Social Interaction

1. Introduction

With rapid advances in technology, participants need to adapt their practices for organizing social interaction (Arminen et al., 2016). This study aims to uncover how a social practice emerges and develops as part of the participants' competence in the complex and context-specific setting of video-mediated interaction. More specifically, it examines the use of chat functions in repair sequences in a longitudinal data set of a particular group's repeated encounters. To date, only a few studies have investigated video-mediated interaction from a longitudinal conversation analytical perspective (for notable exceptions, see Balaman, 2016; Balaman & Sert, 2017; Pekarek Doehler & Balaman, 2021).

Language cafés are arenas where second language users (L2) (or LX user, see Dewaele, 2017) can socialize and practice a new language with first language-speaking (L1) volunteers without a set agenda other than simply to talk, similar to conversation-for-learning (Kasper & Kim, 2015). Because of social distancing during the COVID-19 pandemic, many language cafés and language tutoring sessions moved online (e.g., Malabarba et al., 2022). This study explores a longitudinal set of screen recordings of naturally occurring interactions between first language and second language users of Norwegian in one specific digital language café.

The study shows how participants in the digital language café often establish a mutual agreement to engage the chat function to write linguistic items, such as words or expressions, after other repair attempts have been treated as insufficient. Typically, the focal participant (an L2 user) recruits the co-participant (various L1 users) to write, and in response, the co-participant writes and sends a word in the chat as a potential repair solution. Furthermore, by tracking one specific dyad, the article examines how these recruitment methods develop over time, especially whether they become more fine-tuned and conventionalized (Wagner et al., 2018).

1.1 From not yet competent to competent members

Previous longitudinal conversation analytical studies have shown how participants go from being novices (Schegloff, 1989) to competent members (Goodwin, 2018) in specific interactional settings over time (Deppermann & Pekarek Doehler, 2021). Most research has found that participants' methods diversify over time, for instance, when accomplishing requests (Wootton, 1997) or in repair sequences (Brouwer & Wagner, 2004; Hellermann, 2009; Pekarek Doehler & Berger, 2019; Sert & Balaman, 2018). More recently, studies report on a routinization of speakers' social action formats over time (e.g., Deppermann, 2018; Eskildsen, 2021; Kim, 2019; Pekarek Doehler & Balaman, 2021; Pekarek Doehler & Skogmyr Marian, 2022).

In their study on task-oriented video-mediated interactions, Pekarek Doehler and Balaman (2021) documented how a focal participant goes from drawing on a variety of verbal resources to increasingly using the verb "check" to signal her incipient screen-based activity and account for breaks in the progressivity. Over time, the grammatical construction "let met check" routinizes as the standard format for this specific action. The screen-based activity becomes "became less explicit or inferable and ultimately remained often unspecified" (Pekarek Doehler & Balaman, 2021, p. 197). The authors note that grammatical routines can arguably be motivated by the maintenance of progressivity such as in the video-mediated setting where interactants work towards a specific task. Similar findings were pointed to by Deppermann (2018), who observed changes in recipient design in driving lesson instructions. As "joint action becomes routine and common ground to rely on has been accumulated" (2018, p. 321), instructions become less syntactically complex and elaborate, interactional sequences become more condensed, and the actions to be performed become increasingly presupposed and implicit. These changes appear in learner argument structure, ellipsis, and the substitution of lexical references by more indexical means. Analyzing classroom data over a total time span of eight months, Eskildsen (2021) found how an L2 user learns to volunteer at appropriate moments in the classroom. For example, after the teacher has elicited volunteers, the expression "I can write" becomes a routine for the L2 user when he volunteers to write on the whiteboard. Eskildsen argues that this shows how he is becoming an increasingly competent member in the classroom setting.

While earlier longitudinal studies have recognized the diversification and routinization of social practices and use of linguistic resources, these have mainly been centered on in-person interaction. By analyzing video-mediated interactions in which the participants do not have pre-defined tasks and highlighting the multimodal features of change in all parties concerned (e.g., Greer, 2019; Hellermann & Thorne, 2022), this study aims to contribute with new insights on how participants jointly co-construct social practices in video-mediated settings as part of their context-specific interactional competence.

1.2 Recruitment

The participants in the digital language café often co-construct writing in the chat as an alternative repair method in repair sequences. To analyze how this phenomenon emerges and develops over time, this study draws on the notion of recruitment as defined by Kendrick and Drew (2016). According to the authors, recruitment is the different ways in which one person can convey their need for assistance from another, as well as another's anticipation of someone's need for help and their offering or giving that help without being asked (Kendrick & Drew, 2016, p. 2). While Kendrick and Drew (2016) analyze practical difficulties and solutions, some studies have focused on how L2 users verbally call upon their recipients (L1 users) in repair sequences (e.g., Pekarek Doehler & Berger, 2019; Tůma & Sherman, 2022; van der Ploeg et al., 2022) and how such recruitment methods change over time as part of the L2 speaker's development of interactional competence (Pekarek Doehler & Berger, 2019). The current study shows how the focal participant (an L2 user) verbally recruits help from his co-participants (various L1 users) in repair sequences. In response, his co-participants provide the written version of the linguistic item in the form of a chat message by utilizing the chat function. By eliciting a repair in the written format, the focal participant orients to his role as a language learner, and his co-participants, in turn, take on the role as language experts, providing the repair in the written format. This is constructed by all parties as a relevant and practical solution to an immediate problem that can help them re-establish mutual understanding and also facilitate the potential activity of language learning. This study adds to literature on recruitment by uncovering the multimodal and longitudinal aspects of recruitments in video-mediated L2 interaction.

