Conversation Analysis 153

Conversation Analysis
H Mazeland, University of Groningen, Groningen, with the original recording, the transcription enables
The Netherlands researchers to examine the forms of language use that
ß 2006 Elsevier Ltd. All rights reserved. were available to the participants in the recorded
interaction itself.
A CA transcription is still readable without consid-
Introduction erable expert knowledge. The transcript does not
What kinds of social organizations are used as represent speech production at the level of its me-
resources when people communicate through talk in chanical reproducibility (the etic approach that is
interaction? It is this question that conversation anal- typical of phonetics). Rather, the transcription pro-
ysis attempts to answer. Conversation analysis (CA) vides an empirically reliable approximation of the
studies the methods participants orient to when they interpretative assemblies that participants in talk are
organize social action through talk. It investigates working with (the emic approach). A transcription is
rules and practices from an interactional perspective the combined result of carefully listening to how and
and studies them by examining recordings of real-life where utterances are produced and the interpretative
interactions. work of the transcriber as a competent member of the
Although conversation analytic research may be culture under investigation.
subsumed in typically linguistic disciplines such as Extract (1) exemplifies this way of transcribing
pragmatics, discourse analysis, or (interactional) so- talk. It documents a short episode – just 7 seconds –
ciolinguistics, it started in American sociology. In from a Dutch telephone conversation (the original
particular, the sociologists Erving Goffman (see text is followed by an English translation in italics):
Goffman, Erving (1922–1982)) and Harold Garfinkel
prepared the ground in which CA arose – Goffman Extract (1). Telephone call between brothers. Back-
with his study of cultural rules and rituals in face-to- ground information: Jan is calling his brother Ton
face interaction (Drew and Wootton, 1988), and Gar- from their Rhine barge. Their parents sail a barge as
finkel with his investigations into the situated and well. Ton is in the office of the shipping exchange.
normative character of shared understanding in ev- Caller Jan is inquiring whether their mother is also at
eryday courses of action (Heritage, 1984). Enabled by the shipping exchange.
the spread of recording techniques that opened new
54 Jan: mamah, (0.2) is die d’r ook?
ways of inspecting interactional data, Harvey Sacks mama (.) is she there too?
(y1975) and Emanuel Schegloff established a novel 55 0.4
paradigm for researching the organization of human 56 Ton: HÈ":h?
action in and through talk in interaction (Schegloff, huh?
1968; Schegloff and Sacks, 1973; Sacks, 1992) (see 57 (.)
Telephone Talk; Sacks, Harvey (1935–1975)). Al- 58 Jan: mama:h
though the foundational work in CA focuses on talk mama
in conversations, the framework has gradually been 59 0.6
extended to research of other types of talk such as 60 Ton: waa:r
where
medical and clinical interaction, lessons, or news
61 0.3
interviews. This is why the more general characteri-
62 Jan: is die daa:r?
zation ‘talk in interaction’ nowadays is often pre- is she there?
ferred over ‘conversation.’ 63 0.3
64 Ton: "hie":r?
Studying Transcriptions of Recorded Talk here?
65 (.)
In conversation analysis, the investigation begins 66 Jan: j [ah
with making an audio and/or video recording of nat- yes
urally occurring talk. These recordings are carefully 67 Ton: [nEEj!
transcribed according to specific conventions first de- nO!
veloped by Gail Jefferson (see Jefferson, 2004). The 68 (.)

69 Jan: oh:.
CA transcription notation is designed for rendering
70 0.6
details that contribute to the organization and intelli-
gibility of talk. It helps to retain features of prosody In order to be able to read a transcription like this
and turn positioning in the transcription. Together one, the reader has to know the conventions. Notice
154 Conversation Analysis

