1Studies in multimodality have been advancing rapidly over the past several decades. A theory of multimodality, in the most basic terms, considers different means of making meaning not as separate entities but as forming a whole (Jewitt et al., 2016). As Jewitt et al. (2016) explain, there are three main approaches to multimodality: systemic functional linguistics, social semiotics, and conversation analysis. Systemic functional linguistics, developed by Halliday, views language as a social practice instead of a set of formal rules. Meaning is created by metafunctions (experiential meaning, logical meaning, interpersonal meaning, and textual meaning) and is expressed in two dimensions: the paradigm (the systemic choices) and the syntagm (the structure of language). Social semiotics, developed mainly by Kress and van Leeuwen, considers the social aspect of signs and the agency of the sign maker, and focuses on power relations. Conversation analysis (CA), which emerged in the 1960s at the University of Berkeley, is interested in the sequential organization of talk and in the interaction between actions and materials. There are additional approaches to multimodality, such as multimodal (inter)action analysis, developed by Norris (2004, 2019, 2020) and adopted for this study.
2There is considerable overlap between multimodal approaches; most theories of multimodality consider verbal communication to be an important element, but not the only element, of human interaction. Even written communication is considered to be multimodal, as modes such as layout, disposition, colors and more add meaning to the message that is conveyed on paper (Baldry & Thibault, 2006). In face-to-face interaction, modes such as gestures, facial expressions, gaze, and proximity are understood to be semiotic resources that make meaning and contribute to the interaction (Norris, 2004). As the data used for CA-based studies (Ten Have, 1999) evolved from audio recordings to video recordings, and as distance communication evolved from phone conversations, email, and text chat to videoconferencing and virtual online environments, researching the multimodal aspects of communication became more relevant and researchers had no choice but to update their theoretical models used to analyze such computer-mediated interaction (Develotte et al., 2011). Recent studies have adopted the methodological framework of multimodal conversation analysis for online interactional settings (Balaman & Pekarek Doehler, 2022; Çolak & Balaman, 2022; Ro, 2023; Uskokovic & Talehgani-Nikazm, 2022), and multimodal approaches are being applied to both classroom and non-classroom settings in order to further our understanding of gesture, meaning-making, and human interaction (Brown & Eskildsen, 2024).
3Within the field of CALL (computer-assisted language learning), several studies have adopted multimodal approaches to focus on video-mediated foreign language teaching. For example, the ISMAEL corpus (Guichon & Tellier, 2017) was used to study several of the technical and pedagogical aspects involved in using a videoconferencing platform to teach oral speaking skills in French as a foreign language. These include maintaining social presence (Guichon, 2017), orchestrating lexical explanation sequences (Holt, 2021a; Holt & Tellier, 2017; Wigham, 2017), seeking and receiving feedback (Duthoit & Colón de Carvajal, 2017; Vidal & Wigham, 2017), overcoming technical difficulties (Azaoui, 2017), and providing task instructions (Satar & Wigham, 2017). Studies such as these contribute to the training of language teachers, who are likely to practice their profession at least partially online. To cite another corpus that I am familiar with, the VAPVISIO corpus (Cappellini et al., 2023) was created for the specific purpose of studying teacher training. The multiple studies that were made possible thanks to this corpus were able to compare future teachers, who were enrolled in formal master’s training programs, with non-teachers, who also engaged in video-mediated interaction but without formal pedagogical training. Many of these studies used eye-tracking technology (Cappellini & Hsu, 2018; Conklin et al., 2018) to study different phenomena such as alignment across modalities (Cappellini et al., 2022) or the frequency of negotiation sequences (Holt, 2021b). However, not all corpora focusing on video-mediated foreign language teaching involve pre-service teachers in training. Some corpora involve in-service teachers, such as the one used by Satar and Wigham (2023, 2020).
4The monograph entitled Instruction Giving in Online Language Lessons: A Multimodal (Inter)action Analysis by Müge Satar and Ciara R. Wigham contributes to the growing body of research on multimodal (inter)action in the specific context of video-mediated foreign language pedagogy. In the first pages of their work, Satar and Wigham set the stage for what they will focus on, namely instruction-giving in a video-mediated environment. They set out to analyze instruction-giving as higher-level, lower-level and frozen interactions following Norris’s (2004, 2019, 2020) framework of multimodal (inter)action analysis. A higher-level action, such as ensuring that the learner has received written instructions, is analyzed in terms of its multimodal constituents, for example posture, gaze, gestures, text chat, spoken discourse, etc. The authors set out to answer the following three questions: Does instruction-giving (when measured in terms of higher-level actions and their multimodal constituents) change according to (1) task repetition, (2) number of learners, and (3) type of task? The following sections provide outlines and commentary on each of the chapters.
5The authors begin by reminding the reader of the importance of video-mediated communication. This became especially evident during the COVID-19 pandemic, when lots of teaching was hastily moved online. Unfortunately, due to pedagogical and technical reasons, face-to-face teaching skills do not necessarily or automatically transfer over to online teaching. This is why techno-pedagogical skills (Guichon, 2012), such as the ability to use the right tools at the right time and for the right purposes, are so important. To further illustrate the difficulties involved in teaching online from a multimodal interaction analysis perspective, Satar and Wigham introduce two new concepts: (1) semiotic lag, defined as lag between emission and reception of a message, which can cause silence and overlapping talk, and (2) semiotic misalignment, which can happen when different social actors do not have access to the same resources, layouts or functionalities in their respective environments, potentially leading to difficulties in communication.
6Due to these difficulties, social actors need semiotic awareness (Guichon, 2013; Kern, 2015) while communicating online, which can be broken down into several components. First is visual framing, which can take the form of extreme close-up, head-and-shoulders, or head-and-torso shots (Guichon & Wigham, 2016). Next is gaze, which can either be fixed, free, strategic, averted, or directed (Satar, 2013). Next is text chat, which can be used for providing feedback without interrupting the learner, private communication, providing vocabulary, and can act as a blackboard (Hampel & Stickler, 2012). I would add that according to the redundancy principle in multimodal learning, redundancy is normally harmful (Sweller, 2005), but can be useful for beginners of a language. All of these multimodal elements that teachers must consider while teaching online are part of the four teaching competencies defined by Guichon (2009): socio-effective, pedagogical, semiotic, and technological.
