9

Conversation analysis
Steven E. Clayman and Virginia Teas CiII

Introduction
Conversation analysis is an approach to the study of human interaction in society. Its name might
be taken to imply a concern with informal and purely sociable talk, but the approach encompasses
interactions of all sorts, ranging from informal to formal, from sociable to task-focused, and from
face-to-face to synchronous technologically mediated interactions such as telephone talk and
videoconferences. Although conversation analysis is wide-ranging in scope, the focus on the
organization of conduct within inter u:tion distinguishes this field from other forms of discourse
analysis concerned with narratives, speeches, or texts. Conversation analysis is also disting'iiished
by a methodoloby that exploits the affordances provided by recorded interaction as a form of
data.
Conversation analysis (or CA) was developed by Harvey Sacks in collaboration with Emanuel
Schegloff and Gail Jefferson. It emerged within sociology at a tim the 1960s—when that
discipline was dominated by abstract theorizing and a concern with large-scale structural phe-
nomena. Against the sociological mainstream, certain intellectual cross-currents had begun to
address the specifics of social conduct in everyday life. Erving Goffinan was exploring what he
would later call “the interaction order” (1983): the domain of direct interaction between people.
Goffman argued that this domain is a type of social institution in its own right, one that intersects
with other, more familiar societal institutions but has its own organizational principles, motiva-
tional imperatives, and norms of conduct. In a related but distinct development, Harold Garfinkel
(1967) was examining the procedures of commonsense reasoning that people use to make sense of
one another and the circumstances in which they are embedded. Garfinkel challenged the
mainstream view that social conduct is regulated by internalized norms, arguing instead that
organized conduct emerges through the use of commonsense reasoning practices. These practices
inform how actors implement norms in specific situations, and more generally how they produce
actions and render them intelligible.
CA can be understood as a partial synthesis of these ideas concerning the institution of
interaction, norms of interactional conduct, and the methods of reasoning implicated in the
production and recognition of action. The research enterprise that emerged from this synthesis has
generated a substantial and cumulative body of empirical findings. Some researchers work with
data drawn primarily from ordinary conversation and seek to describe general interactional
practices and systems of practice such as turn-taking, the sequencing of action, the repair of
misunderstandings, the relationship between vocal and nonvocal behaviors, and so on
(e.g. Atkinson and Heritage, 1984; Lemer, 2004; Schegloff, 2007). Others focus on data drawn
from institutional settings—doctors’ offices, courts oflaw, newsrooms—with the aim ofexploring
how generic practices of talk get mobilized and adapted for specific institutional tasks (Boden and
 Steven E. Clayman and Virginia Teas Gill

Zimmerman, 1991; Drew and Heritage, 1992b; Heritage and Maynard, 2006; Heritage and
Clayman, 2010) and how speaking practices affect bureaucratie and professional outcomes
(Maynard, 1984; Boyd, 1998; Clayman and Reisner, 1998; Heritage and Stivers, 1999; Gill,
2005; Stivers, 2007). Still others have addressed the relationship between interaction and racial
and gender identities (e.g. West and Zimmerman, 19b3; M. Goodwin, 1990; Kîtzinger, 2005;
Lerner and Whitehead, 2009; Speer and Stokoe, 2010); cultural difference and historical change
(Houtkoop-Steenstra, 1991; Lindström 1994; Clayman ef al., 2006); and the conduct of social
scientific inquiry itself (Maynard ef nt., 2002; Drew et al., 2006).
The productivity of CA hinges in part on its distinctive methodology, which differs from
both the ethno¡yaphic methods employed by Goldman and the demonstrations favored by
Garfinkel. The aim of this paper is to provide a brief introduction to the methods of
conversation analysis.'

Generating data: recording and transcribing

Conversation analysts work almost exclusively with naturally occurring interaction as it has been
captured in audio and videorecordings and rendered into detailed transcripts.

Natufoll/ OCCUrring Interaction

Conversation analysts avoid role-playing and experimentally induced interactions, as well as
hypothetical and invented examples. Past research has demonstrated that such data yield over-
simplified and misleading representations of interactional processes. Specimens of actual interac-
tion can generate astonishing discoveries, which, in Sacks’ (1984: 25) words, “we could not, by
ima¡;ination, assert were there.”
What constitutes “natural” interaction is, however, by no means straightforward. Because
of the “observer's paradox” (Labov, 1972), a researcher can never know whether an interaction
unfolded as it would have, had it not been externally observed (ten Have, 1999). Indeed,
the recordin@ equipment itself may become a topic of conversation for participants (ten
Have, 1999: 49).
However, such observer effects are less sipmificant than they might seem at first glance.
Sensitivity to being observed is a commonplace and “natural” feature of interaction. As
Goodwin (1981: 44) notes, “participants never behave as if they were unobserved; it is clear
that they organise their behavior in terms of the observation it will receive from their copartici-
pants.” Moreover, thèse eifects tend to be limited to the surface content of the talk, leaving its
underlying interactional structure intact. Thus, while the participants may refer to the presence of
the recording machine, they will do so via processes—ways of taking turns, building actions, and
organizing them into sequences—that are not markedly different from the rest of their talk (ten
Have, 1999). In any case, hyper-consciousness about the recording machine tends to be short-
lived, receding into the background as the participants become enmeshed in the practical concerns
of their daily tives.