first that each speaker contribution – or turn – has a organization of turn taking. Second, the talk in this
separate line. This indicates the turn’s chronological fragment is an interactionally coherent exchange of
position relative to its predecessor and its successor. communicative actions. The episode starts with a
Other notation conventions include: question and it ends with the answer to that question
(see lines 54 and 67, respectively). How are we able to
. 0.6 The length of silences between and
recognize this kind of interactional order in a series of
within turns is measured in tenths of
utterances? How is it achieved? This question is an-
seconds.
swered at the level of sequence organization, that is,
. (.) A dot between brackets (.) indicates a
the way in which participants coordinate actions in
short silence of less than 0.2 seconds.
series of turns in order to effectuate interactional
. [ In the case of simultaneous talk, the
projects.
onset of the overlapping turn is located
by a left square bracket in the over-
lapped turn. Turn Design and the Organization of
. .,? A period indicates a falling final pitch Turn Taking
contour, a comma a slightly rising
A remarkable feature of the interaction in extract (1)
pitch contour, and a question mark a
is that speaker change is coordinated smoothly. Both
strongly rising one.
interruptions (or other kinds of simultaneous talk)
. # " Vertical arrows provide information
and gaps are relatively rare. In a seminal paper first
about local pitch movements within
published in 1974, Sacks, Schegloff and Jefferson
syllables or at the level of a single syl-
formulate these observations in more technical, orga-
lable. A downward arrow signals a
nizational terms; the participants display an orienta-
falling tone movement, an upward
tion to minimization of overlap, while at the same
arrow a rising one.
time, they also orient to minimization of gap. Sacks
. word Underlining signals salient stress.
and his colleagues account for this fine organizational
. wor:d A colon renders a noticeable sound
balance by a description of the systematics that
stretch.
conversationalists orient to when they coordinate
. sto- The hyphen is used as a cut-off marker.
the organization of turn taking.
. >faster< This utterance part is produced with
The basic organizational problem that participants
higher pace than the talk surrounding
have to solve each turn anew is to determine when the
it.
speaker will complete the current turn. The recipient
. <slower> The pace is relatively slower.
is not only figuring out what the turn is about and
. LOud Capitals indicate relative loudness.
what the speaker is doing with it, he also has to be
.  soft The degree sign signals that an utter-
alert for the moment it might become his turn to
ance part is produced more softly than
speak. Recipients anticipate such organizationally re-
the surrounding talk.
levant moments by building expectations as to what
. h Hearable inbreath.
the utterance underway is going to look like. Turns
. hh Hearable aspiration.
are produced linearly in real time, but in the course
. (guess) The transcriber is uncertain about the
of a turn’s production, a recipient can make an in-
utterance part between parentheses.
formed guess about the structure of the whole unit by
Even the transcription of a brief episode such as inspecting – in its environment of use – the part that is
the one documented in extract (1) already displays already there. The turn so far provides cues as to how
very basic features of talk in interaction. Note, for a the unit underway is constructed and when it will
start, how short turns may be. If the reader was possibly be complete.
expecting utterances in turns at talk to consist of The building stones of turns are turn constructional
complete, well-structured sentences, he will be sur- units (TCUs). Each turn is built with at least one turn
prised to find out how little the participants need to constructional unit. The design of a TCU may vary.
achieve meaningful verbal interaction. How do they A TCU can be built as a one-word unit, such as the
do this? This question will be answered by looking at turns with no more than the words here, yes, or no in
two levels of the organization of talk that are central extract (1). Other TCUs have a syntactically more
in conversation analytic research: turn taking and elaborate design, such as the interrogative clause ‘is
sequence organization. The interaction in extract (1) she there too.’ Depending on the unit type the speaker
shows that the participants know where and how to is recognizably using for the construction of a TCU,
change the roles of speaker and listener. How they the recipient will make different predictions as to
manage this is the subject of the section about the when the ongoing turn may be complete.
 Conversation Analysis 155

Figure 1 Prosodic analysis of the TCU in lines 321–322 of extract (2).

The more complex the unit type is, the more facet- information. Whereas the TCU’s construction type
ed the projection of completeness. Compare the ex- ‘nominates’ a place in an ongoing TCU as a syntacti-
tract below. Angela begins a TCU with a particular cally plausible point of completeness, prosody can
type of subordinate clause ((but) if you get them back, ‘second’ the nomination (Schegloff, 1998). For exam-
line 321). This makes the unit underway analyzable ple, a speaker can stretch or reduce the vocalization of
as the first part of a compound TCU with an [if . . . , the intended last syllable of the turn, or mark it with a
then . . .] structure. The turn will not be complete until noticeable tone movement such as the falling pitch
the speaker has finished the then-part that is projected movement in the last word of Angela’s turn in extract
by the if-part: (2). Figure 1 makes the intonation contour of this
TCU graphically visible with the help of Praat, a
Extract (2). Telephone call between two 17-year program for the phonetic analysis of speech.
old Californian girls. Angela has just complained The lower half of the graph shows the fundamen-
that Corey’s friend has not returned three of her CDs. tal frequency – an acoustic correlate of pitch – of
Angela’s TCU in Hertz. Note that the last two
321. Angela: ! hhh (but) if you could get them words of the TCU move toward a final pitch level
322. back, (.) that be great. that is noticeably lower than the base level of the
323. 0.2 preliminary component.

324. Corey: ’kay.
Construction type and prosody are not the only
Note that the recipient does not begin to speak dimensions within which participants negotiate turn
when the speaker has finished the part with the if- rights, however. The issue of whether an utterance
clause. Even the short silence after it is not used as an is possibly complete strongly depends on pragmatic
opportunity for speakership transference. The recipi- factors, most notably on how the ongoing turn is
ent observably orients to the preceding clause as a related to its immediate interactional context (see
preliminary component of a compound TCU with Context, Communicative). Single words such as
a two-part structure. It foreshows a continuation mama, where?, here?, yes, or no, for example, can
with a structurally specifiable type of second part as only function as meaningful independent interaction-
its final component. Only after the subsequent main al moves when uttered in a context that lends them
clause implementing this latter part has reached com- this type of intelligibility.
pletion, does the recipient take over (cf. Lerner, 1991, Each utterance provides an ensemble of various
1996). types of cues that together project a possible comple-
Recipients may locate possible completeness on tion point of an ongoing TCU. The first possible
the basis of the interplay of syntactic and prosodic completion point of a TCU is the place where turn
156 Conversation Analysis

