Introduction and Overview

Through the first nine lessons of HSCI 841 you have treated transcripts mainly as containers of content. You coded for themes, you compared cases, you built typologies, you counted occurrences. The unit of interest was what a participant said — the propositional substance of the talk, the meanings encoded in it. In this lesson the analytic gaze shifts. Discourse analysis treats transcripts as records of action. The unit of interest is what the talk does: how a question opens a space the answer is obliged to fill, how a pause does the work of refusing without saying no, how the preface “okay, this is going to sound dramatic, but” rearranges a listener's expectations before the speaker has actually said anything dramatic at all.

The shift is large. It is not just that we will now attend to features of talk that intelligent-verbatim transcription smooths away — the half-second pauses, the in-breaths, the latching, the overlap, the rising intonation that turns a statement into a question. The bigger shift is theoretical: we will treat language as performative, in J. L. Austin's (1962) sense. Saying something is doing something. The question “How do you spend your days?” is not merely a request for information; it is also an instruction to produce a structured list answer, a demonstration of interest, and a position-taking about what counts as a relevant report of a life. A “weird answer” from the participant is not just an unusual content choice; it is a violation of the conversational contract that the speaker is now obliged to repair.

Bernard, Wutich, and Ryan (2017, Ch. 14) treat this as the third major moment in the analysis of texts, following content analysis (Ch. 13) and the structural narrative methods you encountered in Lesson 9. Their chapter is titled “Discourse Analysis II,” a labelling artefact of the second edition; we treat it as the standalone introduction to discourse analysis it functionally is. The chapter is organized around three strands: conversation analysis (CA), critical discourse analysis (CDA), and discursive psychology. Each strand has a different intellectual ancestry, a different unit of interest, and a different idea of what counts as a finding. Section 1 walks through the three strands and what separates them. Section 2 spends most of the lesson on conversation analysis, because CA is Bernard, Wutich, and Ryan's central exemplar and because its analytic moves transfer most directly to your loneliness dataset. Section 3 adds the Jefferson transcription system, Goffman's dramaturgical framework, and a working orientation to CDA. Section 4 applies all of this to a microanalysis of an interview excerpt. Section 5 is the final assessment.

1.1 Language-in-Use: A Working Definition

The simplest definition of discourse analysis available is the systematic study of language-in-use — that is, language considered not as an abstract system of signs (which would be the province of linguistics proper) but as it is deployed by actual speakers, in actual contexts, to accomplish actual social action (Cameron, 2001). The qualifier “in use” is doing real work. Discourse analysts are not asking whether a sentence is grammatical in English. They are asking what speakers did with that sentence in the particular conversation in which it occurred.

This stance has a philosophical heritage that traces back to Austin's How to Do Things with Words (1962) and his student Searle's elaboration of speech-act theory (Searle, 1969). Austin's contribution was to refuse the framing of utterances as primarily true-or-false propositions. Many utterances — warnings, promises, apologies, declarations, invitations — are not the kinds of things that can be true or false. “I promise to be there” is not a description of a promise; it is the promise. Austin called these performative utterances and argued that, on inspection, virtually all everyday talk has a performative dimension even when it appears descriptive. Saying “the cat is on the mat” in a conversation is not just reporting a fact; it is opening a topic, taking a turn, performing one's competence as a speaker, asserting one's right to be heard. The descriptive content rides on top of the action the utterance is doing.

Discourse analysis builds on this insight. Different strands inherit different versions of it. Conversation analysis takes it to the level of the turn: every turn at talk is doing something for the interaction. Critical discourse analysis takes it to the level of the social formation: large patterns of talk constitute the categories through which power operates. Discursive psychology takes it to the level of the mental life: psychological states are not internal furniture that talk reports on, but achievements that talk produces. All three share a commitment to treating language as the topic of study, not merely the medium for studying something else.

1.2 Strand One: Conversation Analysis

Conversation analysis is the strand Bernard, Wutich, and Ryan treat most extensively and the strand that will absorb most of your effort in this lesson. CA was developed in the 1960s and 1970s by Harvey Sacks, Emanuel Schegloff, and Gail Jefferson, three students of the sociologist Erving Goffman who took Goffman's interest in the orderliness of social interaction and pushed it into the empirical study of recorded talk (Sacks, Schegloff, & Jefferson, 1974; Sacks, 1992). Their intellectual debt was double: Goffman gave them the topic (interaction as a thing with its own structure); Harold Garfinkel and ethnomethodology gave them the method (treating ordinary practices as a topic of empirical study rather than a resource to take for granted; Garfinkel, 1967).

The central CA claim is that ordinary conversation is the bedrock form of human social interaction and that it is structured by orderly, demonstrable, describable practices that speakers and hearers themselves orient to. The order is not imposed by the analyst; it is produced by the participants and is visible in the recording. The CA evidentiary standard is therefore unusually demanding: a claim about conversational structure must be defended by showing that participants themselves treat the structure as relevant — usually by attending to what they do when the expected structure is breached.

The CA tradition has produced a remarkably stable set of findings about how ordinary talk works: how turns are allocated, how questions and answers fit together, how speakers signal preference and dispreference, how trouble is repaired. We will work through these in Section 2. For the moment, the point to register is that CA is the most empirically conservative of the three strands. CA practitioners are reluctant to make claims about psychological interiors, social structures, or even “meanings” in the abstract; they want to stay close to what is demonstrably going on in the talk itself. That conservatism is sometimes criticized as analytically narrow. It is also why CA findings have proven unusually transportable: the structures CA documents in everyday English conversation have been found, with adjustments, in dozens of other languages and settings.

