Social Interaction

1. Introduction

Among many other things, the COVID pandemic changed communication practices at workplaces. Discussions that had previously taken place in meetings, during coffee breaks, and in office hallways were moved to video-mediated environments, and new ways to ensure information transmission and workflow were needed. The change was particularly consequential for employees working in languages other than their strongest language; being a second language speaker often means compensating for limited language skills, both through the physical affordances of the environment and through the diverse and often multimodally coordinated help of co-participants. In video- and audio-mediated meetings where participants are not in each other's immediate co-presence, access to embodied and material resources is limited or restricted, which may be consequential for the ways in which social conduct is organised.

In this paper, we analyse remote meetings where a person who speaks the shared language as a second language (L2) encounters language-related difficulties, and another participant offers assistance to resolve the trouble. Prior literature on multilingual co-present settings has highlighted verbal and embodied resources as key in recruiting and providing different forms of linguistic assistance (e.g., Harjunpää, 2021; Mondada, 2012; see also Kurhila, 2006 on repair). For instance, co-participants have been found to carefully monitor L2 speakers' verbal and embodied conduct, such as body movement and gaze direction, which enables them to detect suitable places to provide help (Harjunpää, 2021). This underlines the need to know what kind of resources participants use in technology-mediated interaction, where access to the affordances of co-participants' physical environments – namely, their spatial, material and bodily configurations – is limited (e.g., Luff et al., 2014; Hutchby, 2014). The present study focuses on practices of third-party assistance, that is, the support provided by a non-primary recipient in ongoing interaction (Goffman, 1981, see also peripheral recipient in Greer & Ogawa, 2021). We address the following research questions: how are language-related problems made relevant in multiparty L2 remote work meetings and how is facilitation through third-party assistance coordinated?

The data consist of video-recorded workplace meetings held by a Finnish non-governmental organisation on the Microsoft Teams platform in 2020–2021. The meetings are analysed using Conversation Analysis (CA; see, e.g., Sidnell & Stivers, 2013), which is an established method for analysing not only in-person but also online interactions (e.g., Koivisto, Vepsäläinen & Virtanen, 2023; Pekarek Doehler & Balaman, 2021; Oittinen, 2020, 2021; Rintel, 2013). The strength of CA lies in its focus on the details of social conduct: the analysis reveals the particularities of the moment-by-moment unfolding interaction at the level of talk and multimodal actions and uncovers how the participants themselves orient to each other's actions. CA thus enables an in-depth investigation of how the participants in the data use language, their bodies and digital objects to solve language-related troubles across distances.

We begin the article with a theoretical overview in Section 2. In Section 3, we describe our data, setting and study participants. In Section 4, we present the analysis, which is divided based on the position of the facilitating turn in a repair sequence. After the analysis, we summarise the findings and conclude the paper in Section 5.¹

2. L2 Remote Work Meetings and Language-Related Problems

Conversation Analysis has a long tradition of investigating institutional interaction (e.g., Antaki, 2011; Arminen, 2005; Drew & Heritage, 1992), but multilingual workplace settings have only recently started to attract attention as a subject of study (e.g., Hazel & Svennevig, 2018). This is partly induced by globalisation and the rapid increase in work-based immigration as a result of the workforce shortage in Western countries. From the viewpoint of workplace interaction, many questions regarding second language-speaking participants' language skills have become relevant.

An important context of research regarding workplace interaction are meetings (e.g., Asmuss & Svennevig, 2009; Boden, 1994). Multilingual meetings have already been explored to some extent, focusing especially on language choice as a means of managing participation (e.g., Hazel & Svennevig, 2018; Markaki et al., 2014; Mondada, 2012; Skårup 2004) or doing identity work (Mondada, 2004; Vöge, 2011). These studies have highlighted multilingualism as a resource that meeting participants can reflexively deploy in international contexts. However, asymmetric language skills of meeting participants have also been found to cause problems in an ongoing interaction. Second language speakers can experience troubles of understanding, which can become manifested either in explicit displays of trouble or in different types of misalignments (Tranekjær, 2018). Furthermore, second language speakers may face troubles in speech production, which are indicated by signs such as pauses, hesitations or word searches (e.g., Svennevig, 2018; Oittinen, 2022).

Problems of speaking, hearing and understanding are typically managed through repair organisation in conversation. Repair is one of the main conversational organisations that CA has identified; it offers participants resources to identify and remedy instances where intersubjectivity between the participants is halted or at risk. Previous research has identified a wealth of repair practices that participants use to overcome interactional trouble, both in first language conversations (e.g., Schegloff, 1992; Drew, 1997; Hayashi et al., 2013; Dingemanse & Enfield, 2015; Haakana et al., 2021), as well as in second language conversations (e.g., Svennevig, 2004; Kurhila, 2006; Lilja, 2014; Theodórsdóttir, 2018).

