CHAPTER 3
Data and Transcription Techniques
Conversation analysis places a great deal of emphasis on the use of extracts from transcriptions of tape-recorded, naturally occurring interactions in its research. This chapter is designed to be both a theoretical and a practical introduction to the style of transcription employed in CA. The focus is on the practice or craft of transcription as well as on the methodological and theoretical bases of the transcription system. This is because, as we emphasize, not only the analytic use, but also the actual practice of transcription is a fundamental part of doing CA.
The transcription of data is a procedure at the core of analysis, in two important respects. First, transcription is a necessary initial step in making possible the analysis of recorded interaction in the way that CA requires. Secondly, the practice of transcription and production of a transcript represents a distinctive stage in the process of data analysis itself.
It is important to stress that, for CA, transcripts are not thought of as 'the data'. The data consist of tape recordings of naturally occurring interactions. These may be audio or video tapes - although clearly, when the people who are being recorded have visual access to one another an audio-only tape will necessarily miss out what could be very salient features involved in the management of interaction, such as gaze direction (Goodwin 1981) and hand gestures (Schegloff 1984). In the early years of CA (and still to a large extent today), researchers tended to restrict their attention to recordings of telephone conversations, precisely because this allowed them to focus purely on the organization of talk in the absence of such factors (Hopper 1992). Since then, video recordings have been more widely used, but CA's explicit focus on the organization of ta/A:-in-interaction
74
Practices
Data and Transcription Techniques
75
means that gesture, body movement and facial expression are not studied in their own right, as may be the case in the field of interactional kinesics (Kendon 1990), but rather in exploring the relationships between speech and body movement. However, it is possible, within CA, to analyse audio-only recordings even when the participants have visual access to one another. For instance, Marjorie Goodwin's (1990) detailed study of the management of disputes among children at play on the street offers a compelling analysis of the role of talk in the social organization of the children's groups using as its data only an audio record supplemented by ethnographic fieldnotes.
Given this conception of the data, the aim in CA is not simply to transcribe the talk and then discard the tape in favour of the transcript. As Hopper (1989a) observes, the latter is often the practice in social psychology where only the transcripts are analysed in terms of the categories of action which interest the researcher. Conversation analysts, by contrast, do not analyse transcripts alone: rather, they aim to analyse the data (the recorded interaction) using the transcript as a convenient tool of reference. The transcript is seen as a 'representation' of the data; while the tape itself is viewed as a 'reproduction' of a determinate social event.
Of course, the tape is only one form of reproduction; and whether it is an audiotape or a videotape, it does not reproduce everything that went on in the vicinity of the recording device during the time it was switched on. (Indeed, we find it difficult to conceive of any way in which such an abstract 'everything' could possibly be recorded.) But conversation analysts take a pragmatic view on this issue. As Sacks once put it, describing why he initially became interested in working with tape-recorded conversations:
Such materials had a single virtue, that I could replay them. I could transcribe them somewhat and study them extendedly -however long it might take. The tape-recorded materials constituted a 'good enough' record of what had happened. Other things, to be sure, happened, but at least what was on the tape had happened. (Sacks 1984a: 26)
Wherever possible, then, the transcript is used in conjunction with the tape during analysis. As both Hopper (1989a) and Psathas and Anderson (1990) point out in their descriptions of CA transcription procedures, repeated listening to the original
recording is central to the CA technique. This allows the analyst to gain an intimate acquaintance with the recording at the necessary level of detail. For this reason, and because analysis is not performed on the transcript alone, it is not standard practice in CA to have transcription done by secretaries or professional transcribers. Rather, transcription is done by the analyst. Transcription thereby becomes an integral part of analysis, since in repeatedly listening to the tape one begins to hear and to focus on phenomena that may subsequently form part of an analytic account.
The process of transcribing a data tape is not simply one of writing down the words that people exchanged. Rather, it is a process of writing down in as close detail as possible such features of the recorded interaction as the precise beginning and end points of turns, the duration of pauses, audible sounds which are not words (such as breathiness and laughter), or which are 'ambiguous' vocalizations, and marking the stresses, extensions and truncations that are found in individual words and syllables. Because CA is concerned with how people manage and accomplish the sequential order of talk-in-interaction, transcription is, first of all, an attempt to capture talk as it actually occurs, in all its apparent messiness. As a result, CA transcripts can often appear formidably complex to the untrained eye.