7The authors then explain the importance of instruction-giving in task-based learning and teaching. From a socio-cultural perspective, language is not just a tool used to relay information but is a means of acquisition, as learners “learn in interaction and not through interaction” (p. 4). This view is embraced by communicative language teaching (CLT), which is realized in the form of task-based language teaching (TBLT), where learners complete tasks that are similar to real-life situations. Instructions are therefore important because they ensure successful completion of the task, create opportunities for authentic communication, are part of task-based teaching competencies, and constitute a significant amount of teacher talking time. Satar and Wigham point out that instructions can be planned ahead of time (“instructions-as-workplan”), but are not always carried out as planned (“instructions-as-process”). The authors explain that they are interested in “instructions-as-process”, and how they are carried out through higher-level actions (HLAs).
8Satar and Wigham conclude their introductory chapter by citing previous studies that lay the groundwork for their book and identify the research gaps they aim to fill. First, Satar and Wigham (2020) examined how experienced language teachers give instructions, extending Markee’s (2015) research on instruction types. They conducted a multimodal analysis of teachers’ HLAs and, through grounded theory (Strauss & Corbin, 1998), identified thirteen in instructional processes. Their findings highlighted the challenge of managing resources in videoconferencing due to the absence of shared physical artefacts. Additionally, they analyzed lower-level actions (LLAs) and observed that visual framing choices affected the use of semiotic resources such as gestures and facial expressions. Next, Wigham and Satar (2021) studied the use of the print mode through the instruction-giving practices of an experienced teacher, while identifying HLAs that this mode was used for and analyzing the ways in which it was embodied and disembodied.
9The research gaps that the authors aim to fill involve methodology and context. Concerning methodology, they attempt to distance themselves from CA-based research, which, as mentioned above, has been abundant. As the authors explain, “CA is guided by recipient design and, thus, does not enable researchers to focus on mediation or the multimodal complexity of a single social actor’s actions. Furthermore, with a specific focus on micro-analysis, it also does not lend itself to the investigation of intermediate and macro actions” (p. 8). Instead, they highlight the need for multimodal (inter)action analysis (Norris, 2004, 2019, 2020), which, in their own words, is “a method that enables a focus on mediation, as well as the modal density (complexity and intensity) of actions taking place in synchronous online language teaching” (p. 8). Finally, concerning context, the authors point to a lack of research focused on synchronous online instruction-giving offering a multimodal analysis of the LLAs that constitute the HLAs.
10In the second chapter, Satar and Wigham outline the context of their study. The corpus consists of three experienced online British English teachers (one male, two female), one of which was recruited from SpeakPlus, a startup company based in Lyon, France dedicated to connecting teachers with learners through a purpose-built online platform for online video-mediated teaching. The teachers selected for this study had a minimum of two years of online teaching experience and regularly used Skype, one of the pioneering desktop videoconferencing platforms which launched in 2003 and was discontinued in 2025. The teachers, Craig, Karen and Sarah, were complemented by nine volunteer language learners (four male, five female) who were studying in Turkey and had a B1–B2 level of English. The conversations recorded for this study, which did not count for any grade, involved up to two learners at a time. The groups were rearranged in order to answer the specific research questions of the study: Karen repeated the activity with a different pair of learners, and Craig repeated the activity with a single learner.
11The teachers led the learners through three different tasks throughout the study, each of which had different amounts of expected negotiation (Pica et al., 1993). The first lesson was an introductory lesson where each teacher led their preferred activities, the second lesson was a convergent task where the learners had to come to an agreement on something, and the third lesson was a divergent task where the learners had to offer their opinions on vegetarianism. Each lesson lasted 25 to 63 minutes.
12The researchers implemented a novel data collection protocol by collecting not only screen recordings from the teacher’s point of view but also from that of an incognito researcher who connected to the meeting for recording purposes. Learners also sometimes took screenshots. Like most studies based on video-mediated interaction, this one had some technical issues, as one user was unable to record. Semi-structured interviews were carried out via Skype with the teachers and the learners up to three days later using an emotionalist approach (Silverman, 2013) in order to understand what the learners thought of the teachers’ instructions. The teachers were asked questions on their background, experience, and practices and the learners were asked questions on their understanding of the instructions.
13The authors adopt Norris’ (2004, 2019, 2020) multimodal (inter)action analysis framework to answer their research questions. This theoretical approach, according to Wigham and Satar (2024: 1), is “complementary to systemic functional (…) and social semiotic (…) frameworks.” They chose this approach because it “offers explanatory and analytical tools for a fine-grained investigation of interconnections between social actors, material objects, and the world surrounding them” (2024: 2). This theoretical framework, from my understanding of it, can be thought of as a prism that is applied to a single mediated action which is the unit of analysis. The prism splits the action into different levels: HLAs, such as engaging in conversation, LLAs, such as utterances or gestures, and frozen actions (FAs), such as a photo on display. HLAs are foregrounded or backgrounded in the social actor’s attention/awareness through modal density; HLAs that have high modal density receive more attention. Modal density is measured through intensity and complexity. If a mode is extremely important for the HLA (i.e., if the removal of the mode would render the action impossible), then it is intense. Modal complexity refers to the interplay of different modes, and the degree to which they are intertwined. This framework makes it possible to take into account the complexity and multimodal nature of mediated actions. Satar and Wigham give an example: “engaging in conversation is an HLA, which constitutes LLAs comprising, for instance, nods (head movement mode), hand movements (gesture mode) and utterances (spoken language mode)” (p. 18). The authors also mention Norris’s concept of interactional attention, whereby interlocutors interpret each other’s mediated actions and perform their own actions accordingly. In order to ensure that the HLAs are mutually exclusive, the authors use grounded theory (Strauss & Corbin, 1998), which allows them to establish a categorization of HLAs one by one until saturation. Using a bottom-up approach, they watched the recordings multiple times and then defined and compared the HLAs.