CA makes a distinction between initiating and performing repair on one hand, and between self and other as the relevant parties in repair sequences on the other. What is called self-repair is when the trouble turn and the repair are performed by the same speaker, whereas other-repair refers to cases when these turns are produced by different speakers (Schegloff et al., 1977, p. 364). The distinction between self and other is relevant for repair initiation techniques as well: self-initiations are typically non-lexical speech perturbations (e.g., cut-offs, sound stretches), while other-initiations can be done for example through question words, (partial) repeats, or candidate understandings (Schegloff et al., 1977, pp. 367–368). In addition, self- and other-initiations occupy different positions in a turn. Other-initiation typically occurs in a turn that is subsequent to the trouble turn, whereas self-initiations are located in the same turn that contains the trouble source. Thus, self-initiated repair is often more compact than other-initiated repair, occupying only one turn, thereby having less impact on the progressivity of conversation.

In second language conversations, however, self-initiated repair sequences can be more extended. If the participants' language skills are limited, they may initiate repair not only to replace a linguistic item with another but also to search for words or constructions, and often this requires more than just a word replacement within one turn (see Kurhila, 2006). Indeed, a word search is an activity that has been observed and investigated in a broad range of interactional situations where people have differing language skills (e.g., Brouwer, 2003; Koshik & Seo, 2012; Svennevig, 2018; Pekarek Doehler & Berger, 2019; Kotilainen & Kurhila, 2020; Oittinen, 2022). An important aspect of word searches is that they can be either self- or other-directed; that is, the speaker can try to retrieve the missing element on their own or they can try to recruit the recipient(s) to provide it. The crucial factor in distinguishing self- and other-directed searches is how the speaker deploys their gaze (e.g., Goodwin, 1987; Hayashi, 2003; Dressel, 2020).

In conversations where participants have differing language skills, initiating repair and word searches can be used as a means to appeal for linguistic assistance (Harjunpää, 2021, Greer, 2015, Kurhila, 2006). Such linguistic assistance can be, for example, language brokering, a form of ad hoc translation and interpretation (e.g., Bolden, 2012; Harjunpää, 2021; Traverso, 2012). Brokering has often been studied in intergenerational contexts (see Morales & Hanson, 2005), but as a role "an interlocutor can enact during the repair activity for the purposes of resolving (or averting) an understanding problem" (Bolden, 2012, p. 114) it can happen in all kinds of interactions. A broker uses their linguistic expertise to translate or explain a problematic segment of talk. Regarding socially distributed rights to knowledge, brokers do not claim epistemic authority over what has been said by the other party, but they claim rights to the knowledge of the language that is being used (Bolden, 2012, pp. 113–114).

In this study, we will examine instances where a L2 speaker displays trouble in remote work meetings, after which a third party offers assistance to resolve the problem, thus acting as a broker. However, in addition to resolving the linguistic problem, the third party may accomplish other functions through their turn, such as fostering the L2 speaker's language learning (e.g., Gardner 2012; Kurhila et al., 2021) or managing the participation framework (see Goffman, 1981; Greer & Ogawa, 2021). What is common to our cases is that the contribution by the third party aims to facilitate the interactional participation of the L2 speaker. We have identified such third-party facilitation in three environments: after other-initiations of repair by the L2 speaker (Section 4.1), after self-initiations of repair by the L2 speaker (Section 4.2), and when assistance is offered even when there is no explicit marking of the trouble by the L2 speaker (Section 4.3).

Previous research (on physically co-present situations) has established that third-party brokering or facilitation is coordinated not only through verbal but also through embodied resources (e.g., gaze, gestures and body movement; see Harjunpää, 2021). In remote interaction, however, we are dealing with a distributed participation framework in which participants are in diverse geographical locations, and their access to mutually shared interactional resources and their possibilities to utilise embodied conduct for interactional work are limited (e.g., Luff et al., 2003; Luff et al., 2014; Hutchby, 2014). Possibilities to indicate and secure recipiency, for example, vary a great deal depending on the technological environment, platforms used and overall configuration. Having only one channel for spoken communication brings limitations: for instance, it is not possible to have parallel discussions (i.e., make use of "schisming"; Egbert, 1997).