1.3 Strand Two: Critical Discourse Analysis

Critical discourse analysis is, by contrast, the most expansive of the three strands. CDA emerged in the late 1980s and 1990s from the work of Norman Fairclough (1989, 1992, 2010), Ruth Wodak (Wodak & Meyer, 2009), Teun van Dijk (1993, 2008), and a small number of others. The intellectual debt to Foucault (1972) on the relationship between discourse, knowledge, and power is explicit across most CDA writing. CDA is openly normative: its commitment is not just to describe how language is used but to expose how language sustains, naturalizes, and reproduces relations of power and inequality. CDA is the strand most often used in health-policy research, in critical public-health work on stigma and racialization, and in the analysis of media coverage of health topics.

Fairclough's three-dimensional framework is the standard exposition of what CDA does (Fairclough, 1992, 2010). On dimension one (text) the analyst examines the linguistic features of a specific instance of discourse — vocabulary, grammar, metaphor, pronoun choice, modality, lexical patterning. On dimension two (discursive practice) the analyst examines how the text was produced, distributed, and consumed — the institutional processes that shape who gets to say what to whom. On dimension three (social practice) the analyst examines the wider social, political, and economic relations that the discursive practices contribute to reproducing or contesting. A CDA analysis of, say, the Canadian federal public-health communications about COVID-19 would move across all three dimensions: textual features (modality of certainty, choice of personification, framing of “essential workers”); discursive practices (who drafted the communications, how they were vetted, which audiences they targeted); and social practice (how the communications related to ongoing patterns of racialized labour, who was constructed as a deserving subject of public-health concern).

CDA is sometimes criticized for the looseness of its evidentiary standard. Where CA can point to a participant orientation as evidence that a structure is real, CDA often relies on the analyst's interpretive judgment that a particular textual feature is doing political work. The defence is that CDA does not pretend to be neutral; its commitments are explicit, and the evidentiary discipline lies in offering enough textual and contextual detail that another competent analyst can evaluate the reading. The trade-off is between empirical conservatism (CA) and theoretical reach (CDA).

1.4 Strand Three: Discursive Psychology

The third strand — discursive psychology — is the smallest of the three but the most theoretically distinctive. It originated in Britain in the 1980s with the work of Jonathan Potter and Margaret Wetherell (Potter & Wetherell, 1987; Edwards & Potter, 1992; Wiggins, 2017). The broader analytic toolkit for treating language as social action is laid out in Gee's introduction to discourse analysis (2014). The discursive-psychology move is to take what looks like a psychological topic — memory, attitude, emotion, prejudice, identity — and ask how it is discursively constituted in talk. Rather than treating “loneliness” as an internal state that participants more or less accurately report on, the discursive psychologist asks what work participants do with the word, in what conversational positions, with what consequences.

Concretely, a discursive-psychology analysis of the loneliness transcripts would look at things like: how participants account for their loneliness in ways that manage blame (theirs? others'? structural?); how they construct themselves as the kind of person who is or is not properly lonely; how they pre-empt skepticism by prefacing their reports with markers like “this is going to sound dramatic, but”; how they perform an identity of being-reasonable about a stigmatized experience. The unit of interest is what loneliness does as a category in talk, not what loneliness is as a state in the mind.

Discursive psychology shares CA's emphasis on what gets done in talk but shares CDA's willingness to make inferences beyond the surface of the conversation, especially about identity, accountability, and the construction of psychological categories. Of the three strands it is the one closest to what your loneliness capstone is most likely to find useful, particularly when you encounter passages where participants are clearly doing identity-work and face-management around the admission of being lonely.

1.5 Why a Loneliness Researcher Should Care

Through Lessons 1–9 you have built a substantial analytic apparatus for the loneliness transcripts: codes, themes, taxonomies, narrative structures, schema-level inferences. What can a discourse-analytic layer add? Three things, concretely.

First, the interview is a conversation. Almost every claim your capstone will make is grounded in a transcript whose features were jointly produced by interviewer and participant. The participant's answer to question seven cannot be properly understood without attending to what question seven did to position them. When Helen (P11) gives a structured list answer to “tell me about yourself” — “Well, dear, I'm 78, never married, no children. I was a librarian at the central branch downtown for 38 years” — she is not merely reporting facts. She is performing a kind of biographical competence that the question solicited. A CA-informed analysis will help you see that the form of her self-presentation is itself analytic data.

Second, loneliness is a stigmatized topic. Multiple participants do recognizable face-work around the admission of being lonely — pre-empting judgment (“this is going to sound dramatic”), accounting for it (“I'm 22, I'm in university, I'm supposed to be having the best years of my life”), distancing themselves from a lonely identity (“a person who sometimes feels lonely, not a lonely person”). A thematic reading will capture the content of these moves but not their interactional achievement. A discourse-analytic reading shows the work the participant is doing to remain a reasonable interlocutor while admitting something they consider potentially discrediting.

Third, the public-health conversation about loneliness is itself a discursive formation. When the UK appoints a Minister of Loneliness or when CDC frames social isolation as a public-health crisis, those moves produce loneliness as a particular kind of object — a measurable population-level phenomenon amenable to programmatic intervention, rather than (for example) a moral signal of community failure or a structural consequence of late-capitalist housing. A CDA-informed reading of your participants' responses to question 23 (“some governments have appointed a Minister of Loneliness, what do you make of that?”) can document how the participants align with, resist, or rework the public-health framing.

