Conversation analysis: meaning is produced in the sequence

The interaction a designer supports is mostly talk, and the tempting unit of analysis is the utterance. Collect what people said, decide what each item means, then group the items. Conversation analysis begins from a finding that makes that unsafe. Participants treat the sense of a turn as settled by the turn that follows it. The unit of analysis is a sequence of at least two turns.

This seventh of ten units argues that in naturally occurring interaction, what an utterance means is a property of its position in a jointly produced sequence rather than of its content. An analysis that codes utterances out of their sequence has changed the object it claims to study. The claim is argued from the mechanism of turn taking and repair, from the premise that makes sequence the right object, and from the settings where systems are used. It is then tested against the disagreement about counting. The two units that follow add the moving record and the plan.

The order of turn taking is the participants' own achievement

Sacks, Schegloff and Jefferson (1974) propose a model of turn taking and test it for compatibility with grossly observable facts about conversation. They conclude that the arrangement is locally managed and administered by the parties themselves, interactionally controlled, and sensitive to recipient design. Each term places the order in the participants rather than in the setting. Nothing above the conversation administers it, and a turn is built for this recipient. The facts the model must fit are the observable ones. Overwhelmingly, one party talks at a time. Transitions between turns occur with a minimal gap and a minimal overlap. The order of speakers varies without being fixed in advance. It follows that a system occupying a turn is treated as acting on the user. The analytic question is what its previous turn made relevant, and what the user did with it.

Repair shows the participants monitoring the sequence

Schegloff, Jefferson and Sacks (1977) describe an organisation of repair addressed to recurrent problems in speaking, hearing and understanding. They explain the mechanism producing a strong empirical skewing in which self-repair predominates over other-repair. Trouble is located where it occurs and repaired there, and the preferred remedy keeps the speaker's authorship of the fix. A system that asks the user to repeat, or silently corrects their wording, occupies the place where the repair would have been done. The analyst then sees a request for repetition, and cannot say which part of the utterance failed, or whether a poor connection caused it.

A turn's sense is fixed by the next turn it makes relevant

Schegloff (2007) sets out sequence organisation, the ways turns are ordered and combined so that actions such as requests, offers, complaints and announcements take place. A turn of a particular kind projects a relevant next, and the next turn is heard against the one it answers. Two identically worded utterances in different positions are therefore not the same action. A voice assistant closes an exchange with "Did that answer your question?" and the user says "Yes". Coded alone, the "Yes" is satisfaction. In its position it is the second part of a question the user did not initiate, so any remaining difficulty is harder to raise.

Ethnomethodology supplies the premise that makes sequence the object

Garfinkel (1967) treats practical activities and reasoning as topics of empirical study in their own right. He holds that the practices through which members produce and manage the settings of everyday life are identical with their procedures for making those settings accountable. Accountable here means observable and reportable to one another. Heritage (1984) draws the consequence for institutions. Context is not a container acting on the talk from outside. It is something participants produce and make accountable in the talk itself. If sense is settled in the setting, no rule applied to the words alone recovers it, and coding assumes a stability the premise denies.

A claim here rests on replayable evidence a reader can inspect

ten Have (2007) presents the practical side: the paradigm, recordings, transcription, analytic strategies and the elaboration of an analysis. Hutchby and Wooffitt (2008) state and defend the disciplinary claim that talk-in-interaction is an object of analysis in its own right. It is not a window onto other social processes or variables. They set out single case analysis as a strategy. Together they anchor a claim to a segment a reader can inspect and dispute. The transcript is itself an analytic artefact, treated in unit 1 of this module.

An interactive system arrives inside sequences the institution already constrains

Drew and Heritage (1992) argue that institutional interaction is not simply conversation with a work topic. Participants orient to a goal, task or identity tied to the institution. The range of contributions available to each party is constrained. An institution-specific inferential framework is in play, so actions that would be disaffiliative elsewhere are unmarked there. Ordinary conversation is the benchmark against which those restrictions show. Heritage and Clayman (2010) work four such domains: calls to emergency numbers, doctor-patient interaction, trials and juries, and news and political communication. Sidnell and Stivers (2013) is the reference for the machinery. Their emergency-call chapter shows the discipline at work. A call to 911 opens without a greeting or a "how are you". The call-taker hears the caller's first substantive turn as a request for help, even when it is ambiguous or silent. The call then moves through an interrogative series that gathers what the institution needs before any assistance is promised.

