Clarification and repair in emergency remote EFL classes

EN During the sudden shift in education onto digital platforms due to the Covid-19 emergency, teachers became streamers and experimented with new tools to involve their students in video-mediated, multi-floor, multiparticipant, and multimodal interactions. In turn, students experienced new ways to participate in lessons and interact with instructors. This study focuses on clarification and repair in videoconferencing as a strategy to address trouble in video-mediated communication and to re-establish mutual understanding. Through participant observation of online classes, the researcher collected data on classroom interactions, which are analyzed through conversation analysis. The findings show how the digital affordances of video-mediated conversation help teachers and students manage intersubjectivity and compensate for the lack of non-verbal cues typical in face-to-face interaction, such as facial expressions or tone of voice. Consequently, this article argues that the wisdom gained during the pandemic can help teachers and lecturers better deal with clarification and repair in digital conversations. Ultimately, it can increase their digital interactional competence, thus giving way to more interaction and learning in EFL classes, both online and in-person.

and broader set of extracts and pursue a twofold goal. On the one hand, it aims to identify how CLA is managed in ERE classes to gain insight into a relevant aspect of Conversation Analysis (henceforth CA) bound to L2 pedagogy. On the other hand, it aims to provide teachers and lecturers with increased awareness of interaction in multimodal digital environments, thus enhancing their interactional competence and their success in teaching a foreign language. To this end, it aims to address the following research questions: RQ1: How do conversational repairs operate in ERE? RQ2: To what extent does the chat, as a means of intersubjectivity, affect student behavior in ERE classes?
To analyze data, it adopts a qualitative analytical approach informed by CA. Data was collected through direct observation of online ERE classes. It gives insights into video-mediated classroom discourse, which could increase teacher and lecturer awareness and improve the quality of interaction in online classes. To this end, first, a brief literature review will be provided. Second, the conceptual framework will be introduced; third, the methodology will be briefly outlined. Finally, the findings and their pedagogical implications will be presented and discussed.

Literature review
This paper on CLA in emergency online classes is informed by previous studies on CLA and videomediated communication. CLA is a repair sequence used to clarify trouble in interaction and achieve mutual understanding (Atar & Seedhouse, 2018;Schegloff et al., 1977;Walsh, 2011). Since it was introduced and developed by Schegloff and colleagues in the late 1970s, this concept has been extensively studied, among others, in the area of classroom interaction, which considers it crucial for its pedagogical relevance (McHoul, 1990;Walsh et al., 2011). In line with the seminal work by Schegloff et al. (1977), studies on CLA unearthed the patterns of repair sequences that allow mutual understanding between students and teachers in and outside of the classroom. The analytic distinction between self-correction and other-correction is central in the study of CLA (see also section 4). The former applies when the speaker of the initial item (trouble source, henceforth TS) corrects it; the latter refers to the correction performed by one of their interlocutors (Atar & Seedhouse, 2018;Koschmann, 2016;McHoul, 1990;Meredith & Stokoe, 2014;Schegloff, 2007;Schegloff et al., 1977).
Addressing problems in conversation through repair is considered a crucial part of interaction and thus an essential skill in second language acquisition (henceforth SLA) (Atar & Seedhouse, 2018;Kasper, 2006). There are several studies on CLA in L2 classes, which adopt a conversation-analytical approach and focus on EFL environments (Atar & Seedhouse, 2018;Mortensen, 2008;Walsh & Mann, 2015;Walsh et al., 2011). On the one hand, CLA and its pedagogical implications are investigated from a teacher-led perspective in terms of strategies to select a willing next speaker (Mortensen, 2008), turn management (Sert, 2019;Waring, 2013), best practices to detect in students' unwillingness to interact (Sert, 2015). Studies on interactions in L2 classes showed that the appropriate use of verbal and non-verbal resources facilitates intersubjectivity and comprehension and encourages students to give longer explanations and express themselves better (Atar & Seedhouse, 2018;Peachey, 2017). CLA should be mastered by teachers to facilitate and enhance students' second-language acquisition (Jenks, 2021;Walsh, 2011). Therefore, it should be part of their Interactional Competence (IC) (Atar & Seedhouse, 2018;Peachey, 2017), which is defined as the appropriate use of linguistic and interactional resources to achieve mutual understanding (Markee, 2008;Walsh, 2012).
On the other hand, studies have shown that in case of an epistemic gap, students also initiate CLA sequences (Hardt, 2018;Koschmann, 2016). Even though the TS (trouble source) was identified by the teachers, the students themselves repaired using different forms of repair strategies, thus enhancing their language skills (Novitasari & Imperiani, 2019). Furthermore, students cooperate to perform a repair (Nakayama, 2013). Cooperation is highlighted as a distinctive feature also in studies on informal teacherstudent talk. Once the purpose of the talk moves beyond controlled production of correct language forms, the interlocutors' roles and relationship shift from expert and novice to coparticipants in managing the talk regardless of their institutional role (teacher-student), which is relevant also from the point of view of classroom discourse analysis (Åhlund & Aronsson, 2015;Jenks, 2021;Nakamura, 2008).
As during the pandemic interactions between students and lecturers were video-mediated, this paper is also partly informed by studies on videoconferencing, which is a branch of video-mediated communication and has attracted increasing interest over the last decades. It is a multifaceted field of research ranging from language-based disciplines (linguistics, sociolinguistics, pragmatics) to social-ethno-anthropological disciplines (interactional sociology, discourse psychology and health studies) (Sindoni, 2014(Sindoni, , 2019(Sindoni, , 2020. For instance, since the introduction of virtual learning environments, videoconferencing has been investigated as a powerful tool of multimodal interaction in online language classrooms (Hampel & de los Arcos, 2013;Peachey, 2017) and suggestions to improve for activity design and implementation have been made (Hampel, 2019;Hampel & Pleines, 2013). This interest is due to the fundamental role of interaction in SLA according to the two main approaches to foreign language learning, namely the interactionist approach and the sociocultural approach, which inform studies on how new technologies can affect L2 classes (Hampel, 2019;Hampel & de los Arcos 2013;Hampel & Pleines, 2013). Research shows that the multiple modalities of the online videoconferencing environment can support learner communication and interaction, thus aiding SLA (Peachey, 2017). Therefore, the affordances of video-mediated communication should be considered while planning activities to foster student interaction (Hampel, 2019;Hampel & Pleines, 2013). More recently, research started to explicitly focus on educational environments from a multimodal perspective. For example, Sindoni investigated the epistemological, theoretical, and educational implications of mode-switching (henceforth MS) in an instructional context (Sindoni, 2020). By paraphrasing the notion of code-switching, Sindoni had already shown in her previous studies how interlocutors re-arrange verbal and nonverbal resources in an attempt to simulate face-to-face conversations and how they alternate speech and writing in video-mediated interaction.
With new proxemic and kinetic patterns, gaze management, and the impossibility of eye contact, the alternation of speech and writing is what makes video-mediated communication peculiar (Sindoni, 2014, see p. 5). This paper borrows the notion of MS to describe how lecturers and students used oral speech and messages in the chat window to achieve a common understanding.