1.3 Video-mediated interaction

The digital language café is a form of video-mediated interaction where multiple synchronous communication modalities such as text (through chat), video, images, and audio are synchronously available to the participants (Thorne & Hellermann, 2022). In the analysis, I will examine how participants do "things" on their individual screens while interacting with their co-participants and hence are occupied in a multiactivity (Haddington et al., 2014). Studies on L2 interaction in video-mediated environments have found that participants coordinate their interactional methods to make their screen-based activity accountable and publicly recognizable for the person on the other side of the screen. For example, by producing verbal alerts in online search sequences (Nguyen et al., 2022) and task-based interaction (Balaman & Pekarek Doehler, 2022; Pekarek Doehler & Balaman, 2021) or by vocalizing the writing—so-called writing aloud (Komter, 2006; Mortensen, 2013)—in collaborative writing interaction (Balaman, 2021). Furthermore, studies have shown how participants interpret each other's bodily-visual conduct in sequentially and interactionally relevant places (e.g., Çolak & Balaman, 2022; Ilomäki & Ruusuvuori, 2020; Licoppe, 2017; Malabarba et al., 2022). For example, a speaker's lip pressing gesture can signal to the co-participant to continue speaking (Malabarba et al., 2022) and finger raising may indicate incipient dictionary lookups (Çolak & Balaman, 2022). Some studies have observed participants' use of bodily-visual resources when engaging with chat messages, such as pointing their fingers to the screen, moving their eyes across the screen, and leaning their head closer to the screen (Dooly & Tudini, 2022, p. 200; Wigham, 2017). This study contributes with new insights into how participants' interactional methods are jointly coordinated as they engage in on-screen activities in video-mediated settings. Furthermore, by investigating how these methods change over time, the study adds new knowledge on how social practices develop as part of participants' context-specific competence in video-mediated interaction.

2. Data and Method

The primary data for this study consist of screen recordings of naturally occurring interactions between the focal participant (Eyad, an L2 user of Norwegian) and various co-participants (L1 users of Norwegian). In sum, 18 different sessions and approximately 23 hours of screen-recorded video have been collected. Including time gaps due to the language café's holiday breaks and changing restrictions during COVID-19, the total time span of the online recordings is one year. The data come from the built-in screen-recorder in the video-conferencing tool used by the organizers of the language café. In addition to screen-recordings, the data consist of saved chat logs from the video-conferencing tool. The chat logs provided information about the content of the chat message, the time at which it was sent and the name of the participant sending the chat message. Furthermore, I conducted short interviews with the participants after each session at the language café and three separate interviews with Eyad. This data set consists of approximately six hours of screen-recordings and was used to obtain background information, such as how many times the participants had taken part in the digital language café before the data collection.

When recording the conversations, the host of the digital language café assigned the participants, who had signed the consent form, and me to a virtual room. After a brief introduction, I started the recording and turned off my video and sound. Thus, I am a non-participant researcher (see Figure A). The recordings from the videoconferencing program capture the meeting as the participant—in this case the researcher—sees the meeting (Zoom, 2022). Considering that the videoconferencing program allows participants to choose their own layout, for example, by only showing the video frame of the talking participant, the recordings do not represent the same layout for all participants. However, according to the participants' comments in the short interviews after the language café sessions, they had a similar layout to the one captured in the recordings.

Figure A. The layout of the recordings

The method used is longitudinal conversation analysis with a multimodal approach (Deppermann & Pekarek Doehler, 2021; Wagner et al., 2018). In order to secure comparability across occurrences and time (Wagner et al., 2018), I narrowed down the analytical focus. In total, 137 chat messages were identified from the saved chat logs. From this collection, I selected, transcribed, and analyzed various examples in which the chat message was sent as a method for resolving trouble related to the production and comprehension of linguistic items such as words or phrases, both in repair sequences and sequences focusing explicitly on language learning. Most of the chat messages were sent by the L1 users (126). A recurrent pattern in all dyads at the digital language café is that Eyad recruits his various co-participants to provide the written form of linguistic items in the chat (48). Based on initial analyses, I narrowed down my collection to prolonged repair sequences initiated by Eyad (other-initiated), in which he recruits his co-participants to write words in the chat after he has treated other repair attempts as insufficient. Because Eyad typically recruits his co-participants to write several times during each conversation, I restricted the examples to the first recruiting turn in the selected conversation. While responses to the chat messages have been investigated to understand their sequential and interactional relevance in repair sequences, these were not analyzed further for reasons of analytical focus and space limitations (for the contribution of chat messages to mutual understanding and orientation to language learning, see Gudmundsen, n.d.