A note on sampling
Unlike many fields, CA addresses a domain of phenomena whose components are not yet fully
known or understood. Sacks (1984: 21) called this domain “the methods people use in doing social
life.” Until these methods are formally described—until their identifying features are catalogued
and their local environments of occurrence are charted—it is premature to ask how prevalent they

2 121
Steven E. Clayman and Virginia Teas Gill

are within some larger population or how they are distributed in relation to exogenous
psychological or sociological variables.
Because the objective of CA is to describe the endogenous organization of interactional
phenomena rather than to determine their distribution, the issue of sampling is approached
rather differently here from other fields. Conversation analysts typically follow the “naturalist's
strategy” of gathering specimens ofphenomena from as many settings ofinteraction as possible, for
the purposes of systematic analysis and comparison (Heritage, 1988: 131; ten Have, 1999: 51).
As sources of data, not all settings are created equal. Ordinary conversation appears to
represent the richest and most varied source of interactional practices, while interactions in
bureaucratic, occupational, and other institutional contexts tend to contain a narrower range of
practices, which are specialized or adapted for those contexts (Drew and Heritage, 1992a). It is
thus important to bear in mind the social context from which data are drawn. For researchers
interested in institutional forms of talk, it is often useful to use ordinary conversation as a
comparative frame of reference (Schegloff, 1987).
While the naturalist's strategy remains primary within CA, quantitative extensions and
applicafions have become increasingly common in recent years (e.g. Clayman et al., 2006;
Heritage et al., 2007, Stivers, 2007). Although not embraced by all within the field, this is a natural
development. Once interactional practices have been thoroughly explicated, this can provide a
foundation for the development of validated measures and for analyses of frequency and
association.

Audio and video recording

The decision to study conversation was originally a practical one for Harvey Sacks, whose main
concern as a sociologist was to formally describe and analyze actual, real-time social events with a
degree of rigor (Sacks, 1984). The availability of audio recording technology in the early 1960s
made it possible to capture and preserve a particular type of social event, namely conversational
interaction. Given the centrality of interaction in the life of society, Sacks' ostensibly practical
decision turned out to be fortunate.
Audio recordings have now been augmented with video, which captures both vocal and
nonvocal behaviors. But recordings still offer the same basic service as they did for Sacks in the
1960s: access to social interaction at a level of detail that approaches what is available to the
participants themselves, and the capacity for repeated examination. The importance of recordings
in CA can be likened to that of slow-motion “instant replay” during televised sporting
events (Atkinson, 1984). While spectators in the stands may have only a vague grasp of the
fleeting events within a particular play, television viewers can—by virtue of instant replay—
achieve a more precise understanding of the specific sequence of behaviors that led to the play's
outcome.

Transcribing data
Transcripts serve both analytical and presentational functions. For the purposes of analysis, when
used in conjunction with the recording itself, a good transcript helps the researcher get a stronger
purchase on the organization of interactional practices. Transcript excerpts, together with video
“framegrabs,” also serve as a resource in publications and presentations. They enable readers to
assess independently the validity of analytic claims by reference to the key empirical instances on
which they are based.
GailJefferson developed the transcription system commonly used within CA (see Appendix).
This system balances two objectives: (1) preserving the details of talk as it was actually produced,
122
 Conversation analysis

while (2) remaining simple enough to yield transcripts that are accessible to a General audience.

123
 Steven E. Clayman and Virginia Teas Gill

Thus a full phonological system was avoided in favor of one that uses standard orthography,
supplemented with additional symbols to capture features such as overlapping speech, silences,
various forms of emphasis, and so on. Over the years, other investigators have built upon
Jefferson’s system, most notably Goodwin (1981), who developed transcription symbols to
represent nonvocal activities such as gaze and gesture.
Audio transcribing has traditionally been done with the aid of a transcribing machine, by using
a foot pedal to start, stop, and rewind a cassette tape. If the original data are on videotape, they can
be inspected later to add aspects of nonvocal behavior. More recently, technological advances
have made it possible to di tize and store data fdes on CD, DVD, or hard drive. A computer can
now serve as a transcribing machine, with software programs enabling the researcher to transcribe
in a word-processing program while simultaneously watching the video. Some programs can also
time silences. The future of data is undoubtedly digital, a medium that is more compact,
accessible, and durable than analog tapes.
The level ofdetail in a CA transcript may initially strike non-CA researchers as excessive.
However, since the objective is to understand how interactants build mutually intelligible courses of
action, any detail that is available to the interactants is potentially relevant for the researcher. for
instance,Jefferson (19b5) demonstrates the importance of seemingly trivial details surroundin g the
articulation of laughter. In the fohowing excerpt, Louise laughs during the utterance “playing with
his organ” (line 7). This transcript simply notes the laughter in line 8 rather than transcribing it
beat by beat.