Introduction and Overview

Conversation analysis is, on one reading, the most ambitious empirical project in twentieth-century social science. It claims that there is a discoverable, language-general, demonstrably real structure to ordinary conversation, and that this structure is producible from recordings of actual talk without any appeal to participants' inner states, cultural categories, or theoretical commitments external to the data. Sacks, Schegloff, and Jefferson spent forty years developing the evidence for that claim. The findings are unusually durable. Many of them appear in every introduction to CA written since 1974 and have not been seriously revised; what has been revised is the granularity at which they are described.

In this section we walk through the core CA findings as a working analyst would deploy them. Four bodies of structure organize most everyday talk: turn-taking, adjacency pairs, preference organization, and repair. A fifth, sequence organization, ties the four together into larger units. We will work through each, with loneliness-transcript examples wherever possible.

2.1 Turn-Taking: The Sacks, Schegloff & Jefferson Model

The Sacks, Schegloff, and Jefferson (1974) paper — usually referred to as SSJ — is one of the most cited articles in twentieth-century sociology, and for good reason. It addressed a question everyone had assumed was uninteresting: how do speakers manage to take turns in conversation? The empirical observation that prompted the question is that despite the apparent looseness of ordinary talk — people don't raise their hands, there is no chairperson — in practice, conversations show remarkably little overlap (typically less than 5% of speech time) and remarkably short gaps between turns (typically less than 200 milliseconds). Something is keeping the system orderly, and the participants are doing it themselves in real time.

SSJ's model has three components. First, talk is organized into turn-construction units (TCUs) — bits of speech that, on their own, count as a complete contribution. A TCU can be a single word (“Yeah.”), a phrase (“After dinner.”), a clause (“I went to the store.”), or a sentence. Whether something is a TCU is decided locally by the speakers, not by grammar: what matters is whether the current speaker has produced something that listeners hear as possibly complete.

Second, at the end of each TCU there is a transition-relevance place (TRP) where speaker change becomes a live possibility. The TRP is not a mandatory turn-end; it is a possibility-point. What happens at the TRP determines who speaks next.

Third, the rule system at each TRP works as follows: (a) the current speaker may select the next speaker (by addressing them, asking them a question, etc.); if not, (b) any other speaker may self-select; if not, (c) the current speaker may continue. These rules apply recursively at every TRP. The model predicts — and the data confirm — that overlaps will cluster around TRPs (where multiple speakers try to take the next turn) rather than randomly throughout speech.

For your loneliness transcripts, the turn-taking model is less directly visible because the transcripts are produced in intelligent-verbatim style, which smooths out micro-pauses, overlaps, and latching. But it is still analytically useful. Notice that Helen's long answers in P11 are not single turns — they are sequences of TCUs that the interviewer chooses not to interrupt at any TRP. The interviewer's restraint is itself analytic data: a more interventionist interviewer would have produced a different transcript by taking different turns.

2.2 Adjacency Pairs

The adjacency pair is the most analytically productive unit in CA. Schegloff and Sacks (1973) defined an adjacency pair as a sequence of two utterances that are (a) adjacent, (b) produced by different speakers, (c) ordered as a first-pair-part (FPP) and second-pair-part (SPP), and (d) typed, so that a given FPP requires a particular type of SPP.

The classic adjacency pairs are: question/answer, greeting/greeting, invitation/accept-or-refuse, request/grant-or-refuse, offer/accept-or-refuse, complaint/apology-or-rejection, assessment/agreement, compliment/acceptance-or-deflection. Each FPP makes a particular SPP conditionally relevant — meaning that the absence of the relevant SPP is itself a noticeable, accountable event. If A greets B and B does not greet back, B has not merely failed to do something; B has done something specific (snubbed, ignored, perhaps not heard). The conditional-relevance property is what gives adjacency pairs their analytic power: it specifies what the next speaker is now obliged to do, and makes any departure from that obligation interpretable.

The loneliness transcripts are saturated with adjacency pairs because they are interviews, and interviews are an organized exchange of question-and-answer pairs. Consider the opening of P11's interview:

I: Thank you for having me, Helen. Tell me about yourself.
P: Well, dear, I'm 78, never married, no children. I was a librarian at the central branch downtown for 38 years... P11 Helen, lines 15–17

The first-pair-part is “Tell me about yourself” — a request-for-self-description, a particular subtype of question. The conditionally relevant second-pair-part is a self-description. Helen produces one. What is analytically interesting is the form of the self-description she produces: a structured list answer (age, marital status, parental status, occupation, length of tenure). The form is not in the question; the question was open. Helen is doing something here — producing a recognizable biographical idiom that signals competent participation in the interview genre. A different participant might have produced a different form: a feeling-state report, a current-moment description, a refusal of the genre. The adjacency-pair frame helps us see Helen's answer as a specific kind of compliance with a specific kind of solicitation, not just as content.

Adjacency pairs also extend: many FPPs allow a range of SPP types, and within that range there is a strong preference structure that the next section addresses.

2.3 Preference Organization

For adjacency pairs that allow more than one type of SPP — invitations that can be accepted or refused, assessments that can be agreed or disagreed with, requests that can be granted or refused — CA documents a robust pattern: one SPP type is preferred and the others are dispreferred. Preference here is not a psychological term about what speakers want; it is a structural term about the form of the turn (Pomerantz, 1984; Schegloff, 2007).

Preferred SPPs tend to be: (a) delivered without delay; (b) structurally simple (no preface, no qualification); (c) contiguous with the FPP (no insertion between them). Dispreferred SPPs tend to be: (a) delayed (often by a pause, a discourse marker like “well” or “uh,” or by an in-breath); (b) prefaced with markers that signal an unwelcome answer is coming (“I'd love to, but...”); (c) accounted for (the speaker explains why the dispreferred answer is being produced); (d) mitigated in form (softened, qualified, hedged).