The total number of participants in the data collection is eight (five L1 users and three L2 users). For this study, one L2 user and three L1 participants are included. All participants have given permission for the publication of video extracts in non-anonymized form. The first subsection includes interactions with Eyad and three different co-participants (Ina, Aud, and Tia). This was done in order to show how recruiting assistance by engaging the chat function emerged as a social practice between Eyad and his co-participants in the digital language café. In particular, I found it analytically relevant to include Eyad's first encounter with the chat function (Excerpt 1) because it appears that the extensive use of the chat in repair sequences influences Eyad's later recruiting turns. In the second subsection, conversations with one specific co-participant (Aud) were selected to illustrate how Eyad and his co-participants developed their methods for initiating writing in the chat in the same setting. This was mainly due to three reasons. First, Aud participated in most of the recordings with Eyad (8 out of 18 recordings). Second, she displayed the most instances of visible change when responding to Eyad's recruiting turns, such as refraining from responding by oral means when providing assistance. Third, Aud is included in both sections in order to warrant comparability between the earlier and later stages.

2.2 Transcription

In the analysis, verbal actions are transcribed according to the conventions developed by Jefferson (2004), while non-verbal behavior is illustrated by references to figures showing screen shots of the participants at the moment indicated (as developed by Mondada, 2019) and references to GIFs showing central movements by the participants at the moments indicated. The chat messages are placed on the right side of the transcript, with red arrows and boxes marking the timing of when the L1 user sent the chat message in relation to the sequential organization of the talk.

Transcription conventions:
*1--> = start of GIF, continues until -->*1 end of GIF.
Audible typing sounds: gray highlighted colour
Chat message (the red box indicates the approximate moment it was sent):

3. Analysis

Below, I provide a series of chronologically ordered examples, in which I first document the observable emergence of recruiting assistance by engaging the chat function and then illustrate the development of this same practice. The analysis is divided into two subsections showing three examples from the initial stage of the data collection (May, October, and November 2020) and three examples from a later stage of the data collection (January, March, and April 2021). The first subsection shows how Eyad and his various co-participants co-construct writing in the chat as a preferred repair method in a specific sequential environment. By tracking the later encounters of one specific dyad, the second subsection shows how the participants in the digital language café have over time changed their methods for accomplishing the same social practice.

3.1 Initial encounters with the chat

The three excerpts in this section show three dyads' first encounters with the chat function. At this point in time, all participants were novices at digital language cafés, and none of the L1 users had interacted with Eyad before. Here, several features indicate that engaging the chat as a resource in repair sequences is not at this point recognized or accepted as a social practice by Eyad and his co-participants. As we will see, Eyad does not specify the repair method when recruiting help (Excerpts 1 and 2), and the L1 users respond to Eyad's recruiting turns and foreshadow their screen-based activity by oral means before they send the chat message.

Excerpt 1. "how we write that"

14.05.20: 1^st recording

This excerpt comes from the first conversation between the participants Eyad and Ina. Three minutes into the recording, Eyad displays problems understanding the temporal expression en stund ["a while"]. Just prior to this excerpt, Ina has provided an explanation of the meaning of the expression.

  Open in a separate window

Gif 1. Line 13–20

At the start of this excerpt, both participants pursue and extend the repair sequence and thereby treat the previous repair attempts as insufficient for reaching mutual understanding. Ina produces a self-initiation of self-repair (line 6) and the start of an offer (line 9), which together may be heard as moving forward by proposing a new solution to Eyad's problem as a method of assistance (Kendrick, 2021). Also, visual clues suggest that she is engaging in a new activity. She shifts her gaze to the right side of the screen, moves her fingers away from her necklace (see Figs. 1–2), and shortly after keyboard sounds can be heard. In overlap with Ina's offer, Eyad initiates a new repair sequence by asking for the written form of the word (lines 10–11), which constitutes a request for assistance (Kendrick & Drew, 2016, p. 10). Thus, both participants orient to the need for recruiting assistance. However, Eyad's repair initiation does not specify how the information about the word should be conveyed, that is, the repair method (Mazeland & Zaman-Zadeh, 2004) that Ina should use to solve the problem. In line 13, Ina reinitiates and completes her offer by proposing to write the word in the chat. Instead of answering Eyad's question orally (e.g., spelling the word aloud) she thus explicitly proposes to change to a different modality (writing) and a different mediational means (chat) as part of her assistance. Her verbal utterance thus takes the form of a meta-comment informing him about the repair method, while her visual orientation towards an on-screen activity displays her engagement in providing the repair proper in a different mode. Furthermore, this comment provides Eyad with access to Ina's perspective, thus making her upcoming action understandable as answering Eyad's request by writing in silence instead of merely being silent after the request.