Contested question: can interaction be coded for comparison?

The first position is that coding is not necessarily foreign to the analysis. Stivers (2015) argues it in a paper that calls itself heretical, and her evidence is argument with examples, drawn from American and British English. She notes that mixed methods studies combining conversation analysis with quantitative methods have been used since the 1980s, holds that formal coding of interaction is not necessarily antithetical to the tradition, and names the cost in the same paper: a massive reduction and flattening of complex conduct into simplistic codes. Her remedy is a distinction between coding grounded in the analysis and coding that is not.

The second position is that quality here is measured on the tradition's own terms. Perakyla (2004) argues it in a book chapter, which is the form that limits it: reliability and validity in research based on naturally occurring interaction are internal to its analytic logic, and cannot be borrowed from variable-based research. His criteria are the transparency of analytic claims, the next-turn proof procedure and the analysis of deviant cases. The third position is the sequence-first claim itself, stated by Sacks, Schegloff and Jefferson (1974) and systematised by Schegloff (2007). It is the tradition's own foundation, so it cannot adjudicate the dispute it defines.

What would settle it is a test the two positions supply. An audit of published studies that code interaction could score each on three questions. Was the category derived from sequential observation, which is Stivers' (2015) distinction, or imported from a variable the analyst brought? Does the study report deviant cases, which Perakyla (2004) makes a criterion rather than an extra? Does the statistic preserve the sequential distinction it names, since a code for a question that cannot separate one which selects the next speaker from one that invites self-selection is a code for a different variable?

The strongest counter-argument to this unit's own claim is that a code applied to a turn in its context is not a code applied to a turn lifted out of it. Coding has coexisted with the tradition for four decades and has produced cumulative findings, so the unit's contrast between position and content is a choice the tradition has not been forced to make, because an analyst can code the sequence.

What to do while the dispute is open

If you code interaction, derive the category from sequential observation and say so. Report the deviant cases, and keep the recording, so a reader can check the code against the sequence. Where a coefficient is used, give it on the sequential coding only, and state which distinction it is unable to preserve.

Where the argument is heading

The direction of travel is towards intervention and measurement, and both carry the risk that the sequence is lost in the move. Antaki (2011) collects applied work in which analysts intervene rather than only describe. The volume opens with six kinds of applied conversation analysis, with cases on aphasic conversation, telephone survey response rates, ethnic monitoring on a helpline and user-centred design. Sidnell and Stivers (2013) show the machinery consolidated into a reference work. Stivers (2015) shows the measurement question argued on the tradition's own terms. The standard the literature supplies: an intervention claim should show the sequence it changed, and a measurement claim should show what the code kept and what it discarded.