1 Ken: And he came home and decided he was gonna play with
2 his orchids from then on in.
3 Roger: With his what?
4 Louise: heh heh heh heh
5 Ken: With his orchids. [He has an orchid-
6 Roger: [Oh heh hehheh
7 Louise: —+ Playing with his organ yeah I thought the same thing!
b ((spoken through laughter))
9 Ken: Because he's got a great big [glass house-
10 Roger: [I can see him playing with
11 his organ ((laughing)).
Jefferson, 1985: 28, simplified transcript)

Such simplification obscures the way Louise employs laughter as a resource. In the more detailed
transcript below, it becomes apparent that Louise precisely places her laughter in the key phrase
“PLAYN(h)W(h)IZ O(h)R’N” (line S), stopping abruptly when she moves on to the next
utterance (“ya:h I thought the same”). Roger subsequently laughs in a strikingly similar way (line
14).
(2)
1 Ken: An'e came hom'n decided'e wz gonna play with iz o:rchids.
2 from then on i:n.
3 Roger: With iz what?
4 Louise: mh hih hih [huh
5 Ken: [With iz orchids.=
6 Ken: =Ee[z got an orch[id-
7 Roger: [Oh:. [hehh[hah.he:h ].heh
b Louise: —+ [heh huh. PLAYN(h)W(h)IZ O(h)R’N
124
 Conversation analysis

9 ya:h I[thought the[same

125
Steven E. Clayman and Virginia Teas Gill

10 Roger: [uh:: [.hunhh.hh.hh

11 Ken: [Cz eez gotta great big [gla:ss house]=
12 Roger: [I c'n s(h)ee ]=
13 Ken: '[( )
14 Roger: —+ =[inn p1(h)ay with iz o(h)r(h)g.(h)n.uh.
Qe erson, 1985: 29, Detailed Transcript)
Deployed in this way, laughter displays recognition of an alternate “obscene" hearing of the
phrase “playing with his orchids,” even as it partially obscures its articulation.
Accordingly, researchers should strive to preserve as much detail as possible. At the same time,
because transcribing is labor-intensive and time-consuming, the amount of time invested in a
transcript will inevitably vary according to the size of the dataset and the interests of the
researcher. One practical strategy is to transcribe in varying amounts of detail, reserving the
highest level of detail for segments that will receive the greatest analytic attention.

Analyzing data

Setting started
Once data have been gathered and prepared, how should analysis begin? Since interaction is
largely uncharted territory whose topography is only partially understood, CA seeks to map this
topography by examining specimens of its contours and analyzing how they were systematically
produced. This type of investigation requires holding in abeyance questions about why a social
activity is organized in a particular way, focusing instead on ir/tot is being done and on how it is
accomplished. Interactional activities can be investigated at different levels of granularity. There
are overarching activit y frameworks that organize extended interactional episodes, such as
“getting acquainted" or “talking about personal problems” or “seeing a doctor for medical help”
or “cross-examining a witness." One step below this is represented by discrete sequences ofaction,
which may be analyzed for their relatively generic sequential properties (e.g. as paired actions,
story-telling sequences, etc.) or for type-specific characteristics (e.g. as question—answer
sequences, invitation sequences, news delivery sequences, etc.). Next come the actions that
comprise sequences, actions commonly accomplished through a single turn at talk such as
questioning, requesting, announcing news, responding to these various actions, and so on. Finally
there are JeniiireS mobilised irif/tirt turns at
talk, such as lexical choices, intonation contours, nonvocal behaviors, etc.
As should be apparent from the preceding list, virtually everything that happens in interaction
is fair game for analysis. While one might be tempted to dismiss familiar details of conduct as
random noise or insignificant “manners of speaking," conversation analysts proceed from the
assumption that all elements of interaction are potentially orderly and socially meaningful for the
participants (Sacks, 1984). This attitude opens up a wealth of possibilities for analysis. But where
to begin? Drawing on Schegloff (1996: 172) we suggest two pathways into the data.

Begin with a "noticing"

One can begin with relatively unmotivated observation. The analyst simply notices how an
interactant says or does something at a given juncture, a bit of conduct that seems in some way
“interesting." Of course, purely unmotivated observation is an unattainable ideal, as experienced
analysts have an established conceptualfoundation, grounded in previous research, which affect
what they are inclined to notice and what strikes them as interesting. Nevertheless, it is possible to
approach data without a specific abenda in mind, thereby remaining open to the prospect of
126
 Conversation analysis

discovery. Having noticed a

127
 Steven E. Clayman and Virginia Teas Gill

given practice, the analyst can then explicate what it might be “doing” or accomplishing. This
typically involves exarriining where it is placed in the stream of interaction and how it operates
within that local context, focusing on the actions that immediately preceded it and the responses it
attracts.
For example, Sacks ([1966] (1992): 256—257) observed that, when children speak to adults,
they commonly begin by asking a question such as “You know what, Daddy?” Anyone who has
been around children for any length of time will be familiar with this recurrent feature of
children's talk. What is accomplished with this practice? One clue can be gleaned from the
response it elicits. Adults typically respond to the “You know what” question with another
question—namely “What?” This type of response not only invites the child to speak again and say
what motivated the original question, but by so doing it simultaneously aligns the adult as one
who is prepared to listen to the ensuing talk. Thus the child's original query sets in motion a chain
of events that culminates in the child gaining a ratified speaking “slot” and an attentive recipient.
When children use this practice, they may be addressing certain basic interactional problems, such
as their diminished rights to talk and adults' preoccupation with other matters.
Heritage (1998) took a similar tack when analyzing a particular way of designing answers to
questions. Heritage initially observed that some answen to questions are prefaced with “oh,” as in
line 6 of the following example, taken from a radio interview with Sir Harold Acton, a noted
British aesthete.