It is possible, however, to learn relatively quickly how to read transcripts of conversation. This is because CA has developed a distinctive style of transcription, involving a comprehensive range of standardized conventions, unlike many other approaches which use recorded talk as their data (see Hopper 1989a; Ochs 1979). This system, developed principally by Gail Jefferson, is in general use by conversation analysts working in many different countries on widely varying forms of recorded interaction. One of the aims of this chapter is to provide the resources by which it is possible to become familiar with conversation analytic transcription conventions.
Approaching transcription: some preliminary issues
Clearly, there are innumerable phenomena in any given stretch of talk which could be transcribed to varying levels of detail. No
76
Practices
Data and Transcription Techniques
11
transcription system exists which is able, or even lays claim to being able, to capture all the possible features of talk that may be observable. As Kendon puts it: 'It is a mistake to think that there can be a truly neutral transcription system, which, if only we had it, we could then use to produce transcriptions suitable for any kind of investigation ... Transcriptions, thus, embody hypotheses' (1982: 478). Similarly, Ochs describes transcription as 'a selective process reflecting theoretical goals and definitions' (1979: 44). This is no less true of CA transcriptions. A CA transcript embodies in its format and in the phenomena it marks out the analytic concerns which conversation analysts bring to the data. These concerns are of two types:
• Dynamics of turn-taking On this level, transcripts seek to capture the details of the beginnings and endings of turns taken in talk-in-interaction, including precise details of overlap, gaps and pauses, and audible breathing.
• Characteristics of speech delivery Here, transcripts mark noticeable features of stress, enunciation, intonation and pitch.
At the most basic level, the central concern with turn-taking is embodied in the very layout of CA transcripts, in which talk is represented in the form of utterances following one another down the page. Conversation analysts are also concerned to transcribe as precisely as possible all of the sounds that are uttered by participants, whether or not these are conventionally recognizable words. This is because CA assumes that any sound may have interactional import and communicative meaning. Hence there are conventions for transcribing such things as audible breathing, because an audible in-breath is often a signal that its producer is about to speak, and so that sound can be consequential for the management of turn transition.
Two particular aspects of speech delivery that are of great importance for doing conversation analytic work are: (1) when spoken syllables are stretched; and (2) basic features of intonation. Both aspects can be closely related to issues of turn-taking. For instance, stretching a sound at the possible boundary of a turn, or possible transition-relevance place, can be a way of 'holding the floor' or preventing another speaker from starting a turn at that point. And different intonation contours used at the boundaries of turn-construction units such as clauses can indicate
whether the speaker may be intending to continue, or, if the intonation is markedly falling, possibly coming to the end of a full turn.
The way these characteristics of speech delivery are marked tends to be relatively gross when compared, say, to transcripts produced by professional phoneticians who have their own sets of technical symbols for representing such phenomena. Indeed, phoneticians such as Kelly and Local, who are basically sympathetic to CA, have questioned the utility of these aspects of CA transcription. They remark that for the details of turn-taking, for example gaps and pauses, overlaps and audible breathing activity, CA transcription is: 'consistent and systematic. At other places, however - in reflecting features of tempo, pitch, loudness, vowel quality and voice quality, for example - the transcriptions [seem] inconsistent and arbitrary' (1989:204).
Kelly and Local suggest that there is much to be gained analytically from paying more serious attention to phonetic phenomena. However, their approach requires of both the analyst and the reader some degree of knowledge of phonetic techniques; whereas in general, the noting of prosodic characteristics in CA transcription is linked to a different aim: 'to get as much of the actual sound as possible into our transcripts, while still making them accessible to linguistically unsophisticated readers' (Sacks, Schegloff and Jefferson 1974: 734).
Before going further into these kinds of issues, we need to introduce the set of transcription conventions used by conversation analysts. We will begin, in the next section, by describing the basic conventions and the rationale underlying their use; then we will illustrate their main strengths with a transcript of a recorded telephone conversation.
