complemented by audio recordings, except for the Landscape class, where audio recordings were primarily relied upon (groups of students each took an audio device to record their small-group interactions while drawing outdoors).There is a risk that the specific classes observed may not have reflected characteristic language use across all sessions of that class or the discipline; however, in the observed classes, a variety of activities were represented.Furthermore, based on two of the authors' first-hand experience of the university (as a student and a teacher), the sample was considered broadly representative.Most of the observed classes included elements of both lecture-and seminarstyle teaching.The Chemistry class was supplemented by separate recitation sessions, the Education class was a practicum (linked to student-teachers' visits to schools as part of their course), and the Landscape Design class was a lab (where the students were learning to draw).The eight class observations and recordings were carried out at the AUB campus in March and April 2019.
final dataset consisted of observation notes for all classes (except Chemistry), video and audio recordings of each class, and data from background questionnaires (students) and interviews (teachers).As a first step in the analysis, the open observation field notes were augmented by re-watching the videos to develop more comprehensive observations of the macro-level features of each class.This step was guided by a structured observation grid, developed specifically for this study, that drew together elements ofDafouz and Smit's (2016) ROAD-MAPPING framework (see alsoEslami et al., 2020), Hymes' (1974)

(
General Education Program or major).During this stage, critical incidents were identified and transcribed using a system based on the VOICE transcription conventions(VOICE   Project, 2007), adapted to suit the needs of our study.Spoken Arabic was rendered with Latin script (see Transcript conventions for more detail).A critical incident was defined as any stretch of talk in which participants (students and/or teacher) appeared to engage in translingual or ELF-oriented practices, which previous research has identified as characteristic of EMI classroom discourse.This process was not restrictive, however, and identification of critical incidents at this stage was guided by the researchers' interpretation of what might be distinctive in the AUB context.These episodes were grouped thematically.
was used in classes to varying degrees, apart from the Civilization class, where only English was observed.Arabic was particularly prevalent in the Education class -on this teacher training course, some students are training to teach content in Arabic and are thus expected to use Arabic.Use of French was limited, with only a few words spoken in the Education and History classes, mostly with reference to specific names or terms in common usage (e.g., Brevet [a national examination for Grade 9]).The English and Arabic observed were spoken with a variety of accents, reflecting the diverse linguistic backgrounds of students and teachers.The classes represented a variety of Academic disciplines and, as anticipated, the use of English in each class reflected variations in disciplinary conventions (e.g., autophosphorylation from the Biology class and perspective view from the Landscape Design class).Observation of Language management suggested a complex picture.Genres of classroom discourse were considered impressionistically against Fortanet's (2005) classification of spoken academic classroom genres.Discourse appeared to span the genres of academic lecture and student presentation (expository) and academic seminar (interactive).
themes: translingual practices, ELF-oriented practices, and meta-commentary around language use.Translingual practices.Translingual practices were common across the five classes analysed at the micro-level.Analysis of instances of translingual practices revealed that Arabic was drawn on principally for three purposes: (a) classroom management, (b) rapport building, and (c) authentic rendering of quotations or names.To exemplify classroom management, Excerpt 1 shows an instance of translanguaging in the Biology class.The context of the extract was that class time was running out, and the teacher digresses from a discussion of content (hedgehog signaling) to list the topics that she will not be able to cover and to check if there is sufficient time to cover one more topic.(In the extracts, translations of speech not in English are included in angle brackets, with the initial "A" indicating that the original is Arabic.Other transcription conventions are given in the Appendix.)a shway <A_a moment one minute> just let me see what's the time now?did we reach eh? SS:

Table 1
Some Salient Features of TOEFL iBT and AUB-EN 4 tasks (3-5 minutes each); at least one monologue; at least one dialogue; up to four speakers in a task textAccents used in Listening sectionNative-speaker English accents from North America, the UK, New Zealand, or Australia

Table 2
Details of Observed Classes a Class typically meets twice a week.b Class typically meets three times a week.c Class typically meets once a week.d Part of the observed session took place outdoors.

Table 3
Language Knowledge Reported by Teacher and Students in Observed Classes * Reported as home language.

Table 4
Observation Grid for Macro-Analysis of AUB Classes These questions represent our re-interpretation of I&G to suit the AUB context better.

Table 5
Summary of Key Observations from the Macro-Analysis Arabic it means the sinore okay?El sinore so eh we're in Elsinore and the players show up what are players?
: … which is what we call the Alliance El Helef El Helef El Thoulathy <A_The Alliance The Triple Alliance> The Triple Alliance of Chamoun Frangieh as well as Edde which was a reaction to a number of things … T While this type of context-sensitive localisation is difficult practically, it is a vital step if