D&D-submitted-250429-copy Dialogue & Discourse 16(1) 91-113 doi: 10.5210/dad.2025.104 ©2025 Jenny Myrendahl This is an open-access article distributed under the terms of a Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0/). Repair of claimed non-understanding of word meaning in online discussion forum interaction Jenny Myrendal JENNY.MYRENDAL@GU.SE Department of Education, Communication & Learning University of Gothenburg Editor: Patrick G.T. Healey Submitted: 25th May 2023, 29th April 2024, April 29th 2025; Accepted 05/2025; Published online 05/2025. Abstract This article describes how participants in online discussion forums manage claimed non- understanding of word meaning, specifically when one participant displays insufficient understanding by requesting meta-linguistic clarification. Claimed non-understanding refers to cases where a participant signals a lack of understanding in a way that invites repair. By engaging in a short word meaning negotiation sequence, the participants collaboratively repair the issue of claimed non-understanding and can move on with the discussion on topic. In some cases, however, participants behave in ways that break the normative pattern of interaction and do not enter into the anticipated sequence of repair dealing with the lack of understanding. The analysis of these deviant cases reveals the participants’ own normative orientations in repair of claimed non-understanding of word meaning, and thus provide evidence that there is an underlying organization of repair dealing with such issues in online discussion forum interaction. Keywords: word meaning negotiation, repair, online interaction, semantic coordination, lexical pragmatics 1 Introduction In communication, interlocutors regularly run into trouble of understanding what the other person has just said or meant. In such situations, interlocutors typically make use of practices of repair, an organized set of techniques that can be used to draw attention to and rectify such problems (Schegloff et al. 1977). This article addresses the issue of how participants in online discussion forum interaction deal with interactional problems that originate in insufficient understanding of word meaning. Interlocutors facing such issues may deal with the problems by entering into a sequence of word meaning negotiation specifically addressing the situated meaning of the problematic word. Word meaning negotiation (henceforth WMN) occurs when participants who are engaged in a discussion about a particular topic remark on a word choice made by another participant, thereby initiating a meta-linguistic sequence in which the meaning of the word is questioned and up for negotiation. In such sequences, participants typically shift their attention from the discussed topic to the words used and their associated meanings. According to Myrendal (2015) WMN can be sorted into two main types, depending on the origin or cause of the negotiation. One type of WMN encompasses sequences that are caused by disagreement between participants concerning what a word can or should mean in the situated interaction (Myrendal 2019). The other type of WMN occurs when there is insufficient understanding between participants regarding the MYRENDAL 92 meaning of a word, and the participants must initiate repair to restore enough mutual understanding regarding the meaning of the word to be able to continue the discussion on topic. This article presents a study that examines this form of repair: WMN caused by insufficient understanding of word meaning in online discussion forum interaction. This type of interaction was chosen for the study for several reasons. Firstly, because the mode of interaction is asynchronous, participants have more time to contemplate and reflect on the words and meanings they exchange. Additionally, participants in online forums usually have little prior knowledge of each other and must work together to ensure that they understand what they mean by the words they choose during discussions. Furthermore, online forums lack the non-verbal cues present in spoken interaction, so verbalizing processes of interpretation and understanding become more important, and explicit, than in spoken interaction where meta-communicative functions such as grounding and turn-taking are performed through gestures, body language, gaze and prosody (Clark & Brennan 1991). The study adopts a conversation analytic perspective, treating insufficient understanding, or non- understanding, as something that participants publicly display or claim through their written contributions – thus making it observable and available for repair. This approach makes it possible to examine how participants manage problems of understanding in an asynchronous, text-based setting, despite the lack of immediate physical or temporal co-presence. This study aims to contribute to a greater understanding of how instances of claimed non- understanding of word meaning are repaired in human interaction that takes place in an online setting. Providing linguistic insight of this kind is important, as it contributes to building knowledge about the linguistic and communicative strategies that people use to repair claimed non- understandings and clarify meaning in online interactions. This knowledge can both inform theories of language and communication and inform the future design of online platforms and support the development of tools and strategies to facilitate communication and mitigate misunderstandings in online interactions. 2 Repair The notion of repair refers to the practice of interrupting the ongoing sequence of interaction to deal with potential problems in speaking, hearing or understanding. Studies of repair have shown that repairs are a pervasive aspect of dialogue and highly prevalent in naturally occurring interaction (Colman & Healey 2011, Dingemanse et al 2015, Purver et al 2018). Hough and Purver (2013) found that repairs occur approximately once every 25 words in conversational speech. Healey et al (2018) have shown that misunderstandings are typically dealt with on the fly and suggest that “running repairs” is a crucial driver for semantic coordination in dialogue. Repair is used so that “the interaction does not freeze in its place when trouble arises, so that intersubjectivity is maintained or restored” (Schegloff 2007). Repair can thus characterize all kinds of problems ranging from failure to hear or be heard, use of a wrong word or failure to come up with the intended word at a particular moment, to more general problems of understanding. Repair can serve as an interface for cross-disciplinary research between CA and cognitive approaches to human interaction (Albert & de Ruiter, 2018). CA studies have shown that the practice of repair in spoken interaction seems to orient towards a preference organization in which self-initiated repair is the preferred alternative, which means that the speaker who causes the initial trouble also acknowledges this and self-corrects by producing a repair solution. In other cases, it is a participant other than the speaker who initiates the repair, which is called other-initiated repair (Schegloff et al. 1977). Schegloff et al. (1977) suggest that mechanisms for other-initiated repair are essentially techniques for locating the trouble source. When other-initiated repair is initiated, the participant can choose a repair initiation form which specifically locates the trouble source by pointing to the repairable item, or they can choose a repair initiation form which addresses the whole prior turn as REPAIR OF NON-UNDERSTANDING OF WORD MEANING IN ONLINE DISCUSSION FORUM INTERACTION 93 problematic (Drew 1997). Schegloff (2007) points out that other-initiated repair can sometimes be used for other issues than repairing instances of miscommunication, for example to offer a pre- disagreement or a pre-rejection to something said in a previous turn. In this way, other-initiated repair sequences can be used in the same way as delays, hedges and accounts to indicate disagreement with a prior speaker’s assessment, and they also offer the speaker a chance of modifying the utterance facing the possible rejection or disagreement before the open disagreement is a known fact (Schegloff 2007: 102-103). 2.1 Computer-mediated communication and repair In recent years, a growing interest has developed around interaction practices that occur in online environments and researchers have worked to adapt CA to online settings to be able to study computer-mediated interaction (Fitzgerald et al 2017; Giles et al 2017; Harris & Church, 2019; Meredith, 2019; Warren & Paulus 2020). Since CA was originally developed for analysis of spoken interaction, one limitation of CA when applied to online interaction is that it works from the assumption that communication is linear, and that turns follow each other in a chronological sequence (Giles et al. 2017). In computer-mediated communication (CMC), sequential coherence tends to be violated due to disrupted turn adjacency which occurs when messages that are related to each other end up being separated by unrelated intervening messages (Garcia & Jacobs 1999; Marcoccia 2004; Lewis 2005; Gibson 2009; Herring 2010). Although this can also be the case in spoken interaction, i.e., that turns that are related in a sequence are interrupted by intervening talk, this happens at a much larger scale in CMC since there are more participants interacting simultaneously and participants are not in the same physical space when interacting. Several studies on CMC have shown that even when chronologically adjacent messages appear unrelated, interlocutors develop strategies for overcoming the problems of interactional incoherence, making sure that it will be possible to interpret how messages are connected to each other relationally even when the chronological adjacency is disrupted (Simpson 2005; Farina 2018). These different strategies are employed collaboratively and typically make use of the discussion platform’s various interaction affordances (Wellman et al. 2003; Arminen et al. 2016; Meredith 2017). The notions phantom adjacency (Schönfeldt & Golato 2003) and virtual adjacency (Garcia & Jacobs 1999) have been used to describe this collaborative achievement of coherence in CMC, stemming from the communication strategies used by participants to form distinct sequences within the multi-party communication. Farina (2018) points out that posts in a discussion thread only make sense when they are interpreted as part of a sequence, and not if analyzed as separate items. Thus far the practices of repair in CMC have mainly focused on synchronous forms of online interaction, i.e., communication that takes place in real time where participants are mutually co- present in the chat or instant messaging system when they are interacting with each other. Jacobs and Garcia (2013) found that participants in chat room interaction mimic procedures used in verbal communication to avoid and repair interactional troubles, and that the procedures of repair are adapted to the affordances and constraints of the CMC system. On a similar note, Jepson (2005) compared repair move patterns in online voice chats and online synchronous text chats and found that the voice chats contained a significantly higher number of repair moves than the text-based chats. Jepson found that in synchronous text-based chats, that repair initiations were often produced by a repetition of the trouble source, for example by producing a clarification request such as “What do you mean by X?” or a confirmation check such as “Did you mean/say X?”