Earlier studies have highlighted not only the challenges of video-mediated environments but also the affordances and participants' creative ways to organise interaction with the resources available to them (e.g., Hjulstad, 2016; Luff et al., 2014; Sert & Balaman, 2018). Recently, some attention has been given to the organisation of problematic moments and repair practices in foreign-language screen-based activities involving "a talking head configuration" (Licoppe & Morel, 2012; see also, e.g., Balaman, 2021). For instance, in their study on word searches in video-mediated German L2 interaction, Uskokovic and Talehgani-Nikazm (2022) found that the speaker may use verbal alerts and suspend the ongoing activity for the duration of the search and restrain help from the co-participant by using a specific gesture: showing their index finger while orienting to the screen in order to carry out a search of the missing item. On the other hand, some enhanced video platforms that emulate co-presence (e.g., via use of large screens) have been found to enable the production and interpretation of trouble-relevant embodied noticing, so that problematic moments become solved without the need to initiate repair verbally (see Oittinen, 2020). This illustrates how the ways to account for one's actions, including silences, are dependent on the resources available in the interactional environment and its situation-specific features (see also Arminen et al., 2016). Against this backdrop, the present study sheds light on both the complexities of attending to language-related troubles and providing third-party assistance and the affordances that the video-mediated environment has in these processes.

3. Data

The data for this paper comes from a Finnish non-governmental organisation (NGO) working in the field of immigration and integration. The study's findings are based on a collection of 44 meetings (about 13 hours in total) recorded during the pandemic period from May 2020 to March 2021. The recordings were conducted by one of the participants, who used the recording feature of Microsoft Teams. The other participants were informed about the recordings and they gave their consent to be recorded and for the data to be used.

The recordings include meetings held either internally within the NGO or with NGO staff and their collaborators. The meetings consist of activities characteristic of modern expert work, such as planning, negotiating and decision-making, and therefore require a relatively high level of language proficiency. They are either daily morning meetings that follow a similar structure (beginning with greetings and progressing through successive dyads between the head of the organisation and each employee) or meetings organised to discuss a specific topic, usually between fewer participants.

The language situation in our focus organisation is diverse (see Kurhila et al., 2021). The main working languages are Finnish and Russian, but English is also regularly used. The staff of the organisation include native speakers of Russian and Finnish. They all have some competence in the other of these two languages but their skills vary from elementary to fluent bilingualism. The examples analysed in this paper come from five different meetings with three to ten participants. A summary of the backgrounds of the participant's relevant for our argumentation is provided in Table 1.

In the following examples, the main language is Finnish. Thus, the participant in need of linguistic assistance is a native speaker of Russian (Jelena, Examples 1, 2, 5, 6; Jegor, Example 3; Ksenija, Example 4). In all these cases, the third party who offers assistance is Katja, a middle manager in the organisation. Due to being fluent in both Finnish and Russian, she had already established a role as a language mediator in the organisation before the change from face-to-face to technology-mediated meetings. This self-chosen role was also brought up in the background interviews we conducted with all employees of the organisation.

4. Analysis

The data show that facilitation of (potential) language-related problems can occur in different sequential contexts. Specifically, the findings indicate that there is notable variability in the ways in which a third party enters the conversation and offers assistance. First, assistance can be offered as a response to other-initiation of repair (Section 4.1). In the data, these cases are usually connected to problems in understanding. Second, facilitation by the third party may be a reaction to a self-initiated repair by the L2 speaker (Section 4.2). These cases are usually related to troubles in speech production, for example, finding a suitable word or a correct grammatical form. Third, possible situations for the facilitator to enter the conversation are instances where they initiate repair before any trouble has been signalled by the L2 speaker (Section 4.3). In a way, the facilitator seeks to prevent possible problems either by checking whether the L2 speaker has understood the preceding talk or by reformulating it in a simpler way. These cases could thus be seen as pre-emptive repairs,³ and they often occur after a somewhat unclear (e.g., lengthy or fragmented) turn. In the following, we offer analysis of two characteristic examples of each of these subtypes.

4.1 Other-initiated repair

Our first example presents a case of facilitation that follows other initiated repair by the second language speaker: the facilitator enters the conversation after an unsuccessful repair attempt by the participant who has produced the trouble turn. The example is from the beginning of a morning meeting for the NGO staff (8 persons present). The participants are engaged in small talk, and Risto produces a verbal noticing concerning Jelena's physical environment that is visible to the others via video (line 01). A phrase Risto uses in his question turns out to be problematic for Jelena.⁴

Example 1. Sloping roof. (IMM2209, 3:34–3:55)

