Framing decides what a video study can claim

This fourth of ten units argues that video changes the kind of claim a study can warrant, and that it drags the recording, framing and consent decisions in front of the reader. The three units before it settled who to ask, how many, and how a field is made.

A record of two people preserves the organisation that audio loses

A camera pointed at one person records a face; pointed at two, it records an organisation. Heath, Hindmarsh and Luff make the second move the discipline of video research. The record preserves gaze, gesture, posture and the timing of turns that audio loses, and that changes what a defensible claim looks like (Heath, Hindmarsh and Luff, 2010). The unit of analysis becomes the sequence, the ordered exchange in which one action makes the next possible. A team cannot report that a participant was confused; it can report that the second speaker waited for the first to finish and that the next question narrowed the topic.

Multimodality changes the object, not the resolution

The easy misreading of multimodal analysis is that it adds detail: transcribe the talk, then a second pass for gesture. Mondada argues against that framing. The body is not a supplementary channel but one of the resources through which action is assembled. Putting it at the centre repositions language as one modality among others, and refuses both a logo-centric account of communication and a visuo-centric account of the body (Mondada, 2016).

The meaningful unit is not fixed in advance. It is whatever configuration of modes the participants treat as doing the action, and it must be shown rather than asserted. The analyst also needs a way to say which mode carries the interaction. Kress and van Leeuwen supply the theory of modes, arguing that meaning is made through several semiotic modes whose design and production each carry their own grammar (Kress and van Leeuwen, 2001). Norris supplies the device: her modal density and foreground-background continuum weight a mode by how participants orient to it (Norris, 2004). Jewitt's handbook surveys social semiotic, conversation-analytic and geosemiotic approaches, so multimodality emerges as a family of competing methods rather than one (Jewitt, 2014). A claim about which mode matters is thus a claim inside a chosen method, and the method has to be named.

Seeing is a socially organised practice

Goodwin's study of professional vision shows how members of a profession learn to see their domain through coding schemes, highlighting and graphical representation. Archaeologists use a colour chart to code soil in an excavation. In a courtroom, the same video footage was coded in competing ways and made to show aggression (Goodwin, 1994). Seeing is socially organised, and the practices that make a phenomenon visible are observable in the record, so an analyst can be asked to show the practice rather than the conclusion. Streeck, Goodwin and LeBaron extend the point to the material world. They collect studies in which language and the body are jointly organised with the objects of a setting, and argue that interaction cannot be analysed from talk alone (Streeck, 2011). Embodiment and materiality are constituents of the interaction, not context around it.

Transcription is a theory of the data

Bezemer and Mavers examine how multimodal transcription is actually done, comparing transcripts from several traditions. They argue that transcription is a theory-laden act that selects what can later be analysed. Different traditions select different features and rarely justify the selection (Bezemer and Mavers, 2011). A transcript that shows only talk has already answered the question of what mattered.

Their comparison is specific. They set transcripts from conversation analysis, discourse analysis, social semiotics and micro-ethnography side by side, and the same stretch of activity reads differently in each. The conversation-analytic transcript covers a few seconds and marks turns and actions. The micro-ethnographic one covers minutes. A social semiotic transcript takes modes as the unit. The differences track the purposes of their makers rather than the recording, so a transcript is an artefact with an argument built into it.

The same discipline applies to the video. Mondada shows that video makes the temporality of interaction observable, and that embodied actions are finely coordinated in time. The warrant rests on sequential visibility: a reader who slows the clip down can see the coordination the analyst claims. What a deaf reader has instead of that check is the transcript, and the reader on a metered connection has one pass at the clip.

A claim is warranted only when a second analyst sees the same thing

The method for turning a record into a warranted claim is collective. Jordan and Henderson set out interaction analysis as the study of learning and work through video records examined in data sessions. In a session a group repeatedly views a clip and argues about what it shows. They argue that disagreement among analysts is the discipline that grounds the claim (Jordan and Henderson, 1995), since a session tests one reading against others of the same clip.

Jordan and Henderson require the analyst to look for other instances of the same event in the corpus, and to check whether the generalisation holds. They prescribe transcripts in which talk and bodily activity run in parallel columns, so that the timing of both can be read at once.