. Lee (2008) studied expert-to-novice synchronous online interaction within a learning environment setting and found that experts used confirmation checks on the form “X?” as a repair initiation device to draw attention to linguistic and grammatical errors made by the novice as an attempt to draw attention to a potential trouble source and influence the novice to make a correction or clarification. Recent work on asynchronous interaction, such as Goddard and Gillespie’s (2025) study of Reddit discussions, has shown that repair initiations are common in these settings but are often left uncompleted. This highlights that while participants orient to the need for clarification, the MYRENDAL 94 completion of repair is not always achieved, especially in environments where responses are delayed or where participants do not remain engaged. Their findings highlight that repair in asynchronous CMC is often unstable and incomplete, emphasizing the need to examine how such sequences develop – or break down – across extended interactional timelines. In summary, research on CMC and repair has concluded that although this form of interaction may suffer from disrupted turn adjacency, participants develop strategies for reconstructing relational adjacency between posts that are not chronologically adjacent. Studies have also shown that participants in CMC use similar repair strategies as in verbal communication, although slightly adapted to the online medium’s constraints and affordances. This study will focus its attention to one source of interactional trouble: issues that originate in insufficient understanding of word meaning. It explores asynchronous online interactions, addressing a gap in the literature that has predominantly examined synchronous online interactions in the context of repair practices. 3 Method and materials This study uses a conversation-analytic approach to analyze forum discussions, a form of asynchronous CMC. According to Meredith (2019), this type of computer-mediated interaction can in some ways be considered more natural than recorded and transcribed verbal conversation, as it can be captured without the researcher’s intervention. The reason the CA approach was chosen to study this form of online interaction is because it is a method specifically designed to examine the structure and organization of talk-in-interaction and has been proven useful for studying processes of interaction and the organization principles on which interaction rests (Sacks et al. 1974; Jefferson 1987; Schegloff 2007; Sidnell 2010). Like face-to-face conversation, online discussions involve turn-taking, repair, and other interactive practices that CA is well equipped to analyze. By using CA to examine online discussions, researchers can gain insight into how these conversations are organized and how participants use language to achieve their communicative goals (Stommel 2008). In addition, since interaction in online discussion forums is asynchronous and involves many participants who interact simultaneously and by overlapping contributions in branching conversations, it can be challenging to analyze online discussions using traditional methods such as discourse analysis, which often assumes a linear sequence of events. CA, on the other hand, is well-suited for analyzing asynchronous discussions because it focuses on the organization of talk and how speakers respond to each other’s actions, and by doing so it offers a micro-level analysis of the structure of talk that the participants themselves organize. CA relies on the assumption that human interaction is a highly organized activity. By closely examining instances of naturally occurring interaction, the researcher attempts to reveal how participants manage their interaction, particularly focusing on how turns are related to prior turns and how the participants themselves display understanding of the communicative practices of which they are a part (Goodwin & Heritage 1990). All discussion posts in the data collection have been analyzed in terms of actions, which is an integral part of the conversation-analytic approach, focusing on how social actions are carried out and interpreted by the participants themselves (Ten Have 2007). CA assumes that turns in interaction are not just serially ordered, but also sequentially organized, which means that they relate to each other in a number of relevant ways (Schegloff 2007). Generally, the primary unit of analysis in CA research is sequences and turns-within- sequences (Heritage 1984). According to Antaki et al. (2005) and van Hooijdonk and van Charldorp (2019) sequentiality in online forums can be considered equal to sequentiality in face-to-face interaction, although clearly manifested in different ways than in verbal interaction. Gibson (2009) suggests that analyzing aspects of sequentiality in asynchronous CMC essentially involves tracking responses through the discussion thread by identifying which posts they relate to in various ways. This study adopts Gibson’s methodological framework, employing a focused approach to investigate how participants navigate word meaning negotiations in online discussions. The study REPAIR OF NON-UNDERSTANDING OF WORD MEANING IN ONLINE DISCUSSION FORUM INTERACTION 95 is further guided by Ten Have’s (2007) specimen-based approach, which supports the detailed collection and analysis of naturally occurring instances that exhibit a particular phenomenon. Some typical CA procedures applied in this study include: an inductive method of data retrieval, identification of regularities and patterns in instances of the researched phenomenon and building analytic accounts of single cases which are both particularized and generalized (Seedhouse 2004). Accordingly, this study carries out a close and in-depth analysis on interaction sequences where apparent non-understanding of word meaning occurs. In line with a conversation analytic perspective, instances of non-understanding in this study are not treated as objective cognitive states but as interactionally displayed or claimed understandings – that is, participants publicly orient to something as a problem of understanding, thereby making it available for repair. The purpose of the study is to examine how these participant orientations to non-understanding are treated as repair in asynchronous, online interaction and will focus on how they are indicated and reacted to, how they are negotiated between participants and how they ultimately are resolved. 3.1 Data retrieval and selection criteria The data used in the study consists of 38 interaction sequences gathered from three large Swedish discussion forums, Familjeliv (www.familjeliv.se), Flashback (www.flashback.org) and Passagen Debatt (www.debatt.passagen.se). The forums were chosen since they, at the time of the data collection, were three of the most popular discussion forums in Sweden. Since then, Passagen Debatt has shut down, but Familjeliv and Flashback are still very well visited and active forums. To identify word meaning negotiation sequences (WMNs) within forum discussions, a systematic search process was employed, using specific utterance-initial constructions as potential indicators of WMN. These constructions, such as “What/How do you mean (by)?” and the repetition of a word as a question, signal a shift in the discussion from the initial topic to a meta-linguistic level. The sequences were retrieved by using five different Swedish variations of the phrase “What/How do you mean (by)?1” as search expressions, specifically targeting these forums. The goal was to identify instances where the conversation shifted from the main topic to a deeper negotiation of the meanings of specific words. The data collection followed a systematic selection method, described in Myrendal (2015, Chapter 3), based on Ten Have’s (2007) specimen-based approach. The searches yielded approximately 1500 initial hits, reflecting 100 results per expression for each of the five Swedish prompt phrases across three forums. Each post was manually reviewed for evidence of a meta- linguistic shift – defined as a transition in the conversation from topical discussion to an explicit focus on word meaning. This shift had to be evident through indicators such as meta-linguistic clarification requests and responses addressing either the meaning potential or the situated meaning of a word. Cases were excluded if they lacked a meta-linguistic shift, such as when responses clarified general context but not the meaning of the targeted word. This ensured that inclusion was not based merely on the presence of a key phrase, but on interactional features indicating actual engagement in word meaning negotiation. Threads were also excluded if they involved highly sensitive or personal content, such as discussions of miscarriage, deceased children, or sexual preferences, which were deliberately omitted for ethical reasons. Although the forums were publicly accessible, the study followed ethical guidelines for internet research (e.g. Markham & Buchanan 2012), which emphasize that public availability does not eliminate the need for caution when dealing with emotionally charged or private material. This approach was adopted to protect participant privacy and minimize the risk of harm or discomfort to forum users. After this filtering, the final dataset comprised 38 analyzed sequences, representing approximately 2.5% of the 1500 1 The five Swedish expressions used in the searches were: Vadå X?, Vad då X?, Vaddå X? (all roughly translating as “What do you mean X?”), Vad menar du med X? (“What do you mean by X?”), and Hur menar du med X? (“How do you mean (by) X?”). These expressions are described in detail in Myrendal (2015: 93). MYRENDAL 96 reviewed posts. The selection of WMN sequences followed the step-by-step procedure outlined below. 1. Locating the post with the problematic word. 2. Identifying the start of the sequence by noting all possible indicators leading up to the meta- linguistic shift. 