14Satar and Wigham used ELAN (Wittenburg et al., 2006) to annotate their data. ELAN is commonly used by gesture researchers because it allows for the creation of multiple annotation tiers, with each tier corresponding to a single mode. In addition to the spoken language mode, they annotated gesture, proxemics, gaze, posture, head movement, print, and layout. For gestures, they used McNeill's (1992) widely used classification scheme consisting of iconics (which describe physical properties of the referent), metaphorics (which depict concepts or ideas), deictics (pointing gestures), and beats (rhythmic gestures with no semantic value). They also added emblematic gestures (Kendon, 1982), which are understood without speech and belong to specific cultures, such as the thumbs-up sign. Proxemics included closeness to the webcam and posture. Gaze was defined according to Satar’s (2013) scheme of fixed, free, strategic, averted, and directed. Head movements were lateral and sagittal movements and directional shifts. Facial expressions included smiles, grimaces, and looks of surprise. The print mode included everything that was written in the text chat or any document shared online. Layout referred to different onscreen windows that were available to the users.
15In the first analysis chapter, Satar and Wigham set out to determine if giving the same instructions twice, to two different learners, has an effect on the ways in which the instructions are given, as analyzed through the prism of HLAs and the LLAs that comprise them. They study two sessions led by Karen to two different pairs of learners: the first pair are two female students and the second pair involve a female and a male student. For the task, the participants need to compare and contrast two gift package deals for a colleague’s departure in order to decide which one to purchase. The chapter is divided into three sections: comparing HLAs, comparing LLAs, and a third section devoted to modal configuration, modal density, semiotic misalignment, foregrounding LLAs, and modal density misalignment.
16Concerning the HLAs, the first observation is that during the second iteration of the task, there are much fewer HLAs, which could mean that the teacher becomes more efficient in giving her instructions. Nearly the same HLAs are performed both times, but during the first iteration, some HLAs are performed many more times, such as checking task understanding and focusing on task accomplishment. There is evidence that the teacher adapts and improves her instruction-giving the second time around. For example, during the first iteration, she responds to a learner request on how the task should be completed, but during the second iteration, she offers this information unsolicited. The authors therefore conclude that it is more efficient to carry out some HLAs proactively before sending the resource and defining roles, such as activating schemata, focusing on study skills and communicating key task information. This adaptation and planning are made possible thanks to the fact that the two pairs of learners have similar backgrounds. The interview with the teacher confirms that she did indeed realize the usefulness in providing instructions before sending the material.
17Next, the authors offer six reasons why the two iterations of the activity are similar from the perspective of LLAs. Although the spoken language mode is intense as expected, in that it is the main mode of communication throughout both iterations, the authors’ framework of LLAs makes it possible to carry out fine-grained multimodal analysis. First, they show that gestures are consciously produced in the center of the webcam frame, especially when explaining ways into the task and when responding to learner requests to clarify the instructions. Screenshots and interview transcriptions show that teacher and learners alike are able to read facial expressions, which can show misunderstanding of instructions, and that gestures can be used as a way to resolve incomprehension, such as the use of metaphoric gestures to explain the concepts of “sameness” and “difference.” Indeed, from a cognitive interactionist approach, corrective feedback is important because it enables learners to focus on the language itself, and gestures can play an important role in this (Inceoglu & Loewen, 2022).
18Second, the authors show that the teacher uses the same gesture types when communicating key task information. In this case, the teacher explains that, in accordance with information-exchange activities, each learner will receive different information. In both iterations, when delivering key task information, the teacher uses a similar combination of iconic gestures to differentiate learner A and learner B, and uses a similar combination of metaphoric and beat gestures to illustrate the concepts of “similarity” and “difference.” This reminds me of Tellier’s (2008) call to establish a common code when using pedagogical gestures, meaning that in order to increase comprehension and efficacy of gestures, the same gesture should be used to refer to the same concept.
19Third, in both iterations, when suggesting ways into the task, the teacher uses a combination of modes to refer to the learners. There is alignment between deictic gestures, spoken language, gaze, and posture, which are all oriented towards the learners’ positions on the screen. The teacher adjusts her gaze, deictic gestures and iconic gestures to refer to learners and allocate their virtual writing spaces for the activity depending on whether the learners are positioned horizontally or vertically on her screen. However, as the authors point out, the learners do not have access to the teacher’s screen.
20Fourth, the researchers show that in both iterations, three modes align when launching the task: the spoken language mode, the gaze mode, and the proxemics mode. In the spoken language mode, the teacher says “okay off you go” both times. This aligns with the gaze mode, as the teacher looks down, showing that she is no longer engaged with the learners and that it is up to them to continue the task. This also aligns with the proxemics mode, as the teacher moves away from the screen. The researchers’ use of screen recordings allows them to observe that the teacher pulls up the task sheet on her screen, and is therefore no longer looking at the learners’ webcam images, which are no longer part of her site of engagement. This analysis further illustrates how transitions between different phases of an activity can be orchestrated multimodally.
21Fifth, the researchers show that the teacher uses the text chat to bimodally carry out various HLAs. One of the advantages of text chat is that it is more permanent than spoken language, as messages remain on the screen, as other researchers have pointed out (Dejean-Thircuir, 2008; Guichon et al., 2012; Lamy, 2006; Nicolaev, 2010; Yanguas, 2010). When the teacher uses the text chat, the authors mention that she “systematically verbalized the content of her textchat messages” (p. 50), which is what I have referred to as meta-semiotic commentary (Holt, 2021a). When managing resources, the teacher uses the text chat to remind the learners what to focus on upon receiving the resource. Analysis of the text chat also allows the researchers to highlight a way in which the teacher improves the second time around: when sending the resource, the teacher first uses the text chat to send the activity files as links, and then updates her method the second time to send them as files through the videoconferencing platform. The authors point out that sending them as links the first time creates semiotic misalignment (Wigham & Satar, 2021) because the teacher has access to them in her personal space before the learners do and the learners do not know what is contained in the link before clicking it. However, sending them as files the second time lets the learners know right away what type of files they are and what they contain. As for study skills and focusing on task accomplishment, the teacher is able to use the text chat to offer strategies on formulating an argument and to provide instructions that remain “frozen” on the screen, such as “talk and compare,” and “write and explain” (p. 51). The text chat was also used to reinforce key vocabulary, which is what I have referred to as intermodal repetition (Holt, 2021: 8). Interviews with the learners revealed that they had some semiotic awareness, as they preferred Skype to Zoom thanks to its ability to conserve the text chat after the session is over. However, when asked whether or not the teacher used the text chat, they responded negatively, exposing a lack of awareness of the teacher’s use of semiotic resources. This highlights one of the pitfalls of methods that rely solely on learner interviews and highlights the robustness of Satar and Wigham’s methodological approach.