  Open in a separate window

After Risto's initial question (line 01), there is a long pause (line 02) that indicates the question as being potentially problematic for its recipient, Jelena. Jelena then repeats the expression vino katto ('sloping roof') and asks in Finnish what it means (line 03). Her repair initiation is an interrogative clause ('what does X mean') that has been reported to occur in Finnish conversation when the trouble source is an unknown concept, and the type of repair it yields is an explanation of the term (Haakana et al., 2016). In his repair attempt, however, Risto reformulates the question by segmenting it into smaller units. First, he topicalises the word katto ('roof', line 04), which Jelena acknowledges with the agreeing particle kyllä ('yes') (line 05). This suggests that Jelena's non-understanding does not concern the noun (katto) but the adjective that Risto has used to describe it. However, the second part of Risto's question (line 06) only reproduces the adjective vino ('sloping') but does not alter it.

As Jelena does not immediately respond to the reformulated question (line 07), Katja joins the conversation. She gives a translation of the word vino in Russian (line 08). Katja thus observes that Risto's repair attempt has not been sufficient to resolve the trouble. Katja chooses an economical way to describe the meaning of the problematic word: she translates it into the language she and Jelena share as their strongest language. The translation resolves the trouble: Jelena starts responding to Risto's original question with an emphatic agreement token, no kyllä ('well yes', line 10). Partly in overlap with Jelena's turn elaboration, Risto continues to explain the meaning of the word vino by placing his hand so that it depicts a diagonal angle (line 12). However, intersubjectivity has already been reached, and Jelena merely acknowledges this with another affirmative response, joo ('yes') (line 13).⁵

In Example 1, a third party takes the initiative and provides a repair after the failed repair attempt by the producer of the trouble source (see also Bolden, 2012, 2018). Our next example illustrates a situation where the repair initiation is directly addressed to someone other than the producer of the trouble turn – by using the private chat function. The first part of the extract shows the problematic turn as it occurs in meeting talk (Example 2a) and the second part the subsequent chat conversation that takes place right after it (Example 2b).

Example 2a (meeting interaction). Jegor's knee. (IMM1706, 2:38–3:01)

The extract shows how Risto delivers the news about Jegor's situation as part of the meeting opening. While some participants can be seen to react to Risto's news by nodding or frowning, other participants' (such as Jelena's) displays are not visible on the screen. A moment later, Jelena sends a private message to Katja through the chat interface:

Example 2b (chat messages, IMM1706). Jegor's knee.⁶

  Open in a separate window In her message, Jelena asks Katja to clarify what Risto has said; in the ongoing spoken conversation, Jelena has given no indication of trouble. With the post, she makes relevant her non-understanding of Risto's prior turn. However, Jelena offers a candidate solution at the end of her turn ('Sick?'), displaying her understanding that Jegor's situation involves a health issue. Katja answers with a concise description of Jegor's situation in Russian. After that, Jelena gives an account of why she asked for clarification in the first place and what the basis of her candidate was: the word 'hospital'.

This instance differs from Example 1 in that Jelena is not the primary recipient of the trouble turn. Risto tells about Jegor's situation as a piece of news at the beginning of the meeting, without specifically addressing Jelena (or anyone else of the other five persons present). Thus, the progress of interaction is not halted by Jelena's difficulty in understanding the turn. When addressing Katja, Jelena creates a private side conversation ('schisming'; see Egbert, 1997) using the affordance of chat (see, e.g., Meredith, 2017) to clarify a language problem without interrupting the ongoing public conversation. Such side conversations also occur in the face-to-face meetings of this workplace, but in those cases the side conversations involve bodily reorganisations (people turning their head, leaning towards someone), gaze shifts and using a lower or softer voice, which are also publicly observable by co-participants. In video-mediated meetings, such bodily reorganisations are not similarly available to shape the participation framework – if someone switches on their mic and asks a question, it is equally heard by all the participants. The chat interface can be used as an alternative or supplementary communication tool to initiate trouble-relevant side conversations with distinct turn-taking structures (Meredith, 2017). The remote meeting environment thus provides both hindrances and affordances for adjusting the interactional framework to accomplish third-party facilitation.

To sum up, this section has presented one way in which second language speakers make linguistic asymmetries and troubles in understanding relevant in ongoing interaction, that is, by initiating repair. The analysis of the extracts shows how the repair initiation can be achieved with a request for help either in speaking (Example 1) or in writing via the chat directly to the third party (Example 2). In the first extract, the third party enters the conversation only after the producer of the trouble turn somehow fails to answer to the repair initiation in a sufficient manner. It is noteworthy that both types presented in this section are related to problems in understanding: the repair initiations are explicit requests for help concerning linguistic elements, but the participants draw on diverse resources in the technological environment to solve them (see Arminen et al., 2016). These distinguish the examples from the cases in the next section, which seem to serve a more pedagogical function.