taking becomes an interactionally relevant issue. The The general characterization of the systematics of
participants of talk in interaction negotiate speaker turn taking has appeared to be very robust across
transition around such transition relevance places. languages. Depending on the structural features of
Conversationalists use specific techniques to allocate specific languages, however, the linguistic practices
next turn. If current speaker selects another partici- deployed to project possible completion points of
pant as next speaker before her turn has arrived at its TCUs may vary. The structure of English, for exam-
first possible completion point (other selection), the ple, allows for early projectability of the design of
selected party has both the right and the obligation to TCUs. Its strict Subject-Verb-Object (SVO) word
begin the next turn at this point. If no other speaker order in full clauses, for example, enforces early
is selected, another participant may self-select as positioning of predicates. Function markers such as
next speaker. If none of these options is used, current question words, imperatives, conjunctions, or quote
speaker may continue. The system then applies again attributions occur in sentence-initial position, just as
as soon as current speaker arrives at the next possible the inversion of subject and auxiliary in yes/no inter-
completion point (Sacks et al., 1974). rogatives enables early recognizability. A language
A TCU can function as an interactional move in its such as Japanese, on the other hand, is said to have
own right, and because of this, it may fill a turn slot an SOV- or OSV-type of word order, an agglutinative
on its own. On the other hand, turns may consist of morphology, and a preference for postpositioning
more than one TCU (multi-unit turns). However, un- over prepositioning of markers of syntactic, semantic,
less special provisions are made to maintain speaking and pragmatic functions. These properties result in
rights over a longer stretch of talk – as is the case a predicate-final design of clauses in TCUs. Conse-
with, e.g., story telling (cf. Sacks, 1974) – each next quently, the construction of TCUs may display a
possible completion point of a subsequent TCU is delayed projectability of possible completeness. On
treated as a place where speaker transition is an the other hand, Japanese has the option of explicit
organizationally relevant, negotiable issue. markers of possible completeness such as final verb
Thus, the organization of turn taking is accounted suffixes or final particles (cf. Tanaka, 1999). The differ-
for by describing it as a set of constructional prac- ences in language structure lead to partially different
tices that enable the co-participants to determine sets of grammatical practices that are deployed for
the place at which speaker transition becomes rele- the interactional organization of turn taking. The
vant and to then deal with that issue according to a general principles of turn construction and completion
structured set of interactional options. This way of projection are nonetheless the same.
modeling the organization of talk is characteristic
for the CA approach. The methods that members
Sequence Organization
orient to are described as formally and as generally
as is necessary to account for the fact that people We now turn to the question of how an exchange like
succeed in managing turn taking in an orderly way, the one in extract (1) is easily understood as a coher-
innumerable times a day, in all kind of situations. ent episode. It is not just the linear temporal order of
At the same time, the description has to explain turns that accounts for our understanding. The series
how participants are able to shape and recognize of turns has a structure. Some turns belong more to-
each time anew the particular context in which the gether than others. The ways conversationalists link
rules for allocating next turn apply. CA thus studies turns to each other as a coherent series of interrelated
the organization of talk as situated, socially organized communicative actions is called sequence organiza-
sets of practices. It describes the methods members tion. A sequence is an ordered series of turns through
use for organizing talk as interactional structures which participants accomplish and coordinate an
that both shape the context in which they operate interactional activity.
and enable its orderly, interactionally coordinated A question followed by an answer is an example
progression. of a sequence. Other examples are a request and the
The general model sketched in the initial paper decision that is made about it, an informative and its
about turn taking has been developed further and receipt, and a criticism and the reply to it. All these
refined in work on systematic practices of overlap different types of two-part sequences are instances of
positioning and overlap resolution (Jefferson, 1986; a very tight type of sequence organization: the adja-
Schegloff, 2000a), collaborative turn construction cency pair (cf. Schegloff, 1968; Schegloff and Sacks,
through anticipatory completion of compound turns 1973). When a recipient of a turn at talk hears the
(Lerner, 1991, 1996), and the role of gaze, gesture, speaker’s utterance as the first part of a particular
and body positioning (Goodwin, 1981) (see Gestures: type of adjacency pair, the appropriate thing to do
Pragmatic Aspects). next is to deliver an utterance that may count as
 Conversation Analysis 157

the second part of the same pair. For example, the

appropriate reaction to a question is to answer it. The
question is treated as the first pair part of a question/
answer pair; the answer is its second part. Requests,
invitations, offers, proposals, informatives, com-
plaints, or accusations establish similar expectations
Figure 2 Sequential structure of the interaction in lines 36–39
with respect to a continuation with a fitting type of
of extract (3).
second pair part in the next turn (see Speech Acts).
Extract (3) documents several instances of ques-
tion/answer pairs. The first one starts in line 33.
The son asks his mother a question, and at the first this is noticeable and accountable (cf. Schegloff,
possible completion point of his turn, the mother 1968; Heritage, 1984). The obligation to deliver the
takes over to answer it. second pair part as soon as possible in next turn may
be suspended, however. Compare, for instance, the
Extract (3). Telephone call between a mother and mother’s response to her son’s next question in extract
her son. Background information: The mother has (3) (‘where all do (we) have to go to’). ‘Huh?’ is not an
called her son from the family’s Rhine barge. The answer. Instead of answering the question, the mother
son is in a boarding school for bargee children. initiates repair (cf. Schegloff et al., 1977; Schegloff,
33 son: nou#: >waar zitteh jullie< nouw. 1992, 2000b). She signals that she is having a prob-
well where are you plural now. lem with prior turn. The son’s subsequent, slightly
0.2 modified repetition of his question apparently solves
34 mother: i:: >Amsterdam.< the problem, because the mother is able to answer
i::(n) Amsterdam. now (‘to Liège’). Unlike the first question/answer
35 0.3 sequence in this episode, the answer is not delivered
36 son: waar moe(we) almal heen. in next turn. A short repair sequence is inserted be-
where all do (we) have to go to.
tween question and answer. Schematically, this can
0.3
37 mother: è":h?
be rendered as shown in Figure 2.
huh? An insertion sequence like the repair sequence in
(.) lines 37–38 locally suspends the interactional expec-
38 son: waar moewe heen tation to deliver the second pair part in the turn
where do we have to go to following the one with the first part. The intervening
39 mother: naar Luik. interaction shows that the participants nevertheless
to Liège. are still oriented to the relevance of the second pair
0.2 part. The repair sequence is recognizably supportive
40 son: naar Lui:k? of the felicitous development of the base sequence
to Liège? in which it is embedded. The delivery of the answer
41 mother: jah.
is still pending. The urgency to answer is only tem-
yes.
porarily postponed. Note that the questioner
When the mother treats the utterance in line 33 as even renews the actuality of getting an answer by
the first part of a particular type of adjacency pair, she redoing his question in the repair itself, thereby creat-
is dealing with it as a specific type of social organiza- ing another opportunity to hold onto the preference
tion. She does not just hear an interrogative sentence for next-positioning of first and second pair parts.
that, under felicitous conditions, may count as a sep- And, by the way, note also that the inserted repair
arate speech act; she hears it as an utterance that sequence is again structured by principles of adjacen-
proposes her engagement in an interactional course cy pair organization!
of action. When an utterance is analyzable as the first Conditional relevance is also the key for under-
pair part of a particular type of adjacency pair, it standing the structure of extract (1). We are able to
locally establishes a normative expectation toward understand the interaction in extract (1) as an orderly,
what its recipient should do in next turn. The first methodically achieved sequential course of action
part makes the delivery of a fitting second part condi- on the basis of the adjacency pair structure. For
tionally relevant. That is, its recipient is expected to the reader’s convenience, the fragment is repeated
deliver the second part in his next turn. If it is absent, below:
158 Conversation Analysis