22Sixth, the researchers show that despite being in different locations for each iteration, the teacher has a similar site of engagement both times, meaning that her webcam framing, gaze patterns, and layout are similar both times. Her webcam framing is a head-and-torso shot (Guichon & Wigham, 2016) each time, which allows for her gestures to be visible. Her gaze is directed at the learners on the screen both times, and her layout is arranged the same way both times according to her personal preference, which was confirmed in an interview. However, the site of engagement differs for the teacher, the learner, and the researcher. The different window sizes for each participant, as well as the mirrored webcam image of the teacher, which render some of her gestures invisible, all lead to semiotic misalignment. In addition, the authors add that “the modes that carry high or low modal density for each participant may differ resulting in instances of modal density misalignment in which the participants do not have access to the same hierarchy of modes to communicate and interpret meaning” (p. 54).
23The last part of the chapter deals with modal configuration, modal density, and semiotic misalignment. The authors show that when focusing on study skills and defining roles, the spoken language, gesture and gaze modes remain moderately intense. The authors then provide an example of semiotic misalignment which occurs when the teacher makes changes to her site of engagement without announcing it in the spoken language mode, which is the only mode available to all participants. Furthermore, the teacher has access to electronic resources that are invisible to the learners, and the learners have access to the teacher’s webcam image, and thus to her gesture, gaze, facial expressions and postures, which are all covered up on the teacher’s own screen. There is also modal density misalignment in that the print mode has high modal intensity for the teacher, but is unavailable to the learners, for whom the spoken language mode has high intensity. This highlights one of the ways in which the ability to teach in a physical classroom does not automatically transfer over to the ability to teach in a video-mediated setting. This also highlights, I would add, the importance of the aforementioned meta-semiotic comments (Holt, 2021a) which can be used to prevent or attenuate semiotic misalignment.
24The fourth chapter examines whether the number of learners affects instruction-giving. The researchers compare two iterations of a task led by Craig on the same theme as in the preceding chapter (deciding on a gift package). The first iteration has two learners, and the second has one. The site of engagement in both iterations is nearly the same, but there are four participants in the first (the teacher, the two learners, and the observing researcher), and only three in the second. Analysis reveals that as with the previous comparison, Craig is more efficient the second time around, with fewer HLAs. However, there are two failed attempts to launch the task during the second iteration with only one learner. According to the teacher, this is because he could not rely on the other learner’s help in explaining the activity. Involving learners is indeed a widespread and useful pedagogical strategy; when studying lexical explanation sequences in the ISMAEL corpus, I found that teachers often involved other learners in the explanation process by asking them to explain or provide unknown words (Holt, 2021a). However, the researchers attribute the failed launching of the task to Craig’s failure to confirm access to the resource after sending it in the second iteration, and remind the reader of the importance of this step in online teaching.
25Next, Satar and Wigham analyze two extracts where the HLAs were the same in both iterations. The HLAs in question are communicating key task information, focusing on task accomplishment, and launching the task. For the first iteration, Craig uses a complex and dense set of semiotic resources including spoken language, gaze shifts between the text chat and the participants, intense facial expressions, vocal intonation, hand gestures, head tilts, and proximity shifts. However, for the second iteration, the print mode is foregrounded in the teacher’s attention/awareness while this resource is not available to the learner. This creates semiotic density misalignment, as only the teacher’s embodied modes are available to the learner and foregrounded in their attention/awareness. Furthermore, compared to the first iteration, Craig’s facial expressions, gestures, and intonation have lower modal intensity when talking to just one learner. As for the final HLA of launching the task, it is similar in both iterations with hand gestures, eyebrow movements, facial expressions, head tilts, and imperatives followed by silence in the spoken language mode.
26Satar and Wigham then explore LLA configurations during two instances of the HLA of managing resources. While the teacher uses the gaze, print and object handling modes to complete the HLAs in both iterations, the researchers find several differences, allowing them to formulate recommendations. First, the HLA of conforming access to the (correct) source was skipped when there was only one learner, and was only implemented later to remedy a problem. Second, the teacher sends the resource as a link when there are two learners, and is actively engaged with the learners while doing so, ensuring modal density alignment. However, when there is only one learner, he sends it as a file attachment and no longer has access to the learner’s image, leading to modal density misalignment. This is exacerbated by the fact that the teacher does not share his screen. Third, when there are two learners, the teacher completes the HLAs of confirming access to the (correct) resource and opening the resource by asking for verbal confirmation, tilting his head and allowing for long silence as the learners find it. He then opens the resource along with the learners and displays semiotic awareness (Guichon, 2013) concerning semiotic lag (Wigham & Satar, 2021) by acknowledging that it could take a while. All of this is done before moving on to the HLA of reading the content of the resource. However, when there is just one learner, the teacher carries out many HLAs such as sending the resource, announcing the next task stage, focusing on task accomplishment, opening, describing, allocating and reading the contents of the resource, and focusing on task accomplishment, among others, all before confirming access to the resource. When the teacher finally realizes that the learner does not have access to the resource, he uses multiple modes including spoken language, object handling, gaze, gestures and posture shifts in order to send the resource and guide the learner in ensuring that he has access to it. Despite these differences, the HLAs and LLAs are similar in both iterations, especially as the teacher uses the object handling and print modes to scroll through the document on his screen which become the focus of his attention/awareness.
27This analysis allows the researchers to offer three suggestions for managing resources. First, the teacher must confirm that the learner has access to the resource before discussing it or launching the task. Second, when sending the resource, teachers should assure modal density alignment by articulating their onscreen actions. Third, when confirming access to the resource, it is helpful to let there be silence after asking a question and to show listening through gaze and head tilt.
28The final part of the chapter deals with modal configuration and modal density misalignment when managing resources. Concerning modal configuration, there is little difference between the two iterations, as the teacher’s screen layout is nearly identical, and he uses the spoken language, gesture, and gaze modes when focusing on task accomplishment. However, there are episodes of modal density misalignment and semiotic misalignment. When sending a message via text chat, there is modal density misalignment at first because the print mode has high modal complexity for the teacher—his gaze follows his typing and the sent message remains visible on screen—whereas for the learner, the print mode has high modal intensity because it is superimposed onto a darkened webcam image. Semiotic misalignment takes place when the text chat disappears from the learner’s window after a short period of time, therefore erasing one of the text chat’s main advantages. There is also semiotic misalignment when a document is sent via the text chat, because the learner only has access to the teacher’s webcam image—and therefore the gaze mode—and not the document, meaning that the sending of the resource is not a foregrounded action for him. The researchers therefore suggest that teachers ensure semiotic alignment between interactional spaces, and not assume that these will correspond automatically. When using the text chat in particular, the teacher should ensure that the learner has access to it and is aware when it is being used.