Indeed, the chat is an essential aspect in studies on computer-mediated communication: it is a close digital match to a spoken conversation (Zitzen & Stein, 2004), which has its peculiarities in terms of turn allocation, disruption of interactional turns and coherence (Berglund, 2009;Garcia & Baker Jacobs, 1999;Herring, 1999Herring, , 2018Ong 2011). By investigating issues of concern in quasi-synchronous conversation such as confusion in turn allocation, overlapping in multiparty and multi-floor conversation, research shows that the chat is not a per se flawed form of interaction. By contrast, it has unique patterns which make it a differentlyabled effective kind of conversation (Herring, 2018;Ong, 2011). Interaction is a fundamental aspect of SLA (Hampel & de los Arcos, 2013;Walsh & Mann, 2015) and an issue of concern in distance learning (Anderson, 2003;Meredith, 2019). Therefore, it seems relevant to investigate how teachers and students adapted to online interaction during the Covid-19 emergency.
All the studies mentioned above account for interaction in digital education and refer to instructional settings that were planned and designed to accomplish that task. So far, investigations on ERE have provided an overview of the coping strategies adopted in different countries Peters et al., 2020). They investigated issues of concern such as the digital divide (Bozkurt & Sharma 2020;Hall, 2020), and analyzed students' and lecturers' perceptions of the challenges and opportunities of ERE (Erickson & Wattiaux 2021;Farrah & Al-Bakry, 2020;Giacosa, 2021b;Luporini 2020;Yoon, 2020). Concerning interaction in ERE, both lecturers and students faced challenges (Appolloni et al., 2021;Cicillini & Giacosa, 2020;Peters et al., 2020). Therefore, it seems appropriate for teachers to develop digital interactional competence as a necessary and timely professional skill (Giacosa, 2021a;Moorhouse, 2020;Moorhouse et al., 2021;Peachey, 2017). Indeed, the knowledge of communication breakdown dynamics could help mutual understanding between lecturers and students. It could enhance lecturer interactional competence and student communicative competence in their target language (Gardner, 2012;Walsh, 2012). Therefore, CLA as a repair mechanism to achieve intersubjectivity in a video-mediated instructional environment seems a relevant and original topic of analysis.

Conceptual framework
Addressing problems in conversation through repair is an indispensable part of social interaction and relevant for all the subjects, but essential for L2 learners (Atar & Seedhouse, 2018;Hampel & de los Arcos, 2013;Hardt, 2018;Novitasari & Imperianni, 2019;Pineda Hoyos, 2018;Walsh, 2012;Walsh & Mann, 2015). While learning how to communicate in a foreign language, students experience trouble in achieving mutual understanding. So, it is crucial to understand how the repair mechanism works to learn how to deal with problems in the classroom and the real world, online and offline. If CLA as a repair mechanism is used systematically, this can enhance interaction and learning in L2 classes Atar & Seedhouse, 2018). The impact of computer-mediated communication on interaction and meaning-making in the language classroom has been extensively investigated to unearth how learners negotiate meaning and co-construct knowledge through interaction in L2 classes (Hampel, 2019;Hampel & Pleines, 2013;Stickler & Hampel, 2015).
This interest is linked to two main theoretical approaches to language learning. On the one hand, the interactionist approach, a cognitive theory of SLA, focuses on interaction as the basis of language development and encourages more active involvement to achieve better results (Hampel, 2019;Hampel & Pleines, 2013). For example, by being exposed to both target-like and non-target-like forms during a lesson (in ERE oral utterances and written posts in the chat), learners may focus on discrepancies and be encouraged to selfcorrect or other-correct. On the other hand, this approach is informed by the social constructivist approach of learning as a situated social activity, which requires interaction with others (Hampel, 2019). In this view, students learn while interacting with others through digital tools. During the pandemic, cohorts of teachers and learners have become familiar with the affordances of the digitally-mediated environment for language learning, whose possibilities and constraints can potentially affect pedagogical action in both in-person and online classes. Therefore, it is worth analyzing how they interacted and collaboratively made meaning through the screen. The lesson learned during the pandemic could affect teaching and learning, both in-person and online. To this end, the following analysis will focus on the pattern of interactions involving the chat feature to identify CLA patterns and contribute to making meaning of wisdom learned in ERE to improve teacher interactional competence in digital instructional settings, as already highlighted in the previous section.
This study investigates CLA in English Linguistics and practical English ERE classes from the perspective of CA and multimodal analysis; it focuses on the coping strategies to manage TS in video-mediated interactions in EFL university classes during the pandemic. To this end, it uses the concept of CLA, a repair strategy adopted in conversation to correct an error made by a speaker or TS to achieve mutual understanding (Atar & Seedhouse, 2018;Drew, 1997;Hardt, 2018;Novitasari & Imperianni, 2019;Pineda Hoyos, 2018;Schegloff, 2007). In line with other studies on CLA and repair, this investigation draws on the distinction between self-repair (the trouble source is addressed by the speaker who utters the mistake) and other-repair (the trouble source is addressed by the speaker who utters the mistake (Novitasari & Imperiani, 2019;Schegloff et al., 1977). Moreover, it considers both teacher-led and student self-repair to investigate the dynamics of repair in video-mediated communication (Åhlund & Aronsson, 2015).
Studies on repair in a naturally-occurring conversation focus on interactional sequences where a turn is perceived as problematic and requires repair (Atar & Seedhouse, 2018;Drew, 1997). Regarding teacher-led repair, it can be noticed that teachers can have trouble understanding what students are saying; in this case, they can ask Type-specific Questions (TSQ), such as what do you mean by that?, which target the TS and locate it in the previous turn. If teachers have trouble hearing, they can use Open Class Repair Initiators (OCRIs), such as Pardon? Sorry? What? (Atar & Seedhouse, 2018;Drew, 1997). This open repair strategy addresses the whole of the prior turn as problematic (Atar & Seedhouse, 2018;Drew, 1997;Schegloff, 1992). If teachers have problems in both hearing and understanding, they opt for a stronger pattern made up of partial repetitions and question words (PR+WHs) (Atar & Seedhouse, 2018). At the same time, studies on classroom interaction have shown that repair sequences are also started by students, who contribute to mutual understanding by using all types of repair strategies identified by Schegloff et al. (1977) in their seminal study. Students can deal with trouble sources in a conversation by self-initiated self-repair (SISR), self-initiated other-repair (SIOR), otherinitiated self-repair (OISR), and other-initiated other-repair (OIOR). In EFL classes, not only can they show their proficiency by initiating a CLA sequence, but also enhance their communicative skills to deal with trouble in interaction, which is valuable while speaking a foreign language (Atar & Seedhouse, 2018;Novitasari & Imperianni, 2019).
As during the pandemic classroom interactions occurred in a video-mediated environment, this study focuses on CLA sequences involving MS, that is the alternation of speech and writing in videoconferencing. Paraphrasing the linguistic concept of code-switching, Sindoni uses the term MS to describe "the alternation from speech to writing and vice versa in the same communicative event, with a general prevalence of the spoken mode with written insertions for specific communicative purposes" (Sindoni, 2020, p. 2). In multipartyvideo interactions, MS is mainly used to manage the conversation flow in terms of repair and self-correction, for specific communicative purposes and to address technical issues (See p. 3).