Extract (1). Telephone call between brothers. answer it, and as such, they account for the answer’s
postponement in each subsequent turn.
54 Jan: mamah, (0.2) is die d’r ook?
The interactions in extracts (1) and (3) clearly illus-
mama (.) is she there too?
55 0.4 trate that the practices through which conversation-
56 Ton: HÈ":h? alists make sense of turns at talk are based upon
huh? sequential reasoning. A turn such as ‘in Amsterdam’
57 (.) in extract (3) can only be interpreted within the
58 Jan: mama:h context of the question it is answering. The expres-
mama sion is not just a place formulation. As a place formu-
59 0.6 lation, it is answering the question where the family’s
60 Ton: waa:r Rhine barge is at the time of asking. The structure of
where the TCU as a lone standing prepositional phrase even
61 0.3
signals the kind of action it is designed to accomplish
62 Jan: is die daa:r?
in its environment of use. The combination of the
is she there?
63 0.3 turn’s position and the composition of the TCU to-
64 Ton: "hie":r? gether signal answerhood. It informs the recipient
here? about how the turn should be related to its local
65 (.) sequential context. The identification of this relation
66 Jan: j [ah is part and parcel of determining what a speaker is
yes saying and doing. Interpretative reasoning goes by
67 Ton: [nEEj! lines of sequential organization.
nO! Utterances in turns at talk accomplish actions that
68 (.) are part of social activities which are sequentially

69 Jan: oh:.
organized and have sequential implications for the
70 0.6
participants. In this section we have only looked at
The question in line 54 urges its recipient to deliver one type of sequence organization – the adjacency
an answer as soon as possible. Its delivery is, none- pair structure – and the way it can be expanded by
theless, suspended three times by the initiation of sequence insertion. Central to this perspective is the
repair from the part of the intended answerer (lines insight that utterances do not just simply count as
56, 60, and 64). Each next repair initiation builds isolated ‘actions.’ Participants in talk in interaction
upon the result of the former one, until the recipient orient to them as moves in contextually situated so-
of the question finally is able to answer it in line 67. cial arrangements. Participants in talk in interaction
The sequential organization of this episode can be do not attribute meaning to utterances by simply
schematized as shown in Figure 3. applying rules that are independent from and external
The interaction in extract (1) is tied together by the to the interaction. They make sense of utterances in
way the utterance in the first turn is sequentially turns at talk by situated, sequential reasoning.
related to the one seven turns later. The question in
line 54 makes an answer conditionally relevant, and Conversation Analytic Methodology
as long as this answer is not given, the participants
work collaboratively toward an occasion in which it Conversation analytic methodology is based upon the
can be delivered. All intervening actions are recogniz- already discussed assumption that the sense-making
ably designed as subsidiary to the task still pending. devices that participants in talk in interaction orient
They should enable the recipient of the question to to can be understood as forms of situated, interactional
reasoning (cf. Heritage, 1984; ten Have, 1999). This
kind of contextual reasoning can only be investigated
from within the interaction. The central requirement of
CA methodology – convergence between the analyst’s
perspective and the perspective of the participants –
attempts to achieve this. The analyst has to make plau-
sible that his or her results are indeed a description of
the methods that the participants themselves orient to.
The Data

Figure 3 Sequential structure of the interaction in lines 54–69 In order to avoid problems with respect to the ecologi-
of extract (1). cal validity of data, naturally occurring interactions
 Conversation Analysis 159