The empirical regularity is striking. Even speakers who do not know each other and have no reason to share preferences in the psychological sense produce dispreferred turns with these features. The form of the turn is what the dispreference status consists in.

The loneliness transcripts provide a perfect case. Consider Maya's closing line:

P: ...I just — I wish we could opt back in. Like a button you could wear that says I'm okay to talk. [pause] Sorry, that's a weird answer. P01 Maya, line 57

The pause and the explicit account (“Sorry, that's a weird answer”) are textbook dispreferred-turn features — except that this is not a second-pair-part to an adjacency pair, but rather a self-evaluation of Maya's own answer. Maya is treating her own utterance as if it were dispreferred and preempting the dispreference by doing the apology and account in advance. This is what discursive psychology would call face-work: managing the social acceptability of her own talk before the interviewer can negatively evaluate it. The CA frame gives us the vocabulary to name the form of the move; the discursive-psychology frame helps us understand why she is making it (loneliness is a stigmatized topic; suggesting that strangers should wear buttons inviting conversation is potentially read as “weird”).

2.4 Repair

Repair is CA's term for the practices speakers use to deal with trouble in talk — trouble in hearing, in understanding, in producing the right word, in maintaining the conversational alignment between participants (Schegloff, Jefferson, & Sacks, 1977). Repair is everywhere in conversation. Sacks, Schegloff, and Jefferson estimated that in ordinary English conversation, some form of repair occurs roughly every minute and a half.

Repair is classified along two dimensions: who initiates the repair (the speaker of the trouble-source = self; the recipient = other) and who performs the repair (self or other). This produces four types:

1. Self-initiated self-repair — the speaker notices their own trouble and fixes it. “I went to the — I drove to the store.” The most common form.
2. Other-initiated self-repair — the recipient signals trouble (“Sorry, who?” “What did you say?”) and the original speaker fixes it. The second most common form.
3. Self-initiated other-repair — the speaker signals trouble (“The one with the — you know, the thing”) and the recipient supplies the fix. Common in long-running relationships where one party has stronger lexical access on certain topics.
4. Other-initiated other-repair — the recipient both signals and supplies the repair (“You mean the Cambie store”). The rarest form, because it is socially marked: it has the force of a correction.

Repair is also organized by position relative to the trouble source. First-position repair occurs within the same turn as the trouble source. Second-position repair occurs in the very next turn. Third-position repair occurs in the turn after the trouble-source speaker has had a chance to respond. Fourth-position repair occurs even later. The pattern is that the closer to the trouble source the repair happens, the less interactionally costly it is — first-position self-repair is almost invisible; fourth-position repair is dramatic and disruptive.

Repair in interview data is unusually rich because interviews involve sustained interpretive work by both parties about what is being said and what is being asked. Consider Robert's account of his ex-wife:

P: My ex caught me drinking by myself last summer on a Wednesday afternoon and that was, uh, that was a bad moment. P04 Robert, line 25

The “uh, that was” sequence is a first-position self-initiated self-repair on the assessment he is producing. He has begun “that was”; he hesitates, deploys an “uh,” and re-starts with the same words. This isn't a repair on a word he forgot; it's a repair on the kind of evaluation he is about to give of the event. The hesitation is the work of choosing how to characterize the moment. CA reads the hesitation as a sign that more than one characterization was available and the speaker is selecting among them. The selection (“a bad moment”) is itself a reduced form — an understatement — and the repair sequence is the speaker calibrating that understatement in real time.

2.5 Sequence Organization: Pre-Sequences and Insertion Sequences

Adjacency pairs are the smallest organized units of conversation, but they rarely appear in isolation. They are nested and chained into larger sequences. Two of the most analytically productive larger structures are pre-sequences and insertion sequences (Schegloff, 2007).

A pre-sequence is an adjacency pair that prepares for a subsequent action. The classic example is the pre-invitation: “What are you doing Saturday?” — this is not actually a question about Saturday; it is a check on whether the conditions for an invitation are favorable. If the response is “Nothing, why?” the invitation will come next. If the response is “Oh, I'm busy,” the invitation will not be issued, and the would-be inviter has been protected from the social cost of a refusal. Pre-sequences are a major device for managing potentially face-threatening action.

An insertion sequence is an adjacency pair that is embedded between the FPP and SPP of another adjacency pair. The classic example: “Want coffee?” / “Decaf or regular?” / “Regular.” / “Yes please.” The middle pair (decaf-question, regular-answer) is inserted between the offer and the acceptance because the acceptance depends on the answer to the inserted question.

In interview data, insertion sequences are everywhere because interviewers routinely ask clarifying questions before the participant has completed their answer. Pre-sequences are less common in interview talk — the genre licenses direct questioning — but pre-faces that do similar work (“Can I ask you something?”) are recognizable. Robert's “okay, this is going to sound dramatic, but...” (line 25 of his transcript) is functionally a pre-sequence: it is a framing turn that conditions how the upcoming utterance will be received. It is not formally a question/answer pair, but its work is analogous — preparing the recipient for an utterance that needs interpretive scaffolding.

2.6 Putting It Together: A CA Reading of an Adjacency-Pair Sequence

Let me close this section with a slightly longer worked reading, before Section 3 introduces Jefferson notation and Section 4 asks you to do this work yourself. Helen's answer to “how do you spend your days?” in P11 is a structured list, but the structure is informative. Reading the question/answer pair in the original transcript:

I: When you hear loneliness, what comes to mind?