1 Act:.............hhhh and some of thuh- (0.3) some of my students

2 trans@ted Eliot nto Chip::se. I think thuh very
3 first.
4 (0.2)
5 Har: Did you learn to speak (.) Chine[:se.
6 Act: —› [.hh Oh yes.
7 (0.7)
8 Act: .hhhh You ::n’t live in thuh country without speaking
9 thuh 1ang[uage it's impossible .hhhhh=
10 Har: [Not no: cour:se.
(Heritage, 1998: 294)
This practice, far from being random or insignificant, turns out to be socially meaningfiil and
consequential. By prefacing an answer with “oh,” the answerer implies that the prior question
“came from leG field” and is thus of questionable relevance. In this particular case, the ongoing
discussion concerns Acton's experience teaching modem poetry at Beijing University, and it is in
the context of this discussion that he is asked (at line 5) if he learned to speak Chinese. He treats
the answer to the question as obvious or self-evident. He expresses this explicitly at lines 8—9, but
he also does so implicitly in his initial response to the question (line 6) via the oh-prefaced
affirmative answer. With this pathway into the data, an initial noticing is “pursued by asking what
—if anything— such a practice of talking has as its outcome” (Schegloff, 1996: 172). Not every
observed practice will turn out to have a systematic import, but many core findings of CA
have their origins in
unmotivated noticings ofjust this sort.

Begin with a vernacular action

Another pathway into the data is to focus on a particular type of action that is already part of
the vernacular cultur asking questions, wing advice, delivering news, and so on. Here the
128
 Conversation analysis

analytic challenge is to transcend what is already intuitively known about the action in question.
This can

129
Steven E. Clayman and Virginia Teas Gill

be accomplished by exploring specific ways that a given action is designed and implemented,
focusing again on the sequential environments in which speakers deploy each form and on the
responses they receive.
For example, using announcements of news as a starting point, Maynard (2003) has uncovered
a range of practices associated specifically with the telling ofbad news in both everyday and
clinical settings, while also demonstrating that they operate as solutions to specific problems
associated with this difficult interpersonal task. One set of practices involves forecasting the news
in advance of its delivery. Maynard demonstrates that forecasting, by providing some advance
warning of what is to come, maximizes the likelihood that recipients will be prepared to register
and accept the news. In a similar vein, studies have explored various methods for designing
requests (Curl and Drew, 2008), presenting medical symptoms (Halkowski, 2006), and offering
explanations for illness to doctors (Gill, 1998; Gill and Maynard, 2006). In each case, the analyst
explores how participants deploy and respond to familiar actions and their varying forms.

Crounding an analysis
Once a possible phenomenon has been located, how should analysis proceed? In the broad
tradition of interpretive social science, CA seeks analyses that are grounded in the understandings
and orientations of the participants themselves. Within interaction, the understandings that matter
most are those that participants display, act on, and thus render consequential for the interaction's
subsequent development (Schegloff and Sacks, 1973).

The response as an analytic resource

One central resource for tapping into such understandings is embodied in how recipients respond
to the phenomenon in question. Consider that interaction ordinarily unfolds as a series of
contributions or “moves,” each move being normally addressed to, and to some extent condi-
tioned by, the move that preceded it. Given this, each move will normally display that speaker's
understanding of what came before (Sacks ct al., 1974). Interactants themselves rely on such
retrospective displays of understanding to ascertain whether and how they were understood, and
this “architecture for intersubjectivity” (Heritage, l9b4b) is also a resource for analysts.
To illustrate, consider the utterance: “Somebody just vandalized my car.” As Whalen and
Zimmerman (1987) have observed, while the lexical meaning of this utterance is transparent, the
type ofaction that it implement what it is “doing” from the standpoint of the participants—cannot
be determined by considering the utterance in isolation. It could be a straightforward announcement
o its, with no agenda other than that of conveying information to an uninformed recipient. If this
were the case, one would expect it to generate an iriitial response along the lines of “Oh” or “Oh
really” or “My goodness”—that is, a response that attends to it as new and perhaps surprising
information (Jefferson and Lee, 1981; Heritage, 1984a). Alternatively, this item of news could be
a vehicle for requesting help or assistance of some sort, in which case one would expect a response
that either accepts or rejects the request, or at least proceeds in that direction. In actuality, the
utterance was produced by a caller to a 911 emergency service, and it was responded to as follows
(arrowed below).
(4)
1 Dispatcher: Midcity Emergency
2 Caller: Um yeah (.) somebody jus’ vandalized my car,
3 Dispatcher: —+ What's your address.
4 Caller: Sixteen seventy Redland Road.
(VVhalen and Zimmerman, 1987: 1 74a
130
 Conversation analysis

Notice that the dispatcher's response in line 3—a question about the caller's address—is a purely
instrumental query and a necessary prerequisite for sending assistance. It clearly treats the prior
utterance as a request for help rather than a news announcement, an interpretation that is routine
in the institutional environment of a 911 helpline (Whalen and Zimmerman, 1987). The general
point is that recipients' own understandings are displayed publicly in their responses, and are thus
available as an analytic resource.
Responses can also be informative in more subtle ways. Beyond revealing participant under-
standings of the type of action produced previously, they can also shed light on its valence. For
instance, news announcements may be understood by recipients as either good or bad, and this too
is displayed though subsequent talk (Maynard, 1997). Thus the following birth announcement is
receipted not only as news (“Oh”), but specifically as good news (“how lovely”).
(5)
1 Carrie: I: thought you'd l_ike to know I've get a little gran'daughter
2 Leslie: —› thlk Oh: how lovely.
(Maynard, 1997: 111)

In other cases the valence of a given news announcement may be unclear to the recipient,
resulting in a more cautious mode of receipt. Contrast the birth announcement sequence
above with a similar announcement in the next example. This time the announcement
generates an initial response (“Oh my goodness” at line 2) that registers it as surprising, but
specifically avoids evaluating the news in an explicit way.