are strongly preferred. They are recorded and tran- case eventually provides some kind of second-order
scribed according to the conventions discussed earlier. validation of the basic pattern. The latter is frequently
The transcriptions are used to generate initial ideas the case with conversational repair. Conversational-
about how people communicate in talk in interaction. ists frequently restore departures from the methods
These ideas are worked out by looking at other they use. Such cases convey how and to what extent
instances of the same phenomenon. As a result, the the participants are oriented to the principle that the
description is gradually broadened and cyclically analyst tries to establish as a pattern. Remember, for
refined by falsifying or validating evidence. Addition- instance, how departures from the principle of con-
ally, transcripts allow the researcher to make her data ditional relevance in extracts (1) and (3) eventually
available to the scientific community. The data are re- confirmed a participant orientation to that very same
trievable for the audience. Other students are enabled device. The participants maintain the rule of having
to redo and to check the analysis. to provide a fitting second pair part by solving the
troubles they encounter in trying to obey to the
Analyzing Data rule. Usually, phases 1 and 2 are repeated recursive-
ly. The ideal is to achieve an exhaustive description
Conversation analytic methodology is strongly data that accounts for all instances of the phenomenon in
driven. There are two kinds of studies. In a single case question in the corpus.
analysis, the researcher develops an analysis of the Quantification may play a role in determining the
interaction in a single episode with respect to some distribution of the observed pattern. Quantitative pro-
interesting or relevant aspect. In a collection study, cessing of the data, however, is subsidiary to qualitative
the analyst generalizes the results of a cumulative exploration of the phenomenon in phase 1 and to qua-
series of single case analyses with respect to a specific litative validation of the candidate description in phase
aspect. All cases are compared with respect to some 2. Distributions that confirm a hypothesis primarily
feature by describing how, and the degrees to which establish regularities. The researcher still has to provide
they are the same, similar, or different. qualitative evidence that the participants observably
Single case analyses serve purposes such as generat- orient to such a pattern as a normative interactional
ing ideas that have to be grounded in a collection study, rule. If the analyst is able to demonstrate that conver-
testing and applying the results of collection studies, or sationalists orient themselves by the principle in ques-
exploring the interplay of constellations of practices in tion and to give a plausible account of how they do
episodes of talk in interaction. Collection studies have this, this accounts for the regularities.
two phases (cf. Heritage, 1995). In the first phase, the The next section gives a short demonstration of a
analyst describes regularities with respect to some par- conversation analytic way of working. A single case
ticular aspect of the data and develops a description will be discussed in order to explain how it confirms a
of a candidate pattern by going from case to case in a general pattern. The case analysis describes an aspect
corpus of transcripts. The description of phenomena of sequence organization that is called preference
is both formal and situated. A description is formal organization. In this case, the preference of agreeing
when it is formulated at a level of generality that over disagreeing assessments will be looked at. The
allows for a characterization of the recognizability analysis shows that the orderly packaging of system-
of a device across contexts. A description is situated atic sequential alternatives allows for inferences that
when the context is specified. We can discern dis- are part of the interpretative procedures participants
tinct levels of description: the level of turn design – use to make sense of talk.
practices of turn construction or ‘packaging’ – and
the description of the kind of social action that is
Preference Organization
implemented by practices of turn construction at a
sequential level. The excerpt below is taken from a telephone conver-
In a second phase, the analyst attempts to meet the sation between spouses. The wife has called her hus-
requirement of convergence between the participants’ band late at night in his restaurant bar. The episode
and the analyst’s perspective, and has to prove that starts with the husband introducing a new topic. He
the participants observably orient to the candidate tells how business is doing tonight. This report is
pattern. A keystone procedure to do this is deviant concluded with a summary assessment: ‘so u:h it’s
case analysis (cf. Schegloff, 1968); the analyst exam- much better tonight than expected’ (lines 41–42).
ines cases where some departure from the described When his wife parsimoniously confirms this evalua-
pattern can be observed. Examination of deviant tion by only saying yes, the husband is not content
cases either results in a modification of the theory with this response. It is challenged almost immediate-
developed so far, or it can be shown that a boundary ly (‘so what?,’ line 46):
160 Conversation Analysis

Extract (4). Telephone conversation between 65 Bert: dat hoe[:-

you don’t-
spouses. Background information: In the preceding
66 Ans: ! [ALS IK zeg JA: van nou
episode, Ans has blamed her husband for failing to
if I say yes like well
answering the phone earlier that evening. 67 >met ander woordeh< da’s dan
27 [ 3.2 in other words that’s
28 [( (cutlery sounds in background) ) 68 PRIma eh dan¼ehm: hHh
29 Bert: nou e:::::h ‘t is (nog eh) lekker druk. fine then uh then uhm hhh
well u:::::h it’s (still uh) pretty busy. 69 0.4
30 0.6 70 Bert: "neeh dat hoe:ft #nie:t,h
31 Ans: m:. no you don’t have to,
32 0.5
33 ‘k hoor ‘t jah The wife does the kind of metatalk we all know
So I hear uh huh from our own quarrels. Acting as a competent lay
34 0.5 linguist, she formulates explicitly what she meant

35 Bert: "jah: with her response: ‘if I say yes, (. . .) in other words,
yes that’s fine (. . .).’ In her version, the husband has mis-
36 1.2 understood her completely. When reacting with yes,