Transcription conventions
The first necessary step in doing transcription is to understand the transcription conventions that CA uses, and to have a sense of which features of talk to concentrate on when listening to tape recordings. It is often said that CA transcription procedures are designed to make for more and more 'accurate' transcripts of naturally occurring talk (for instance, Graddol, Cheshire and
78
Practices
Swann 1994:181). However, this is only partly true. The principal features of talk represented in CA transcripts are almost all involved with particular analytic issues: issues to do with analysing the local production of order in naturally occurring talk-in-interaction. The features that should be listened for during transcription, as we have already noted, fall into two broad categories: those to do with turn-taking and those to do with what we loosely call 'speech delivery'. In this section, we introduce the principal conventions used to transcribe these features.
Turn-taking and overlap
Most research in CA is based on the 'simplest systematics' model outlined in Sacks, Schegloff and Jefferson (1974). That ideal model of conversational turn-taking stressed the exchange of turns with minimal gap and overlap between them. Of course, in actual conversations both gaps and overlaps are frequent; and so they must be marked even if only for representational adequacy. However, overlaps in particular are bound up with the management of turn-taking and the observable achievement of mutual understanding in talk-in-interaction. As we saw in the previous chapter, Jefferson (1986) has shown that, on close inspection, much overlapping talk which appears interruptive is in fact closely coordinated with the occurrence of transition-relevance places.
Overlap onset is marked in transcripts by the use of left-hand square brackets:
(1) [SBL:1:1:115]
1
2
3->
4
5
6
7-> 8
B: D:
B: D:
Uh huh and I'm soirry I didn' get Mar.gret I reallyr've been wan'ing to.
IW'U I think" she mu:st've stayed out'v
to:wn (6.2)
I thijnk so too.
•in Fre:sno sh- see she'n Pe::g (0.7) dro:ve over to 'er sister's 'oo lives in Fresno::
As we see with this extract, the aim is to be as precise as possible about marking the point at which overlap begins: even when, as in lines 6 and 7, that is in the middle of a word. By marking the
Data and Transcription Techniques
79
precise points of onset in overlapping talk, very close calibrations in the understandings that speakers display of each other's talk become available for analysis.
Other forms of overlap may also be marked. For instance, speakers may start a turn simultaneously. This is marked by the use of a double left-hand bracket:
(2) [Heritage:I:II:3]
1 I: Well .h I a-always feel it's best to get it
2 all over at rtEe same time yiou know.
3 N: IWell ye::sJ
4 N: Ye:s.
5- * I: frit's uh:
6- > N: "And and who did you go to.
While it is not always strictly necessary to mark the end of a stretch of overlapping talk, studies such as Schegloffs (1987a) account of 'recycled turn-beginnings' show how, by focusing on the point at which overlap ends, important aspects of the ongoing management of conversation can thereby be revealed. When the end of overlap is marked, that is done by a single right-hand square bracket:
(3)    [Schegloff 1987a: 75]
1
2
3
4
5-> 6
R:
K:
in fact they must have grown a culture, you know, they must've-1 mean how long- he's been in hospital for a few days right? Takes arbout a week to grow a culture]
ll don't think they grow a) I don't think they grow a culture to do a biopsy.
It is noticeable here that K, finding himself beginning to speak in overlap with R's continuation of his turn about growing a culture for a biopsy, does not drop out of the competition for the floor but keeps going until R comes to a next recognizable completion point. At that stage, however, rather than simply carrying on talking, K, with remarkable precision, restarts his turn, thereby displaying how speakers can coordinate turn transitions with transition-relevance places even when a first attempt to do so has failed.