4.2 Self-initiated repair

In this section, we examine cases in which the second language speaker makes trouble in speech production relevant and uses self-initiated repair while searching for a word or a correct form of a word. The first of our two examples comes from a morning meeting with ten participants. The extract involves three participants: Risto, Jegor and Katja. Risto has asked Jegor about his current activities, and Jegor has started telling about the marketing of a new online publication. The national broadcasting company, Yle, is about to publish a piece of news on the topic.

Example 3. Announcement. (IMM1405, 11:43–12:00)

  Open in a separate window

Jegor's turn (lines 01–05) contains signs of trouble related to speech production: there are pauses, hesitations, and restarts. His embodied actions also contribute to flagging problems: he turns his gaze upwards twice (Figure 1; Figure 4), he frowns and smiles (Figure 5), and he makes various hand gestures (line 03, Figure 3; line 07, Figure 7). One problematic element in the turn seems to be the right form of the name of the Finnish broadcasting company, Yle. Jegor produces it first in the basic form (nominative case, line 03), a bit later again in the nominative case, and then in the allative case ('to YLE') with slightly rising intonation and a gaze shift towards the camera, frowning slightly (Figure 6). After this, two participants respond: Katja nods, and Risto, who has asked the question, acknowledges Jegor's turn with a confirming response particle joo (line 06), thus claiming to have understood Jegor's turn and encouraging Jegor to continue (see Sorjonen, 2001).

Jegor continues by verbalising his lack of knowledge (line 05) and producing once again the name of the broadcasting company in the nominative case. He then moves on to the next lexeme while making a hand swish. After the gesture, however, Katja enters the conversation, which suggests that she has interpreted the gesture as an indication of giving up (Figure 6, line 07). In overlap with Jegor's turn continuation, she produces the broadcasting company in the adessive case ('on/at Yle'), thus completing the word search (see Lerner, 2004) and repairing Jegor's choice of cases (lines 08–09). Jegor repeats Katja's repair and confirms it with the particle joo (line 10). After this, Jegor directs his gaze forward and moves on to tell about another activity (lines 10–13).

In this example, Jegor displays having trouble in inflecting the word 'Yle'. Word searches combined with a gaze towards the interlocutor often invite them to produce the repair solution (Goodwin & Goodwin, 1986, pp. 53–54), but in the remote meeting environment it is not possible to effectively use gaze alone for targeting recipients (e.g., Luff et al., 2003). However, the hand swish through which Jegor displays that he gives up may trigger Katja to enter the conversation and produce the repair in Finnish at this specific moment (see also Dressel, 2020). The linguistic assistance is slightly delayed, since Jegor has already continued his turn after the problematic moment. In this case, the repair itself is not about achieving intersubjectivity: Risto has already displayed his understanding of the prior turn, and Katja herself has nodded after Jegor's attempt to find the right form. By providing the sought-for form, Katja orients not only to the intersubjectivity between the participants but also to the grammatical correctness of the utterance, which could be seen to introduce a pedagogical aspect in conversation (see Kurhila, 2006). Also, Jegor accepts Katja's help from the position of a less knowledgeable participant: by repeating the correct form he explicitly displays having registered the new form.

The following example comes from another remote morning meeting (eight participants), and it illustrates another instance of repair as a response to hesitancy-marking. The extract begins with Ksenia's explanation of her current duties to Risto's prior question: Ksenia is gathering information for an article about an ecological way of life, which will then be published on the organisation's online forum. The extract shows how, although being a non-primary recipient of the ongoing turn, Katja still attends to Ksenia's trouble in finding the correct form for the word 'environmentally friendly'.

Example 4. Environmentally friendly. (IMM0506, 4:48–5:32)

  Open in a separate window

Even though some delay occurs due to a bad connection, Ksenia is also struggling with speech production: she is fumbling for words (displayed in hesitation sounds in lines 03–06). In particular, inflecting the word ympäristöystävällinen ('environmentally friendly') causes challenges for Ksenia. She utters the word in chunks: first the main part of the word (ympäristöystävä), then the derivational suffix (lii) and finally the plural partitive case ending (siä). After Ksenia has ended her turn in a rising intonation (line 08), both Risto (the primary recipient of the turn) and Katja nod, which can be seen not only to acknowledge Ksenia's turn but also to encourage her to proceed despite the difficulties. Ksenia produces a discourse particle (joo) with a rising intonation, followed by laughter and a hesitation sound (line 10). This prompts Katja to produce a confirming particle, which along with her thumbs-up gesture, results in being interpreted as a go-ahead display by Ksenia (line 11).