1 Audi: hhhh! Bob and I are going to have a baby.

2 Betty: —+ Oh my goodness hhow- (1.0)
3 did you have a reversal- he have a reversal?
4 Audi: Yea:h.

5 Audi: It was [very succe ful,][very quickly] hh::h.hhh

6 Betty: —› [OH I'M SO ] HAPPY. ]
(Maynard, 1997: 116, simplified)

In this case the announcement is being issued by the expectant mother (Audi) whose husband
(Bob) had previously undergone a vasectomy, raising the spectre of an unplanned pregnancy.
Moreover, the recipient of the news (Betty) is aware of this fact, as evidenced by her subsequent
query about a reversal (line 3). Only when subsequent talk reveals that the husband's vasectomy
had indeed been reversed and that the pregnancy was fully planned does Betty receipt it
unequivocally as good news (“Oh I'm so happy,” line 6).
At a still more subtle level, responses can shed light on the import of momentary silences
in interacfion (Davidson, 1984; Pomerantz, 1984). In the next example, C invites B and a third
party to stay with him at the beach (line 1). This invitation makes relevant a response that either
accepts or declines the invitation, but what iriitially follows is silence (line 2). A silence at this
juncture is ordinarily understood as “belonging” to the recipient of the invitation (Sacks ct al.,
1974) and it could, in principle, arise for a number of reasons. B may have a problem hearing or
understanding the invitation, or B may have heard/understood the invitation but is having some
problem with accepting it. The difficulty, in short, could be either in the intelligibility or in the
acceptability of the invitation.
131
Steven E. Clayman and Virginia Teas Gill

(7)
1 C: Well you can both Sta:y.
2 (0.4)
3 C: —› [Got plenty a’ roo:m, hh[hh
4 B: h I- [Oh(h)o(h)o,
5 (.)
6 B: Please don't tempt me,
(Davidson, 1984: 105, simplified)
C's response to the silence (line 3) clearly treats it as indicating the latter type of problem. Instead
of repeating or reformulating the invitation—the usual way of handling a problem of intelligibility
(Scheglolf ct al., 1977)—C offers an argument for accepting it. This move presupposes the
invitation's intelligibility and displays C's understanding that B is reluctant to accept. Moreover,
the substance of C's argument displays his inference regarding the reason for B's reluctance
(concern about insufficient room and the inconvenience this might entail)—a reason that he
counters in an effort to nudge her toward an acceptance.
At varying levels of detail, then, successive contributions to interaction shed light on how the
participants understand preceding events. Of course, it is possible for a respondent to misunder-
stand what a speaker originally intended, and such misunderstandings may come to light through
subsequent repair efforts (Scheglolf, 1992). More often, subsequent talk implicitly confirms
previously displayed understandings. In any event, the sequential organization of interaction
provides a running index of the participants' own mutual understandings and is thus a key
methodological resource.

Deployment as an analytic resource

The response to an utterance is an extremely usefiil resource, particularly when analyzing
utterances that initiate sequences (e.g. news announcements, requests, invitations). However, it is
not always a sufficient basis upon which to build an analysis. Responses may be less than
transparent, and at times designedly opaque in the understandings they exhibit. Fortunately, other
analytic resources are available that center not on the reâpient but on the producer of the talk in
question. Examining in detail how speakers recurrently deploy a given practic in particular
sequential environments and in particular positions within the speaker's own turn, and in
conjunction with other practices— can provide important clues about that practice's meaning and
import.
To illustrate, consider the various bits Of talk that are used to receipt prior talk—items such as
mm hm, yeah, oh, and okay. These were long assumed to comprise an undifferentiated set of
“acknowl- edgment tokens” or “backchannel” communications. However, it has been
demonstrated, largely on the basis of the selective manner in which these tokens are deployed, that
each performs a somewhat disfinct interactional function (Heritage, 1984a; Jefferson, 19b4;
Beach, 1993). The contrast between rum /trn and yeah provides a usefiil case in point Jefferson,
19h4). In the following excerpt, notice how B deploys these receipt tokens (arrowed) in the course
of M's extended telling.
(8)
1 M: and she's been very thri_fty.
2 B: —+ Mm hm,
3 M: .hhhhh So: (.) I said it- it a:dds up to one thing
4 money somepla:ce
5 B: —+ M hm,
132
 Conversation analysis

6 M: .hhhh=

133
 Steven E. Clayman and Virginia Teas Gill

7 B: —+ Mm [hm,
b M: [But ish (.) she tn- transacts all h business in
9 Lo:s Angeles you know and people 1i_ke this are so secretive
10 it's a(m) really it's almost a mental state
11 B: —› Yeah .hh Well .hh uh:m (0.9) y- there's mething wrong too
12 if she doesn't pay her bills .. ..
(fefferson, 1984: 205)
Although B uses both forms of receipt, she deploys them in different ways. One point of
difference is the prior sequential environment: the mm lms tokens (lines 2, 5, and 7) appear in the
midst of M's extended telling as it unfolds, while the yeah token (line 11) appears at the
completion of the telling. Correspondingly, there are differences in what B does next. Each
rust Atm token stands alone within B's turn at talk, while B follows the yeah token with further
talk that responds more substantially to M's telling. Accordingly, these tokens embody different
interactional stances, mm /im displaying “passive recipiency” and yen/t displaying “incipient
speakership” (|efferson, 1984). This conclusion is based on the systematic manner in which
they are deployed.
The dissnct functions of such tokens are further revealed when the tokens are used in sequen-
tially incongruous ways. Thus, when speaker G finishes an extended telling and explicitly marks it
as complete (“So that's the story,” line 10, below), B receipts the story with “Mm hm” (line 11).