3. Reviewing the entire thread to determine relevant posts, including those directly discussing the identified word and others contributing indirectly (e.g., through pronouns referring to the word). 4. Determining the end of the WMN sequence as the last post that significantly adds to the negotiation before the conversation concludes or shifts away. This systematic approach for retrieving WMNs ensures thorough coverage of how participants negotiate and clarify word meanings in online discussions, drawing on methods for spotting miscommunication events as detailed by Linell (1995: 185-187). In the excerpts presented in this article, the original Swedish discussions have been translated into English, and the names of the participants have been anonymized into P1, P2 etc. Each turn is labelled in brackets with the ID number assigned to the post by the discussion forum, displaying its consecutive order in the discussion thread. In total, the 38 interaction sequences comprise 149 turns produced by 90 participants. The data collection comprises approximately 12 000 words. The careful selection and anonymization process ensures the integrity and confidentiality of the participants’ contributions while facilitating a detailed analysis of the conversational dynamics present in the forums. The sequences have been manually identified as instances of apparent non-understanding WMNs (Myrendal, 2015), which means that they all have in common that they originate in participant orientations to a lack of understanding about the situated meaning of a particular word used in the discussion. Consequently, all sequences involve one turn in which a word is used that is retroactively identified as problematic, and at least one turn in which another participant requests meta-linguistic clarification regarding the meaning of the word, thus initiating repair. Following the initial use of the word, and the indication that the word is not fully understood in the repair initiation, the discussion unfolds into a short repair sequence focusing on the meaning of the word and how it should be understood. Most of the sequences in the data collection consist of either three or four turns. In the sequences, the first turn is the one containing the trouble source, the second turn is the one initiating repair, the third turn addresses the issue of repair and in cases where there is a fourth turn in the sequence, this generally contains a reaction to the clarification provided in the third turn. All repair initiation posts in the data collection bear great similarities with each other as they contain a meta- linguistic clarification request targeting the perceived meaning problem of the trouble source (mirroring the specific search expressions employed to identify such instances). Drawing on the classification of clarification requests outlined by Purver (2004), the clarification requests observed in our data predominantly consist of non-reprise clarifications. This form involves participants restating or reformulating the information being clarified, often taking the shape of questions like “What do you mean by X?”. This pattern is exemplified in many of the dialogue excerpts in this article, for example in the dialogue of Excerpt 1 (post #29) and Excerpt 2 (post #2), where participants seek further explanation or definition. Excerpt 1 P1 (27): Men ska som sagt till husläkaren och ska säga att jag vill göra en helkroppsscanning och börja med medicin. P1 (27): But as I said, I’m going to the doctor and will say that I want to have a full body scan and start on medication. P2 (29): Vad menar du med helkroppsscanning? P2 (29): What do you mean by full body scan? P1 (31): Alltså en slags röntgen där dom ser inflammationerna. REPAIR OF NON-UNDERSTANDING OF WORD MEANING IN ONLINE DISCUSSION FORUM INTERACTION 97 P1 (31): It’s like an X-ray where they see the inflammations. P2 (33): Jag har aldrig fått nått sånt, var nyfiken på vad det var. Jag har däremot fått händerna undersökta med ultraljud P2 (33): I have never had anything like that, was curious about what it was. I have had my hands examined with ultrasound though. Excerpt 2 P1 (1): Priset brukar väl ofta vara utan resning. Ca 2-3 miljoner för standardhus. P1 (1): The price is normally without putting up. About 2-3 million for a standard house. P2 (2): Vad menar du utan resning? Ursäkta ifall frågan låter knas är inte så duktigt på sånt. P2 (2): What do you mean without putting up? Sorry if the question sounds weird, am not good at this. P1 (3): Dvs pris på själva huset från leverantören, men det är ju en byggsats. Det ska byggas på plats, elektriker, rörmokare osv osv osv. Kostar ca 1 miljon att resa huset. Så plussa på ca 1 miljon på huspriset från leverantören. P1 (3): That is to say the price of the house from the manufacturer, it’s modular. It has to be constructed on site, electricians, plumbers etc etc etc. So plus about 1 million on top of the supply costs. P2 (4): Jaha nu förstår jag bättre. Men hur funkar det med avlopp och allt sånt. För det finns redan en sådan. P2 (4): Oh, I see, now I understand better. But how does it work with plumbing and those kinds of things? Because we already have that. The third turn in the sequence generally determines if the sequence turns into a repair sequence or not. Up until this point, a word has been used by P1 that has been perceived as problematic in some way by P2. In 34 out of the 38 sequences, the requested meta-linguistic clarification is in fact provided in the third turn, which means that the repair initiation is addressed, and repair is carried out in the third turn of the sequence. In the four cases where the expected clarification is not provided in the third turn, the participants overtly address this in the negotiation, making the expectation explicit in the interaction itself. This will be discussed in more detail in the section concerning deviant cases. The fourth turn of the sequence is generally used to wrap up the side-sequence of repair and resume the main sequence of the discussion. In the sequences where the fourth turn is present, this turn is generally used for grounding purposes (Clark 1996), which manifests itself in two main functions: • Confirming understanding of the meaning of the trouble source, which means that the fourth turn is a part of the repair negotiation. • Returning to the main discussion, i.e., exiting the micro-sequence of repair and continuing the discussion on topic. P2’s post #4 in Excerpt 2 contains both functions. Here, P2 first acknowledges that an understanding of the word has been grounded in the ongoing communication (“Oh, I see, now I understand better.”), and then moves on in the discussion on topic (“But how does it work with plumbing and those kinds of things?”). MYRENDAL 98 3.2 Sequentiality in discussion forum interaction One issue to address when analyzing how repair is carried out in asynchronous CMC is to deal with the problems of interactional incoherence and disrupted turn adjacency brought up in previous research. To identify how turns relate to each other and form distinct sequences, there needs to be some way of interpreting how posts are connected to each other relationally even in cases where unrelated posts end up intervening the chronological adjacency of posts (Gibson 2009; Farina 2018). This study therefore investigates the different practices adopted by participants to show how posts are sequentially related to each other in this form of multi-party communication, unveiling how “virtual adjacency” (Garcia & Jacobs 1999) or “phantom adjacency” (Schönfeldt & Golato 2003) manifests itself and how chronologically unrelated posts form a distinct repair sequence in the interaction. One strategy for displaying how posts are related in asynchronous CMC involves explicitly quoting the post, indicating that the new post is responding to the previous one. Many discussion forums have this feature built into the communication interface. Both Familjeliv and Flashback have quoting buttons placed next to every discussion post in each thread, facilitating the quoting practice amongst the participants. Passagen Debatt does not have a quoting button, but instead uses a direct-reply function which makes it impossible to write “the next post” in a discussion thread without explicitly directing that post at a previous post. Another way of making explicit that a post is relating to someone else’s post is by referring to that participant by name (alias) or by use of a pronoun, typically ‘du’ (you, second person singular) or ‘ni’ (you, second person plural). A third way of knowing how posts are sequentially related to each other is by looking at parts of adjacency pairs (Schegloff & Sacks 1973) which are expected to be produced next to each other, although there is no absolute requirement for the two parts to be strictly adjacent in all cases (Hutchby & Wooffitt 2008), since adjacency becomes a matter of interpretation in CMC (Giles et al. 2017). When a discussion post constitutes the second part of an adjacency pair, that post is functionally dependent on a specific previous post, i.e., it is part of a two-part exchange and relates to the first part post, regardless of how far back in the chronological flow of discussion that the first part post is located. When producing the first part of an adjacency pair, such as a question, there is an underlying anticipation that the addressed participant(s) will respond in the expected way, by producing the second part of the adjacency pair. Consequently, a first pair part projects a prospective relevance, establishing certain conditions for how the following turn will be perceived (Schegloff 2007). Most of the repair initiation posts in the second turn component of the 38 sequences contain the pronoun ‘du’ (you), explicitly addressing the person who produced the turn containing the trouble source, requesting clarification regarding the word. A little more than half of the repair initiation turns contain quotes or direct-replies of the post containing the trouble source, making explicit how these turns are related in a sequence. As all repair initiation turns contain meta-linguistic clarification requests pointing out the trouble source, they are all sequentially related both to the post where the trouble source is located, as well as to the response turn expected to clarify the meaning of the word. By providing an answer and repairing the issue of claimed non-understanding of the trouble source (in the initial post), the second turn component (the repair initiation post) and the third turn component (the repair post) are explicitly sequentially related to each other in a kind of adjacency structure. Most of the repair posts and reaction posts also contain quotes or direct-replies of the previous posts in the WMN, explicitly displaying how the posts are related in a sequence. Some repair and reaction posts also include the pronoun ‘you’ or an alias addressing the name of the person who the post is responding to. In summary, participants establish and maintain sequentiality between posts in discussion forum interaction by making use of affordances of the system, for example by quoting, direct- REPAIR OF NON-UNDERSTANDING OF WORD MEANING IN ONLINE DISCUSSION FORUM INTERACTION 99 replying or addressing particular interlocutors by name or personal pronoun, or by posts being sequentially linked in an adjacency structure. 