To analyze CLA sequences involving MS, a concept typical of CA will be used: the disrupted adjacency pair. It is a typical phenomenon of text-based communication, which occurs in multi-participant interactions because messages in the chat appear linearly, in the order in which they are received by the system. A response (the second pair part of an adjacency pair, henceforth SPP) may be separated in linear order from the previous message (the first pair part of an adjacency pair, henceforth FPP) it is responding to if another message or messages happen to have been sent in the meantime (Herring, 1999;Ong, 2011). It makes it difficult for interactants to establish coherence between the messages, thus hindering mutual understanding. Therefore, lecturers must be aware of CLA dynamics in video-mediated communication to improve interaction in their EFL classes not only to achieve mutual understanding but also to scaffold their students to deal with troubles in interaction (Atar & Seedhouse, 2018).

Methodology
This paper deals with the micro-analysis of both teacher-and student-initiated CLA in English Linguistics and practical English ERE classes from the perspective of CA. It focuses on the interplay of oral and written speech in video-mediated communication. To this end, it deals with exchanges involving the chat window as a crucial element of CLA and employs the concept of MS, which is typical of multimodal studies (Sindoni, 2014(Sindoni, , 2019(Sindoni, , 2020. However, this article concentrates on language and does not report on multimodal aspects such as gestures, gaze, postures, framing. Therefore, it does not adopt a multimodal conversation analytical approach, which focuses on both language and gesture and considers the former as one crucial element among others (Mondada, 2019). By contrast, it carries out a qualitative analysis of oral and written exchanges between teachers and students dealing with aspects typical of CA, such as turn organization and repair, as highlighted in the previous section. Moreover, it also considers MS, an element borrowed from multimodal analysis, but it focuses on logocentric aspects, namely the alternation of oral and written speech (see previous section). The lack of microanalysis of CLA in ERE and the combination of CA with an aspect of multimodal analysis make the perspective of this study original.
The study uses qualitative analysis since in previous studies on CA the qualitative analysis has proved to be the most suitable approach for the micro-analysis of interactional phenomena to address research questions exploring how certain phenomena occur (Atar & Seedhouse, 2018;Sindoni, 2020). Moreover, microanalysis is a recommended approach while dealing with the sequential organization of interactions in CA (Giles et al., 2017;Jenks, 2021). Even though the results can hardly be generalized, it seems relevant to describe the phenomena that can trigger teacher awareness and increase their interactional competence thus facilitating interaction and learning in L2 classes (Markee, 2008;Walsh, 2012). With these reflections in mind, this study has adopted a qualitative approach to the micro-analysis of data.

Research context and participants
This paper draws on data collected for a PhD dissertation on interaction in ERE English Linguistics and practical English ERE classes. Data was collected through the participant observation of online synchronous 2hour classes taught via videoconferencing tools (WebEx, Microsoft Teams and Zoom) in the spring and autumn of 2020 in 14 Italian universities, as already mentioned in the introduction. At the beginning of the pandemic, English Linguistics and EFL lecturers were randomly selected from the official webpages of several Italian universities to have an overview of ERE classes taught using different tools. 27 out of over 300 who were contacted participated in the study and allowed the researcher to join their online classes and collect data about them. Regarding English Linguistics classes, they were attended by 70-130 students and were taught in English by an Italian lecturer. Practical English classes focused on the English language, were attended by 30 students and taught by native speakers. This article will not deal with differences between the two kinds of classes, because it would go beyond its scope. Concerning our informants, 96% of the lecturers kept their cameras on and shared their screen with the students to show their slides, internet webpages or files they had prepared in advance. Few (4%) used the interactive whiteboard provided by the videoconferencing tool, whereas 100% typed at least a message in the chat window. As this paper draws on an ongoing study, data is provisional. Nonetheless, as the situation is constantly changing due to the regulations to curb the virus, it is fundamental to take a snapshot of the first year of ERE to gain an insight into changes, opportunities and challenges involved in ERE.

Data collection
Data was collected through participant observation of online classes, a method increasingly adopted in qualitative research (Kawulich, 2005;Sindoni, 2020). The affinity of CA and ethnographic methods such as participant observation has been identified as one of the emergent methods to ask and answer new research questions (Hesse-Biber & Leavy, 2008;Kyprianou et al., 2015). However, participant observation is not widely adopted for data collection in CA research since it relies on recordings and video recordings (de Kok, 2008). As far as this research on ERE is concerned, the constraints of video recordings required the researcher to adapt theoretical research methods (presence or absence of the observer) to these new forms of online lessons. Though very useful for integrating field notes, the researcher could not rely solely on recorded classes. In contrast to the streamed ones, they did not display the chat interactions simultaneously with the audios. As the objectives of the present study necessitated tracking how the oral and written modes interact instantaneously, participant observation was adopted as a data collection method. Therefore, even if CA methodology has an emic perspective, participant observation was chosen as the most suitable way to collect data for this study.
While observing the lessons, the researcher benefitted from the same affordances as the other attendees. The researcher could see the public chat window and could send and receive private messages but did not have access to the lecturers' private interactions with other attendees. This is not considered a limitation to the study, as the focus is on interactions in the chat involving the lecturers and the class. To limit unspontaneous behavior, neither the lecturer nor the students knew that the focus of the observation was on the interactions. At the beginning of the lesson, the researcher was introduced to the attendees. They were informed about the presence of a participant observer and accepted to attend the class. The researcher filled in a grid to collect general information about the setting (number of attendees, the main topic dealt with, resources and tools), and took notes about the interactions between the lecturer and the students by writing down the verbal interventions and copy pasting the posts from the chat window. While attending the online classes, the researcher took notes about the interactions and copied the messages from the chat. This study relies on four 2-hour-lessons, that is three English Linguistics lectures and one practical English class to mirror the dataset collected during the observations (3/4 lectures and 1/4 practical English classes 2 ); each of the analyzed classes was taught by a different teacher in spring 2020 (see the table below). This article reports on the CLA sequences which involved the chat and occurred during the lesson. Since the aim is to provide a qualitative analysis, quantitative data will not be discussed as they go beyond the purpose of the present paper. Regarding the amount of data analyzed in this paper, an 8-hour dataset is in line with studies on CA, which recommend 5-10 hour datasets as the required amount for reliable studies (Atar & Seedhouse, 2018). Fourteen extracts composed of interactions involving the chat were identified and analyzed using a micro-analytic lens. Nine extracts were used in this paper to provide a preliminary overview of teacher-and student-initiated CLA involving the chat.

Data transcription and analysis
To increase data reliability, the researcher integrated the manual transcription with missing details from the lesson recordings, which were available on the Moodle course page and provided the timing and the exact content of the spoken utterances. The posts reported in the transcription objectively correspond to written intervention and are introduced by the symbol #. Identifying elements were anonymized to comply with privacy regulations.