(9)
1 G: I'd 1i:ke to have the mirrors. But if she wants them? (.)
2 .hh why that's: I-th-tha:t’s fi::ne.
3 B: Mm hm,
4 G: If she's going to use them you kno:w.
5 B: Mm [hm,
6 G: [.hhhhhh I'm not going to uh,hh maybe queer the dea:1
7 just by wanting this that and the othe[r (you know),
b B: [NO:.
9 (0.2)
10 G: .hhhh s:So: uhm,h (.) tha:t’s the story.
11 B: Mm hm,
12 (0.2)
13 G: An:d uh (0.6) uhm,hhh (1.0) .hhhh u-Then I have a ma:n
14 coming Tue:sday...
(feffer.‹on, 1984: 209)

This display of passive recipiency appears strikingly misfitted to such an obvious story
completion. And yet it seems to have been produced and understood as embodyingjust such a
passive stance subsequently, B falls silent and offers no further talk dine 12), whereas G searches
for and eventually finds something further to say (lines 13—14). Here, then, an interactant exploits
the passivity of “rum /tm” as a resource for resisting the speakership role, which in turn prompts
the prior speaker to continue.
The analytic resources sketched here are based on the insight that the import of a given
practice is observable in the manner in which it is deployed and responded to. By exploiting
these resources, the researcher moves beyond speculation to generate analytic claims that are
grounded in the displayed understandings and orientations of the interactional participants
themselves.

134
 Conversation analysis

Working through ColleCtions

The primary objective of CA is to elucidate the methods people use to build interaction together.
Although analysis often begins by examining a single fragment of talk, this is normally the first
step in a deeper analysis, which transcends that particular fragment and sheds light on practices
that operate across a range of participants and social contexts. As Sacks has observed:
Thus it is not any particular conversation, as an object, that we are primarily interested in.
Our aim is to get into a position to transform . .. our view of “what happened,” from a matter
of a particular interaction done by particular people, to a matter of interactions as products of
a machinery. We are trying to find the machinery. In order to do so we have to get access to
its products.
(S«:ks, 1984: 26—27)
This requires the systematic analysis of numerous cases. Working with cohections can flesh out
and enrich an analysis initially arrived at through a single case, illuminating such matters as the
practice's various forms, the boundaries that separate it from related practices, and its scope and
normativity.
When building a collection of candidate instances ofa given phenomenon, it is useful to begin
by casting a wide net. One should include what appear to be clear cases of the phenomenon in
question, cases in which the phenomenon is present in an atypical form, and also what appear to
be outright negative or “deviant” cases. Analyzing such cases rather than dismissing them as
random error almost always yields a richer and more powerful analysis.
Once a collection is assembled, analysis proceeds on a case-by-case basis, with the aim of
developing a comprehensive account that encompasses all relevant instances in the collection. The
process is roughly analogous to analytic induction (Katz, 1983), although in CA the objective is
not causal explanation but an analysis that will encompass a practice's varying occurrences across
a range of interactional contexts and exigencies.
Central to this process is the analysis of problematic or deviant cases. Some such cases are
shown, upon analysis, to result from interactants’ orientation to the same considerations that
produce the “regular” cases. We've already seen an illustration of this in the discussion ofexcerpt 9
above, in which an rntn firn token was placed in an unusual sequential environment, but was
nonetheless shown to function much like other such tokens as a display of passive recipiency.
Cases of this sort are, in elTect, exceptions that prove the rule.
In other instances, deviant cases can prompt the researcher to revise the initial analysis in favor
of a more general formulation, one that encompasses both the regular cases and the anomalous
departure. Perhaps the clearest example of this process can be found in ScheglofFs (1968) analysis
of telephone call openings. In a corpus of 500 telephone calls, Scheglofffound that a straightfor-
ward ml “answerer speaks first” —adequately described all but one of the call openings. In that
one unusual case, the caller speaks first (line 3):

(10)
i ((^•a))
2 ((receiver is lifted, and there is a one-second silence))
3 Caller: Hello.
4 Answerer: American Red Cross.
S Caller: Hello, this is police headquarters. ...
(Scheglojf, 1968: 1079)