P: Days where I do not say a word out loud to another person. That happens regularly. I will get up, make my breakfast, listen to the radio, do my exercises, eat my lunch, do a crossword on the audio thing my niece set up, and go to bed. And I will not have spoken to anyone. Three or four days in a row, sometimes. And then I will go to the grocery store and the cashier will say "have a nice day" and I'll feel — I'll feel almost giddy from the contact. P11 Helen, lines 19–21

What CA notices: Helen's answer to a request for definition (“what comes to mind?”) is delivered as a temporal-narrative list (get up — breakfast — radio — exercises — lunch — crossword — bed). The list-form is the answer's primary structural feature. Then there is a first-position self-initiated self-repair: “I'll feel — I'll feel almost giddy from the contact.” She begins with “I'll feel,” cuts off, and re-starts with the same phrase followed by the intensifier “almost giddy.” The repair is on the strength of the affective claim, not on its content. She is calibrating the magnitude of the feeling she is reporting — choosing “almost giddy” over the unmarked “giddy” or the milder “a little happy.” The repair is doing the work of producing a precise, defensible self-assessment in real time.

What thematic coding sees: “Helen describes loneliness as days of no verbal contact; cashier interactions produce intense feelings of connection.” True, but flat. What CA adds is the demonstration that Helen is doing precision-work in her self-report, and that the list-form is how she gives her loneliness empirical content (these specific activities, in this order, with this absence at the centre). The CA reading does not contradict the thematic reading; it adds a layer about how the report is being constructed in real interactional time.

Introduction and Overview

Section 2 gave you the analytic vocabulary of conversation analysis. This section gives you three more pieces of equipment. First, the Jefferson transcription system: the notational conventions Gail Jefferson developed across the 1970s and 1980s to render the features of talk that ordinary orthography erases. Jefferson notation is what makes CA possible as a public, defensible practice; it is what gives a transcript enough granularity to do interactional analysis. Second, Erving Goffman's dramaturgical framework: a complementary apparatus, developed before CA crystallized, for thinking about social interaction as performance, and about the management of self-presentation in front-stage versus back-stage settings. Third, an operating orientation to critical discourse analysis with Fairclough's three-dimensional framework as the working scaffold.

3.1 Jefferson Transcription: Why and What

The transcripts you have been working with all term are produced in intelligent verbatim style. The conventions: include all words the participant said; remove false starts that obscure meaning; remove filler words that do not contribute (“um,” “like,” “you know,” though we have kept many of these for narrative texture); standardize punctuation; smooth pauses into ellipses or em-dashes; ignore overlap; ignore in-breaths and out-breaths; ignore prosody. Intelligent verbatim makes transcripts readable and is appropriate for thematic, narrative, and content-analytic work. It strips out almost everything CA needs.

For CA, transcription is itself an analytic act. The conventions Jefferson developed across her career (collected most authoritatively in Jefferson, 2004) try to render in writing the features of talk that matter for interactional analysis: how long the pauses are, where speakers in-breathe before a difficult turn, where two speakers overlap, where emphasis falls, where volume rises and falls, where pitch is marked. The full Jefferson system has more than thirty conventions; a working subset of about a dozen captures most of what an analyst needs.

For your capstone, you do not have access to the original audio — the transcripts are intelligent-verbatim synthetic composites, with pauses marked only as “[pause]” or “[long pause]” and many micro-features unrendered. This is a real constraint, and it is a teaching opportunity. Part of the Week 10 milestone is to take a passage and re-transcribe it in Jefferson style, inferring pauses, emphasis, and prosodic features from the intelligent-verbatim source plus your reading of the passage in context. The act of re-transcription is itself part of what discourse analysis teaches: it forces you to attend to features of the talk that thematic reading slides past, and it makes you aware of how much intelligent-verbatim erases.

A Worked Re-Transcription

Here is the intelligent-verbatim version of an exchange from P01 (Maya, lines 23–25):

I: How do you know when you're lonely? Like what does it feel like?

P: It's — okay, this is going to sound dramatic, but it feels like being hungry. Like, in my chest. It's a physical thing. I get this feeling, especially at night, where my chest just feels — hollow isn't the right word, it's like an ache. Um. P01 Maya, intelligent-verbatim transcript

And here is a plausible Jefferson-style re-transcription, inferring conservatively from the intelligent-verbatim source. Pauses are estimated from the context and the em-dashes:

The re-transcribed version makes several things visible that the intelligent-verbatim version smoothed. The 0.8-second pause before the answer is the marker of a dispreferred turn — she is preparing to say something she anticipates will be received as excessive. The mid-turn cut-off “It's-” followed by a 0.5-second pause and the preface “okay, this is going to sound dramatic, but” is a self-initiated pre-emptive face-management move. The emphasis on “chest” (rather than on “hungry” or “physical”) localizes the felt experience anatomically — she is doing the work of giving loneliness a body part. The repair “hollow isn't the right word, it's like an ache” is a first-position self-initiated self-repair on her own lexical choice, and the emphasis on “right” treats word-choice itself as something that can be more or less right.

The intelligent-verbatim transcript records what was said. The Jefferson re-transcription records how it was said and what the saying was doing in real time. The two are different analytic artefacts.

3.2 Goffman: Performance, Front-Stage, Back-Stage

Erving Goffman (1922–1982) is the figure CA's founders learned from and then exceeded. His Presentation of Self in Everyday Life (1959) is the foundational text of the dramaturgical perspective: the claim that social interaction is best understood through the analogy of theatrical performance, in which actors stage carefully managed versions of self for audiences whose readings of those performances constitute the actor's social reality. His later Frame Analysis (1974) and Forms of Talk (1981) provide the vocabulary — frame, footing, face, face-work — that is now standard equipment in discourse analysis even outside formal CA.