3.3 Ethical considerations Technically, it is easy to collect interaction data from online platforms. However, the collection and processing of such data must be preceded by a series of ethical considerations and reflections on the part of the researcher. At present, there is an international guideline for online research published by the Association of Internet Researchers, but the recommendations regarding an ethical approach in relation to the study of online interaction are relatively general. Therefore, the individual researcher must make his or her own ethical considerations in relation to conducting research on online data (franzke, Bechmann, Zimmer, & Ess 2020). The ethics of conducting research on online discussion forums without participant consent depends on several factors, including the type of research, the potential harm to participants, and the safety measures in place to protect participants’ privacy. The ethical considerations can vary based on the purpose of the study, the method of data collection (such as active engagement with participants vs the use of existing archival data), and the type of online space accessed (Roberts 2015). When it comes to ethical considerations in relation to the collection and processing of online interaction data, there are mainly two aspects that the researcher needs to consider, partly whether the interaction takes place in an open or private space, and partly how sensitive the content of the interaction data is (Markham & Buchanan 2012). In general, it can be said that interaction that the participants themselves perceive as private, for example personal interaction between a low number of participants in a closed group, should not be collected or used without the participants’ consent. In more open spaces, such as public discussion forums where a larger number of participants interact under pseudonyms instead of their real names, participants are likely not perceiving the interaction to be private or closed. In such cases, the researcher can show ethical consideration towards the participants in other ways, for example by avoiding collecting material that deals with sensitive topics and by anonymizing the material as far as possible (Smedley & Coulson 2021). In this study, discussion threads dealing with sensitive issues have been excluded and all names of the participants have been anonymized. The membership agreements of the three discussion forums used in this study allow for data collection for non-commercial purposes and inform users that the forums may be used for private or educational use. Familjeliv specifically reminds users that their written contributions are made in a public space and may be used in ways beyond their control. 4 Analytical findings This section will present the findings of the study, by analyzing how participants initiate and carry out repair in instances of insufficient understanding of word meaning in online discussion forum interaction. 4.1 Repair of claimed non-understanding of word meaning This section analyzes how participants initiate and carry out repair in instances of insufficient understanding of word meaning in online discussion forum interaction. In 34 of the 38 sequences, repair is accomplished in the third turn component, following a recognizable interactional pattern. Four sequences deviate from this pattern and will therefore be treated separately as deviant cases. While the majority of sequences follow a recurring structure, the presence of deviant cases shows that this structure is not imposed by the analysis but emerges from the participants’ own orientations. These deviant cases will be analyzed in Section 4.2. MYRENDAL 100 4.1.1 Initiating repair indicating claimed non-understanding of word meaning In all of the sequences, the repair initiation turn overtly addresses the issue of apparent non- understanding of word meaning by explicitly referring to the trouble source in a meta-linguistic clarification request, typically a variation on the form “What do you mean by X?” as seen in the repair initiation turns in Excerpt 1 and 2. Occasionally, the repair initiation turn can also contribute to the repair by displaying partial understanding of the meaning of the trouble source, for example by suggesting alternative interpretations of the word. This practice is in line with what Heritage (1984) calls proposing “candidate understandings”, which is a common way of initiating repair in spoken communication. Excerpt 3 P1 (1): Vad är dina tankar om du skulle träffa en ogift 39 kvinna som inte har barn? P1 (1): What are your thoughts on if you would meet an unmarried 39 woman who doesn’t have children? P2 (6): Vad menar du med ”Ogift”. Många är ju sambos, särbos, lever i partnerskap etc. Menar du verkligen bokstavligen gift i din fråga eller menar du snarare att kvinnan ifråga är utan partner? Finns ju mängder av folk som lever som sambos och har barn. P2 (6): What do you mean by ”Unmarried”. Lots of people are in relationships and live together, or live apart, or in partnerships etc. Do you really mean married in your question or do you mean that the woman in question does not have a partner? There are lots of people who live together and have children. P1 (16): Hej, jag menar enbart ogift. P1 (16): Hi, I mean strictly unmarried. In the repair initiation post in Excerpt 3, P2 shows that the word ‘ogift’ (unmarried) is partially understood by proposing two candidate understandings (post #6). The clarification request thus targets whether the intended meaning of the word in the current context should be interpreted as literally unmarried or rather as “without a partner”. All repair initiation turns contain a meta-linguistic clarification request addressing the need to repair the issue of claimed non-understanding of word meaning. In addition, some also contribute to the repair by displaying partial understanding of the word by proposing candidate understandings of the word. The following excerpts offer additional instances of repair initiation posts, illustrating a variety of approaches beyond the singular example of an ‘or’/disjunctive question presented in Excerpt 3. 4.1.2 Responding to and repairing instances of non-understanding of word meaning This section will investigate how issues of displayed non-understanding of word meaning are addressed and repaired in the third turn component of the sequence. In sequences where the repair initiation turn includes candidate understandings of the problematic word, P1 typically responds by confirming which of the interpretations is the intended one, as seen in the response posts in Excerpt 3 (post #16). In cases where the repair initiation turn does not provide any partial understanding of the word, the P1 typically clarifies the meaning of the word by adopting one of the following repair practices: explicification or exemplification. One way of responding to a clarification request targeting the the meaning of the trouble source is by producing an explicification, a concept described by Ludlow (2014) that denotes the introduction of an explicit definition-like component to a word meaning under negotiation. Through REPAIR OF NON-UNDERSTANDING OF WORD MEANING IN ONLINE DISCUSSION FORUM INTERACTION 101 acts of explicification, participants may clarify the meaning of the word by introducing a definition- like component that foregrounds aspects of the word’s semantic properties, such as in Excerpt 4 (post #4). Excerpt 4 P1 (1): Jag är antisexist, vilket betyder att jag är emot sexism I samhället! Fråga mig vad ni vill! P1 (1): I’m anti-sexist, which means that I’m against sexism in society. Ask me anything! P2 (3): Vad menar du med begreppet “sexism”? P2 (3): What do you mean by the concept “sexism”? P1 (4): Att människor behandlas olika pga sin könstillhörighet. P1 (4): That people are treated differently because of their gender. In the response post in Excerpt 4, P1 repairs the issue of claimed non-understanding by producing an explicification. P1 here provides a definition-like clarification about the meaning of the word, drawing upon semantic affordances tied to the word itself. Another way of responding to the meta-linguistic clarification request is by providing an example of what could be meant by the word, in the current situation or in situations beyond the current conversational context. This practice is called exemplification (Myrendal 2019), and entails exemplifying or enumerating what the word can denote in the current conversational context, but also in other possible situations. Excerpt 5 P1 (23): Hursomhelst finns det en ordningslista för hur man designar en byggnad och jag tror inte TS greppar det riktigt. P1 (23): Anyway there is an order list for how you design a building and I don’t think TS completely gets this. P2 (35): Vad menar du med ordningslista? P2 (35): What do you mean by order list? P1 (39): Tänk prioritetslista. ”Mitt hus SKA ha en pool” vs. ”Mitt hus KAN ha en pool om förhållandena tillåter”. Det finns vissa saker som man löser först och vissa saker man löser senare. P1 (39): Think priority list. ”My house SHOULD have a pool” vs. “My house COULD have a pool if the conditions allow.” There are some things you solve first and other things you solve later. In the response turn in Excerpt 5, P1 repairs the issue of claimed non-understanding first by producing an explicification (“Think priority list”) and then by producing an exemplification, enumerating what could be meant by the word “ordningslista” (order list). Note the difference between introducing a definition-like component in an explicification and exemplifying what the word can mean. In exemplification, there is no definition of the word that draws upon aspects of the word’s own semantic properties, instead an account of what the word may denote in different situations is provided. The continuation of P1’s utterance can be seen as a part of the initial explicification, as this is interpreted more as a definition-like clarification about the meaning of the word itself, by drawing upon the word’s semantic properties such as the implication of prioritization. 