37 éen twee drieh (0.6) nou drie she was giving an agreeing, even approving response.
one two three (0.6) well three
The question is, however, whether the husband in-
38 (garnetborde) en twee drie vier zit-(.)
deed did misunderstand her. Did he have any conver-
(plate services) and two three four sit-
39 vijf zes: zitt"n "r nog. sational evidence for an alternative interpretation
five six still sitting there. when he threw doubt upon his wife’s response?
40 2.5 In order to be able to answer this question, we have
41 dus e:h dat valt to go back to the sequence that occasioned the dis-
so u:h it’s pute. The utterance ‘so u:h it’s much better tonight
42 vana:vond reuze mee than expected’ (lines 41–42) is an assessment. It is an
tonight much better than expected interactional property of first assessments that when
43 (.) its recipient is also knowledgeable about the evalu-
44 Ans: ! thHH j:Ah. hhh ated object, a second assessment is expected from the
yes.
part of that party. Second assessments have the prop-
45 0.2
erty that they find their measure in the assessment
46 Bert: ! wat dan:?
so what? they are responding to. Second assessments are never
47 0.7 neutral; they either agree or disagree with the first
48 Ans: *nou ja:h e:hh* one. Disagreeing assessments are more delicate ac-
well yes u:hh tions than agreeing assessments. Participants in talk
in interaction treat a disagreeing second assessment
In the aftermath of this exchange, a short dispute
usually as a less preferred type of next action than its
develops between the spouses. The wife’s initial reac-
agreeing alternative. They are nonequivalent alter-
tion is rather resistant. Yet when her husband keeps
natives. Agreement is preferred and unmarked;
pushing her, she finally bursts out in an angry,
disagreement is dispreferred and marked. Preferred
reproaching tirade (lines 59–70): second pair parts are delivered without delay and for-
mulated in a frank, concise mode. Dispreferred sec-
Extract (5). 22 seconds later in the same episode
onds, on the other hand, are frequently delayed,
(including a long, awkward silence).
mitigated, hesitantly produced, hidden away, put in
58 Ans: je:zus a roundabout way or accounted for (Pomerantz,
jesus 1984; Schegloff, 1995).
59 ik moet elk woor:d >wat ik zeg< The ranking of sequential alternatives with respect
I have to every word that I say to their relative degree of preference is called prefer-
60 moet ik verantwoordeh ence organization. The preference for agreement is
I have to account for
just one type of preference organization. A related
61 moet ik vier keer uitLEGgehH
I have to explain four times
type of preference organization is the preference for
62 (.) project success. It accounts for the preference for
63 Bert: nee dat hoeft niet. second pair parts of adjacency pairs that bring about
no you don’t have to. the result targeted in the interactional project that is
64 1.3 initiated with a first pair part, e.g., getting an answer
 Conversation Analysis 161

to a question, accepting an invitation, granting a another aspect of sequence organization: preference

request, affiliating with a complaint, etc. Participants’ organization. Second, it shows one more time how the
orientations to preference organization is a major meaning of utterances is constituted along lines of
source for sequence expansion of adjacency pairs sequential reasoning. An utterance in a turn at talk is
(Schegloff, 1995). Participants may probe and try to not just what it says, but what it does in a particular
preempt the likeliness of a dispreferred second pair sequential context. Third, the analysis illustrates that
part in a presequence (Schegloff, 1980). They may the methods by which participants make sense of their
initiate repair on first pair parts in insertion sequences talk may be (re-)specified and (re-)negotiated in
in order to provide an opportunity to adjust the pref- the course of the interaction. Finally, the discussion
erence structure of the preceding first pair part, or to demonstrates some aspects of CA methodology dis-
at least delay the delivery of a dispreferred second cussed in the former section. The knowledge that first
pair part. And they may try to revise or to accommo- assessments invite a second assessment from its recipi-
date a preceding dispreferred second pair part in var- ent stems from a collection study (Pomerantz, 1984).
ious types of post-expansion (Schegloff, 1995) – as The interaction in extract (4) seems to contradict this
in the kind of post-expansion that can be observed pattern. Instead of continuing with a second assess-
in extracts (4–5). The relational implications of types ment, the recipient responds with only an acknowl-
of interactional alignment and disalignment that edgement token. However, when this observation is
are governed by preference organization are also combined with insight into the ways how preference
investigated from the perspective how participants organization operates, the interaction can be explained
negotiate epistemic rights (Heritage and Raymond, in terms of the very same mechanism. The analysis of a
2005; Stivers, 2005). For example, a co-participant deviant case eventually provides a kind of second-order
may claim or give evidence for primary epistemic validation of the theory developed so far.
authority by the way he responds to a first assess-
ment, although the response itself has a kind of sec-
Extensions and Applications
ondness because it is done in a sequentially second
position. The basic theoretical, analytical, and methodological
The notion of preference does not refer to psycholog- framework of CA has been developed further into
ical dispositions. It is a description of interactionally various domains and directions. Studies in the area
observable orientations of participants. Preference or- of interaction and grammar explore the relationship
ganization provides the participants with a subtle and between language structure, linguistic practices, and
powerful apparatus for making interpretative infer- the organization of turn taking and of sequences in
ences. In extract (4), the husband uses it as a resource talk in interaction (Ochs et al., 1996; Selting and
when he challenges his wife’s response. In a context in Couper-Kuhlen, 2001; Ford et al., 2002; Couper-
which an agreeing second assessment is preferred, his Kuhlen and Ford, 2004).
wife avoids taking a stance. When she reacts with ‘yes,’ The work of Charles Goodwin has given a major
she merely acknowledges her husband’s statement ‘so impetus to the study of the multimodal and embodied
u:h it’s much better tonight than expected.’ She does not character of the organization of human action in talk
affiliate with her husband, but responds in an evasive in interaction – not only the role of gaze, gesture, and
manner instead. Her husband’s reaction attends to pre- body positioning, but also the use of tools and other
cisely this aspect of her response. He challenges a weak- features of the setting (Goodwin, 2000; see also
ly agreeing response in a sequential environment in the workplace studies in Heath and Luff, 2000).
which another alternative is more preferred. So, despite Goodwin (2003) provides a collection of CA stud-
his wife’s subsequent (re-)formulation of the meaning ies of the ways in which people with one or another
of yes as simply agreeing, the man nevertheless has form of language impairment use various types of
good reasons to hear her response as a sign of reserva- sequential and situational reasoning in ordinary com-
tion or even foreboding disagreement. From a sequen- municative situations (see Institutional Talk; Family
tial perspective, the response is not just acknowledging Speak; Computers in Lexicography).
prior speaker’s assessment. Saying yes in this context is Talk in institutional, professional, or work settings
rather deployed as a device to avoid agreeing. As a is also studied by describing how it is constrained or
contextually specifiable selection of another alternative modified in comparison to conversational interaction
than the preferred one, it legitimizes the interpretation (Drew and Heritage, 1992). The research in this area
that is subsequently challenged by the husband. has frequently the shape of studying genres or activity
The analysis of the origins of the argument in ex- types in a specific domain, e.g., the news interview
tract (4) demonstrates several aspects of the con- (Clayman and Heritage, 2002), judicial interaction
versation analytic approach. First, it introduces (Atkinson and Drew, 1979), emergency calls (Whalen
162 Conversation Analysis