Goffman's core distinction is between front-stage and back-stage performance. Front-stage is where the actor is visible to the audience and works to maintain a coherent, presentable, role-appropriate self. Back-stage is where the actor relaxes from the demands of the performance, prepares for the next front-stage moment, and is available to a different (often smaller) audience of fellow performers. The hotel dining room is front-stage; the kitchen is back-stage. The clinic consulting room is front-stage; the staff room is back-stage. The interview itself is front-stage; what the participant says to a friend after the interview is back-stage.

For interview research, this distinction is consequential. Participants in an interview are always to some degree performing — producing a presentable version of self for a sociological audience whose evaluations they cannot directly perceive but whose existence they know about. The performance is not insincere; sincerity is itself one of the things being performed. Goffman's insight is that the performance-character of the encounter is not a flaw to be controlled for but a constitutive feature to be analyzed. What kind of self is the participant trying to present? What audience are they presenting it to (you? the imagined readers of your paper? the institutional ethics board?)? What back-stage versions might exist?

The related Goffmanian concept of face (Goffman, 1955, 1967) is the public self-image a person claims in an interaction. Face-work is everything the participant does to maintain that face against threats — their own missteps, the recipient's potential disapproval, the difficulty of the topic. Reading the loneliness interviews through a face-work lens makes a great deal of the transcript material newly visible. Maya's “sorry, that's a weird answer” is face-work. Robert's “okay, I'm not a feelings guy” preface to a discussion of his emotional response is face-work (specifically, footing-work: he is repositioning himself relative to the topic before producing the emotionally laden utterance). Helen's structured list answer in lines 19–21 is also face-work: she is presenting a person whose loneliness is empirically grounded in describable daily activity, not a person who is performing a complaint.

Footing (Goffman, 1981) refers to the participant's alignment to the utterance they are producing — the relation between the speaker and what is being said. A speaker can be the animator (the one physically producing the talk), the author (the one whose words they are), or the principal (the one whose position the talk represents). In direct quotation, animator and author come apart: I am the animator but the original speaker is the author. In citations of others' views (“my GP says...”), the speaker is animator but the GP is the author and principal. Footing shifts are diagnostic of how the speaker is relating to their own talk. Robert's “I'm not a feelings guy” is partly a footing claim: he is positioning himself as an animator who does not normally produce the kind of talk he is about to produce, and inviting the interviewer to read what follows as out-of-character.

3.3 Performance in Clinical and Public-Health Encounters

Goffman's framework has been particularly productive in the discourse-analytic study of clinical interaction. Heritage (1984, 2005) and Heritage and Maynard (2006) have produced a body of work on the conversational organization of medical encounters that uses both CA and Goffmanian categories. West (1984, 1990) examined the specific patterns of physician-patient interaction with attention to gender and racialization. The accumulated finding is that the clinical encounter has a distinctive turn-taking structure (the physician asks; the patient answers; the physician evaluates the answer and asks the next question — a three-turn IRE pattern borrowed from the analysis of classroom talk) and a distinctive distribution of face-threats (patients are routinely required to discuss embarrassing, painful, or stigmatized topics; physicians routinely deliver bad news or refuse requests).

Several specific clinical interaction events have been the subject of sustained CA research: the opening of the consultation (Robinson, 1998); the patient's presentation of the chief complaint (Heritage & Robinson, 2006); the announcement of a diagnosis (Maynard, 1991, 2003); the delivery of bad news (Maynard, 2003); the informed-consent conversation (Roberts, 2002); and the closing of the encounter. Each event has documented conversational properties — pre-sequences, dispreferred turns, repair patterns — and each provides a model for what a discourse-analytic look at non-clinical health interactions (like your loneliness interviews) can document.

The contemporary moment adds telehealth, which has its own emerging discourse-analytic literature (Stommel & Stommel, 2021; Shaw et al., 2022). The conversational dynamics of a video consultation differ from those of an in-person consultation in identifiable ways: the visual channel is partial, gaze is altered by camera position, side-talk by other people in the room is more frequent, technical trouble interrupts turns. These features have implications for how clinical CA findings transfer to telehealth contexts and for the design of telehealth services in public-health programs.

3.4 Critical Discourse Analysis: Fairclough's Framework in Operation

The third equipment item in this section is an operating orientation to critical discourse analysis. Section 1 introduced Fairclough's three-dimensional framework conceptually; here we walk through it in operation.

Dimension one — text. The textual analysis is the most CA-like part of CDA, but it works at a different grain. The CDA analyst examines linguistic features at the level of the text rather than the turn: vocabulary choices and their connotations; transitivity (who is positioned as agent and who as patient in the grammatical structure of sentences); modality (the degree of certainty or obligation expressed by modal verbs and adverbs); intertextuality (which other discourses the text draws on); presupposition (what is taken as given rather than asserted); metaphor; pronoun choice and the construction of in-groups and out-groups; nominalization (turning processes into things, which often hides agents).

Consider a single sentence from a typical Canadian federal public-health document on loneliness: “Loneliness has been identified as a significant public health concern.” A CDA reading attends to the passive construction (identified by whom?); the nominalization (loneliness as a thing, rather than as something experienced by particular people in particular conditions); the modal-adjective “significant” (significant to whom, by what standard?); the framing as “concern” (an emotional-administrative category, rather than as suffering or as injustice). The textual choices add up to a particular construction of what loneliness is and who is positioned to act on it.