Then, the manual transcription of the interactions was divided into sequences regarding the management of TS. Only the exchanges involving CLA are analyzed. After their identification, CLA sequences were analyzed by locating the initiation (CLI), the repair and the closing of the sequence (CLS). Finally, the repairs were analyzed by using the concepts self-initiated self-repair (SISR), self-initiated other-repair (SIOR), other-initiated self-repair (OISR), and other-initiated other-repair (OIOR) (Schegloff et al., 1977). At the same time, turns were analyzed by identifying MS, and the components of the adjacency pairs (FPP, SPP, see p. 5), and disrupted adjacencies (Herring, 1999;Ong, 2011;Sacks, Schegloff, & Jefferson, 1974). Exchanges uttered in the speakers' L1 (Italian) are reported in the original version, translated into English and written in brackets. The annotations of the researcher on relevant aspects of the interaction are written in italics and reported in brackets 3 .

Findings
In this section, preliminary qualitative results on conversational repairs in ERE classes will be presented regarding the two research questions posed earlier in this article.

How conversational repairs operate in ERE
The transcriptions of ERE classes show that the chat plays a fundamental role in maintaining mutual understanding between lecturers and students in many respects.
First, students could start CLA sequences by typing in the chat to reach out to the lecturer for help while experiencing troubled hearing and/or seeing, as shown in extract 1. The extract shows an example of OISR, where Student1 initiated a CLA sequence by describing a problem, rather than asking a question. Almost synchronously, the lecturer saw the question and addressed it orally by partially repeating the student's sentence. Then he provided a suggestion to fix the problem and a comment. It is not clear whether the TS was produced by the lecturer or due to a technical problem. However, the lecturer provided the repair to respond to the CLA initiated by the student. In ERE, a very frequent TS regards the quality of the rendering of shared materials on students' devices. It often happened because the lecturers did not get any form of automatic feedback from the videoconferencing tool on the poor reception of the displayed and shared material. So, this example shows that the role of the chat is twofold. On the one hand, it facilitates learning and teaching. On the other hand, it provides valuable feedback to teachers, thus allowing them to help students. Though incomplete, data shows that a CLA sequence plays a crucial role in restoring mutual understanding, as displayed in Extract 2.
Extract 2. (Lesson 2, lecture, teacher B, 123 students) In this extract, the lecturer was explaining new concepts by referring to the slides, but he did not notice that the slideshow was stuck. He continued referring to slides that the students could not see, thus making it difficult for them to follow the lesson (TS). At this point, two students simultaneously started a CLA by writing in the chat (lines 4 and 6). Even if it took one minute for the lecturer to notice the messages, the chat window played a crucial role in re-establishing mutual comprehension. After reading the postings, the lecturer fixed the freeze (line 7) and performed an OISR by repeating the concepts and showing the slides that the students had not been able to see. Then, the lecturer introduced new notions without asking for feedback on the repair. This CLA sequence has no closure, but it is interwoven into the rest of the lesson. Nonetheless, similarly to extract 1, this incomplete CLA sequence fostered an epistemic change (Heritage, 1984) by moving from K-(not Knowing) to K+ (Knowing) (Heritage, 2012). After that, the lecturer could cover the planned steps of his lesson.
Second, lecturers could start CLA sequences by asking questions orally and additionally typing messages in the chat. While streaming classes, it was not easy for lecturers to collect feedback from their students, which caused insecurity. When they sensed a TS, they felt compelled to interrupt the lesson and reach out to students orally while writing a message in the chat, as shown in extract 3. In this example, the lecturer was not receiving feedback from students and assumed they were having problems following the lesson. First, they orally asked and then wrote a question in the chat (MS) to address a possible TS, thus initiating a CLA. It is an example of SIOR, where the teacher sensed a TS and started a CLA sequence while a student provided a repair. Student 1 reassured the lecturer about the maintenance of intersubjectivity, while Student 2 orally addressed the TS. By accident and unknowingly, the lecturer had muted and then unmuted the participants; the repair provided by Student 2 made the lecturer realize what had happened, thus targeting an epistemic gap. The lecturer replied orally (MS) by using the change-of-state marker "Oh", which is a strong indicator of epistemic change (Heritage, 1984) from K-(not Knowing) to K+ (Knowing) (Heritage, 2012). In this extract, the lecturer used the chat as an additional resource to identify TS and reach out to students to maintain intersubjectivity. So, even if the oral channel is the preferred means for lecturers to ask for a repair, the chat can help restore mutual comprehension. In turn, for students, the chat appears as an effective way to provide feedback and actively participating in the lesson, even though oral speech is a quick and decisive means of providing repair and feedback. Third, CLA sequences involving the chat can be complex. For example, they can contain disrupted turn adjacency, caused by the fact that messages appear in the order received by the system, without regard for what they are responding to (Herring, 1999) The teacher asked a question verbally and Student 1 answered in the chat. Her answer was split into two turns, which shows that the student added a part of the answer after sending the message. At this point, the teacher looked at the chat, noticed and addressed orally a previous question a student had written in the chat (disrupted adjacency pair). Then the teacher read out Student 1's answer, identified it as a TS, and initiated a CLA by asking a type-specific question (TSQ) and eliciting an OISR. Student 1 activated her microphone to provide a verbal repair (MS referring to her previous turn). The teacher closed the CLA by giving feedback. While the teacher was talking, two messages appeared in the chat window. Another student (Student 2) wrote a comment in the chat whereas Student 1 closed the sequence by typing a thank you message (MS and disrupted adjacency pair).
Finally, messages in the chat could contain a TS, which required CLA, as shown in extract 5. In this example of OISR, the CLA sequence occurs in a multiparty and multi-floor conversation, which shows how complex interactions can be in video-mediated communication. In line 1, the lecturer asked a question addressing the whole class. Student 1's answer was not consistent with the question (line 2), thus creating a TS. In line 3, the lecturer initiated the CLA sequence by partially repeating the question and locating TS. This utterance is the first part of an adjacency pair (FPP). Even if CLI was addressing Student 1's answer, other students answered. Their posts (lines 4, 5, 6) appeared before Student 1's repair (line 7), namely the second pair part of the adjacency pair (SPP), thus providing an example of disrupted adjacency turn. As other messages appeared between Student 1's repair (line 7) and the lecture's oral closing of the CLA sequence (line 13), we could talk about a disrupted CLA sequence. CLS (line 13) has a double function: on the one hand, it provided feedback to the students' answers; on the other hand, it closed the CLA. It seems relevant that the chat allows students and lecturers to interact while using two modes: the lecturers interacted orally while the students typed in the chat (written mode).

To what extent the chat, as a means of intersubjectivity, affects student behavior in ERE
The extracts analyzed in the previous section shed light on how CLA sequences involving the chat can operate to maintain or restore mutual comprehension in ERE classes. The examples described in this section will delve into some of possible effects of the chat window as a means of intersubjectivity on student behaviour in ERE classes.