29Finally, there is some semiotic misalignment when sending the resource and when reading the resource. When sending the resource, there is semiotic misalignment in both iterations; the teacher has access to the object handling and print modes, but the learners only have access to the teacher’s webcam image. The difference in available modes in the participants’ site of engagement leads to semiotic misalignment between their interactional spaces. In the first iteration, the researchers take note of modal shifts that occur in the spoken language, gesture, facial expression, and gaze modes, which make the HLA of receiving the resource appear to be foregrounded in the learners’ attention/awareness. However, during the second iteration with one learner, the researchers only notice shifts in the gaze mode. As a result, the teacher’s facial expressions carry less modal density. When reading the content of the resource, there is shared attention/awareness in the first iteration but semiotic misalignment leading to modal density misalignment in the second iteration. During both iterations, the teacher has access to the object handling mode (scrolling through the resource) which is unavailable to the learners. However, in the first iteration, one of the learners connects to the document and has access to it, and it is foregrounded in her attention/awareness. In the second iteration, the learner does not have access to the teacher’s object handling modes since the teacher does not share his screen. The partial alignment in the first iteration, however, is not linked to any specific pedagogical action on the teacher’s part. This reinforces the researchers’ recommendation that document handling should be commented on in the spoken language channel, further emphasizing the importance of meta-semiotic commentary (Holt, 2021a). In summary, the comparison allows the researchers to highlight the need for semiotic alignment.
30In the final chapter of analysis, Satar and Wigham compare instruction-giving during two different types of tasks: convergent and divergent. The researchers observe three different teachers—Craig, Karen and Sarah—who each carry out a convergent task and a divergent one. The analyses are broken down into two parts—higher-level actions and lower-level actions—in order to see whether or not task type has an effect on these. The convergent task, which is analyzed in chapters 3 and 4, involves selecting a gift package for a colleague. The divergent task involves exchanging opinions on vegetarianism.
31The authors begin the chapter with a discussion of task-as-workplan, meaning what is planned, and task-as-process, meaning how the task is actually carried out, for the divergent task. This task, which was created by researchers and shared with the teachers beforehand, features three micro-tasks: a pre-task, a main task, and a post-task. The pre-task is carried out similarly by the three teachers, whereas the post-task is carried out differently: one teacher assigns homework, another creates a role-play activity, and another elicits verbal summaries that are added to an interactive whiteboard. The focus of this chapter is on micro-task 2, where teachers elicit opinions for and against vegetarianism.
32The chapter then focuses on the teachers’ perspectives on the activities, which are revealed via post-lesson interviews. All three teachers agree that the divergent (opinion-based) task is more conversational and easier to administer, offering a clearer path forward. One teacher reveals that she sent the lesson plan to her learners ahead of time. The teachers agree that using the interactive whiteboard is difficult for the learners, so two of the teachers just used it themselves, adopting the role of scribe.
33Next, the researchers offer a comparison of the HLAs used in each type of task. For the convergent activity, they focus only on the HLAs before the learners begin the task, and for the divergent activity, they focus on the HLAs before and during the task, because the teachers continue to provide instructions throughout the activity, often in the form of questions (Meskill & Anthony, 2015, 2018). Before the start of the activity, several HLAs are common between the two task types: managing resources, focusing on task accomplishment, formulating task stages, focusing on study skills, suggesting ways into the task, and launching the task. There are also some differences in HLAs between the two pre-tasks. For example, in the information-gap (convergent) task, the HLA of communicating key task information: identifying learners have different information is important, whereas in the opinion (divergent) task, this HLA is not employed at all. For the convergent task, stating the outcome is present during the instruction-giving phase because there is only one possible outcome (the choice of a gift package and writing a collaborative email), whereas for the divergent opinion activity, multiple outcomes are possible. The HLAs of defining roles, checking understanding (of task), and allocating time are all absent in the divergent task, which allows the teachers not to restrict themselves during the activity and to respond authentically, intervene in case of misunderstanding of the task, and manage the rhythm of the task. Finally, launching the task is different because in the convergent information gap activity, the learners have to understand all of the instructions beforehand, whereas in the divergent activity the teacher can guide the activity as it progressed. Two of the three teachers take much longer to explain the instructions before launching the convergent task, but launch the divergent task quickly because they can offer more instructions as the activity unfolds. The analysis also reveals differences between the teachers’ use of HLAs. Some HLAs are shared between two teachers but not all three, such as formulating task stages and launching the task, and some are observable with only one teacher, such as focusing on task accomplishment, managing resources, focusing on study sills, and suggesting ways into the task. Craig is not like the others in that he spends more time on instructions before the divergent task than before the convergent one, and includes some HLAs that the other two do not.
34Satar and Wigham then move on to the multimodal configuration of HLAs and LLAs in the different task types. Their teacher-by-teacher micro-level analyses focus on the divergent task, because the convergent task was already analyzed in Satar & Wigham (2020).
35Sarah uses text chat to send instructions beforehand for both the convergent and divergent tasks. During the conversation, her own webcam image is small and the text chat window is not visible, similar to the layout she used during the convergent task. She engages in the HLA of managing resources before both the convergent and divergent tasks, but uses fewer HLAs before the divergent task so that she can get into the task more quickly. The micro-level analysis of LLAs is difficult to carry out on Sarah’s interaction because her screen recording is not available. However, the researchers use the other recordings to focus on her head movements and direction of gaze as she moves through various HLAs including eliciting opinions from learners and listening to their responses.
36Karen foregrounds the print mode in her attention/awareness as she launches the divergent task and asks questions. She then uses gaze and proxemic shifts away from the screen to adopt a listening posture. She also uses smiles and nods to suggest ways into the task. Some differences between the convergent and divergent tasks are that one of the gestures that is only used at the end of the convergent task is used throughout the divergent task, even though it is low in modal density and does not add anything to the HLA of suggesting ways into the task. Another difference is that when suggesting ways into the task, the teacher directs her instructions at one specific learner during the divergent task, instead of at both of them as is the case for the convergent task. For both tasks, gestures are used when suggesting possible responses to her questions. Another difference is that during the convergent task, the teacher clearly states that the instruction-giving phase is over by using spoken language, gestures, gaze, facial expression, posture, and proxemics, whereas for the divergent task, she embodies the print mode by reading the instructions aloud and the conversation flows organically.