Dimension two — discursive practice. The discursive-practice analysis examines the processes by which the text was produced, distributed, and consumed. Who drafted the public-health document? Which experts were consulted? Which were not? What other texts does the document cite, and which does it not cite? Who is the imagined reader? Through what channels is the document distributed (web, press release, internal memo, briefing book)? How is it taken up by news media, by clinicians, by the populations it claims to address? The discursive-practice level is where CDA most departs from CA: instead of staying close to the text, it traces the text's institutional life.

Dimension three — social practice. The social-practice analysis situates the discursive event in its wider context: the political-economic conditions, the demographic patterns, the historical trajectories of how the topic has been constructed in public discourse. For loneliness, the social-practice analysis would attend to the housing market, the labour market, the privatization of care, the post-2008 economic patterns that have made multigenerational living more economically necessary and more culturally fraught, the differential burdens on immigrant and racialized populations, the gendered distribution of unpaid care work. The text is a moment in a much larger field of social practice; CDA reads the text against that field.

An operating CDA analysis usually does not give equal weight to all three dimensions; it builds an argument that moves through them. A short paper might examine a single text in detail (dimension one), trace its production and uptake briefly (dimension two), and locate it within a broader social-practice argument that other scholars have developed (dimension three). The discipline is to make all three layers visible enough that another analyst could disagree with you on specific points.

Introduction and Overview

Sections 1–3 have given you the analytic vocabulary: three strands, four CA structures, a Jefferson notation, the Goffmanian framework, and a working orientation to CDA. This section puts the vocabulary to work on an extended worked example, then sets up the Week 10 capstone milestone, in which you produce your own microanalysis.

4.1 The Excerpt

The passage I will work through is from P04 Robert, lines 21–25 of his transcript. It is short — the canonical CA microanalysis works on small chunks of talk — and it contains several of the features we have built vocabulary for.

I: When you hear the word loneliness, what comes to mind?

P: The week the kids aren't with me. That's it. That's the whole answer. The weeks they're with me, the apartment's loud, there's noise, there's homework, there's me cooking dinner and them complaining about it. And then Sunday afternoon I drop them at Jenny's and I come back to this — this silence. And the apartment is a tomb. I'll sit on the couch for like six hours and not move. So that's loneliness for me. It's the off-weeks.

I: How does it feel physically?

P: Yeah. [pause] So, look, I'm not a feelings guy. I don't talk like this normally. But — I get this thing in my throat. Like a tightness. Like I want to cry but I can't. And I don't sleep. I drink more on those weeks. I know I do. My ex caught me drinking by myself last summer on a Wednesday afternoon and that was, uh, that was a bad moment. So yeah. Throat, sleep, the drinking. That's the body part. P04 Robert, intelligent-verbatim transcript, lines 21–25

4.2 Jefferson Re-Transcription (Inferred)

Here is a plausible Jefferson re-transcription, with pauses and emphasis inferred from the intelligent-verbatim source plus the dashes and pacing cues:

4.3 The Microanalysis

What does this re-transcribed sequence let us see? I will work through five observations.

Observation 1: The first answer is a closed list, then opens up. The opening four turn-construction units — “The week the kids aren't with me. (0.4) That's it. (.) That's the whole answer. (0.8)” — constitute a closure-claim. Robert produces a one-clause definition, immediately marks it as complete (“That's it”), and then re-asserts the completeness with emphasis on “whole.” The 0.8-second pause that follows is a checkpoint: he is offering the interviewer the chance to accept the short answer and move on. When the interviewer does not take the turn, Robert continues. The continuation is structurally different from the closure: it is a contrastive narrative pair (loud weeks / silent weeks) that ends with a return to the definitional formulation (“So that's loneliness for me. It's the off-weeks”). The structure is closure → checkpoint → elaboration prompted by the interviewer's restraint → closure-recapitulation. Helen does something similar in P11 lines 19–21; this looks like a common pattern in our dataset.

Observation 2: The body-question answer is heavily prefaced. The transition from emotional content (“loneliness is the off-weeks”) to physical content (“how does it feel physically”) produces a long delay (1.2 seconds), a discourse marker (“So, look”), a footing claim (“I'm not a feelings guy. I don't talk like this normally”), and a contrastive connective (“But”). This is a textbook dispreferred-turn structure, with the dispreference here being not that he is refusing the question, but that he is going to produce a kind of talk (emotional, embodied) that his self-presentation as “not a feelings guy” would normally rule out. The footing claim accomplishes specifically Goffmanian work: he is establishing that the upcoming utterance is out-of-character, which preserves his front-stage self as “not a feelings guy” while permitting the feelings-talk that the question solicits.

Observation 3: A repair on a self-assessment. “My ex caught me drinking by myself last summer on a Wednesday afternoon and that was, .hh uh, that was a bad moment.” The in-breath, the “uh,” and the re-start (“that was, ... that was”) are a first-position self-initiated self-repair. He has begun an assessment, hesitated, audibly in-breathed, and then re-started with the same syntactic frame and the qualifier “bad.” The repair is not on a referential error; it is on the strength and form of his own evaluation of the event. The eventual word — “bad” — is understated relative to what a drinking-discovery by an ex-spouse during the day on a weekday might warrant. The repair is the speaker calibrating the magnitude of his own admission downward, perhaps because admitting to a more severe characterization would threaten the “not a feelings guy” front-stage performance.

Observation 4: The metaphors are anchored. “The apartment is a tomb.” “A thing in my throat. Like a tightness.” The metaphors are doing definitional work that the literal vocabulary cannot do, and they are doing it through embodiment (tomb = dead space; throat tightness = physical correlate of suppressed crying). The metaphor of tomb specifically maps the lived space onto a death-related image — in a dataset of widowed and bereaved participants (P05 Linda; P19 Rose) this is a striking choice from a divorced (not widowed) participant. CDA would notice that the tomb metaphor invokes a death-frame for what is structurally an absence-frame; discursive psychology would notice the affordance for legitimating the magnitude of the felt experience (a tomb is a serious thing to be in).