In response to Ina's offer, Eyad produces a series of receipts with considerable delay, long after Ina has started writing and informed him that she is starting with a word that had been a trouble source in a previous part of the conversation (lines 14–17). First, Eyad produces a minimal receipt token, ja ["yes"], in line 18, then a more elaborate receipt in the form of a full sentence echoing the words in her offer (line 19), and finally a positive evaluation (line 22). This delay may be understood as related to his process of discovering the chat function in the videoconferencing program and thereby understanding Ina's offer. As Ina starts offering to write the word in the chat (line 13), Eyad's gaze begins to move around the screen as if searching for something (GIF 1). Only when he focuses his gaze on the lower part of the screen does he produce a receipt (line 18). It thus seems that he discovers the chat function at this point and that this is what occasions his delayed response. This interpretation is also corroborated by information provided by Eyad in a post-recording interview, in which he reported that this was the first time he was introduced to the chat function.¹ In all, the participants' explication of the repair method and multimodal orientation towards the screen-based activity halt the progressivity of the talk (Heritage, 2007; Schegloff, 2007).

This example shows how writing words in the chat function is co-constructed as a possible solution after Eyad has treated Ina's repair attempts as insufficient. Several observations indicate that the writing practice in the chat is not established as a conventional repair method at this point in the participants' history of online interaction. Ina's offer is explicit and non-elliptical, taking the form of a meta-comment on her choice of repair method. Moreover, Eyad's delayed response seems to display his online processing of the import of the offer and his discovery of the chat function in the program.

Excerpt 2. "can we write here"

06.10.20: 2^nd recording

The sequence in Excerpt 2 occurs five months after the previous one. Meanwhile, there has been a summer break and three physical sessions at the language café. In Excerpt 2, the participants Eyad and Aud meet online. After they have established that Eyad does not recognize the word bistand ["aid"], Aud starts to explain its meaning. Aud's turn starts in overlap with Eyad's minimal response to Aud's previous attempt to explain the word.

  Open in a separate window

Gif 2. Line 4–11

Gif 3. Lines 11–16

After interrupting Aud's explanation attempt in line 4, Eyad restarts his turn in line 7 and asks kan vi stud- kan vi skriver her ["can we stud- can we write here"]. This turn can be heard as a composite action (Rossi, 2018), in which he asks about the possibilities, or affordances (Hutchby, 2014), of the video-mediated setting and proposes to write the word as a joint activity, "put forth as something speaker and recipient(s) might do together" (Thompson et al., 2021, p. 125). Thompson et al. (2021) found that proposals using modal interrogatives are used "in environments in which the action proposed is an unprojected next step in the larger activity and thus has not been discussed yet" (p. 145). Moving on, Eyad adds eller det går ikke ["or it does not work"], an increment indicating that Eyad orients to potential trouble. Together with his looking behavior on the screen (GIF 2), these turns shift the focus from the verbal repair to the on-screen resources in the video-mediated environment. By doing this, Eyad treats Aud's previous repair attempts as insufficient for reaching mutual understanding. Furthermore, it shows that Eyad orients to Aud as the participant with most knowledge (Heritage, 2012) with regard to using the chat function in the video-mediated setting and conveys a need for assistance (Kendrick & Drew, 2016).

In response, Aud answers his questions (lines 9 and 10), indexing high epistemic stance (Heritage, 2012) concerning the possibilities of using the chat to write in the video-mediated environment. In combination with her visual orientation to the keyboard and on-screen resources (hand and gaze down, GIF 2), she treats Eyad's question as a proposal (Thompson et al., 2021) to write the word in the chat. In lines 12–14, she provides meta-comments of her actions while leaning forward towards her screen (GIF 3). By doing this, she turns her individual on-screen writing activity into a collaborative one, making salient for Eyad what she is doing outside the visual frame (Balaman & Pekarek Doehler, 2022). After Eyad has accounted for his recruiting turn (lines 17–18), Aud sends the chat message "bistand" ["aid"] in line 19 as a written repair in the form of a chat message. Thereby, the participants utilize the chat function as a resource in their co-constructed and multimodal work of recruiting assistance with the problematic word. In sum, the participants' foreshadowing of the repair method halts the progressivity of the talk.

In comparison with Excerpt 1, Eyad's recruiting turn is more directed towards the repair method when he produces a question and proposal about writing in the video-mediated environment. However, Eyad orients to Aud as the technological expert in the digital language café, and both participants show that this practice is not yet familiar in this particular setting. Taking into consideration that Ina had been using the chat function actively in the session occurring five months earlier in the same setting, it appears here as if Eyad recalls that the chat function can be used to provide him the written versions of unfamiliar linguistic items. From this point, Eyad initiates writing in the chat on a regular basis with the different L1 users at the digital language café.

Excerpt 3. "can you write it"

17.11.20: 4^th recording

This excerpt comes from a conversation between the participants Eyad and Tia. Prior to this excerpt, Tia has used the Norwegian expression det er en bra innstilling (similar to "that's the spirit" in English). When Eyad says he does not recognize the expression, Tia starts to explain its meaning.