Utterances may also sound 'latched' together: they may occur right next to each other with absolutely no gap, but also no overlap, between them. This is marked by equals signs, thus:
m
80
Practices
Data and Transcription Techniques
81
(4) [NB:IV:10]
1 E:   Is the swimming pool enclosed with the
2 gla:ss bit?=
3 N:  =N"o::, it's uh: ou.ts- (.) eh no outside
Equals signs are also used to deal with a problem inherent in the attempt to represent naturally occurring talk on the page: the fact that page lines are strictly limited in length, whereas conversational turns are not. This means that a single utterance may need to be broken up, if, for instance, another speaker says something during its course. In such a case, the two parts of the longer turn are connected by equals signs, with the embedded utterance transcribed on a line between:
(5) [NB:IV:14]
1 N:  But eh- it's- it's terrible to keep people
2 ali:veand you [know and just let them=
3 E: 'Right.
4 N:  =suffer day in and day out,
In a similar sense, a protracted spate of simultaneous talk may require a combination of left and right brackets and equals signs, if the spate extends across a number of lines on the page:
(6) [NB:IV:14]
1 E: Well, we don't know what it's all about
2 TgT- ((sniff)) Don't get yourself=
3 N: =rrOh I'm not.  I just- you know I wishi
4 E:    "Honey you've got to get a hold of your-1 knowJ=
5 N: =I'd-I'd kind of liked to gone out there but
6 I was afraid of the fog
It is also worth noting that an early device for transcribing the onset of overlap was the double slash (//). This is no longer widely used in transcription. Occasionally, though, it will be used to mark the point of overlap onset when a single line from a longer data extract is being discussed in the text. The longer extract might be the following:
(7) [SBL:2:1]
1 B:   I still haven't my dishes done, I'm
2 right in the middle of doing them, but
3 I stopped rto call you.
4 J: iWell I worked on my-medicine
5 cabinet again, I'm so mad at that painter,
Then B's turn might be cited in this way: T still haven't my dishes done, I'm right in the middle of doing them, but I stopped//to call you.'
Gaps and pauses
From the beginning, conversation analysts have timed intervals in the stream of talk relatively precisely, in tenths of a second. Again, this is not just a matter of accuracy. Work on dispreferred responses (Davidson 1984; Pomerantz 1984a) has demonstrated that pauses even as short as two or three tenths of a second can have some interactional, and therefore analytic, significance. Slightly later, Jefferson (1989) produced an extensive exploration of the interactional significance that is attached to silences of one second in length during conversation. It is clear that using a catch-all device such as writing 'pause' in transcripts would not have enabled many of the finer analytic points contained in these papers to be made.
Timings of pauses, then, are important features of transcripts. The timings, which are usually done with a stopwatch, are inserted in the transcript at the precise point of their occurrence in the recording. They may occur within a turn:
(8) [SBL:2:7:20]
1 A:  I-if you want to uh(b) (1.1) maybe get up a game .
Or they may occur between turns:
(9)    [SBL:2:7:20] (continuation of (8))
1 A: I-if you want to uh(b) (1.1) maybe get up a game
2 some morning while you're out the:re,=why that's
3 always fu:n,
4 B: Mm hm. 5-> (0.5)
6 A: So let me kno:w.
Pauses that are detectable, but run for less than 0.2 of a second, are indicated by a period within parentheses:
82
Practices
(10) [SBLOnM] (Continuation of (9)) 6      A:  So let me kno:w.
7-> (•)
8      B:   Yah will do:.
Breathiness
Breathiness which is audible to the transcriptionist is marked by 'h' for exhalation and '.h' for inhalation. As we have already remarked, this feature of recordings is transcribed because (among other reasons) audible in-breaths may be involved with the management of turn-taking, in as much as an open-mouthed in-breath may mark a participant's attempt to start a turn. Notice how in the following extract, speaker E draws in a long breath at line 3, which overlaps substantially with P's invitation. This in-breath signals her attempt to take the floor and respond to the invitation; although the reasons for it may be analytically complex. One relevant factor is that P's invitation has multiple possible completion points ('you wanna go to the store/ er anything/ over at the Market Basket/ er anything?'), so that E's in-breath may be because she has decided on her response but is waiting for the invitation to come to an end (see Davidson 1984). As soon as the final 'er anything?' has been produced, she stops her in-breath and begins to speak:
(11) [NB:52:2.-66]
IP:   Oh I mean uh: you wanna go to the store er anything
2 over at the Market [Basket  er anything?]