and Zimmermann, 1990), meetings (Boden, 1994), Drew P & Wootton A (eds.) (1988). Erving Goffman. Ex-
telling good and bad news in clinical settings ploring the interaction order. Cambridge: Polity Press.
(Maynard, 2003), and gossip (Bergmann, 1993). Drew P & Heritage J (eds.) (1992). Talk at work. Inter-
The research here is sometimes called applied CA action in institutional settings. Cambridge: Cambridge
University Press.
(see ten Have, 1999) (see Institutional Talk).
Ford C E, Fox B & Thompson S A (eds.) (2002).
Wootton (1997) is an example of a CA study in the
The language of turn and sequence. Oxford/New York:
area of early language acquisition. The study of for- Oxford University Press.
eign language use in talk in interaction focuses on the Gardner R & Wagner J (eds.) (2004). Second language
organization of repair (see Schegloff, 2000b; Gardner conversations: Studies of communication in everyday
and Wagner, 2004). settings. New York/London: Continuum.
A primarily British group of social psychologists Goodwin C (1981). Conversational organization. New
approaches typically psychological topics such as at- York: Academic Press.
titude, identity, and cognition from an interactional Goodwin C (2000). ‘Action and embodiment within
perspective, using CA as a central theoretical and situated human interaction.’ Journal of Pragmatics 32,
methodological framework (Te Molder and Potter, 1489–1522.
Goodwin C (ed.) (2003). Conversation and brain damage.
2004).
New York: Oxford University Press.
Journals that regularly publish CA papers are Re- Have P ten (1999). Doing conversation analysis. London:
search on Language and Social Interaction, Discourse Sage.
Studies, Human Studies, Journal of Pragmatics, Lan- Heath C & Luff P (2000). Technology in action. Cam-
guage in Society, Pragmatics, and Text. Important bridge: Cambridge University Press.
centers of CA research are UCLA (Schegloff, Heri- Heritage J (1984). Garfinkel and ethnomethodology. Cam-
tage, Clayman), UCSB (Lerner, Raymond), the Uni- bridge: Polity Press.
versity of Wisconsin (Maynard, Ford), The University Heritage J (1995). ‘Conversation analysis: methodological
of York (Drew, Local, Wootton), the University of aspects.’ In Quasthof U (ed.) Aspects of oral communica-
Helsinki (Sorjonen), Odense (the Graduate School tion. Berlin/New York: Walter de Gruyter. 391–418.
of Language and Communication), Bielefeld (Berg- Heritage J & Raymond G (2005). ‘The terms of agree-
ment: Indexing epistemic authority and subordination in
mann), and Max Planck Institut Nijmegen, Language
talk-in-interaction.’ Social Psychology Quarterly 68(1),
and Cognition Group (Stivers). 15–38.
Discussion lists with CA-oriented discussion are Jefferson G (1986). ‘Notes on ‘latency’ in overlap onset.’
the Language-use list, the Ethno-Hotline, the German Human Studies 9(2–3), 153–184.
Gesprächsanalyse-list, or the Danish MOVIN-list. Jefferson G (2004). ‘Glossary of transcript symbols with
The Ethno/CA News website of Paul ten Have an introduction.’ In Lerner G (ed.) Conversation analy-
announces conferences, publications, and other news. sis. Studies from the first generation. Amsterdam/
Philadelphia: Benjamins. 13–31.
See also: Computers in Lexicography; Context, Communi- Lerner G (1991). ‘On the syntax of sentences-in-progress.’
cative; Family Speak; Gestures: Pragmatic Aspects; Goff- Language in Society 20, 441–458.
man, Erving (1922–1982); Institutional Talk; Sacks, Harvey Lerner G (1996). ‘On the ‘semi-permeable’ character of
(1935–1975); Speech Acts; Telephone Talk. grammatical units in conversation: conditional entry
into the turn space of another speaker.’ In Ochs et al.
(eds.). 238–276.
Bibliography Maynard D (2003). Bad news, goods news. Conversational
order in everyday talk and clinical settings. Chicago/
Atkinson J M & Drew P (1979). Order in court. London: London: University of Chicago Press.
Macmillan Press. Ochs E, Schegloff E & Thompson S (eds.) (1996). Gram-
Atkinson J M & Heritage J (eds.) (1984). Structures of mar in interaction. Cambridge: Cambridge University
social action. Studies in conversation analysis. Cam- Press.
bridge: Cambridge University Press. Pomerantz A (1984). ‘Agreeing and disagreeing with assess-
Bergmann J (1993). Discrete indiscretions. The social orga- ments: some features of preferred/dispreferred turn-
nization of gossip. New York: Aldine/de Gruyter. shapes.’ In Atkinson J M & Heritage J (eds.) 57–101.
Boden D (1994). The business of talk: organizations in Sacks H (1974). ‘An analysis of the course of a joke’s telling
action. Cambridge: Polity Press. in conversation.’ In Bauman R & Sherzer J (eds.) Explo-
Clayman S & Heritage J (2002). The news interview. ration in the ethnography of speaking. Cambridge:
Journalists and public figures on the air. Cambridge: Cambridge University Press. 337–353.
Cambridge University Press. Sacks H (1992). Lectures on conversation. Jefferson G (ed.),
Couper-Kuhlen E & Ford C E (eds.) (2004). Sound patterns Introduction by Schegloff E. Oxford/Cambridge, MA:
in interaction. Cross-linguistic studies from conversation. Blackwell.
Amsterdam/Philadelphia: Benjamins.
 Conversational Agents, Synthetic 163