  Open in a separate window

Gif 4. Lines 12–21

After Tia has provided two reformulations of the meaning of the expression (lines 1–4), she repeats the expression with falling intonation in line 5, finalizing her repair and inviting Eyad to respond. Following a pause, Eyad utters kan du skriver det ["can you write it"]. This turn is a request for Tia to write the problematic expression, that is, a request for assistance (Kendrick & Drew, 2016, p. 10), specifying the repair method. By not responding to Tia's earlier repair attempt and initiating a new repair sequence involving the written mode, Eyad treats her previous turns as inadequate for reaching mutual understanding. In response, Tia affirms (↑ja) and says skal vi se ["let's see"]. Simultaneously, she shifts her gaze downwards on her screen and removes her hand from her chin (Figs. 1'–2), orienting to her on-screen resources. Next, she produces an offer in line 10, jeg kan sende til deg ["I can send to you"], in which she displays her commitment to carrying out the requested action, and furthermore specifies the repair method. Similar responses with modal adverbs have been found by Steensig and Heinemann (2014) as responses to remote requests, and occur in instances where the recipient indicates that the relevance of the requested action was not recognizable as being part of a larger, jointly established activity. Similarly, Tia's responses (lines 9'–10) suggest that the writing practice is not yet established between the two participants in this particular setting. Furthermore, these responses function as floor-holding devices and verbal accounts for her screen-based activity (Balaman & Pekarek Doehler, 2022). Although Tia makes salient that she is occupied in an on-screen activity through bodily-visual orientation towards her screen, and even specifies her upcoming action through an offer, Eyad starts an informing sequence in which he explains where the chat is located on the screen, while gazing and pointing at his own screen (lines 12'–16, GIF 4). Rather than waiting silently for Tia to write, Eyad treats Tia's multimodal on-screen orientation as insufficient interactional proof of writing in the chat. Through his verbal utterances Eyad positions himself as the knowing participant (Heritage, 2012) in the video-mediated environment. However, the epistemic hedges ("I think it is here" and "this thing") modify his expert role in the technological domain.

Tia in turn, responding minimally and in overlap (line 18), does not seem to orient to Eyad's turns as relevant for moving on with her task at hand. While a continuous typing sound is audible in lines 22–26, Tia vocalizes the expression, dividing the components in the expression into smaller chunks (Svennevig, 2018), suggesting that she is "typing aloud" (Komter, 2006, p. 207). In line 27, she sends the chat message At du har en bra instilling ["That you have a good spirit"]. Thus, she has provided Eyad with repair in two modes: oral and written. A general observation is that both participants verbally negotiate and foreshadow the repair method before Tia sends the chat message: Tia by making her upcoming actions understandable for Eyad, and Eyad by making comments about the chat function. Together, their back-and-forth multimodal on-screen navigation halts the progressivity of the talk, which implies that writing in the chat is an unfamiliar practice between Eyad and Tia.

This example shows one of the first instances of Eyad's use of the request format "can you write x," which over time becomes his most common method of recruiting his co-participants to write new words in the chat. Together with Tia's elaborate response to Eyad's request, their verbal foreshadowing and negotiation of the writing activity and the chat function indicate that this practice is not yet established as a conventional repair method.

3.2 Later encounters with the chat

As Eyad and his co-participants become more acquainted with each other and the various affordances of the video-mediated landscape, they appear to recruit assistance by engaging the chat in a more conventionalized manner in similar sequential environments. The following section will focus on interactions between Eyad and Aud. The analysis shows how the request format "can you write x" has become the common method of recruiting chat-based repair, how the chat-based activity is no longer explicated, and how, accordingly, the participants recognize this activity without it being explicated.

Excerpt 4. "can you write it"

19.01.21: 6^th recording

Before this excerpt, Eyad has explained that he believes Donald Trump does crazy things because he wants to be famous. In response to Eyad's assertions about Trump, Aud uses the expression all PR er god PR ["all PR is good PR"], which Eyad just prior to this excerpt has treated as unknown.

  Open in a separate window

Gif 5. Lines 6–9

This excerpt begins with Aud repeating the expression and providing an explanation of its meaning (lines 1–4) as a response to Eyad's repair initiation (not shown in the transcript), which are characteristic features of a word explanation sequence (Mazeland & Zaman-Zadeh, 2004). After a long pause, Eyad requests Aud to write (kan du skrive det ["can you write it"]) in line 6. Thus, he reuses the modal interrogative form from earlier encounters with Aud and other L1 users in similar sequential environments (see Excerpt 3). By not providing a receipt to Aud's previous turns, Eyad indicates that he has not understood her previous repair attempts and instead elicits a repair in the written format. Immediately after, Aud gazes down, shifts posture slightly backwards and moves both arms so that they are placed in front of her device, into a typing position (Figs. 1–2), followed by leaning forward towards her screen (Fig. 3). With this embodied maneuver (see also GIF 5), Aud shows Eyad that she is about to commence with the screen-based assistance of writing in the chat. In line 9, Aud sends the chat message All PR er god PR ["all PR is good PR"]. Thus, only by using bodily-visual resources and the chat function, Aud has granted Eyad's request and provided him with a written repair.

While Aud is writing in the chat, Eyad says han vil bli en kjendis skjønner du ["he wants to be a celebrity you see"] and laughs (line 8). This turn refers to his assertion about Donald Trump before the repair sequence started and functions as a skip-connector back to the main activity. Instead of commenting on the word, the writing activity or the chat function in the video-mediated environment, Eyad shows that he is more oriented towards moving on with the topic. Furthermore, his response suggests that Eyad treats Aud's embodied orientation towards her screen as accountable for providing assistance. In sum, the participants display that they deal with the chat function in repair environments in a more implicit manner, hence fostering the progressivity of the talk (Heritage, 2007; Schegloff, 2007).