4.1.3 Resolving instances of claimed non-understanding of word meaning As mentioned, the fourth turn component (the reaction to the response post) is typically used to ground understanding of meaning, i.e., to acknowledge that there is enough mutual understanding concerning the the meaning of the trouble source so that the repair sequence can be resolved and MYRENDAL 102 the discussion on topic can be resumed. In the data, the reaction post is only rarely used to continue the negotiation of meaning or by issuing another meta-linguistic clarification request, urging the other participant to continue negotiating the word. However, there are a few instances where the reaction turn is used to continue the sequence of repair, and these instances are found in the sequences constituting the deviant cases which will be described in detail in the next section. 4.2 The deviant cases From the sequential pattern found in most of the sequences, it is clear that P2’s repair initiation post establishes an expectation that P1 will respond with meta-linguistic clarification. In most cases, P1 fulfills this expectation by elaborating on the meaning of the word and thereby resolving the perceived non-understanding. However, in four sequences, no clarification is provided in P1’s third-turn response. Sidnell (2013: 80) writes: “Of course, a recipient sometimes responds in a way that might not have been predicted by the prior turn, or, indeed, does not respond at all. What can we make of such examples? As it turns out, such deviant cases often provide the strongest evidence for the analysis because it is here that we see the participants’ own orientations to the normative structures most clearly.” The four sequences that do not contain repair actions in the response post are treated as deviant cases in this study. They will be described in detail below, focusing on how participants orient to the normative structures by responding to their interlocutors’ unexpected behavior in different and observable ways. Although this study does not aim to quantify occurrence rates, it is relevant to note that these 38 cases were identified from a larger pool of roughly 1500 forum posts. Thus, while word meaning negotiation is not frequent, it is a recognizable and recurrent interactional phenomenon in asynchronous CMC. 4.2.1 Unwillingness to provide meta-linguistic clarification In two sequences, P1 delivers a response post in the third turn component, but without providing the requested meta-linguistic clarification regarding the meaning of the trouble source. In these response posts, the expected clarification is missing, which means that there is a reply to the repair initiation post, but repair is not carried out. In Excerpt 6, P1 is also called TS, thread starter. Excerpt 6 P1 (20): Ja påhittade räkningar vägrar jag ju såklart betala är det konstigt på något sätt? P1 (20): Yes of course I refuse to pay made up invoices is that strange in some way? P2 (21): Vad menar du med påhittade räkningar? Har du en skuld hos fogden? P2 (21): What do you mean by made up invoices? Do you have a debt with the bailiffs? P1 (22): Antar att jag menar det jag säger såklart (-: P1 (22): Guess I mean what I say of course (-: P1 (24): Hade men byt ut faran mot räkningarna så ska du se resultatet.2 P1 (24): Used to but switch the danger with the invoices and you will see the result. P2 (25): Ursäkta men jag förstår inte vad du skriver? Byt ut faran?? P2 (25): Sorry, but I don’t understand what you are writing? Switch the danger?? P3 (26): Hopplös tråd gillar inte när TS inte förklarar ordentligt utan man ska gissa sig till vad hon menar. 2 This utterance does not make sense in Swedish and has therefore been translated into English word for word. REPAIR OF NON-UNDERSTANDING OF WORD MEANING IN ONLINE DISCUSSION FORUM INTERACTION 103 P3 (26): Hopeless thread don’t like it when TS doesn’t explain properly and you have to guess what she means. P1 (27): Sorry fakta ska det stå. P1 (27): Sorry it should say facts. P4 (30): Haha! Vilken rolig tråd... En TS som inte kan skriva och man får dessutom inga rediga svar.. Bara flum.. P4 (30): Haha! What a funny thread… A TS who cannot write and you don’t get any clear answers.. Just fluff… P5 (31): Denna TS har för vana att då och då skapa trådar som ingen begriper och som hon sen inte ger någon mer förklaring till och så vanligen avslutas de med att TS vägrar svara på något mer av en eller annan anledning. P5 (31): This TS has a habit of occasionally creating threads that noone understands and that she doesn’t explain so usually they end with TS refusing to respond for one reason or another. The sequence dealing with the meaning of the word ‘påhittade’ (made up/fictitious) in Excerpt 6 starts out like the other sequences. However, the expected pattern is disrupted when P1 does not provide the requested meta-linguistic clarification in the response post. Instead, P1 responds evasively: “I guess I mean what I say” (post #22), which acknowledges the clarification request but avoids explaining the meaning of the problematic word. This failure becomes a communication problem, which is openly addressed by several other participants. P1 makes another response post, which contains both grammatical and spelling errors, and appears to further confuse the other participants (post #24). A side-sequence clearing up a spelling error is launched, and some participants comment on P1’s reluctance to clarify or engage properly with the thread (posts #25, #30, and #31). This deviant case is analytically important because it shows that the expected pattern – problematic word, repair initiation, clarification – is not a researcher-imposed structure but something participants themselves orient to. The reactions from several of the other participants in Excerpt 6 show that there is a normative structure in place, with the expectation that P1 should provide the requested clarification in the response turn following the repair initiation (post #21). When the expected clarification is not delivered, this is overtly addressed by the other participants, revealing their orientations to the normative structures of what constitutes the expected behavior. While participants clearly treat the absence of clarification as a communicative issue, it should also be noted that the responses from P2, P3, and P4 appear to exhibit signs of irritation, sarcasm, or even strategic exploitation of the repair expectation. The clarification requests and critical comments could thus be interpreted not only as sincere efforts to re-establish understanding but also as actions that portray P1 as uncooperative or unreasonable. This possibility highlights that some cases of claimed non-understanding may blur into more conflictual or mocking uses of the repair sequence, especially when the trouble source concerns sensitive or contested matters. Although the analysis primarily focuses on the organization of repair following displayed non- understanding, these more adversarial dimensions cannot be fully ruled out. This strengthens the claim that word meaning negotiation follows a recognizable sequential pattern and that departures from it are socially marked and made relevant by participants. As such, cases like Excerpt 6 offer insight into the interactional variability and social accountability at play in WMNs. They show that while a recurring structure is empirically observable, it is not presupposed in the analysis. This conclusion can be drawn because the structure is not only followed in most cases but also becomes visible through its disruption. In Excerpt 6, multiple participants explicitly react to the missing clarification, expressing frustration and meta- commenting on the lack of cooperation. This shows that the sequential pattern is treated by participants themselves as normative and accountable. It is through such participant orientations, both to the pattern and to its violation, that the analyst can identify the structure as an emergent MYRENDAL 104 property of the interaction rather than an imposed coding scheme. In this way, the case illustrates how patterns in conversation analysis are not assumptions, but findings grounded in participants’ own displayed understandings of what constitutes appropriate interaction. Next, one participant re-uses the word in a form of comprehension check directed at P1, in another attempt to elicit a repair containing a meta-linguistic clarification (post #32). Excerpt 7 P6 (32): Okej, så något företag har skickat ut påhittade räkningar till dig, som du inte har betalat men inte heller bestridit, och nu har ärendet gått till kronofogden. Är det så det har gått till? P6 (32): Okay, so some company has sent you made up invoices, which you have not paid, but which you have not contested either, and now the matter has gone to the bailiffs. Is this how it went? P1 (33): eller skriver överenskomna summor det roliga är att jag vet inte om några överenskommelser eller vad det avser och har inte skrivit sådana i heller. P1 (33): or write agreed-upon sums the funny thing is I don’t know about any agreements or what they refer to and haven’t signed such agreements either. The participants have still not made any progress regarding the meaning of the word, as P1 has not provided a response post clearing up the issue of claimed non-understanding. However, a few turns later, P1 quotes an earlier post commenting on her unhelpful behavior and contests that she has not been clear about what is meant (post #35). Excerpt 8 P1 (35): Jag sa ju det det är räkningar som sägs vara överenskomna och jag har ingen aning har dock inte gjort några överenskommelser och vet heller inte vad det avser. P1 (35): I told you they are invoices that are said to be agreed upon and I have no clue have not made any agreements and don’t know what they refer to. P5 (38): Varför har du då inte valt att bestrida räkningarna? Man kan inte bara låta räkningar ligga, inte ens om de är felaktiga, för då hamnar det ju hos kronofogden till slut. P5 (38): Why haven’t you contested the invoices then? You can’t just let invoices sit, not even if they are incorrect, because they will end up with the enforcement authority eventually. In Excerpt 8, P1 finally touches upon aspects of meaning potential of the word ‘påhittade’ (made up/fictitious) (post #35). It has to do with agreements that are said to be arranged but that have not been agreed upon by both parties, i.e., the arrangements do not correspond to reality, according to P1. This is in fact fairly close to the dictionary meaning of ‘made up’ or ‘fictitious’, and the issue of claimed non-understanding of the word likely has to do with P1’s creative use of the word in the current conversational context. The word ‘påhittade’ (made up/fictitious) is typically not used to describe invoices. In the reaction to P1 in Excerpt 8, P5 displays enough understanding of the scenario to produce a follow-up question regarding the topic (post #38). P5 substitutes the word ‘påhittade’ (made up/fictitious) with ‘felaktiga’ (incorrect). Beyond this point in the discussion, the word ‘påhittade’ is not used again. Instead, the word ‘felaktiga’ is used in the remainder of the discussion to refer to the erroneous invoices. Another sequence in which a participant does not provide the expected meta-linguistic clarification in a response to a post initiating repair is found in the negotiation of the word REPAIR OF NON-UNDERSTANDING OF WORD MEANING IN ONLINE DISCUSSION FORUM INTERACTION 105 ‘självständig’ (independent). In this sequence, there is no meta-linguistic shift in the third turn as P1 in his response does not deal with the the meaning of the trouble source (post #3). Again, this is overtly addressed in the reaction to the response post in the fourth turn (post #4). Unlike in the case of ‘påhittade’, the meta-linguistic shift comes in the reaction turn, by a third person entering into the sequence clarifying the meaning of the word in P1’s place (post #4). Excerpt 9 P1 (1): Min önskekvinna. Hon ska vara självständig men ändå tillgiven. Hon ska vara utmanande men bara för mig. Hon ska ha värme men visa kyla och avståndstagande mot utomstående. Hon ska vara en hemmafru som pysslar om hemmet och gör det hemtrevligt som bara en kvinna kan. P1 (1): My ideal woman. She should be independent yet devoted. She should be challenging but only to me. She should possess warmth but show coldness and distance towards outsiders. She should be a housewife who takes care of the home and makes it cozy in a way only a woman can. P2 (2): Vad menar du med självständig men ändå tillgiven? Vad lägger du in för kriterier vad gäller självständighet hos en kvinna kontra tillgivenhet? Det är jag nyfiken på. P2 (2): What do you mean by independent yet devoted? What criteria are you considering regarding a woman's independence versus her devotedness? I'm curious about that. P1 (3): Att min önskekvinna ska ha dessa egenskaper. Jag vill ha tillbaks den trygghet man hade som barn av att ha en kvinna helt för sig själv och som man vet känner något alldeles extra för en. P1 (3): That my ideal woman would possess these qualities. I want to reclaim the security one had as a child, having a woman all to oneself who you know feels something special for you. P3 (4): en självständig kvinna = en kvinna som tjänar egna pengar och kan ta hand om sig själv vilket krockar med din önskan om en hemmafru för en hemmafru får DU ta hand om mer än vad hon tar hand om dig. P3 (4): an independent woman = a woman who earns her own money and can take care of herself, which contradicts your desire for a housewife because a housewife requires YOU to take care of her more than she takes care of you. In Excerpt 9, many of the criteria on P1’s list are listed in pairs, with the first element in the pair in some way contradicting the second element since they are contrasted with each other by the word ‘but’ (post #1). This applies to the first criterion, about being independent yet devoted. P2 requests clarification about the meaning of the word ‘självständig’ (independent), which he seems to regard as difficult to combine with the word ‘tillgiven’ (devoted). Here, it seems that there are aspects of meaning potentials of the two words that P2 cannot merge into one personality trait in a woman. P2 hints that the two words pull into different directions, that they are opposites (post #2). P2 asks P1 to provide meta-linguistic clarification about the the meaning of the trouble source, which is signaled in the repair initiation post asking about which criteria must be met for a woman to be considered independent (post #2). However, P1 does not shift focus to the meaning of the word in the response post, and thus does not enter into a repair sequence concerning the meaning of the word (post #3). Even though P1 attends to the repair initiation post by replying to it, there is nothing said about the meaning of the word, which means that no repair action is carried out in the response post (post #3). P1 simply repeats that these are the qualities he is looking for in a woman. He does not further explain what he thinks these qualities entail, or what the words he is using to describe them mean. MYRENDAL 106 When the meta-linguistic clarification request is not attended to in the response post, another participant enters into the discussion and puts forward an explanation of the word ‘självständig’ (independent). P3 touches upon aspects of meaning potential in the sense that being independent means making money and being able to support oneself. Both sequences that break the mold for what is the typical pattern in cases of claimed non- understanding of word meaning presented in this section manifest the expectations of what is considered “normal behavior” when engaging in a repair sequence dealing with issues of claimed non-understanding. When the meta-linguistic clarification is not provided in the third turn, the participants orient to this lack of clarification by explicitly addressing it in the discussion. Through the collaborative efforts of the participants, they end up finally dealing with the issue of claimed non-understanding of word meaning, just not in the third turn component (as expected). 4.2.2 Being vague in response posts, playing on words The two sequences presented in the previous section were treated as deviant cases since there was no meta-linguistic clarification provided in the third turn in each sequence. Next, two other deviant cases will be described and analyzed. In these cases, there are attempts to take repair action in the third turn, but they are not fully successful. Again, the failure to provide meta-linguistic clarification is overtly addressed in the discussion that follows, highlighting how participants orient to the underlying normative structures when repair has been initiated but not carried out as expected. The first sequence is about the word ‘evighetsgaranti’ (eternity-guarantee) which is used in a religious discussion where one participant uses it as a promotional piece on why believing in God is a good thing. Excerpt 10 P1 (1): Tron på Gud är som en utgångspunkt för allt jag är med om. Inte som en vag gissning utan som det mest verkliga jag vet: livet. Gud som livets ursprung och mål är stabila grejer, ja som jag kan se den enda hållbara utgångspunkten för livet. Evighetsgaranti medföljer ju. P1 (1): Faith in God is the starting point for everything I experience. Not as a vague guess, but as the most real thing I know: life. God as the origin and goal of life is stable stuff, yes as I can see it the only sustainable starting point for life. Eternity-guarantee included. P2 (4): Orsak, mening och evigheten. Det är den där vissheten om orsaken, meningen och evigheten som känns märklig för mig. Vad då evighetsgaranti? Garanti??? Varför är det så säkert att det finns en orsak eller en mening? DET är tankekonstruktioner om något. P2 (4): Cause, meaning and eternity. It’s the certainty about the cause, meaning, and eternity that feels strange to me. What eternity- guarantee? Guarantee??? Why is it so certain that there is a cause or a meaning? THOSE are constructions of thought if anything. P1 (5): Vissheten känns märklig för dig skriver du. Fast jag skulle tycka det var ännu mer märkligt om en människa inte var förvissad om sin egen tro. Evighetsgaranti var bara en liten lek med våra ord, ta det inte så bokstavligt. P1 (5): You write that certainty feels strange to you. But I would find it even more strange if a person was not sure of their own faith. The eternity-guarantee was only a little play on words, don’t take it so literally. P2 (6): Hur ska jag ta ordet evighetsgaranti om inte bokstavligt? Du säger ju att du är förvissad. Är du förvissad (känner att du REPAIR OF NON-UNDERSTANDING OF WORD MEANING IN ONLINE DISCUSSION FORUM INTERACTION 107 har garanti) för att du ska få evigt liv, eller är du inte förvissad om det? Eller är du bara förvissad om att du tror att du kommer att få evigt liv? Det är faktiskt en viktig distinktion. P2 (6): How should I take the term eternity-guarantee if not literally? You say you are sure. Are you sure (feel that you have a guarantee) that you will have eternal life, or are you not sure about it? Or are you just sure that you believe you will have eternal life? It’s actually an important distinction. P1 (8): Tack för dina synpunkter. Orden är allas vårt redskap att skapa begriplighet med allvar och lek och den här gången blev mitt försök med evighetsgaranti ingen hit. Beklagar det men jag har ingen möjlighet att fortsätta diskussionen ikväll. P1 (8): Thank you for your point of view. Words are everyone's tool to create understanding with seriousness and play, and this time my attempt with an eternity-guarantee wasn’t successful. I regret that but I am unable to continue the discussion tonight. In Excerpt 10, P2 requests clarification about the word ‘evighetsgaranti’ (eternity-guarantee), indicating the need to specifically address the meaning of this word (post #4). P1 responds to the post but does not take repair action by providing meta-linguistic clarification concerning the meaning of the word. This excerpt exemplifies a complex case that navigates the boundaries between a genuine request for clarification, and potential disagreement about the appropriateness of a chosen word and its meaning. P2’s request indicates a need to specifically address the meaning of ‘evighetsgaranti’ (post #4), prompting P1 to respond. However, P1 does not provide a meta- linguistic clarification but states that the situated meaning of ‘evighetsgaranti’ should not be taken literally (post #5). This response acknowledges the word’s literal interpretation but dismisses its relevance in the current context, suggesting instead that it was merely a “little play on words”. P2 seems dissatisfied with the lack of clarification and repeats the request, emphasizing a literal interpretation as the only conceivable meaning (post #6). This insistence on a literal interpretation can be seen as P2 potentially contesting the adequacy of P1’s initial explanation rather than simply seeking clarification. Here, the example highlights a nuanced negotiation about word meaning, underscored by P2’s repeated questions which may suggest a deeper, possibly critical engagement with the underlying assumptions of P1’s faith-based assertions rather than lack of clarity. This dynamic is reinforced when P1 ultimately retracts the use of ‘evighetsgaranti’ altogether (post #8), signaling a concession that the term might have introduced more confusion or controversy than clarity. The discussion straddles the line between insufficient understanding and disagreement, reflecting the dialogical nature of WMN where words serve not just as means of communication but as tools for negotiating deeper ideological and existential dimensions (Myrendal 2019). The repeated request for clarification in post #6, the use of multiple question marks in post #4, and the challenges to the provided explanations