First, the possibility to address TS by typing in the chat box can encourage student participation. Extract 6 shows that the chat allows students to actively contribute to mutual understanding during online classes by integrating the lecturer's repair. Student 1 was reading a sentence (line 2) when she came across a TS (the word invoice). Within the same turn, she initiated a CLA sequence (sequence 6a) by asking orally a Type-specific question (TSQ) (what's invoice?, line 3) and the lecturer provided repair by translating the word into Italian (line 4). As the perceived TS was in the book, we will arbitrarily consider it related to the teacher and this sequence an example of OISR. Student 1 closed the CLA sequence by apologizing and explaining the TS (line 5), which initiated another CLA sequence (sequence 6b), namely an SIOR, as the lecturer provided a repair (line 6). However, the repair performed by the teacher (line 6) was partly incorrect and perceived as a TS by Student 2 (sequence 6c), who provided a repair in the chat (OIOR in line 7). This CLA sequence is closed by two interventions. The lecturer repeated Student 2's repair providing an example of OISR, while Student 2 typed an interjection and the adverb okay in the chat, typical of the oral exchanges. Therefore, a CLA sequence can benefit from different repair strategies, which shows the crucial role of the chat in intersubjectivity, even though it does not play a leading role.
Conversely, when lecturers do not address the CLA required by a student, this can negatively affect student participation and involvement, as shown in extract 7. In this extract, the lecturer asked to translate a phrase (line 1) and received an oral answer from Student 1 (line 2), which he perceived as partial. So, he encouraged the student to provide the complete sentence by either writing in the chat or speaking up again (lines 3-4, FPP). At this point, different messages appeared in the chat window, as other students typed their answers (lines 5, 6, 7, 8, 9, 10), Student 1 included (line 11, MS). Due to the order of appearance, Student 1's answer (SPP in line 6) was preceded by other messages (disrupted adjacency pair). In lines 13 and 14, Student 9, who perceived the phrase in line 1 as a TS, initiated a CLA sequence (OISR) with a Type-specific question (TSQ) for the lecturer (what does mean "a mala pena"?). Then he apologized and provided further information to clarify his question. The lecturer, who was sharing his screen, overlooked the question and carried on with the lesson without repairing it. So, this CLA sequence is incomplete since no repair was provided. As annotated by the researcher, Student 9, who had typed a message every ten minutes before this turn, did not type any more messages. Even if technical issues cannot be ruled out, it is relevant to notice the negative effect of the overlooked question on Student 9's participation and involvement.
Second, the possibility to provide repair through the chat increases the cooperation among students and with the lecturer, as shown in extract 8. This extract shows that while not sharing the same environment, TSs can be unexpected and not related to the lesson content. The lecturer orally announced a 10-minute break and switched off his camera (line 1). Student 1 initiated a CLA sequence by typing a Type-specific question (TSQ) on the chat addressing the whole class (line 2). Student 2 provided a repair-OIOR (line 3)-and Student 1 closed the sequence by typing a reaction (ah ok), a change of state token which typically shows an epistemic change from K-to K+ (Heritage, 1984(Heritage, , 2012. Moreover, Student 1 typed a thank you message and an emoticon (the red heart, line 5). At that point, the lecturer realized there had been trouble understanding and announced the break in the chat, thus providing a repair through MS (line 6, OISR). Students were active in initiating CLA sequences and providing repairs, thus helping lecturers to achieve intersubjectivity. At the same time, the sequence shows that thanks to the chat, CLA can be both OIOR and OISR, thus revealing the complexities of repair in video-mediated communication.
Finally, students use the affordances of the videoconferencing tool like emoticons to compensate for the distance. Finally, given that words fly, and writings remain, students seem willing to make their messages clear by self-correcting their mistakes in future posts, as shown in extract 9. It is an example of SISR, where Student 1 reached out to the lecturer (line 1) regarding the topic of the class (Corpus Linguistics applied to literary texts): he wanted to join in the group dealing with Shakespeare. The interpretation of TS is debatable. The student might have referred to a spelling mistake (plural instead of singular, line 1). Alternatively, they might have felt compelled to avoid a misunderstanding as they exclusively referred to the previous lesson. Regardless, Student 1 perceived the post as a TS and performed a SISR by typing the correction and an apology without waiting for a CLI. It is a self-initiated, self-completed, same turn repairs, an element described in both face-to-face conversations (McHoul, 1990;Schegloff et al., 1977) and computermediated communication (Garcia & Jacobs, 1998;Meredith, 2019;Schönfeldt & Golato, 2003). A few minutes later, the lecturer read the message and addressed it. Even if the lecturer might not have detected a TS in the student's post, it is relevant that the student used the chat to ensure intersubjectivity.

Discussion
Regarding the first research question (How do conversational repairs operate in ERE?), this study shows that the chat helped speakers reach intersubjectivity in many respects. Concerning the repair initiator, data shows that both students (extracts 1, 2, 6, 7, 8) and lecturers (extracts 3, 4, 5) started a CLA 9 . Classroom conversation in emergency streamed classes appears to be a multiparty collaboration, as both parties contribute to maintaining or restoring intersubjectivity. It is in line with studies on SLA in brick-and-mortar classrooms (Åhlund & Aronsson, 2015;Koschmann, 2016;Mc Houl, 1990) and online environments (Licoppe & Morel, 2018;Melander Bowden & Svahn, 2020;Meredith, 2019;Meredith & Stokoe, 2014), which question the predominance of teachers in CLA sequences and shed light on student agency. Indeed, during a streamed lesson, students could start a CLA sequence by typing comments and questions in the chat (for example, extracts 1, 2) or orally asked for CLA (extract 6). In turn, lecturers could also exploit the different affordances of the medium and use written and oral speech, even if data shows a preference towards oral CLA. They preferred starting a CLA orally (extracts 4, 5) and used the chat as an additional means of communication (extract 3). By contrast, students preferred to start a CLA by typing a question in the chat, which allowed them to ask for clarification without interrupting the lesson (extracts 1, 2, 6, 7, 8). The possibility of initiating a CLA via chat had the additional benefit of providing lecturers with feedback on what was shown or played (extracts 1, 2). As lecturers and students did not share the same physical environment, the former did not have any information about the quality of hearing and seeing on the devices of the latter. So, mutual comprehension was enhanced thanks to postings in the chat as they provided lecturers with feedback and allowed them to address TS.
Regarding the type of repairs analyzed in this study, self-repair is preferred to other-repair: in the examples, self-repair (either OISR or SISR) occurred in all the extracts but one (extract 3). In line with studies on CLA (McHoul 1990;Schegloff et al. 1977;Schönfeld & Golato, 2003), OISR appears to be the main pattern of interaction in ERE. Concerning the repair organization, data regarding ERE is in agreement with the outcomes of previous studies on computer-mediated communication (Herring 1999;Schönfeldt & Golato, 2003). The constraints of the medium have a considerable impact on the turn-taking system. For example, each message is considered a turn, which makes the concept of transition relevance place (TRP) inapplicable (Meredith & Stokoe 2014;Schönfeldt & Golato, 2003). As interactants have no control over the order of appearance of their posts, also the notion of next-turn repair is hardly applicable in this context. Furthermore, the same TS can generate different types of repairs performed by different interactants using different modes. In extract 6c, for example, the repair is performed by a student via chat (OIOR) and by the teacher orally (OISR). In extract 8, the repair is performed by a student (OIOR) and the teacher (OISR) via chat. The differences between mundane and online interactions go beyond the scope of this paper. What is relevant is that, despite the difficulties and the peculiarities of the new setting, the extracts show that TSs were addressed but in one example, and mutual comprehension was restored.