This example shows how the practice of writing and sending words in the chat is no longer foreshadowed by verbal comments and explicitly accounted for as an on-screen activity. On the contrary, the practice centered around the chat function is embedded as an integrated part of the prolonged repair sequence; they "just do it" in order to resume the topic of the conversation.

Excerpt 5. "you can write it"

02.02.21: 8^th recording

When telling Eyad about her recent hike, Aud says that one gets a nice view because it is steep (in Norwegian: bratt). Before this excerpt, the participants have unsuccessfully attempted to establish mutual understanding concerning the meaning of the word in the context of a hill being steep (bratt bakke).

  Open in a separate window

Gif 6. Lines 13–14

Gif 7. Lines 15–17

After Aud explicitly conveys difficulties with explaining the word, she starts a new multimodal explanation attempt (lines 5 and 7), with her hand in a tilted position (Fig.1). Together with the deictic utterance sånn her ["like this"], the iconic gesture highlights the shape of the hill and conveys the semantic meaning of the word. In overlap with Aud's repair attempts, Eyad produces a similar utterance on two different occasions: du kan skriver ["you can write"] in line 6 and du kan skriver det ["you can write it"] in line 9. Subsequently, however, Aud's open-class repair initiator he (similar to "hæ", see Svennevig, 2008) and her leaning towards the screen (Fig. 2) indicate that she has not heard Eyad's previous turns. In response, Eyad says du kan skriver det ["you can write it"] with falling intonation (line 13). By providing a repair in the form of a repetition, he treats Aud's repair initiator as a problem of hearing. The turn, syntactically formatted as a modal declarative, consists of a possible solution for Aud to implement. Because the sequential placement of the turn comes after Aud has displayed trouble with explaining the word, it seems to function as advice (Kendrick, 2021). In this case, the problem is twofold: Eyad's lack of understanding of the meaning of the word (the general and main problem) and Aud's problem with explaining the meaning of the word (the specific and most urgent problem, the repair method). By formatting his turn as advice rather than a request, Eyad claims increased epistemic authority with respect to the proper solution to their problem. In response, Aud gazes down at her screen and moves her upper torso slightly backwards, entering typing position (GIF 6). Thus, without verbally confirming that she has now heard and understood Eyad's turn, she uses her body to display to Eyad that she is about to implement his proposed solution to their problem(s).

In line 15 Eyad utters "e kan du skrive det" ["uh can you write it"]. Interestingly, Eyad self-repairs from the modal declarative form to the interrogative form, which happens to be his most common request format in similar sequential environments at the digital language café (for an overview of Eyad's recruiting turns, see Table 1 in the Appendix). Together with the questioning intonation, the format of the request, a modal interrogative with can, initiates a sequence in which writing and sending the word in the chat becomes an especially relevant action (Kendrick & Drew, 2016, p. 10). This indicates that Eyad treats Aud's non-verbal response as an insufficient response to Eyad's turn in line 13. Given that Eyad has not received an oral response from Aud that she has heard, understood, and accepted his turn (both a repair and advice), the advice seems to have lost sequential relevance and might explain the choice of an utterance that to a higher degree invites a response from Aud.

Moreover, by switching to his standard construction and producing an explicit request, Eyad orients to this as a more conventionalized format for recruiting assistance, which in prior interactions has mobilized Aud and the various L1 users to write in the chat. In the following pause, Aud produces two quick head nods, leans forward towards her screen (GIF 7) and typing sounds can be heard from the keyboard. While her visual on-screen orientation accounts for her writing activity, the head nods display to Eyad that she has heard, understood, and accepted the proposed solution in the chat. She sends the chat message Bakken er bratt ["The hill is steep"] during Eyad's turn in line 17, a written repair in the form of a chat message. Thus, without oral means, Aud has provided assistance. Of note is Eyad's turn in lines 17 and 18 "du kan skrive denne tre (.) tre forskjellig" ["you can write this three three different"], additional advice to Aud to also write the two other words she used when explaining the meaning of the word (not shown in the transcript). This shows that Eyad takes the role of an expert with regard to the repair method specifically, and furthermore it highlights his preference for eliciting repair in the written form in a specific sequential environment.

In contrast to the previous excerpt, this example shows how Eyad elicits a chat-based repair by producing advice, displaying a strong epistemic stance towards the repair method. Following hearing problems, he reformulates his turn into a request, using the common construction "can you write it." In both instances, Aud orients to Eyad's proposed solution as a known procedure by visually orienting to her screen.

Excerpt 6. "can you write it I do not understand"

13.04.21: 15^th recording

Prior to this excerpt, Aud has told Eyad that her student group will work with a research project about alienation (In Norwegian: "fremmedgjøring"). After Eyad's lack of response, Aud starts a reformulation of her turn. Here, Aud presents the word a second time, here as an adjective rather than a noun (fremmedgjørende—"alienating").