suggest that the sequence may not solely be about seeking understanding, but also about contesting the word, its meaning and implications. However, the example shows that participants orient to otherwise invisible structures when their interlocutors behave in ways that break the normative pattern of communication. When a response post that may be anticipated to provide clarification regarding a trouble source instead contains vagueness or word play, the issue of claimed non-understanding remains to be cleared up as the participant requesting clarification is essentially left without a proper response and therefore adopts alternative strategies for coping with the persisting issues of claimed non-understanding. In this case, we see that both strategies from the two deviant cases described earlier are employed, namely reiterating the clarification request and supplying an answer oneself regarding what the word can (must) mean in the given context. MYRENDAL 108 Yet another example of a deviant case containing word play in a response post is found in the repair sequence of the word ‘mannar’ (menfolk). The discussion is about a person, P1, who wants to start up a tattoo removal business abroad and who has a few questions regarding the funding of that business. Excerpt 11 P1 (1): Får jag bidrag av svenska mannar om jag öppnar upp företag utomlands, eller ska ja registrera företaget i Sverige o flyga runt o ta bort tatueringar… P1 (1): Will I get financial support from Swedish menfolk if I open up the company abroad, or should I register the company in Sweden and fly around and remove tattoos… P2 (2): Vad menar du med mannar? P2 (2): What do you mean by menfolk? P1 (3): Givetvis menar jag män i skor.3 P1 (3): Of course I mean men in shoes. [Bold font in original] P2 (5): Menar du alla män i landet som bär skor, eller vadå? Går det inte lika bra att du får bidrag ifrån kvinnor med höga klackar? P2 (5): Do you mean all men in the country who wear shoes, or what? Wouldn’t it be just as good if you got funding from women in high heels? In Excerpt 11, P2 reacts to the use of the word ‘mannar’ (menfolk) and requests clarification (post #2). P1’s reply in post #3 does not resolve the confusion but instead introduces a pun, playfully complicating the conversation. The Swedish phrase ‘män i skor’ translates directly to ‘men in shoes’, but phonetically it resembles ‘människor’, meaning ‘people’. This linguistic twist merges the literal translation with a humorous nod to the more inclusive term ‘people’. For the second time, P1 highlights a distinction between the gender-neutral word ‘människor’ (people) which could have been used and the two word choices specifically focusing on men (menfolk and men in shoes). What P1 appears to be asking, is if he will receive financial support from the Swedish government, i.e., the Swedish people, so explicitly asking about Swedish ‘mannar’ (menfolk) or ‘män i skor’ (men in shoes) seems to draw attention to a difference between the word choices foregrounding male citizens and the neutral ‘människor’ (people). This is also picked up in the reaction post, as P2 explicitly asks if financial support coming from women would not be just as good (post #5). At this point, P1 has been vague in the response post, by employing a play on words, and it is still not clear if ‘mannar’ (menfolk) should be interpreted as ‘people’ or ‘male people’. In the reaction post, P2 makes the other, more specific clarification request, drawing attention to the distinction between male and female taxpayers. The reaction post indicates the need to keep repairing the issue of claimed non-understanding, until sufficient understanding of the trouble source has been established. However, there is no uptake from P1 with regards to the second clarification request. In fact, the reaction post is the very last post in the entire thread. In this case, we see yet another example of how a participant addresses behavior that does not follow the expected pattern of interaction. When the requested meta-linguistic clarification is not provided and the issue of claimed non-understanding is not attended to in the expected place, this issue is overtly addressed, and the clarification request is repeated. In this case, it is made more specific and simultaneously adds to the negotiation of the word, asking if it should be understood as ‘male people’ or just ‘people’. It could be argued that this sequence is also a borderline case between a WMN originating in claimed non-understanding and a disagreement, following Myrendal’s categorization (2015), seeing that the meta-linguistic clarification request may be 3 The Swedish phrase ‘män i skor’ is a word play on ‘människor’ which, when written as one word, means ‘people’. However, when split into three separate words, it literally translates to ‘men in shoes’. REPAIR OF NON-UNDERSTANDING OF WORD MEANING IN ONLINE DISCUSSION FORUM INTERACTION 109 interpreted more like a meta-linguistic objection (commonly found in WMNs originating in disagreement), seeing that P1 uses “givetvis” (of course) before employing the word play “män i skor” (men in shoes) and P2’s somewhat sarcastic response about “women in high heels”. 5 Concluding discussion This study shows how participants in online discussions manage to organize their interaction, in instances where there is insufficient understanding regarding the meaning of particular words, by participating in WMN repair sequences. By adopting practices of establishing explicit relations between posts that may chronologically be very far apart in the discussion flow, participants are able to form distinct sequences of repair to deal with the issues of claimed non-understanding. Establishing this kind of explicit sequentiality is crucial in overcoming the problems of disrupted turn adjacency and interactional incoherence often said to be characteristic of this form of communication. Most repair sequences in this study display a similar three- or four-turn pattern. In the first turn component, a word is used that is later recognized as difficult to understand. In the second turn component, the word is addressed as a trouble source and repair is initiated. In the third turn component, the issue of claimed non-understanding is typically addressed and repaired, and in cases where there are four turn components in the repair sequence, this typically contains a reaction confirming understanding of the word, thus exiting the sequence of repair, and resuming the main discussion on topic. The analysis reveals that most of the issues of claimed non-understanding are solved quickly and easily between participants. When faced with a meta-linguistic clarification request, most participants who produced the trouble source provide a response that quickly clears up the issue of claimed non-understanding, either by defining the meaning of the word, or by exemplifying what the word can denote in the world. Consequently, participants seem willing to cooperate concerning the meanings of the words used in communication, both when producing repair initiators and when agreeing to elaborate on and explain the meaning of a word indicated by someone else as problematic. It is interesting to note that when participants do not provide an explanation of the meaning of the trouble source, subtle yet revealing insights into participants’ orientations to the normative structures underpinning repair sequences emerge. The findings from the analysis of the deviant cases indicate that when the expected meta-linguistic clarification is not provided in the third turn of the repair sequence, the participants overtly address this as a communicative issue that needs to be resolved. This demonstrates that participants seem to anticipate a certain response when a clarification request is directed towards someone; specifically, the expectation is that the recipient of such a request will offer an answer to clarify the issue of claimed non-understanding. This anticipation, and the way participants address the absence of clarification, aligns with the broader concept of preference organization (Goodwin & Heritage 1990). While Goodwin and Heritage primarily discuss the notion of preference in the context of agreements (being produced promptly) and disagreements (being marked by delay and mitigation), the expectation for clarifying responses can be seen as part of the conversational machinery that participants navigate. This machinery, illustrated by preference organization, is underpinned by rationales that favor certain conversational outcomes over others. In the context of repair sequences, the anticipation of a clarifying response reflects a preference for resolving ambiguities and misunderstandings, embodying the normative expectation that communication should progress towards mutual understanding. The findings show that when the expected meta-linguistic clarification is not provided in subsequent turns, participants employ several strategies to address the lack of clarification as an interactional problem. These strategies include producing repeated repair initiations, prompting the other participant to collaborate on restoring enough mutual understanding, and speculating about what the trouble source word must mean in context. Participants’ perseverance in wanting to stay MYRENDAL 110 in the repair sequence until sufficient mutual understanding has been reached demonstrates the importance of such repair work for maintaining coherence in communication. Thus, the findings reveal an order of preference governing online repair of claimed non-understanding of word meaning: once participants have entered a repair sequence targeting the meaning of a word, they prioritize staying within it until the issue is resolved. In the deviant cases, if the participant expected to clarify the meaning of the trouble source fails to do so, other participants seem to prefer to advance the repair themselves rather than allow the interaction to stall. This highlights an orientation towards maintaining the interaction’s progressivity, as participants continually take necessary actions to resolve claimed non- understandings and resume the discussion on the main topic. These findings resonate with earlier CA research on spoken interaction, underscoring how participants prioritize the progression of talk. Previous studies have explored how participants focus on progressivity, for instance, by displaying that a repair move has made progress towards the solution of the trouble addressed (Schegloff 1979) and by showing a preference for answers over non-answers in responses to questions seeking specific information (Stivers & Robinson 2006). Moreover, the deviant cases sometimes reveal a complex interplay between claimed non-understanding and potential disagreement, where the nature of the interaction may initially appear to be