Broth, Laurier and Mondada collect studies of video practices in operating theatres, television production and video games. They show that professionals make sense of events through video, and that the medium is both a research instrument and an object of practical work (Broth, Laurier and Mondada, 2014). The analyst's video competences are continuous with those of the settings studied: a resource for interpretation and a warning against treating the camera as neutral.

Every framing decision fixes what can later be claimed

Heath and colleagues are candid that the difficulties begin before the recording does: where to place the camera, what to frame, and whether to collect in one intense period or in phases. They also ask what happens to the material afterwards, and how consent can be obtained for events that have not happened yet (Heath, Hindmarsh and Luff, 2010). Each decision fixes what can later be claimed, so each is a method decision rather than an administrative one. Pink presses the point, arguing that visual methods require attention to the sensory and embodied dimensions of research and addressing reflexivity and the ethics of representing participants (Pink, 2013). The record is made of people who can be recognised in it, so who consented to what is part of what it means.

The coding strand answers from the other side

A second literature has codified video content analysis rather than contesting its warrant. Fazeli, Sabetti and Ferrari set out a six-step Visual-Verbal Video Analysis method for qualitative content analysis of video, integrating visual and verbal coding into one procedure. They argue that video needs a method that treats the visual and verbal channels together, rather than transcribing one into the other (Fazeli, Sabetti and Ferrari, 2023). It answers a real gap: frameworks from actor-network theory and picture theory give little guidance on how to select and examine the multimodality of video. It also takes a side: defined steps assume an object stable enough to receive codes, which the post-qualitative strand disputes.

Contested question: is the sequential record a warrant or a construction?

The literature agrees that video changes the object of analysis and disagrees about what grounds a claim made from it. One position holds that the warrant is the participants' own displayed orientation, made visible in the sequential record. A claim is disciplined by what participants treat as relevant. The analyst earns it by showing the same coordination a reader can see in the clip (Heath, Hindmarsh and Luff, 2010; Mondada, 2013; Jordan and Henderson, 1995). The opposing position holds that the record is constructed at least as much as it is found. Menning and Murris rework video through filmmaking practice and a post-qualitative cinematic analysis. They use Deleuze's film philosophy to argue that framing, tracking, speed changes, reverse motion and sound shape what counts as data. Their analysis opens to more-than-human bodies and relies less on language (Menning and Murris, 2024). The analyst's choices are then part of the account, not a contaminant. The social-semiotic strand sits closer to the second position, treating meaning as designed across modes (Jewitt, 2014).

What would settle it is a demonstration that separates the two grounds. One test would ask independent analysts to code the same clips and report where their accounts diverge and whether the divergence tracks the clip's sequence or the analysts' framing. A second would publish raw and edited records side by side, so a reader can try to reconstruct the claim from the unedited material. If the claim survives, the warrant is in the participants' orientation. If it does not, the construction is doing the work. Menning and Murris offer one such demonstration on a single ordinary event (Menning and Murris, 2024); the evidence that would generalise is a body of re-readings checked the same way.

The two grounds are not equally evidenced. The sequential warrant is demonstrated on corpora (Mondada, 2013; Jordan and Henderson, 1995) and codified in a handbook (Heath, Hindmarsh and Luff, 2010). The constructive warrant rests on one re-reading of a single ordinary event (Menning and Murris, 2024) and on a theory of modes (Kress and van Leeuwen, 2001). The coding strand answers with a procedure, not a warrant (Fazeli, Sabetti and Ferrari, 2023). While the dispute is open, a designer should keep the unedited record, document the framing decisions and publish the coding scheme. The strongest objection is that sequential visibility is not self-interpreting, since a reader cannot see the coordination without accepting the analyst's transcription of it.

Where the argument is heading

The direction of travel is from the transparency of the record to its composition. Fazeli and colleagues proceduralise video content analysis and make its steps inspectable (Fazeli, Sabetti and Ferrari, 2023). Menning and Murris treat filmmaking itself as analysis, loosening the boundary between the recording of an event and the making of one (Menning and Murris, 2024). Jewitt explains why both can claim the field: multimodality never was one method, and its traditions collect and warrant data differently (Jewitt, 2014). Which demonstration the field accepts will decide the next phase of video research.

The sources this unit rests on

Each note states what the cited document did and found rather than quoting it. The DOI records were checked against registered metadata on 30 October 2026.