Rather than ignoring this instance or explaining it away in an ad hoc fashion, Schegloffreturned to
the drawing board and developed a more general analysis, which accounted for all 500 cases and
135
 Steven E. Clayman and Virginia Teas Gill

revealed the organization of(what would later be termed) adjacency pairs (Schegioffand Sacks,
1973). Schegloff realized that the ringing of the telephone launches a special kind of adjacency
pair sequence, namely a summons—answer sequence. The rule “answerer speaks first” actually
reflects the more general principle that, once a summons (here, a ringing phone) has been issued,
an appropriate response is due. The deviant case also can be explained by reference to summons—
answer sequences. The ring dine 1 above) was followed by silence dine 2), during which the caller
heard the relevant response to be absent. Caller then spoke first (line 3) as a way of renewing the
summons, soliciting the missing response, and thereby completing the incomplete sequence. The
end result is a more analytically powerful account, which encompasses both re$nilar and atypical
cases.
Finally, some deviant cases may, upon analysis, turn out to fall beyond the parameters of the
phenomenon being investigated. Such cases are not genuinely “deviant” at all, and clarifying how
this is so furthers understanding of the core phenomenon and its boundaries. For instance,
consider how personal troubles are discussed in conversation ([efferson and Lee, 1981; Jefferson,
1988). When speakers disclose their troubles, recipients commonly respond with alfiliative
displays of understanding. However, in contrast to this typical pattern, recipients may instead offer
advice and thereby transform the situation from a “troubles-telling” to a “service encounter.” This
line of analysis, unlike the previous one, does not result in a single analytic formulation
encompassing “regular” and “deviant” cases. Rather it recognizes differences between cases and
the phenomena they instantiate.

Discussion
CA addresses a domain of phenomena, the endogenous organization of talk-in-interaction, in a
manner that has proven to be both illuminating and productive. Much has been learned about the
basic objects that comprise this domain.
Progress on this front has made it possible for researchers to use CA methods and findings
to address questions extending beyond the orpariization of interaction per se, including
questions about how this domain intersects with, and can illuminate, other aspects of the social
world. As we noted at the beginning of this chapter, some researchers have examined the impact
of interactional practices on bureaucratic and professional decision-making in medical, legal,
educational, journal- istic, and other contexts. Others have done comparative analyses of
interactional practices to elucidate large-scale cultural differences and processes of historical
change. Some of this work involves formal quantification, correlating interactional practices with
other variables of interest. The utility of CA in this context is that it identifies previously unknown
practices, establishes and validates their meanin g and import, and thus provides a solid foundation
for analyses of frequency and association.
As progress is made in these various applied areas, it is important to bear in mind that such
work would not be possible without the basic research on which it rests. Talk in interaction
remains a rich and compelling phenomenon in its own right, one in which human agency is
exercised, intersubjectivity is achieved, and contexts of the social world are brought to life.
Notwithstanding what has already been accomplished, much remains to be discovered about how
human interaction actually works.

Transcription conventions
[]
Square brackets show beginning and ending of overlapping talk
(0.5)
Numbers in parentheses are silences timed to tenths of a second
(.)
Period in parentheses is a very brief silence (less than .1 sec.)
3 30
 Conversation analysis

((quiet))
Transcribers' comments are enclosed in double parentheses
Empty parentheses denote indecipherable utterance
Text within parentheses is transcriber's “best guess” as to a speaker's utterance
Period indicates downward intonation, not necessarily the end of a sentence
?
Question mark indicates upward intonation, not necessarily a question
Commas indicate slightly rising or “continuing” intonation
Colon(s) indicate that a sound is stretched. The more colons, the longer the sound
.hh
h's with precedirig period indicate audible inbreath; the more h's, the longer the
hh
inbreath h's with no preceding period indicate audible outbreath; the more h's, the
longer the outbreath
(h)
Parenthesized “h” indicates plosiveness, often associated with laughter, crying,
breathlessness, etc.
>word<
WoRD Enclosed talk is spoken more quickly than surrounding talk
Upper case indicates greater loudness than surrounding talk
°Yes
Words inside degree signs are spoken softly or whispered
°
very
Underlines indicate sounds that are stressed
Yes:: Colons indicate stretching of the preceding sound
n- Dash indicates a cut-off of the preceding sound
Equal sign indicates utterances before and after have no intervening silence.
(Adapted from]efferSoH, 1974)

Further reading
ten Have, P. (1999) Doing Conversation Analysis: A Prmtical Guide. London: Sage.
Provides a comprehensive discussion of the methodology of CA.
Heritage, J. (1984b) Gn abet rind Ethnomethodology. Cambridge: Polity Press.
Surveys the theoretical background to CA in the work of Harold Garfinkel and provides a useful overview
of somc of the main areas of research.
Schegloff, E. A. (2007) Sequence Organization in Intention. Cambridge: Cambridge University
Press. Offers a focused analysis of a central feature of interactional organization.
John Heritage and Steven Clayman's Talk in Action (2010).
Surveys research on interaction in a variety of institutional settings.

Note
1 For a much more elaborated discussion of CA methods, see ten Have (1999).

References
Atkinson, ]. M. (19H4) Our Masters’ Voices: The Language and Body Language of Politic:s. London:
Methuen.
Atkinson, J. M. and Heritage, J. (eds.) (1954) Stmctures oJ Social Action: Studies in Conversation Analysis.
Cambridge: Cambridge University Press.
Beach, W. A. (1993) ‘Transitional regularities for casual “okay” usages’, Journal of Pragmatics, 19:
325—352.
Boden, D. and Zimmerman, D. H. (ed.) (1991) Talk and Social Stmcture. Cambridge: Polity Press.
Boyd, E. (199*) ‘Bureaucratic authority in the “company of equals”: initiating discussion during medical peer
review’, American Sociological Review, 63: 200—224.
Clayman, S. E., Elliott, M. N., Heritage, J., and McDonald, L. (2006) ‘Historical trends in questioning
presidents 1953-2000’, Presidential Studies Qtiotterfy, 36: 561-583.
 Steven E. Clayman and Virginia Teas Gill