Thereafter, Ksenia starts reformulating her utterance: she replaces the word ympäristöystävällisiä ('environmentally friendly') with ekologian (literally 'of ecology') (line 12). Another pause ensues, after which Katja repeats the word ympäristöystävällisiä ('environmentally friendly') (line 14). The prosody of the word makes it sound like a confirmation: there is a rise-fall pitch contour within the word (Hellermann, 2003). (The pause in the middle of the word seems to be related to the bad connection.) By repeating Ksenia's original version, Katja treats the original word choice as adequate and the reformulation as unnecessary.

Katja's confirmation (line 14) comes in a situation where intersubjectivity between the participants is not halted – both Risto and Katja have displayed understanding through embodied means (i.e., nods; lines 09 and 11). Notwithstanding these responses, Ksenia seeks to replace the term by another ('environmentally friendly' -> 'ecology'). Katja seems to react to this replacement, providing Ksenia information about the adequate word choice, and thereby adopting the position of a linguistically knowledgeable participant. It can be argued that both Ksenia and Katja orient not only to the progress of interaction but also to the correctness of linguistic choices, treating the interaction as a potential site for language learning (e.g., Brouwer, 2003; Gardner, 2012; Kurhila et al., 2021).

This section adds to Section 4.1 by presenting another sequential environment for an L2 speaker to make their limited language skills relevant in interaction: by using a self-initiated repair through a word search. In these cases, the trouble is connected to speech production and it is attended to in specific ways. Many observations of word searches in co-present face-to-face interaction also apply to video-mediated interaction (e.g., Hayashi, 2003), but the fractured ecology of remote conversations makes the use of gaze direction for soliciting help from a specific co-participant challenging. However, as shown here, co-participants can still offer help to varying degrees, such as by producing go-ahead gestures, so that their actions are visibly adjusted to the screen (see Uskokovic & Taleghani-Nikazm, 2022). Examples 3 and 4 illustrate moments when intersubjectivity has already been reached before the linguistic assistance; the correction or the confirmation of the word/form thus orients to linguistic correctness, not to understanding, and can be seen as connected to language learning/teaching and encouraging L2 use. Although the participants themselves do not make language learning/teaching explicit but instead focus on the progression of work-related tasks, it is still an additional dimension of the main activity (see also Kurhila et al., 2021).

4.3 Pre-emptive facilitation

The last group of examples features situations in which the third-party facilitation is not responsive to the signs of trouble as clearly as it was in other- or self-initiation of repair. In these instances, no explicit markers of trouble have been produced, but a non-primary recipient of the ongoing talk anticipates that some turn(s) may be challenging for the second language speaker and accomplishes an entry to mediate the conversation. It is worth noting that in our data these "pre-emptive" turns typically occur in conversations that include only three participants and they manifest an orientation to the chosen technology as a potential contributor to the understanding issue.

Example 5 is from a meeting between Jelena, Katja and Maria, a representative of a collaborating museum. They are planning an event involving Russian-speaking museum assistants. Maria, Jelena and Katja try to come up with some question(s) which the assistants would then answer during their presentation. Most of the meeting has been held in Finnish, but Jelena has used English in her more substantial contributions (as in this example, beginning from the line 16). Jelena and Maria have their cameras on, whereas Katja participates in the discussion with her camera turned off.

Example 5. Did you understand? (ECM2209, 29:39–30:50)

  Open in a separate window

Maria's proposal for the question emerges in her long and somewhat fumbling turn (lines 01–12). Maria ends her turn with an explicit question, seeking an assessment of her proposal (line 12). A pause follows (line 13), after which Katja takes the turn. She first addresses Jelena in the third person (mitäs Jelena sanoo 'what does Jelena say'), after which she directs a question to Jelena in the second person (line 14). Katja's turn can be seen to achieve two functions: she makes explicit that Jelena is the right person to answer the collaborator's question, and she checks whether Jelena is able to do it. Thus, Katja's turn shows that a contribution is expected from Jelena. However, Katja's turn not only demonstrates that the collaborator's question belongs to Jelena's area of responsibility, but it also implies that the question was not easy to understand. By verbalising the possibility of Jelena not understanding Maria's (long and fragmentary) turn (lines 01–12), Katja paves the way for Jelena to ask for help if needed. At the same time, the turn (line 14) makes relevant Jelena's potential linguistic limitations.