  Open in a separate window

Gif 8. Lines 12–13

The start of this excerpt shows how Aud and Eyad treat the word fremmedgjørende ["alienating"] as problematic: Aud reformulates it in lines 5–6 and Eyad initiates repair in line 8. Aud repeats the word in line 9, treating Eyad's turn as a problem of hearing ("trying the easiest solution first," see Svennevig, 2008). Interestingly, Aud does not repeat the word format she produced in line 4 but presents the word form she used earlier. After another partial repeat of the word in line 10, Eyad requests Aud to write it by using the interrogative format kan du skrive det ["can you write it"] and initiates a new repair sequence (line 12). By adding an account to the request ("I do not understand"), Eyad treats Aud's repair attempts as insufficient for him to understand the meaning of the word. In response, Aud gazes down at the screen, moves her upper torso as if she is placing her arms in front of the keyboard (Figs. 1–2) and leans forward towards her screen (Fig. 3, see full movement in GIF 8). Thus, by multimodal means, she orients to Eyad's request as an urgent matter which needs to be resolved. A few seconds afterwards, Aud sends the chat message fremmedgjøring ["alienation"] in line 13, a textual and screen-based repair. Similar to Excerpt 4, Aud does not provide an oral response to his request, for instance, by producing an affirmation token, but goes directly to the task at hand: locating the chat function on her screen, writing, and sending the word.

This example, taken from one of the last recordings in the data, shows that Eyad uses the request format "can you write it" (similar to Excerpts 3 and 4). Again, Aud does not account for her on-screen activity by oral means but simply starts writing in the chat. This shows how the social practice of recruiting assistance by orienting to the chat function has become more routinized.

3.3 Summary of main findings

This article demonstrates how recruiting assistance with words by orienting to the chat function emerges and develops as a social practice through participants' interactional encounters in the video-mediated setting of a digital language café. Below follows a summary of the analysis.

The first part of the analysis shows how Eyad and three different L1 users (Ina, Aud, and Tia) establish the practice of writing and sending words in the chat as a repair method. First, Eyad's recruiting turns are less specific about writing in the chat as a repair method. For example, they are designed to seek assistance with the written form of the linguistic item ("how we write that," Excerpt 1) and the technical possibilities of the video-mediated environment ("can we write here," Excerpt 2). This shows how Eyad treats the chat function as an unfamiliar resource in the video-mediated environment. The third excerpt shows how Eyad produces a request ("can you write it") that more specifically orients towards the chat as a relevant repair method. Secondly, Eyad's co-participants tend to explicate their assistance. In the first excerpts, they respond to Eyad's recruiting turns by oral means (offers, Excerpts 1 and 3, and answers, Excerpts 2 and 3) and verbalize or vocalize their ongoing activity on their individual screens before they send a message in the chat function. Through these responses, they inform the other participants about the repair method (Excerpt 1: "I can write on chat for you"; Excerpt 2: "I shall write to you"; and Excerpt 3: "I can send to you") or in other ways make their chat-based activity hearable for Eyad, for example, by vocalizing their writing (Excerpt 3). The analysis illustrates how these responses are closely coordinated with bodily-visual conduct, such as gazing towards specific parts of the screen, leaning forward towards the screen and placing the arms downwards into a typing position. Third, Eyad tends to display low recognition of his co-participants' screen-based assistance. In Excerpt 1, Eyad's responses to Ina's offer are delayed, and together with his gaze behavior across the screen this indicates that he discovers the chat function. Excerpt 3 shows how Eyad starts an informing sequence in which he explains to Tia where the chat is located, thus implying that he does not treat her offer and bodily-visual orientation to her screen as sufficient displays of providing assistance in the chat.

The second part of the analysis zooms in on how one specific dyad (Eyad and Aud) has changed their interactional methods for accomplishing the same practice over time. First, Eyad's recruiting turns become more specific regarding the repair method as a possible solution to a problem, requesting the L1 user to write ("can you write x," Excerpts 3, 4, 5, and 6). As shown in the analysis, the interrogative request format "can you write x" quickly became his standard format for recruiting help with linguistic items (see Table 1 in the Appendix). Excerpt 5 showed how Eyad designed his recruiting turn as advice ("you can write it"). In comparison to the request format, the advice format displays an increased epistemic authority with regard to the specific repair method. Secondly, his co-participants decreasingly draw on verbal talk and rely more on embodied means when providing assistance. Excerpts 4, 5, and 6 show how the L1 user, instead of explicitly granting the recruiting turn or verbally informing about her screen-based assistance, displays her compliance in a more implicit way by writing and sending the chat message. In sum, these responses show that the writing activity in the chat is presupposed and taken for granted. Third, Eyad tends to recognize his co-participants' screen-based assistance more over time. For example, Eyad resumes the topic they talked about before the repair sequence started while Aud is "doing writing in the chat" (Excerpt 4) and he remains silent (Excerpt 6).