Clayman, S. E. and Reisner, A. (1998) ‘Gatekeeping in action: editorial conferences and assessments of
newsworthiness’, American Socialagical Review', 63: 178—199.
Curl, T. S. and Drew, P. (2008) ‘Contingency and action: a comparison of two forms of requesting’, Research
on Language and Social Interaction, 41 (2): 129-153.
Davidson, J. (19*4) ‘Subsequent versions oFinvitations, offers, requests, and proposals dealing with potential
or actual rejection’, in J. M. Atkinson and J. Heritage (ed.) Stmctures of Social Action. Cambridge:
Cambridge University Press, pp. 102—12S.
Drew, P. and Heritage, J. (1992a) ‘Analyzing talk at work: an introduction’, in P. Drew andJ. Heritage (eds.)
Tally at Work. Cambridge: Cambridgc University Press, pp. 3—65.
Drew, P. and Heritage, J. (eds.) (1992b) Talk at Work. Cambridge: Cambridse University Press.
Drew, P., Raymond, G., and Weinberg, D. (2006) Talk and Interaction in Social Research Methods. London:
Sage. Garfinkel, H. (1967) Studies in Ethnomethodology. Englewood CliRs, NJ: Prentice-Hall.
Gill, V. T. (1998) ‘Doing attributions in medical interaction: patients’ explanations for illness and doctors’
responses’, Sacial Psycholagy Quarterly, 61 (4): 342—360.
Gill, V. T. (2005) ‘Patient “demand” for medical interventions: exerting pressure for an offer in a primary care
clinic visit’, Research on Language and Social Interaction, 3S (4): 451-479.
Gill, V. T. and Maynard, D. W. (2006) ‘Patients’ explanations for health problems and physicians’ respon-
siveness in the medical interview’, in J. Heritage and D. W. Maynard (eds.) Communication in Medical
Care: Interaction Between Primary Care Physicians and Patients. Cambridge: Cambridge University Press,
pp. 115—150.
Goffman, E. (19*3) ‘The interaction order’, American Sociological Review, 4*: 1-17.
Goodwin, C. (19h1) Conversational Organization: Interaction Between Spealeers and Hearers. New York:
Academic Press.
Goodwin, M. (1990) He-Said-She-Said: Talk as Social Organization among Black Children. Bloomington, IN:
Indiana University Press.
Halkowski, T. (2006) ‘Realizing the illness: patients’ reports of symptom discovery in primary care visits’, in
J. Heritage and D. W. Maynard (eds.) Communication in Medical Care: Intention between Primary Care
Physicians end Patients. Cambridge: Cambridge University Press, pp. S6—114.
Heritage, J. (1954a) ‘A change-of-state token and aspects of its sequential placement’, in J. M. Atkinson and
J. Heritage (ed.) Structures of Social Action. Cambridge: Cambridge University Press, pp. 299-345.
Hcritage, J. (1954b) Garfinkel and Ethnomethodology. Cambridge: Polity Press.
Heritage, J. (1988) ‘Explanations as accounts: a conversation analytic perspective’, in C. Antaki (ed.)
Analyzing Everyday Explanation: A Casebook of Methods. London: Sage, pp. 127—44.
Heritage, J. (199*) ‘Oh-prefaced responses to inquiry’, Language in Society, 27 (3): 291-334.
Heritage, J. and Clayman, S. (2010) Talk in Action: InteractiOHS, Identities, and Institutions. Oxford: Wiley-
Blackwell.
Heritage, J. and Maynard, D. W. (eds.) (2006) Cammunication in Medical Care: Interaction between Pñmary
Care Physicians and Patients. Cambridge: Cambridge University Press.
Heritage, J., Robinson, J. D., Elliott, M. N., Beckett, M., and Wilkes, M. (2007). ‘Reducing patients’ unmet
concerns in primary care: The difference one word can make’, Journal of General Internal Medicine, 22
(10): 1429—1433.
Heritage, J. and Stivers, T. (1999) ‘Online commentary in acute medical visits: a method of shaping patient
expectations’, Social Science and Medicine, 49 (11): 1501—1517.
Houtkoop-Steenstra, H. (1991) ‘Opening sequences in Dutch telephone conversations’, in D. Boden
and D. H. Zimmerman (eds.) deft and Social Stricture. Berkeley, CA: University of California Press,
pp. 232-250.
Jefferson, G. (1974) ‘Error correction as an interactional resource’, Longuuge itt Society, 2: 1b1-199.
Jefferson, G. (1964) ‘Notes on a systematic deployment of the acknowledgement tokens “Yeah” and “Mm
hm” ’, Paper5 in LinguistitS, 17: 197-216.
Jefferson, G. (1955) ‘An exercise in the transcription and analysis of laughter’, in T. A. Dijk (ed.) Handbook of
Discourse Analysis, Volume 3. New York: Academic Press, pp. 25-34.
Jefferson, G. (1988) ‘On the sequential organization of troubles-talk in ordinary conversation’, Social Problems,
35 (4): 41h—441.
Jefferson, G. and Lee, J. R. E. (1981) ‘The rejection of advice: managing the problematic convergence of a
“troubles-telling” and a “service encountcr” ’,]oumal of Pragmatics, 5: 399—422.
Katz, J. (1983) ‘A theory of qualitative methodology: the social system of analytic fieldwork’, in
R. M. Emerson (ed.) Contemporary Field ReSearch. Boston, MA: Little Brown, pp. 127—14h.

133
132