Sacks H, Schegloff E & Jefferson G (1974). ‘A simplest Selting M & Couper-Kuhlen E (eds.) (2001). Studies
systematics for the organization of turn-taking for con- in interactional linguistics. Amsterdam/Philadelphia:
versation.’ Language 50, 696–735. Benjamins.
Schegloff E (1968). ‘Sequencing in conversational open- Stivers T (2005). ‘Modified repeats: One method for assert-
ings.’ American Anthropologist 70, 1075–1095. ing primary rights from second position.’ Research on
Schegloff E (1980). ‘Preliminaries to preliminaries: Language and Social Interaction 38(2), 131–158.
‘‘Can I ask you a question? ’’’ Sociological Inquiry Tanaka H (1999). Turn-taking in Japanese conversation.
50(2), 104–152. A study in grammar and interaction. Amsterdam/
Schegloff E (1992). ‘Repair after next turn: The last struc- Philadelphia: Benjamins.
turally provided defense of intersubjectivity in conversa- Te Molder H & Potter J (eds.) (2005). Talk and cognition:
tion.’ American Journal of Sociology 98, 1295–1345. discourse, mind and social and interaction. Cambridge:
Schegloff E (1995). Sequence organization. Cambridge: Cambridge University Press.
Cambridge University Press. (ms., Department of Sociol- Whalen M & Zimmerman D (1990). ‘Describing trouble:
ogy, UCLA. An updated version appears in 2006 as: A practical epistemology in citizen calls to the police.’ Lan-
primer in conversation analysis. Sequence organization. guage in Society 19(4), 465–492.
Cambridge: Cambridge University Press.) Wootton A J (1997). Interaction and the development of
Schegloff E (1998). ‘Reflections on studying prosody in mind. Cambridge: Cambridge University Press.
talk-in-interaction.’ Language and Speech 41(3–4),
235–263.
Schegloff E (2000a). ‘Overlapping talk and the organization Relevant Websites
of turn-taking for conversation.’ Language in Society 29,
1–63. http://ecampus.bentley.edu/dept/bps/emca – Ethnomethod-
Schegloff E (2000b). ‘When ‘others’ initiate repair.’ Applied ology and Conversation Analysis section of the American
Linguistics 21(2), 205–243. Sociological Association.
Schegloff E, Jefferson G & Sacks H (1977). ‘The preference http://www.institut-gespraechsforschung.de – German
for self-correction in the organization of repair in conver- Gesprächsanalyse-list.
sation.’ Language 53, 361–383. http://www.conversation-analysis.net – Danish MOVIN-list.
Schegloff E & Sacks H (1973). ‘Opening up closings.’ Semi- http://www.paultenhave.nl – Ethno/CA News website of
otica 8(4), 289–327. Paul ten Have.

Conversational Agents, Synthetic

J Cassell, Northwestern University, Evanston, IL, USA the system is designed to teach Arabic in such a way
ß 2006 Elsevier Ltd. All rights reserved. that soldiers going into an unfamiliar culture learn
appropriate body language as well as the necessary
foreign words and phrases.
Embodied conversational agents (ECAs) are virtual
ECAs also allow linguists to model human linguis-
humans (often life-size) who are capable of carrying
tic behavior and to evaluate competing theories of
on conversations with real humans. These virtual
language use by observing them in action. Figure 3
humans may serve many functions, both practical
shows an ECA that gives directions by speaking, ges-
and theoretical. In the practical vein, they may act
turing, and tracing a route on a map. This system has
as the interface to a computer so that instead of
allowed researchers to discover the role of nonverbal
choosing commands on a menu, one can carry on a
behaviors, such as eye gaze and head nods, in ground-
conversation.
ing, or the establishment of information as shared
Figure 1 shows an ECA named REA who acts as the
between two participants.
interface to a database of houses in the Boston area.
In all of these cases, the embodied conversational
Rather than having to type in search terms, users can
agents are modeled on human face-to-face conversa-
tell REA what kind of property they are looking for,
tion and therefore get their meaning across by employ-
and REA will nod, reflect, and then find appropriate
ing not just text (as do regular computers with a
properties and describe them using a combination
mouse, keyboard, and screen) but also spoken speech
of descriptive hand gestures, head movements and
with intonation, hand gesture, head movements, and
spoken language.
facial expressions. Embodied conversational agents
ECAs can also serve as autonomously acting char-
are defined by the following:
acters in video games. Figure 2 shows an ECA that
plays the role of a village leader that reacts to the . The ability to recognize and respond to verbal and
soldier character played by the user. In this instance, nonverbal input