Observation 5: The closing recapitulation organizes the answer for the analyst. “So yeah. (.) Throat, sleep, the drinking. That's the body part.” Robert ends the body-question answer with an enumeration (throat, sleep, drinking) and a closure-claim (“That's the body part”). The enumeration is structurally a list, the same form Helen used in P11 line 21. The closure-claim parallels the closure he produced after the first answer (“That's it”). The list-then-closure structure appears to be a recurrent feature of his self-presentation across the interview. A grounded-theory or thematic reading might note this as a stylistic feature. The CA reading allows us to specify the function: closure-claims are turn-yielding devices that hand the floor back to the interviewer, marking the speaker's contribution as complete and the next move as the interviewer's.

4.4 What the Discourse-Analytic Reading Adds

The thematic reading of this excerpt (Lesson 5 vocabulary) would code roughly as follows: “Loneliness defined as absence of children; bodily symptoms include throat tightness, sleep loss, increased alcohol use; explicit identification of a moment of trouble (ex-spouse discovering daytime drinking); self-characterization as 'not a feelings guy.'” Those codes are accurate. They are also flat: they record what Robert said but not how the saying was structured or what work the structure was doing.

The discourse-analytic reading adds the following:

• A specific account of how Robert's self-presentation as “not a feelings guy” is interactionally accomplished — via footing claims, dispreferred-turn prefaces, and repair-mediated downgrading of self-assessment.
• A pattern (closure-then-checkpoint-then-elaboration) that recurs across his answers and may recur across the dataset.
• A specific reading of his metaphor choices as doing legitimation work — the “tomb” framing licenses the magnitude of feeling that the “not a feelings guy” identity would otherwise undercut.
• Evidence that loneliness is, for this participant, a topic that requires sustained face-management even in a research interview that explicitly invites the talk.

None of these readings contradict the thematic reading. They add a layer that lets us say things about how Robert produces a credible account of his loneliness, not only what the content of that account is. For a public-health audience interested in the experience of loneliness among divorced men — a population with documented elevated risk for severe loneliness and for related health outcomes — the “how” is methodologically important. It tells us what kinds of supports or framings would or would not allow this population to be reached.

4.5 The Hand-Work of CA

Conversation analysis is hand-work. There is no software that does CA the way Taguette does code-and-retrieve or the way quanteda does content analysis at scale. The reasons are not technical but conceptual: the CA analyst is producing a transcript and a reading that are themselves the analytic product, and the production cannot be automated without losing what makes the analysis CA. There are tools for transcript management — CLAN (the Computerized Language Analysis system from the CHILDES project; MacWhinney, 2000) supports timed alignment of audio with transcript and is widely used in developmental and clinical CA work; ELAN (the EUDICO Linguistic Annotator from the Max Planck Institute) is the standard for multimodal video work with multiple tiers of annotation. Both are free. Both have substantial learning curves.

For your capstone you do not need either. The transcripts are intelligent-verbatim text. The audio does not exist. Your tools are: a text editor (any will do; a monospaced font helps because Jefferson notation aligns better in monospace), a printout of the Jefferson conventions, and your reading attention. Print the passage. Read it aloud at the pace you imagine the participant spoke. Mark pauses where the punctuation and dashes suggest. Mark emphasis where the lexical and contrastive content suggests. Write the analysis on the same page as the re-transcription. This is the workflow CA practitioners have used for sixty years and it is the workflow your Week 10 milestone uses.

4.6 The Week 10 Capstone Milestone

The Week 10 milestone is a one-week deep dive into discourse analysis as a complement to the predominantly thematic and structural analysis you have done through Weeks 5–9. The milestone is designed to teach you what a discourse-analytic layer would add to your eventual paper, without requiring you to convert the whole capstone into a CA study.

Bringing It All Together

Lesson 10 has reframed the transcript as a record of conversational action rather than (only) as a container of content. The shift is consequential. It opens the analytic gaze to features of the talk — turn-taking, adjacency pairs, preference, repair, sequence, face, footing — that the thematic and narrative methods of earlier lessons did not directly address. It introduces a transcription system (Jefferson notation) that is itself part of the analysis. It situates the work in a larger family of three strands — CA, CDA, and discursive psychology — whose different commitments give the analyst different things to look for.

The lesson's central practical move is the microanalytic dive into a small interactional moment. The 20–40 line excerpt — rather than the full corpus — is the natural unit of discourse-analytic work, and the depth of the reading is what licences the claims. For your capstone paper, the Week 10 deliverable need not produce a full CA study; it produces a layer that demonstrates you can do interactional analysis, and that, when combined with your thematic and structural analysis, gives your eventual reader a richer picture of how your participants do the work of being lonely in talk.

Lesson 11 picks up a different thread: analytic induction, qualitative comparative analysis, and decision models — the model-testing side of qualitative work. Lesson 12 closes the course with computational text analysis and large language models. Together with Lesson 10, those two final lessons round out the full Bernard-Wutich-Ryan toolkit.

The final reflection below asks you to consolidate the lesson by naming what discourse analysis adds to your eventual capstone paper. There is no single correct answer; the goal is to articulate a stance that you could defend in the methods section.

This glossary collects the key concepts, people, and methodological stances introduced in Lesson 10. Use it as a reference while you work through the material, or as a review before the final assessment. Type in the search box to filter entries.