about clarifying issues of insufficient understanding but can subtly transition into a borderline disagreement concerning the underlying meaning or use of the word (Myrendal 2019). This study’s findings resonate with Goddard and Gillespie’s (2025) large-scale analysis of Reddit interactions, which showed that while repair initiations are widespread online, many go unanswered, particularly in short threads. However, the present study demonstrates that when a thread continues and the participants remain engaged, interlocutors treat clarification requests as socially accountable actions that call for a response. Unlike cases where the thread simply dies out, a non-answer in an ongoing thread is seen as a marked absence, prompting meta-comments and criticism. This comparison highlights that even in environments where uncompleted repairs are statistically common, active non-responses are interactionally dispreferred – reinforcing the normative expectation to engage in clarification when misunderstandings are made explicit. The significance of investigating online interaction and its linguistic practices lies not only in its widespread prevalence in our daily lives, but also in the fact that digital tools and platforms play a critical role in shaping communication of today. These tools enable interaction to take place asynchronously, opening new ways for discussion, negotiation, and sense-making. This study sheds light on repair practices in asynchronous online interaction, showing that they play a crucial role in the overall organization of the interaction and facilitate the progression of the discussion on topic. Knowledge of these practices is important not only for advancing linguistic theory based on empirical studies of online interaction, but also for the future development of online platforms. Such knowledge can inform the design of tools and strategies that more effectively support successful communication and mitigate misunderstandings in online settings. Acknowledgements This work was supported in part by Swedish Research Council grant 2022-02125, Not Just Semantics: Word Meaning Negotiation in Social Media and Spoken Interaction. References Saul Albert and J.P. de Ruiter (2018). Repair: The Interface Between Interaction and Cognition. Topics in Cognitive Science 10: 279-313. REPAIR OF NON-UNDERSTANDING OF WORD MEANING IN ONLINE DISCUSSION FORUM INTERACTION 111 Charles Antaki, Elisenda Ardévol, Francesc Núnes and Agnès Vayreda (2005). ““for she knows who she is:” Managing Accountability in Online Forum Messages.” Journal of Computer- Mediated Communication 11(1): 114-132 Ilkka Arminen, Christian Licoppe and Anna Spagnolli (2016). “Respectifying Mediated Interaction. Research on Language and Social Interaction 49(4): 290-309. Herbert H. Clark (1996). Using Language. Cambridge: Cambridge University Press. Herbert H. Clark and Susan Brennan (1991). Grounding in communication. In L. B. Resnick, J. M. Levine & S. D. Teasely (Eds.), Perspectives on Socially Shared Cognition, 127-147. Washington: American Psychologist Association. Marcus Colman and Patrick G.T. Healey (2011). The Distribution of Repair in Dialogue. In: Proceedings of CogSci 2011 33rd Annual Meeting of the Cognitive Science Society, 1563- 1568. Mark Dingemanse, Seán G. Roberts, Julija Baranova, Joe Blythe, Paul Drew, Simeon Floyd, Rosa S. Gisladottir, Kobin H. Kendrik, Stephen C. Levinson, Elisabeth Manrique,Giovanni Rossi, and N.J. Enfield (2015). Universal Principles in the Repair of Communication Problems. PLoS ONE, 10 https://doi.org/10.1371/journal.pone.0136100 Paul Drew (1997). Open' Class Repair Initiators in Response to Sequential Sources of Troubles in Conversation. Journal of Pragmatics 28: 69-101. Matteo Farina (2018). Facebook and Conversation Analysis. The Structure and Organization of Comment Threads. London: Bloomsbury Academic. Richard Fitzgerald, William Housley, and Sean Rintel (2017). Membership Categorisation Analysis. Technologies of Social Action. Journal of Pragmatics 115: 51-55. aline s franzkeAnja Bechmann, Michael Zimmer, and Charles Ess (2020). Internet Research: Ethical Guidlines 3.0. https://aoir.org/reports/ethics3.pdf Angela C. Garcia and Jennifer B. Jacobs (1999). The Eyes of the Beholder: Understanding the Turn-Taking System in Quasi-Synchronous Computer-Mediated Communication. Research on Language and Social Interaction 32: 337-67. Will Gibson (2009). Intercultural Communication Online: Conversation Analysis and the Investigation of Asynchronous Written Discourse. Forum: Qualitative Social Research 10: 1- 18. David, Giles, Wyke Stommel, and Trena Paulus (2017). The Microanalysis of Online Data: the Next Stage. Journal of Pragmatics 115: 37-41. Alex Goddard and Alex Gillespie (2025). Conversational repairs on Reddit: Widely initiated but often uncompleted. PLOS ONE, 20(1): https://doi.org/10.1371/journal.pone.0316618 Charles Goodwin and John Heritage (1990). Conversation Analysis. Annual Review of Anthropology 19: 283-307. Jess Harris and Amelia Church (2019). Methodological Insights from Ethnomethodology and Conversation Analysis. Journal of Pragmatics 143: 201-204. Patrick G.T. Healey, Gregory J. Mills, Arash Eshghi, and Christine Howes (2018). Running Repairs: Coordinating Meaning in Dialogue. Topics in Cognitive Science 10: 367-388. John Heritage (1984). A Change of State Token and Aspects of Its Sequential Placement. In Structure of Social Action, ed. by J. Atkinson and J. Heritage, 299-345. Cambridge: Cambridge University Press. Susan Herring (2010). Who’s Got the Floor in CMC? Edelsky’s Gender Patterns Revisited. Language@Internet 7. Julian Hough and Matthew Purver (2013). Modelling expectations in the self-repair processing of annot-, um, listeners. In: Proceedings of the 17th SemDial Workshop on the Semantics and Pragmatics of Dialogue. Amsterdam, Netherlands, December 16-18. 92-101. Ian Hutchby and Robin Wooffitt (2008). Conversation Analysis. Cambridge: Polity Press. MYRENDAL 112 Jennifer B. Jacobs and Angela C. Garcia (2013). Repair in Chat Room Interaction. In Pragmatics of Computer-Mediated Communication, ed. by S. Herring, D. Stein and T. Virtanen, 565-87. Berlin: Mouton de Gruyter. Gail Jefferson (1987). On Exposed and Embedded Correction in Conversation. In Talk and Social Organisation, ed. by John R.E Lee and G. Button, 86-100. Clevedon: Multilingual Matters. Kevin Jepson (2005). Conversations – and Negotiated Interaction – in Text and Voice Chat Rooms. Language Learning and Technology 9: 79-98. Lina Lee (2008). Focus-on-Form through Collaborative Scaffolding in Expert-to-Novice Online Interaction. Language Learning & Technology 12: 53-72. Diana M. Lewis (2005). Arguing in English and French Asynchronous Online Discussion. Journal of Pragmatics 37: 1801-18. Per Linell (1995). Troubles with mutualities: towards a dialogical theory of misunderstanding and miscommunication. In Mutualities in Dialogue, ed by I. Marková, C. Graumann and K. Foppa, 176-213. Cambridge: Cambridge University Press. Peter Ludlow (2014). Living Words: Meaning Underdetermination and the Dynamic Lexicon. Oxford: Oxford University Press. Michel Marcoccia (2004). On-Line Polylogues: Conversation Structure and Participation Framework in Internet Newsgroups. Journal of Pragmatics 36: 115-45. Annette Markham and Elizabeth Buchanan (2012). Ethical decision-making and internet research: Recommendations from the AoIR Ethics Working Committee (Version 2.0). https://www.aoir.org/reports/ethics2.pdf Joanne Meredith (2017). Analysing Technological Affordances of Online Interaction Using Conversation Analysis. Journal of Pragmatics 115: 42-55. Joanne Meredith (2019). Conversation Analysis and Online Interaction. Research on Language and Social Interaction, 52:3. 241-256. Jenny Myrendal (2019). Negotiating Meanings Online: Disagreements About Word Meaning in Discussion Forum Communication. Discourse Studies 21: 1-23. Jenny Myrendal (2015). Word Meaning Negotiation in Online Discussion Forum Communication. Ph.D Dissertation. Bohus: Ale Tryckteam. Matthew Purver (2004). The Theory and Use of Clarification Requests in Dialogue. Ph. D Dissertation, University College London. Matthew Purver, Julian Hough, and Christine Howes (2018). Computational Models of Miscommunication Phenomena. Topics in Cognitive Science 10:425-451. Lynne D. Roberts (2015). Ethical Issues in Conducting Qualitative Research in Online Communities. Qualitative Research in Psychology 12: 314-325. Harvey Sacks, Emanuel A Schegloff, and Gail Jefferson (1974). A Simplest Systematics for the Organization of Turn-Taking for Conversation. Language 50: 696-735. Emanuel A. Schegloff (1979). The relevance of repair for syntax-for-conversation. In T. Givon (ed.), Syntax and Semantics 12: Discourse and syntax, 261–88. New York: Academic Press. Emanuel A. Schegloff (2007). Sequence Organization in Interaction: A Primer in Conversation Analysis. Vol. 1, Cambridge: Cambridge University Press. Emanuel A. Schegloff, Gail Jefferson, and Harvey Sacks (1977). The Preference for Self- Correction in the Organization of Repair in Conversation. Language 53: 361-82. Emanuel A. Schegloff and Harvey Sacks (1973). Opening up Closings. Semiotica 7: 289-327. Juliane Schönfeldt and Andrea Golato (2003). Repair in Chats: A Conversation Analytic Approach. Research on Language and Social Interaction 36: 241-84. Paul Seedhouse (2004). Conversation Analysis Methodology. Language Learning 54: 1-54. Jack Sidnell (2010). Conversation Analysis. In Sociolinguistics and Language Education, ed. by N.H. Hornberger and S.L. McKay, 492-527. Bristol: Multilingual Matters. Jack Sidnell (2013). Basic Conversation Analytic Methods. In The Handbook of Conversation Analysis, ed. by J. Sidnell and T. Stivers, 77-100. Malden, MA: Wiley-Blackwell. REPAIR OF NON-UNDERSTANDING OF WORD MEANING IN ONLINE DISCUSSION FORUM INTERACTION 113 James Simpson (2005). Conversational Floors in Synchronous Text-Based Cmc Discourse. Discourse Studies 7: 337-61. Richard M. Smedley and Neil S. Coulson (2021). A practical guide to analysing online support forums. Qualitative Research in Psychology 18: 76-103. Tanya Stivers and Jeffery D. Robinson (2006). A preference for progressivity in interaction. Language in Society, 35: 367-392. Wyke Stommel (2008). Conversation Analysis and Community of Practice as Approaches to Studying Online Community. Language@Internet, 5: 1-22. Paul Ten Have (2007). Doing Conversation Analysis: a Practical Guide. London: Sage Publications. Charlotte van Hooijdonk and Tessa van Charldorp (2019). Sparking Conversations on Facebook brand pages: Investigating Fans’ Reactions to Rhetorical Brand Posts. Journal of Pragmatics 151: 30-44. Amber N. Warren and Trena M. Paulus (2020). Postgraduate students’ accomplishment of epistemic positioning through personal experience in online discussion forums. Classroom Discourse 11: 11-40.