The sources this unit rests on

• Antaki, C. (2011). Applied Conversation Analysis: Intervention and Change in Institutional Talk. Basingstoke: Palgrave Macmillan. doi:10.1057/9780230316874 Antaki's collection shows analysts working with practitioners to change institutional talk, ranging across aphasic conversation, survey response rates, ethnic monitoring on a helpline and user-centred design. It is the unit's evidence that the tradition has moved towards intervention.
• Drew, P. and Heritage, J. (1992). Talk at Work: Interaction in Institutional Settings. Cambridge: Cambridge University Press. https://archive.org/details/talkatworkintera0000unse Drew and Heritage found the study of talk at work, distinguishing institutional interaction by participants' goal orientation, by constraints on the contributions available to each party, and by institution-specific inferential frameworks. It gives the unit its reason to read a service setting as a describable interactional order.
• Garfinkel, H. (1967). Studies in Ethnomethodology. Englewood Cliffs: Prentice-Hall. https://www.wiley.com/en-us/Studies+in+Ethnomethodology-p-9780745600055 Garfinkel founds ethnomethodology, treating practical activities and reasoning as topics of empirical study and holding that members' practices for producing the settings of everyday life are identical with their procedures for making those settings accountable. It is the unit's root source for the claim that sense is produced in the setting.
• Heritage, J. (1984). Garfinkel and Ethnomethodology. Cambridge: Polity Press. https://www.politybooks.com/bookdetail?book_slug=garfinkel-and-ethnomethodology--9780745600604 Heritage gives the standard exposition of Garfinkel's programme and draws the consequence the unit needs: institutional context is not a container acting on the talk from outside but something participants produce and make accountable in the talk itself. It is the bridge from ethnomethodology to the sequential analysis of talk.
• Heritage, J. and Clayman, S. (2010). Talk in Action: Interactions, Identities, and Institutions. Chichester: Wiley-Blackwell. doi:10.1002/9781444318135 Heritage and Clayman introduce the approach through four institutional domains, emergency calls, doctor-patient interaction, trials and juries, and news and political communication, with chapters on talking social institutions into being. It shows the apparatus bearing on the encounters systems are built to support.
• Hutchby, I. and Wooffitt, R. (2008). Conversation Analysis. 2nd edition. Cambridge: Polity Press. https://www.wiley.com/en-us/Conversation+Analysis%2C+2nd+Edition-p-9780745638669 Hutchby and Wooffitt introduce the tradition, its history and its method, and defend the claim that talk-in-interaction is an object of analysis in its own right rather than a window onto other social processes or variables; they set out single case analysis as a strategy. It gives the unit its account of what a checkable claim consists of.
• Perakyla, A. (2004). Reliability and validity in research based on naturally occurring social interaction. In D. Silverman (ed.), Qualitative Research: Theory, Method and Practice, 2nd edition. London: Sage, 283-304. https://archive.org/details/qualitativeresea0000unse_n8j9 Perakyla argues that reliability and validity in research on naturally occurring interaction are internal to its analytic logic, and sets out transparency, the next-turn proof procedure and the analysis of deviant cases as the criteria. It is the position the unit places against coding-based measurement. Its limit is its chapter form.
• Sacks, H., Schegloff, E. A. and Jefferson, G. (1974). A simplest systematics for the organisation of turn-taking for conversation. Language 50(4), 696-735. doi:10.2307/412243 Sacks, Schegloff and Jefferson propose a model of turn taking and test it against grossly observable facts about conversation, concluding that the arrangement is locally managed and administered by the parties themselves, interactionally controlled, and sensitive to recipient design. It is the unit's foundation, moving the analytic object from what a turn means to how the participants produce the order in which it is heard.
• Schegloff, E. A. (2007). Sequence Organisation in Interaction: A Primer in Conversation Analysis. Cambridge: Cambridge University Press. doi:10.1017/cbo9780511791208 Schegloff sets out sequence organisation, the ways turns are ordered and combined so that actions such as requests, offers, complaints and announcements take place. It supplies the mechanism by which a first turn makes a relevant next, and so the reason a turn's sense is positional. Its limit is the primer form.
• Schegloff, E. A., Jefferson, G. and Sacks, H. (1977). The preference for self-correction in the organisation of repair in conversation. Language 53(2), 361-382. doi:10.2307/413107 Schegloff, Jefferson and Sacks describe an organisation of repair addressed to recurrent problems in speaking, hearing and understanding, and explain the mechanism producing a strong skewing in which self-repair predominates over other-repair. It shows participants monitoring the sequence for trouble and preferring that the speaker fix it, which is the resource a system preserves or removes. Its limit is that it analyses ordinary talk in one language.
• Sidnell, J. and Stivers, T. (2013). The Handbook of Conversation Analysis. Chichester: Wiley-Blackwell. doi:10.1002/9781118325001 The handbook consolidates the field: turn design, turn-constructional units and the transition-relevance place, turn allocation, sequence organisation, preference and repair, with chapters on epistemics and the applied domains of medicine, the classroom, the courtroom and psychotherapy. It is the unit's reference for the machinery and the reach of its applications.
• Stivers, T. (2015). Coding social interaction: A heretical approach in conversation analysis?. Research on Language and Social Interaction 48(1), 1-19. doi:10.1080/08351813.2015.993837 Stivers notes that mixed methods studies combining conversation analysis with quantitative methods have been used since the 1980s, argues that formal coding of interaction is not necessarily antithetical to the tradition, and states that formal coding risks a massive reduction and flattening of complex conduct into simplistic codes. That trade-off is the unit's contested question.
• ten Have, P. (2007). Doing Conversation Analysis: A Practical Guide. 2nd edition. London: Sage. doi:10.4135/9781849208895 ten Have's guide works through the paradigm, the production of recordings, transcription, analytic strategies and the elaboration of an analysis, and on to institutional interaction. It shows what the analysis consists of in practice and how a recording becomes a claim.