37Craig’s instructions for the divergent task take longer than those of the first two teachers and include more HLAs, which more closely resembles instructional conversations (Meskill & Anthony, 2015, 2018). The teacher uses multiple modes to focus on task accomplishment including written dashes in the text chat to separate tasks, iconic gestures to show how to use the online whiteboard, gaze towards the screen with wide open eyes, and enunciated lip movements. He simultaneously completes the HLAs of managing resources and communicating key task information by looking down to the Skype window, using it to copy and paste parts of his lesson plan, and vocalizing his use of Skype to his learners. He also uses multiple modes to focus on key vocabulary and core content with increasing modal density, which is similar to the modal configuration he used when communicating key task information in the convergent task. During the divergent task, his instructions follow a style closer to the techniques of instructional conversations, implying that Craig is able to generate higher social presence (Satar, 2024). For example, the word “moral” is interspersed throughout the instruction-giving progress in order to elicit responses from learners.
38The researchers then compare the modal configuration of the LLAs during the convergent and divergent tasks, focusing on one HLA per teacher. Sarah’s modal configurations, compared across the HLA of launching the task, are the same for both tasks, with the spoken language carrying the highest modal density and her webcam framing choices remaining consistent. The close-up shot makes her gestures unavailable, but she is able to use gaze, posture and head movement shifts while giving instructions. The authors also add a timeline to compare multiple screenshots, as a single moment is not representative of an entire lesson. The expanded timeline shows that although the modal configuration is similar, the modes of head movements and gaze shifts become slightly less dense during the divergent task. Karen, who is compared across the HLA of suggesting ways into the task, has the same framing throughout both tasks with gaze, posture, head movements, and visible gestures. The print mode is also similar for both. An expanded analysis shows that as with Sarah, the speed and frequency of Karen’s modal shifts are greater during the convergent task, especially in the modes of gesture, head movement and gaze shifts. Craig’s modal configurations, compared across the HLA of communicating key task information, are also similar across tasks with gaze, hand gestures and facial expressions for the convergent task, and gaze and print for the divergent task, in addition to spoken language for both. Further analysis reveals that as with Sarah and Karen, Craig displays more modal shifts during the convergent task than during the divergent task. The authors conclude that “the divergent task HLAs are more conversational in style,” that the teachers do not need to be as animated or display as many modal shifts during the divergent task, and that “while modal configurations (intensity and complexity) of the teachers’ LLAs are similar in both task types, modal density is higher in the convergent task through a higher number of and more frequent modal shifts” (p. 135).
39In the final section of the chapter, the researchers present new secondary-level HLAs observed for managing resources during the divergent task. These are all instructions-as-process and relate to whiteboard use: identifying the resource type, facilitating use of the resource, and editing the content of the resource. To illustrate this, they present micro-level analyses of Karen and Craig’s use of the whiteboard during the divergent task when discussing vegetarian eating habits. For the HLA of facilitating the use of the resource, Karen vocalizes her sending of the resource, invites the learners to make themselves recognizable by changing their usernames in the platform, and reads aloud the system notifications. She then edits the content of the resource (the online whiteboard) using the drawing and typing tools. Craig also engages in the HLAs of identifying the resource type and editing the content of the resource, first by sending the URL and saying “Here is the whiteboard” in the spoken language mode, and then by copying and pasting text, all the while vocalizing his actions. This micro-level analysis shows the importance of carefully guiding the use of tools during online teaching.
40Overall, this chapter shows that the LLAs of the convergent task have higher modal density caused by more brisk modal shifts and that the divergent task, more conversational in style, has a fewer number of HLAs overall but features new HLAs having to do with resource management.
41Satar and Wigham use the final chapter to summarize the contributions that their book makes to the field, offer pedagogical reflections, and look forward to future research. One of the main contributions of this book—and of multimodal approaches in general—is to show that when communication is video-mediated, some affordances are lost (e.g., embodied actions such as gaze, gestures, and facial expressions), while others are gained, such as text chat and the interactive whiteboard, depending on the software used. Therefore, new skills are necessary (Cappellini & Combe, 2017), and online teachers need to capitalize on the modes that are offered to them such as proxemics, gesture, gaze, posture, and so on. Another contribution of this book has to do with the research gaps that it fills. Although this book is not the first to address video-mediated foreign language teaching, it focuses specifically on the instruction-giving practices of experienced teachers as opposed to new or pre-service teachers, and focuses on task repetition, task type and number of learners as factors.
42This book offers several contributions to methodology and knowledge. First, the authors use a qualitative approach, which is less common in applied linguistics. Many researchers have used multimodal conversation analysis to study the sequential organization of talk-in-interaction, but Satar and Wigham, by combining multimodal interaction analysis and mediated action as their fundamental theoretical units of analysis, are able to focus on higher- and lower-level actions simultaneously and not just on the sequentiality of conversation. In the words of the authors, “Our overarching argument in this book is that effective instruction giving in online language lessons is a multimodally complex process, understanding the nature of which fully requires attendance to both higher- and lower-level mediated actions” (p. 142). Multimodal (inter)action analysis (MIA) was originally developed by Norris (2019) for in-person interaction, but Satar and Wigham respond to the need to apply innovative approaches to online foreign language teaching and learning. Another methodological breakthrough is the authors’ use of grounded theory to identify and bundle HLAs to ensure that they are mutually exclusive. Concerning multimodal transcription, which is time-consuming and can lead to widely different outcomes across different researchers, the authors use modal shifts in the LLAs as basic units of analysis, which should increase replicability and consistency. Satar and Wigham continue to develop the concept of frozen actions, proposed by Norris (2004) to mean material objects in a physical interactional setting, and extended in Wigham & Satar (2020) to mean the print mode in an online setting. They also offer the new concepts of semiotic lag, semiotic misalignment, and modal density misalignment, developed in previous chapters and summarized in Wigham & Satar (2024). Finally, their semi-controlled corpus allows for the study and comparison of instruction-giving among experienced teachers in naturally occurring data concerning task repetition, number of learners, and type of task.