Concerning the repair performer, it can be noticed that also in this case both teachers and students provided repair. As far as students are concerned, they provided OISR in extracts 4 and 5. Extract 4 is particularly revealing, as the student accepted the lecturer's suggestion and orally repaired to clarify a TS that he had produced by writing in the chat. It confirms lecturers' preference for oral repairs and the supporting role of the chat in repair. Lecturers identified TS in the chat but encouraged students to orally self-repair as they would not be hindered by the constraints of computer-mediated communication and could express themselves more extensively. Despite their preference for oral repairs, lecturers adapted to the setting and used the chat to clarify, as shown in extract 8. This observation is in line with studies on repair in online environments, which underline that online interaction is an adaptation of an oral speech-exchange system (Meredith, 2019;Meredith & Stokoe, 2014). Participants adjust the repair practices from ordinary conversation to suit the affordances and constraints of the online medium (Schönfeld & Golato, 2003).
Regarding the distribution of MS, it can be noticed that there are six occurrences of mode-switching out of 58 turns (10.3%). This rate is lower than the findings of Sindoni's studies, which suggest that the average rate of mode-switching is between 11% and 15% (Sindoni, 2020). This difference can be explained by the fact that Sindoni investigated interactions either in informal contexts or in an instructional context but within a study on MS, which might have produced a higher number of occurrences. In line with Sindoni's studies on MS, it appears that "mode switching is neither casual nor erratic in these texts. This does not mean that it has been possible to determine specific and recurrent patterns of mode switching so far, but this notion has proved useful in explorations of the spoken/written continuum in digital habitats" (Sindoni, 2014, p. 332). Indeed, from the preliminary findings on ERE presented in this study, it can be observed that both lecturers and students performed it. The former mode-switched from oral to written speech to initiate a repair (extract 3) and provide it (extract 8). In the first case, the lecturer mode-switched to written speech to support a CLI performed orally and then went back to the oral mode to close the CLA sequence. In the second case, the lecturer mode-switched to adapt to the mode chosen by students. Both examples confirm that the lecturers use the chat to support their oral utterances and not as a primary means of CLA. Regarding students, extract 4 contains two occurrences of MS. The first one is from written to oral speech. The student wrote something that the lecturer perceived as a TS and was asked to explain it orally. After the CLS was orally performed by the lecturer, the student thanked by typing in the chat (MS oral > written). This example confirms that the preferred means of CLA is the oral one. Extract 7 is an example of MS from oral to written; in this case, the lecturer encouraged the student to use an adverb in a complete sentence either orally or in the chat. At this point, the student decided to mode-switch and used the chat to write a full sentence. Further research is needed, but it is worth observing that the chat appears to be a useful tool for short answers, but it is not suitable to provide longer explanations.
Even though data shows that the chat can facilitate interaction and mutual comprehension, its peculiarities encourage students to choose conciseness over clarity. Since the chat embedded in the videoconferencing tools does not indicate if someone is typing a message, the interlocutors tend to be concise to answer as soon as possible to allocate their turn to detriment of clarity, as already shown in studies on computer-mediated communication (Herring, 1999(Herring, , 2018Schönfeld & Golato, 2003). For this reason, the messages in the chat can be a TS, as shown in extracts 4 and 5. At this point, the lecturer tended to orally initiate a CLA sequence by reading out the trouble source and ask a question to help the student clarify (Pr+Whs), in line with the findings on CLA in EFL face-to-face classes (Atar & Seedhouse, 2018;Schegloff, 2007). Extract 5 shows that not only the addressee answered, but also other attendees wrote the repair in the chat. This is typical of multiparty communication, which makes it challenging for the participants to identify pair adjacency (Herring, 1999;Meredith & Stokoe, 2014;Ong, 2011). Concerning CLA, it requires an effort from the lecturer to identify the repair when it appears in the chat with other messages (extract 5). Moreover, in multiparty conversations involving the chat, lecturers can miss information, which is crucial in the case of a CLI (extract 7), as it fails to address a CLA, thus affecting mutual comprehension. Finally, it can be time-consuming and distracting to manage the interplay of oral and spoken conversation. As extract 2 shows, the time-management of an online lesson can be affected by technical issues; the lecturer performed the repair but then went on with the explanation without closing the CLA sequence. Even though CLAs could be incomplete, what matters is that mutual comprehension could be achieved in ERE. Indeed, this study's data confirm that in ERE classroom, too, interaction is not per se flawed, but just a differently-abled kind of conversation (Herring, 1999;Ong 2011). However, further studies could compare the rate of failures in CLA in ERE and face-to-face environments to have a deeper insight into possible problems and solutions regarding the former.
Concerning the second research question (To what extent does the chat as a means of intersubjectivity affect student behaviour in ERE classes?), the possibility to write in the chat window actively involved other students in the CLA sequence, thus increasing their participation at various degrees. For example, in extract 1 a student having trouble hearing started a CLA by typing in the chat, thus actively contributing to ensure his understanding. In extract 2, students reached out to the lecturer starting a CLA via chat to restore intersubjectivity for the whole class. As one of the feelings reported by the lecturers was the sensation of talking into an empty space, these examples show that ERE classes were not deserted places per se. They could be made livelier and more interactive by encouraging the students to use the tools embedded in the medium, as will be discussed in the following section. Despite not being able to always collect feedback from students' facial expressions, extract 3 shows that in ERE, too, lecturers could activate student participation by asking them questions, as it happens in a face-to-face class (Jenks, 2021;Wong & Waring, 2020). In this setting, they could offer students more options: they encouraged them either to write in the chat or to unmute the microphones and speak up (extract 7) (Giacosa, 2021a). This option dramatically increased student participation in terms of the quantity and quality of the interventions. Regarding the former, extracts 5 and 7 show that an open question to the class could get several answers. For example, the question in extract 7 ("a mala pena", how can you translate it?) receives eight answers and a CLI. It would be interesting to compare in further studies the answer rate regarding a similar question to quantify the impact of the chat in terms of the number of responses to an open question. For the scope of this qualitative study, it is relevant to notice that in virtually streamed classes the presence of the chat positively affects student participation. Concerning the quality of student participation, it seems relevant that despite the constraints and challenges, students could actively participate by commenting on the topic via chat, as in extract 4 (line 2-3 and 14-15). Even if messages are concise, students wrote comments in the target language by using complete sentences, which is valuable from the SLA point of view (Atar & Seedhouse, 2018;Walsh, 2011). Therefore, lecturers should be aware of the importance of the chat to encourage student participation thus maximizing their opportunities to interact also in the foreign language. As extract 7 shows, when lecturers failed to address a CLA, this could have consequences on a student's participation.