• Heath, C., Hindmarsh, J. and Luff, P. (2010). Video in Qualitative Research. A guide to recording, ethics, transcription and analysis at the level of the sequence. It argues that the record preserves the gaze, gesture and timing audio loses, and that the difficulties begin before recording. https://us.sagepub.com/en-us/nam/video-in-qualitative-research/book233902
• Mondada, L. (2016). Challenges of multimodality. doi:10.1111/josl.1_12177 Examines the methodological challenges of multimodality, including the temporality of embodied action and the coordination of talk with bodily conduct. It argues that multimodality is a different object of analysis, not an additive layer.
• Kress, G. and van Leeuwen, T. (2001). Multimodal Discourse. Develops a social semiotic account in which meaning is made through several modes whose design and production each carry their own grammar. It contends that analysis must address the mode as well as the text. https://archive.org/details/multimodaldiscou0000kres
• Jewitt, C. (2014). The Routledge Handbook of Multimodal Analysis (2nd ed.). Surveys social semiotic, conversation-analytic and geosemiotic approaches and how each collects and annotates data. It shows that multimodality is a field of competing methods rather than one. https://www.routledge.com/The-Routledge-Handbook-of-Multimodal-Analysis/Jewitt/p/book/9781138241206
• Norris, S. (2004). A methodological framework for multimodal interaction. doi:10.4324/9780203379493 Proposes modal density and the foreground-background continuum as ways of weighting a mode, arguing that weight follows how participants orient to a mode.
• Goodwin, C. (1994). Professional vision. doi:10.1525/aa.1994.96.3.02a00100 Analyses how members of a profession learn to see their domain through coding schemes, highlighting and graphical representation, in an archaeological excavation and in a courtroom where the same video was coded in opposing ways. It shows that professional vision is socially organised and analysable in the interaction.
• Streeck, J. (2011). Embodied Interaction: Language and Body in the Material World. Assembles studies showing how language and the body are jointly organised with the material world. It argues that interaction cannot be analysed from talk alone, since the material environment is part of the action's organisation. https://www.cambridge.org/gb/academic/subjects/psychology/cognition/embodied-interaction-language-and-body-material-world
• Bezemer, J. and Mavers, D. (2011). Multimodal transcription as academic practice. doi:10.1080/13645579.2011.563616 Examines how multimodal transcription is done, comparing transcripts from several traditions, and argues that transcription is a theory-laden act that shapes what can be analysed, with the selection rarely justified.
• Mondada, L. (2013). Video analysis and the temporality of interaction. Shows that video makes the temporality of interaction observable and that embodied actions are finely coordinated in time. It argues that the warrant rests on the sequential visibility of the phenomenon. https://www.degruyter.com/document/doi/10.1515/9783110261318/html
• Broth, M., Laurier, E. and Mondada, L. (eds) (2014). Studies of Video Practices: Video at Work. doi:10.4324/9781315851709 Collects studies of video as a practical resource in operating theatres, television production and video games, showing that professionals make sense of events through video, and treats it as both instrument and object of work.
• Pink, S. (2013). Doing Visual Ethnography (3rd ed.). A guide to visual ethnography covering photography and video as methods of research and as objects of study, arguing that visual methods require attention to the sensory and embodied dimensions of research. https://archive.org/details/doingvisualethno0000pink
• Jordan, B. and Henderson, A. (1995). Interaction analysis: foundations and practice. doi:10.1207/s15327809jls0401_2 Sets out interaction analysis as a method for studying learning and work through video records examined collectively in data sessions, arguing that the disagreement among analysts grounds the claim.
• Menning, S. F. and Murris, K. (2024). Reconfiguring the use of video in qualitative research through practices of filmmaking. doi:10.1177/14687941231206755 Reworks video through filmmaking practice and a post-qualitative cinematic analysis, using Deleuze's film philosophy to argue that framing, tracking, speed changes and sound shape what counts as data.
• Fazeli, S., Sabetti, J. and Ferrari, M. (2023). Performing qualitative content analysis of video data. doi:10.1177/16094069231185452 Sets out the six-step Visual-Verbal Video Analysis method, integrating visual and verbal coding into one procedure. It argues that video needs a method that treats the two channels together.