43The authors’ analysis of instruction giving and task repetition leads to the conclusion that teachers are more efficient in giving instructions the second time, with fewer HLAs and better management of resources. In terms of modal aggregates for the LLAs, the teachers use gestures, “particularly when repeating the instructions following learners’ lack of alignment with the instructed activity and their signalling of trouble” (p. 145), and the teachers make sure that their gestures are visible to the learners, “demonstrating critical semiotic awareness.” This reminds me of my qualitative analysis (Holt, 2020) that showed that when the learner displays incomprehension, the teacher tends to make his or her gestures more visible to the webcam. This was confirmed statistically in Holt (forthcoming). Furthermore, this responds to the need, pointed out by Inceoglu & Loewen (2022), for more analyses on corrective feedback and videoconferencing. The authors also find that the print mode has serious uses, including presentation of new vocabulary, as I demonstrated in Holt (2021a), so long as it is permanently visible to the learner. Webcam framing, which the authors think is intentional, just like Sindoni (2013), allows for particular modal configurations. However, teachers need to remain vigilant about semiotic misalignment and modal density misalignment.
44Concerning the number of learners, analysis shows that the HLA of managing resources is very complex, because if not done correctly it can lead to semiotic misalignment. This can happen, for example, if the teacher does not confirm access to the resource. Furthermore, semiotic density misalignment can lead to the foregrounding of different HLAs in the social actors’ attention/awareness, for example if the HLA of sending the resource is foregrounded for the teacher but the HLA of activating schemata is foregrounded for the learner. Another finding is that instruction-giving is more efficient when there are multiple learners—which aligns with socio-cultural approaches to language learning (Vygotsky, 1978)—and during the second iteration a session. Finally, the number of learners appears to have little effect on site of engagement, hierarchical organization of LLAs, modal configurations, and gaze direction. Comparison of one versus two learners shows that when there are two learners, there tends to be more modal density alignment when sending resources. As a result, the authors recommend engaging in just one HLA at a time, using spoken language, gaze, gesture, and facial expressions to address learners, vocalizing any onscreen actions not visible to learners, and ensuring that the same modes are foregrounded in each social actor’s attention/awareness. This is especially important when confirming access to the resource, since the same modes are not available during video-mediated and in-person interaction. Allowing for silence in the spoken language mode gives the learner enough time to signal access to the resource or lack thereof.
45Regarding the effects of task type (convergent and divergent) on instruction giving, analysis reveals that several HLAs are common to both task types, whereas some are only found in the convergent task. Three new HLAs relating to managing resources, and more specifically to using an interactive online whiteboard, are also defined. The main difference between the two task types is that the convergent task requires more instructions up front because the learners have to reach a single outcome. The divergent task requires less instruction-giving up front and therefore has fewer HLAs; the instructions are given in a stepwise fashion throughout the activity. Similar modes and modal configurations are used for LLAs in both task types, and modal density increases for important parts of the instructions. Finally, the LLAs have higher modal density for the convergent task. According to Norris, modal density happens when there is modal complexity (modes are inextricably linked) and when a mode is important (modal intensity). These authors go further in suggesting that “modal density can also be achieved through brisk modal shifts that foreground the LLA in the teacher’s attention/awareness” (p.151), which may help direct the learner’s attention/awareness and increase cognitive engagement.
46When offering pedagogical reflections for language teachers, Satar and Wigham are careful to avoid best practice recommendations since not much research has been done on the relationship between practices and outcome. Instead, they offer reflections and categorize their recommendations. Their first recommendation for managing electronic resources is to reflect on how documents should be shared, and recommend simplifying the process by using single folders, clear file names, using the learners’ names when naming the files, etc. Teachers should confirm that the learner has access to the text chat and to the electronic resources themselves. Resources should be prepared ahead of time in order to reduce the number of HLAs required during the lesson, and teachers should ensure that the learner has access to everything and knows how to use the resource in question such as the online whiteboard. The contents should be clearly organized. Teachers should give clear instructions on how to use a resource and avoid deictic pronouns such as “this or that,” and screen sharing can ensure that the learner has access to everything that the teacher is doing. If the learner has trouble, then the teacher should move to another type of resource that the learners are more familiar with. When teaching writing, teachers can write down the learners’ contributions when they are unable or unsure of how to do it themselves. In terms of layout, multiple and/or large screens can enable teachers to have access to their own image and to shift gaze between resources more easily. Saying where each window is can help learners understand what the teacher is looking at, and vocalizing attentional shifts between windows can ensure that the same HLAs are foregrounded in each social actor’s attention/awareness. Finally, using software that is specifically made for teaching, such as Big Blue Button, can ensure that the learners have a specific layout.
47Concerning management of semiotic resources, Satar and Wigham highlight the importance of planning use of different modes. They recommend a head-and-torso shot so that learners can see gestures, and gestures should be centralized and sustained long enough to be seen by learners. They recommend clear articulation in the spoken language mode, free gaze, and a relaxed posture. Instructions can be given bi-modally with voice and text chat, and the learner’s response can be monitored on screen. Teachers can back away from the screen and use gaze shifts to signal the end of the instruction-giving process, just as learners back away from the webcam to signal the closing of a lexical explanation sequence (Wigham, 2017). When giving complex instructions, modal intensity and complexity should be taken into account; for example, the webcam can be switched off in order to increase the intensity of the print mode if it is of particular importance. Learner involvement can be increased by asking one learner to read the resource sheet in the print mode before asking questions to the other learner.
48Finally, in terms of instructions-as-workplan, Satar and Wigham offer recommendations on how to multimodally orchestrate instructions depending on the task type. For information-exchange tasks, building on their previous comments on the importance of modal shifts, they recommend using the latter to punctuate complex instructions as they unfold, and to think about how to use the different modes for scaffolding. If the learner doesn’t understand, then teachers should keep in mind that instructions tend to improve with each iteration.
49In terms of limitations and future research, the authors recommend studying different types of convergent tasks, for example reasoning gap activities and problem-solving activities. Differences in modal density between tasks with different levels of complexity are a potential area for future research. Using a larger or uncontrolled corpus could compensate for the small dataset of the current study. Another area for future study could be to explore interactions with learners of lower levels of proficiency in order to analyze the possible effects of learner proficiency on modal configuration or HLAs. Using eye-tracking technology could help researchers know with certainty what the social actors are attending to on the screen. The authors point out that since the COVID-19 pandemic has normalized large group sessions, it would be interesting to compare the effect that group size has on instruction-giving. Finally, future studies could continue to investigate task facilitation where teachers scaffold the instructions and help the learners along after the task has been launched.