Second, the possibility to write in the chat encouraged the collaboration with other students and the lecturer to achieve intersubjectivity. Data shows that despite the challenges and constraints, students actively participated in ERE classes by helping target epistemic gaps for the lectures and other students' benefit. This complies with the socio-cultural approach to foreign language teaching and learning, which highlights the benefits of co-constructing meaning in L2 classes (Hampel, 2019). Compared to a face-to-face lesson, the chat virtually increased the number of possible interlocutors: if in a physical classroom a student would ask his/her desk mate or a nearby person for CLA, by writing in the chat Student 1 was virtually addressing any other attendee (there were 123 participants in the lesson plus the lecturer). Even though Student 2 had not been specifically addressed by Student 1, he/she felt compelled to provide repair. Moreover, by using the emoticon Student 1 could compensate for the lack of facial clues in the digital setting and make the message more personal, as shown also in other studies on ERE classes (Luporini, 2020). At the same time, the chat encourages more interlocutors to provide repair as is the case in extract 6, where the repair was provided by both the lecturer and a student. While the former orally provided a repair, the latter typed a repair in the chat. So, without interrupting the lecturer, a student used the chat by integrating the lecturer's answer thus collaborating to tackle a TS.
Finally, as extract 9 shows, the possibility to write in the chat encourages students to self-correct both in terms of accuracy and mutual comprehension. Even if the example is debatable, as the motivation of the student for self-correcting is not clear, it is relevant that students care for the accuracy and consistency of their messages and self-correct. This aspect, which is in line with studies on computer-mediated communication (Meredith & Stokoe, 2014;Schönfeldt & Golato, 2003), seems relevant from an SLA point of view, as it could be exploited to encourage student attention towards accuracy and clarity in the target language, two fundamental aspects of SLA (Walsh, 2011;Wong & Waring, 2020).

Pedagogical implications
From the analysis of the extracts, it can be assumed that teachers should encourage students to use the chat by asking questions and eliciting feedback to double-check their understanding. The chat maximizes the possibilities for students to ask for CLA as they can write their questions without interrupting the lesson, which shy students can find handy. The questions in the chat can provide feedback and help the lecturer identify trouble sources regarding the contents and provide repair, namely OISR (extract 2). Furthermore, as the posts are often in the target language, lecturers can use them to help students focus on common mistakes regarding the English language and even ask them to self-correct. It is in line with studies on the role of interaction in SLA: students can learn in interactions by being exposed to target-like and non-target-like forms to understand the difference, thus improving their proficiency (Hampel 2019;Hampel & Pleines, 2013). As studies on computermediated communication show (Herring, 1999;Zitzen & Stein, 2004), it is not infrequent that students and teachers amend their own mistakes by rewriting the correct version of their post and adding an asterisk * (SISR). The chat can also be used for pedagogical purposes by asking students to provide answers and translations; lecturers could read them out and correct them (OIOR, extracts 6c and 8) or use them to elicit students' self-repair (OISR) (extract 5). To achieve the same result in a face-to-face class, the lecturer would have to ask single students either to write their answers on a blackboard or to show him/her their notebook, which would be more time-consuming. By contrast, in an online lesson, lecturers can copy students' answers from the chat and paste them into a shared document; later, they can provide learners with lists of common mistakes and corrections. The pandemic has made every teacher and student more familiar with the affordances of digital platforms which could enhance CLA and repair during face-to-face classes as well. ERE has blurred the borders between brick-and-mortar universities and online environments: for example, both students and lecturers could benefit from the possibility to write questions and answers in a shared chat during an in-person class as well, to collect and give feedback, provide and elicit CLA and repair.
Data shows that providing students with more options to interact (orally and writing in the chat) in the target language increases their participation and engagement during the lesson impacting their proficiency (for example, extract 7). It is especially crucial in higher education, as classes can be large and the possibility for students to interact in L2 is limited (Aoumeur, 2017;Todd, 2013). Moreover, by encouraging students to write their contributions in the chat, lecturers can increase cooperation. In turn, by typing messages in the chat, not only can students ask for CLA, but they can provide OIOR, thus actively contributing to achieving mutual understanding for everyone's benefit (extract 8). It can improve the classroom atmosphere, which is crucial in L2 classes, where students need to feel at ease to express themselves in a foreign language (Hampel & Pleines, 2013). Interactions in the chat can be more informal, which encourages students to participate more actively (Hardt, 2018;Luporini, 2020). Moreover, student involvement in chat interactions can help teachers to maintain mutual understanding by tackling TS. It partially relieves the teachers of being the sole CLA providers, which is challenging in an online environment, as shown in this study. This aspect is relevant in the social constructivist view of language learning (See p. 4). In an emergency remote class, lecturers and students cooperate and co-construct meaning while interacting to achieve mutual understanding through the screen, adjusting to the constraints and affordances of the medium. As the extracts show, the chat is a powerful tool that allows teachers and students to cooperate and contribute to meaning-making. It seems a relevant aspect of video-mediated communication to be treasured for face-to-face classes as well.
In line with studies on new technologies and SLA (Hampel, 2019;Stickler & Hampel 2015), the extracts show that the chat feature is an effective tool to reduce the teacher talking time and tutor dominance. Thanks to the chat, English Linguistics lectures can be more similar to a conversation, even if this can involve speech and written messages. Provided that the lecturer accepts students' contributions, lessons could be more interactive and learner-centered, as studies on L2 classes recommend (Hampel, 2019). Moreover, students will have to engage by using the foreign language more actively than in a traditional in-person lecture. After typing a message in the chat, they can double-check for accuracy, as they know it will be on display and everybody will read it. Unlike in face-to-face classes, where proficient students are likely to interact the most (Atar Seedhouse, 2018;Hampel & de los Arcos 2013), the chat in video-mediated communication encourages everybody to express themselves in the foreign language. As the written message can be a trouble source (extract 9), students try to be as clear as possible to make the readers understand their point. If the message is not clear, the lecturer can ask the student who provided the trouble source to activate their microphones and provide repair (OISR) by expressing themselves more clearly in the foreign language (extract 4). As shown in studies on SLA, this will positively affect students' proficiency by providing them with tools and habits to deal with misunderstandings and CLA in L2 (Atar & Seedhouse, 2018;Walsh 2011). Moreover, the genuine need to achieve mutual understanding by using a foreign language is very close to the actual needs of a real conversation. By making the lecturer notice that they are having connection problems or asking the lecturer to provide CLA, students train their skills to interact in L2 for their needs. It scaffolds their abilities and empowers them as speakers (Atar & Seedhouse, 2018;Kasper & Wagner, 2011;Walsh, 2011). Therefore, lecturers should encourage them to actively participate in the lesson by writing messages in the chat.