Jelena answers in English that she, indeed, has not understood the prior turn(s) (line 16), but she continues by verbalising the source of her trouble: the bad sound quality (lines 17 and 19), which Jelena has also oriented to during Maria's turn by touching her earpiece (line 09) (see Oloff, 2018). By identifying the trouble in understanding as stemming from the technology, Jelena can downgrade the role of her (limited) Finnish language skills (see Rintel, 2013). Furthermore, to remedy the situation, Jelena asks Maria simply to repeat the question (lines 21–22), not to reformulate it in English. In other words, Jelena admits the understanding problem but indicates with her actions that it originates from the chosen modality. Nevertheless, as a response to Jelena's turn, Maria reformulates the question (proposal) in English (lines 23–29). Thus, Maria is open to the possibility that Finnish language may be a source of trouble, too, and adapts her language choice to that of Jelena. During Maria's turn, Jelena claims understanding by nodding (line 27 and 28) and then starts evaluating the proposal in English (line 32).

In Example 5, the facilitation concerns the interactional aspect more than the linguistic aspect: by directing the question to Jelena, Katja makes it easier for her to take a turn, and by checking Jelena's understanding, Katja allows her to acknowledge her possible linguistic problems. Checking about understanding, however, might be face-threatening, since it makes visible Jelena's limited language skills. In this case, the technology has a double role as a problem-source and a face-saving resource: Jelena can refer to poor sound quality instead of non-understanding (see also Rintel, 2013). The potential delicacy of topicalization of limited language skills is also reflected in the fact that talking about language problems often includes laughter or smiling (see, e.g., Kurhila, 2006, Vöge, 2011) – including in this case (line 22), even though the trouble is attributed to technical problems instead of linguistic ones.

Our final extract illustrates another case of pre-emptive facilitation. This time the facilitator provides a simplified version of the previous turn. The example comes from a meeting between Jelena, Katja and Risto. They need to agree on a time for their next meeting and are discussing Friday as a potentially suitable day. In the extract, Jelena's camera is off, which makes it difficult for the others to interpret whether she is on board with the joint decision-making.

Example 6. At ten or at one? (IM1506, 5:00–5:39)

  Open in a separate window

Katja and Risto start by discussing another existing commitment for Friday (lines 01–03). Katja then begins to make a proposal (line 05), but cuts this off to ask for Jelena's schedule (line 06). Jelena and Katja jointly establish Friday as the suitable day for the meeting (lines 08–12), after which Risto suggests two alternative times ('at ten or at two o'clock', lines 14–15). However, Risto continues his turn by repairing the later time option from 'two' to 'one' (line 16). After Risto's turn, no immediate response comes from Jelena. She does not request help in any explicit way (i.e., verbally flagging a trouble), but her camera is turned off, which makes it impossible for the other participants to interpret her embodied displays or actions related to the proposal. Katja then reformulates the alternatives in a simplified form, articulating them clearly and pausing between the two alternatives (line 19; see Kunitz & Majlesi, 2022). She thus transmits Risto's proposals to Jelena and points out that an answer is expected from her. A pause follows (line 20), and then Jelena says she prefers the later time (one o'clock) (line 21). Then Katja recycles Jelena's answer and starts giving her own rationale for the preference (line 22).

In Example 6, Katja facilitates the conversation for Jelena by clarifying Risto's turn even before any trouble has emerged. By repeating the key elements (the suggested times), Katja both clarifies the final options and makes explicit that she, at least, expects a response from Jelena. Katja's turn could be considered a case of pre-emptive facilitation that resembles intra-language brokering (see Harjunpää, 2021).

In this last section, the (possible) troubles in interaction are not manifested by the L2 speaker as clearly as in previous sections (see Kunitz & Majlesi, 2022). However, the assisting turns are preceded by features that can contribute to the facilitator's interpretation that clarification might be needed: the preceding turn is either long and fragmentary or contains self-corrections. In addition, the L2 recipient's facial expressions or gestures may be seen to indicate trouble in reception (Example 5, especially lines 9–10). This resonates with Harjunpää's (2021, p. 154) observation on how "brokers anticipate others' needs for 'linguistic assistance' based on embodied conduct". However, if embodied conduct is not available, as is the case in Example 6 where the L2 speaker's camera is not on, the facilitator is left with her own interpretation of the potential complexity of the verbal turn. Pauses in responses could indicate trouble, but in remote interaction the timing in turn-taking is not so accurate: lags occur and connections may be unstable. In other words, at least in our data, if visual information is not available, the facilitator decides if linguistic assistance is needed based on her own interpretation of the previous turn rather than on the actions by the L2 speaker. The facilitation need not be direct language help such as translations or corrections; it can manifest an attempt to make it easier for the L2 speaker either to participate in the conversation or to bring up troubles that might hamper participation. Such action is close to how Mondada (2012) describes the work of a mediator/facilitator: giving advice about how to speak or what to say.