50This book advances our understanding and provides breakthroughs in terms of data collection, our conceptual understanding of multimodal interaction, and online language teaching, while leaving several avenues open for future research.
51Researchers interested in video-mediated communication often collect as many different types of data as possible in order to carry out complementary analyses from different angles, for example screen recordings, external camera recordings, interviews, and eye-tracking data (Cappellini et al., 2023; Guichon, 2017b). Although recording a participant’s screen is a non-intrusive way of collecting data (Guichon, 2013), this method is not always available, especially if the participant is geographically distant and his or her screen is not accessible to researchers. This is why Satar and Wigham’s method of using the researcher’s screen as a proxy for what the learners can see is innovative. The researchers’ combination of different types of data allows them to have a bird’s eye view of the interaction, and to show coordination across modes that the learners can access, such as proxemics and gaze, as well as modes that they cannot, such as the teacher pulling up a task sheet, thereby excluding the learners’ webcam images from their site of engagement. These analyses allow the authors to formulate practical recommendations. Instead of trying to collect longitudinal data in order to search for changes according to pre-defined criteria, Satar and Wigham observe the same teachers carrying out the same tasks under different circumstances, thereby letting the differences and improvements emerge by themselves. By observing which HLAs are carried out proactively during the second iteration of an activity, the researchers allow the teachers to show which changes make teaching more efficient instead of trying to predefine them ahead of time. This has important implications for teacher training, as teacher trainers should focus specifically on the HLAs that were found to be carried out more efficiently during the second iteration, such as those involving resource management.
52Satar and Wigham’s work in defining HLAs and LLAs in instruction-giving is an important methodological contribution. They use grounded theory to define the different HLAs involved in instruction-giving, all the way to theoretical saturation. This allows them, as well as other researchers, to structure their analyses around HLAs and dive deeper into the LLAs that compose them. This framework, which allows for micro- and macro-level analyses, has the merit of being flexible, as Satar and Wigham have shown by removing HLAs that are seldom used and adding new ones as they emerge, for example while using an online interactive whiteboard.
53Perhaps the greatest contribution of this book has to do with our conceptual understanding of multimodality and multimodal approaches. Satar and Wigham expand Norris’s definition of modal density to include modal shifts, which enhances our understanding of which modes and actions are foregrounded in social actors’ attention/awareness while carrying out HLAs related to instruction-giving. The researchers also develop an innovative way of annotating and representing these modal shifts: after showing overall modal configuration with a representative frame, they add a timeline in order to focus on the number of modal shifts and on the resulting modal density. In addition, they develop two new concepts—semiotic misalignment and modal density misalignment—offering researchers a conceptual tool to make sense of different problems that can arise during video-mediated communication. I have come across both of these in my own studies. For example, I encountered semiotic misalignment when the learner’s webcam image, which was covered up by a browser window, was unavailable to the teacher (Holt, 2024). I also came across modal density misalignment when an image was so modally intense that a lexical explanation could not be performed orally (Holt, 2021a). In this example, the teacher relied so heavily on the image that she had no words to explain the lexical item in question. The solution that I proposed for overcoming this was for teachers to develop multimodal flexibility, meaning the ability to switch from one mode to another when a given mode was either unavailable or, I would now add, carried too much modal intensity in relation to the other available modes. Satar and Wigham’s methodological tools lead to practical recommendations for avoiding semiotic misalignment. For example, regarding document handling, “it is important for teachers not to presume that the learner can receive, locate, open, access, and use the electronic resources without confirmation of guidance” (pp. 147-148). The ability to verify access to resources and deploy modal shifts at opportune times, as well as the ability to comment on the multimodal configuration of one’s LLAs and HLAs via meta-semiotic commentary, all the while displaying multimodal flexibility (Holt, 2021a), should all be part of the techno-pedagogical skills (Cappellini & Combe, 2017) that teachers must develop during their training and throughout their careers.
54Satar and Wigham’s multimodal analyses contribute to our understanding of various pedagogical aspects involved in video-mediated language teaching. First, their macro and micro-analyses confirm that any time teachers are compared, there is a high degree of inter-individual differences, as pointed out by Tellier et al. (2014). Next, they add to our growing list of what the text chat can be used for, such as reminding learners what to focus on upon receiving an electronic resource. Concerning webcam framing, they confirm the importance of head-and-torso shots (Guichon & Wigham, 2016) so that gestures are visible to the interlocutor. Their finding that teachers display semiotic awareness by making their gestures visible during key moments corroborates previous research on lexical explanation sequences (Holt, 2021a, forthcoming; Wigham, 2017). In the case of instruction-giving sequences, gestures are produced in the center of the webcam frame especially when explaining ways into the task and when responding to learner requests to clarify the instructions. This contributes to the notion that gestures produced in front of the webcam have communicative purpose (Cosnier & Develotte, 2011). Multimodal analyses presented in this book also show that learner involvement can be increased by asking one learner to read the resource sheet in the print mode before asking questions to the other learner. The finding that facial expressions can be used to show misunderstanding, and that gestures can be used to help overcome incomprehension, confirm the importance of multimodal approaches to video-mediated language teaching and learning.
55This study leaves several avenues open for future research. As pointed out by the authors, eye-tracking technology would make it possible to know what the teachers are looking at, and therefore cognitively attending to (Conklin et al., 2018) when they direct their gaze towards different parts of the screen, including their learners’ webcam images. Further studies could also investigate learners’ ability to tell when teachers are looking at them. A rich corpus such as this one would also make it possible to test for alignment (Costa et al., 2008) across different modes (Cappellini et al., 2022) and across different social actors, or to apply multimodal (inter)action analysis to other pedagogical phenomena other than instruction-giving sequences. The expansion of the definition of modal density to include modal shifts warrants further empirical testing in order to investigate the ways in which modal shifts are perceived by learners. Finally, Satar and Wigham’s framework of HLAs and LLAs could be applied to novel techno-pedagogical contexts, such as virtual reality environments, in order to understand the multimodal composition of pedagogical actions including but not limited to instruction-giving.