However, data shows that lecturers need to adjust the timing of CLA to the medium. The most common videoconferencing tools such as WebEx, Zoom, Microsoft Teams and Google Meet do not provide information about messages being typed (Sindoni, 2014(Sindoni, , 2020. So, lecturers should give enough time to students to ask for CLA or provide repair (either self-repair or other-repair) by typing messages. It could help students provide more complex postings with main, subordinate clauses and linkers, thus positively impacting their proficiency. As questions in the chat can be distracting, they could agree with their students on how to deal with them. For example, the chat could be used during the lesson for urgent matters, such as asking for CLA or providing repair, or for more complex answers, whereas simple feedback such as "yes" and "no" could be provided by using the reactions button (for example, thumbs up or down). Moreover, lecturers could plan to devote a part of the lesson to encourage students' contributions and questions and scroll down the messages from the chat to provide repair. As shown in extracts 5 and 7, this is important to involve students more actively. Besides, it could offer the possibility to tackle linguistic issues concerning vocabulary or grammar consolidation. This paper argues that the possibility to ask for CLA via chat could be offered also in face-to-face classes to integrate the feedback provided by students' facial expressions and their whispering. Moreover, it could help teachers have a quick overview of the trouble sources and difficult points, to keep track of questions and trouble sources, for which they could provide repair at a later time.
Finally, the extracts show lecturers read postings and paste links in the chat, but they prefer providing CLA orally. Even if more extensive investigations are needed, the extracts show that lecturers do not paste explanations or definitions from their slides in the chat. As it is an interactive space, where materials can be exchanged, it seems that lecturers should use the chat more to provide examples and explanations from the slides. Like in a face-to-face class, the attendees of a video-mediated lesson can read what is displayed, but they cannot copy and paste sentences from the screen into their notes. Even though lecturers usually upload their slides on their webpage before or after the class, students could benefit from the possibility to copy sentences from the chat into their devices to focus on relevant points and consolidate their knowledge.
Extracts from ERE classes discussed in this paper show that the pandemic opened the traditional faceto-face classroom revealing to cohorts of teachers and students the opportunities of video-mediated interaction for educational purposes. Not only had the students more options to interact and prove themselves in the foreign language, but they actively contributed to meaning-making even more than in face-to-face classes, thus limiting tutor dominance. Studies on SLA had already highlighted the importance of including new technologies in SLA to adopt a more learner-centered approach and involve students more actively, thus facilitating learning (Hampel, 2019;Sert, 2019;Walsh, 2011). For example, besides regularly checking the messages, lecturers should incorporate the chat into their teaching, by encouraging students to use it and even planning activities involving it. Despite the challenges and complexities, ERE has revealed ways to make L2 lessons more interactive which should be incorporated in L2 classes, both in-person and online. Therefore, instead of going back to in-person classes as they were before the pandemic, lecturers should try to incorporate affordances of video-mediated communication such as a shared chat window into face-to-face lessons, which could encourage student participation and increase interaction.

Conclusions
This paper has addressed CLA in ERE classes from a CA and multimodal perspective to account for video-mediated interactions between students and lecturers under unprecedented circumstances. This qualitative analysis shows that lecturers and students adjusted to the new instructional context. The preliminary findings highlight some recurrent behaviors: for example, lecturers would orally interact, while students would write in the chat. This resulted in increased student participation, which confirms SLA studies (Hampel, 2019;Walsh, 2011).
Whereas the chat enhanced the interaction in EFL classes, it made it more demanding for the lecturers to carry on with the lesson and address students' many requests and comments. Lecturers could not rely on the students' facial expressions, as usual, to detect trouble in understanding. They controlled the chat while sharing their screen and delivering new content; they decided how to properly address a TS. It was not always easy due to the constraints of the medium: during screen sharing, the chat window became so small that lecturers could hardly keep track of the messages. On the one hand, data shows that the chat has a tremendous potential to enhance interaction and L2 acquisition. On the other hand, it requires the lecturers to know how to handle it and use it for pedagogical purposes, as is the case with CLA. The more lecturers manage aspects of video-mediated interaction such as CLA, the better their students will deal with trouble interacting in L2, in the classroom and real world (Sert, 2015(Sert, , 2019Walsh, 2011Walsh, , 2012. Therefore, digital interactional competence should become part of teacher training for multiple reasons. First, data shows it is an essential and specific skill. Second, given that lessons have already had to quickly move online due to public health and safety concerns (for example, during the Sars outbreak in 2003), remote teaching cannot be ruled out, so teachers must be ready (Barbour et al., 2020). Third, before the pandemic, distance learning was increasingly spreading, as studies on trends of Higher Education showed (Gaebel et al., 2014;Gaebel & Zhang, 2018). Therefore, teachers will likely be required to respond to the increasing demand for digital interactional competence. Fourth, by treasuring the wisdom learnt and the awareness gained during the pandemic, they could embrace features and strategies to open up the physical classroom. The more lecturers can open their brick-and-mortar classes to the affordances provided by technology, the more frequently their students will interact and practice the target language. To this extent, teachers and lecturers should be trained not to fear change. In this smart new world, the borders between physical and virtual places can be blurred. So, teachers should employ technology originally typical of virtual environments to enhance the possibilities of the physical classroom. Studies on second-language acquisition highlighted the crucial contribution of CLA to L2 learning (Atar & Seedhouse, 2018;Jenks, 2021;Kasper & Wagner, 2011;Walsh, 2011). Not only do teachers evaluate the epistemic gap and choose the appropriate strategy to handle it, but they can help students make longer repairs, thus improving their language skills. In video-mediated communication, lecturers must choose between different strategies and different modes to deal with CLA. Data shows that this can affect both mutual understanding and student participation (which, by the way, is an issue of concern due to distance). Therefore, it seems strategic for lecturers to reflect on their interaction in ERE classes. They should be trained on how to appropriately use the affordances provided by the digital tools in the possible future scenarios for education, namely online and hybrid environments (Barbour et al., 2020;Peters et al., 2020). For instance, further research could employ Walsh's SETT (Self-evaluation of Teacher Talk in Walsh, 2006Walsh, , 2011 framework to enhance video-mediated communication. This validated tool for helping teachers create better interactions in L2 classes Walsh & Mann, 2015) could benefit from microanalysis of video-mediated interactions to enhance teacher digital interaction. It seems a relevant topic of study as videoconferencing has been massively adopted for online classes in the pandemic and online and hybrid classes may be a widespread instructional setting in the future.
Concerning the limitations of this study, this paper deals with a relatively small group of informants, who belong to the same field (EFL studies), which makes the outcomes hardly generalizable. It investigates CLA sequences in video-mediated communication and their possible effects on student behaviour leaving aside other fundamental elements like triadic dialogue. This study has a logocentric approach and does not report on fundamental aspects of videoconferencing. Data on multimodal sources (video and audio materials), proxemic, kinetic elements and gaze management in ERE could enrich teacher interactional competence (Sindoni, 2014(Sindoni, , 2020. This study addresses interaction in ERE as a complex and multifaceted phenomenon. However, it does not deal with the alternation of L1 in L2 classes per se and with the mode, which could prove to be a thoughtprovoking area of research. Finally, despite recommending that teachers increase their digital interactional competence, it does not provide any self-analytical tool.
In conclusion, this study provided a snapshot of the dramatic changes that lecturers and students experienced and their possible pedagogical implications. Further investigations on ERE are needed to unearth ways to enhance interaction in face-to-face, online and hybrid classes thanks to the unprecedented experience of videoconferencing on a massive scale. However, the study shed light on the opportunities and challenges of ERE video-mediated communication, its possible impact on SLA and on the need for specific training to help teachers successfully deal with a kind of interaction that could be playing a relevant role soon.