5. Conclusions

This study has investigated third-party entries into an ongoing interaction in technology-mediated meetings. Conversation analysis (CA) was used to examine moments in which a person, who was not the primary recipient of (potentially) problematic talk provided linguistic assistance for the primary recipient, who had less language proficiency. Our analysis shows that third-party entries are carefully coordinated and accomplished through the participants' verbal and multimodal conduct, calling also for reflexive ways to deploy the resources of the social-digital environment.

We have shown that in remote work meetings, third-party facilitation may take place on at least three different levels. First, the facilitator can use repair practices to ensure intersubjectivity and the progression of the conversation in essentially the same way as when problems of understanding occur in interaction (e.g., Schegloff, 1992, 1997). Secondly, third-party entries may have pedagogical objectives, as evidenced by instances where the facilitator repeats and/or corrects words and word forms even in situations where this would not be necessary for the progress of the interaction. The embodied signs of encouragement may also be seen as serving pedagogical aims (e.g., smiling or giving a thumbs-up). Thirdly, facilitation can be observed to modify or clarify the participation framework (e.g., Goodwin & Goodwin, 2004), even when there are no clear repair initiators or indications of a problem. Examples of this include the simplification of previous turns and topicalisation of a potential language problem. These findings illustrate how verbal strategies may hold special relevance for third-party facilitation of remote interaction, especially when not everyone has their camera on, but multimodal resources still have a role in the process. L2 speakers can flag troubles partly in similar ways as in co-present settings (e.g., via hesitations and gazing up during halts in an own turn).

Our study shows how recruiting and offering linguistic assistance in remote meeting environments manifests a complex process, and it is accomplished through the participants' reflexive ways to draw on the resources available in the setting. One aspect contributing to the complexity is the limited opportunity of participants to use their bodies and gaze to display recipiency and indicate troubles in implicit ways (e.g., Luff et al., 2003; see Oittinen, 2020). Whereas in co-present situations body movement, gestures and shifts in gaze direction can be easily deployed to draw attention to problematic turns and/or to invite specific people to contribute to their collaborative resolution (e.g., Bolden, 2006; Greer, 2015; Harjunpää, 2021), here inviting co-participation in trouble-relevant trajectories calls for more work and orientation to the affordances of the setting. In cases where the L2 speaker's trouble is attended to retrospectively (e.g., word searches), the resources may include asking for clarification via the chat interface and producing other-oriented gestures (i.e., adjusting them to the screen). In cases where troubles are attended to pre-emptively, both the L2 speaker and the facilitator orient to the chosen technology as a contributing factor in the potential misunderstanding. The analyses thus illustrated how the participants oriented to the chosen technology as a resource for flagging troubles and recruiting help from specific people (e.g., Example 2), but also as a source for troubles or an affordance for accounting for understanding issues (e.g., Example 5).

Overall, noticing and assessing the need for help and accomplishing a timely entry is a practical problem for the facilitator, which they must solve by interpreting locally produced actions in and across modalities. This also depicts a methodological challenge faced by the analyst: it is not unquestionable what counts as self- or other-initiated repair in problematic sequences, such as in word searches, because not all actions of the participants are necessarily visible to the analyst when they occur in remote interaction. Therefore, more work is needed to substantiate some of the views presented in this paper.

^{1 The article has been written with equal collaboration by the authors. The order of the authors was decided by lots.↩}

^{2 The first language is the participant's native language; the others are ordered according to how much the participant uses them in data.↩}

^{3 Our cases of "pre-emptive repair" are similar to those reported in prior research in that they are forward oriented, that is, "dealing with a potential problem of how a turn will be understood or received by the interlocutor" (Svennevig, 2023, p. 252). However, prior research on pre-emptive repair focuses mainly on practices in a single speaker's turn, that is, the speaker's way to enlarge their turn beyond the projected TCU (Svennevig, 2010, p. 2023). In our cases, it is not the "self" but the "other" who employs pre-emptive practices. This "other" is the third party – a person who is neither the producer of the potentially problematic turn nor its primary recipient. ↩}

^{4 The names in the extracts are pseudonyms. The data have been transcribed following the CA conventions. Words in bold are produced in Russian/Spanish/English.↩}

^{5 The timing of the explanations may be affected by the remote connection, as it is difficult to know if people have lags in their internet connection.↩}

^{6 We present the example messages as they were sent (in Russian, in Cyrillic). The English translation is not verbatim but gives a sense of what the messages are about. ↩}