4. Discussion and Conclusion

The practice of recruiting assistance with the written form of words by engaging the chat function is constructed as a relevant solution to an immediate problem, which can help the participants to re-establish mutual understanding and facilitate language learning, benefitting both parties at the digital language café. Typically, the recruiting turns come in response to the L1 user's oral (or multimodal, Excerpt 5) repair attempts, indicating the focal participant's preference for a written repair in a specific sequential environment. The results show that recruiting assistance with linguistic items by engaging the chat is a multimodal and carefully coordinated collaboration that needs to be established within each dyad (Excerpts 1–3), building up their common ground, which is the foundation for all joint actions (Clark, 1996, p. 14).

By focusing on one specific dyad's later encounters, the study documents how the recruitment methods of the participants at the digital language café become more fine-tuned and conventionalized over time. The analysis shows how the focal participant displays increased knowledge of the chat function being a relevant tool for solving specific problems in the video-mediated setting, which to a higher degree presupposes that his co-participant would recognize his call for assistance and provide a written repair in the chat. Furthermore, it shows how specific constructions (in this case: "can you write x") become routine in context-specific environments to fulfill certain interactional purposes (Pekarek Doehler & Balaman, 2021). This illustrates how the focal participant increasingly manages to draw on the affordances (Hutchby, 2014) of the video-mediated context, thereby becoming more competent in the technological domain. This is in line with previous findings that show how L2 users develop new techniques and take increasing responsibility in repair sequences over time (Brouwer & Wagner, 2004; Pekarek Doehler & Berger, 2019; Sert & Balaman, 2018; Siegel, 2015). By no longer responding to Eyad's recruiting turns by oral means or explicating the screen-based activity before sending the chat message, his co-participant increasingly treats the writing activity in the chat as an established practice. Correspondingly, this conduct appears to be more recognized and accepted by Eyad as being relevant for the task at hand. Thus, over time both Eyad and his co-participant show an enhanced ability to interpret each other's multimodal actions and coordinate their on-screen activity in meaningful ways in the video-mediated setting, thereby displaying an increased competence in the domain of mediated interaction.

In sum, these changes lead the participants to move more quickly towards the preferred repair solution, fostering the progressivity of the talk. Their conduct is competent because the participants display that the same social practice becomes more recognizable and acceptable, providing no grounds for comment or repair (e.g., Mori & Koschmann, 2012). These findings corroborate those of previous studies that have identified a progressive simplification of social action formats when the same participants perform the same joint activity in specific settings over time and become more competent (Deppermann, 2018; Pekarek Doehler & Balaman, 2021). Furthermore, as previous research has pointed out, larger processes of socialization is inseparable to the process of development of interactional competence (see e.g., Pekarek Doehler & Berger, 2016). In addition, the findings shed new light on how the multimodal features of change in the conduct of all parties concerned are part of speakers' routinization of social practices in the complex and multi-layered setting of video-mediated L2 interaction.

The study has several methodological limitations. The first pertains to the suitability of a set of different co-participants for analyzing change in interactional practices (see discussions on similar methodological challenges in Deppermann & Pekarek Doehler, 2021; Wagner et al., 2018). Given that the three co-participants each have their separate interactional history with Eyad and their experiences with video-conferencing programs and use of chat vary, Aud's conduct is not representative of the other participants' practices. This may weaken the comparability in the study. However, similar to previous longitudinal studies (e.g., Clayman & Heritage, 2002), it can be argued that Eyad's co-participants belong to the same member type. First, they were all L1 users of Norwegian and volunteers at the same digital language café. Second, because the digital language café moved online in April, they were all relatively new to this specific setting. Third, none of them had interacted with Eyad in this specific setting prior to the data collection. Hence, all members shared the same starting point in the transient setting (Lønsmann et al., 2017) of the digital language café. The second methodological challenge is related to the fact that the study only analyzes data from one perspective. Analyzing data from multiple perspectives (e.g., Ilomäki et al., 2021; Hansen, 2020; Rusk & Pörn, 2019; Seuren et al., 2021) or even from the perspective of one of the participants (Olbertz-Siitonen, 2015) would have provided a more detailed view of how the participants themselves perceive the other's multimodal and situated actions. While the study recognizes that a strict emic perspective is challenged in the analysis, the actions are relevant from an emic perspective in that they are treated as observable and accountable by the participants. In order to achieve generalizability, a larger data set is needed to investigate how participants in similar settings deal with repair sequences. Nevertheless, the results show how a possible social practice (Peräkylä, 2004, pp. 283–304) in the specific context of the digital language café emerged and developed as part of the participants' context-specific interactional competence in a video-mediated setting. This study offers empirical descriptions of what participants in video-mediated L2 interaction can do when they encounter interactional trouble.

Acknowledgement

I am most grateful to the participants at the language café for their permission to let me record the video material and analyze and publish selected parts. I would also like to thank the two anonymous reviewers for extensive and useful comments to previous versions of this paper. This work was partly supported by the Research Council of Norway through its Centres of Excellence funding scheme, project number 223265.

Appendix

Table 1. Total overview of recruiting turns

^{1 The chat log was not saved as part of this Zoom recording. However, immediately after this excerpt, Eyad reads aloud the expression for en stund siden ["a while ago"] while gazing at a specific part of his screen. Thus, in addition to Aud's visual on-screen orientation and audible typing sounds, one can assume that Ina sent a chat message with this text at the end of this excerpt.↩}