3 E: ~~      l.hhhhhhhhhhhhhhhhhhlh Well honey...
In transcripts, the longer the breath, roughly, the longer the line of 'h* or '.h' provided in the transcript:
(12) [HG:28]
1-
2
3
4
5-
N
N
H
.hhh (0.5)
A::nywa::y,
eh-eh .hhhhhhh Uh::m,
Data and Transcription Techniques
83
The lengths of these breaths are not timed in any strict sense. Rather, the length of a breath is assessed impressionistically, relative to the general tempo of the surrounding talk (although more precise measurement is possible, of course, when the entire breath takes place within overlap, as happens in line 3 of extract (11) above).
Transcripts also mark plosive aspiration within a word: this happens, for instance, when someone speaks 'through' laughter (that is, laughs while enunciating a word or phrase). This is indicated by placing the 'h' in parentheses. However, due to the nature of the phenomenon, this can be very hard to do, and the transcriptionist may need to listen to the same word or group of words many times before he or she can achieve a satisfactory textual rendering of the sounds that occur on the tape:
(13) [EB:1]
1 S:   I hope by next semester it'll be a bi(h)t
2 b(h)edd(h)er heh heh heh heh .hh
(14) [H:2.2.89:4:3]
1
2
3
4
5
6->
H:
C:
Yes but you ca:n't actually:, take anybody to la::w, .h jus:t on:, an accusation. (1.1) "
.p .hhh (.) No I kno:w I'm not just making accusations I've got proof of my owri" ej;(h)(huh huh hih)es!Tilihh
Laughter
In a series of papers on laughter, Jefferson (1979; 1985; Jefferson, Sacks and Schegloff 1987) showed that laughter, as it occurs in talk-in-interaction, is a finely coordinated interactional phenomenon. This in turn meant that laughter, which may previously have been represented descriptively by the transcriber simply writing '(laughs)', now should be transcribed as literally as possible in the form of onomatopoeic renditions of laugh particles: 'ha ha', 'heh heh', 'hih hih' and so on. The particles are designed to represent as closely as possible the sounds emitted by the participants:
84
Practices
(15) [Goodwin:GR:40]
1
2
3-> 4
5-» 6
7-»
J:   So I said look Gurney, yer just a big ass
kisser, (04) en iyer getting yer wa:y,i B: lAAHh hah-uh hah-uh huhJ:
J:   =fl(h)    ju(h)sti laird it a:ill on, B:    l.hhhhhhhhhhhl IhhabJ (•)
B:  ehh huh uh-huh uh-huh
One of the difficulties with transcribing laughter is that participants frequently laugh together. This can make for a highly complex transcribing job, especially when the recorded talk involves more than two people. In the following fragment, there are only three participants; but their extended stretch of laughing together leads to a rather daunting-looking section of transcript:
(16) [Goodwin:AD-.56:r]
1 B
2 C:
3
4
5
6
7
8
9
10
11
he:uhthe-uh-hai
'Keiha:   :i|ah ha-ha-ha-ha-ha-ha-a-hai lah!ah!lah!^Ei!äh!ah!ah!ah!ah!ah! •= =[.hhh [he|hei:h= lah!a!a!J
[ah hhj =her:h he-ehh-eihe- rhe-l Oo:::::ps, ' n'he
huh	
	eh!
=[ e::::a::yee: 1 [uh!ah!ah!ah!J
ee::
We have used these features of speech production to illustrate some of the reasons why conversation analysts have developed the transcription system currently in use. However, this is not an exhaustive description; other symbols are explained in the glossary of transcription conventions found at the front of the book.
One final thing to mention is that the transcriber may sometimes be in doubt as to what actually occurs on the tape. For instance, it may be that a speaker evidently says something, but it may not be clear precisely what it is he or she has said. (One place in which this is likely to occur is during stretches of overlapping talk.) In such cases, the standard convention is to enclose one's best hearing of what was said in brackets (if no actual words are discernible, the brackets may enclose empty space,
Data and Transcription Techniques
85
signifying at least that some sound occurred at that point). On other occasions, the transcriber may want to record some descriptive properties of the speaker's voice, such as that it is 'gravelly', or 'smiley'. This is usually done as an aid to the memory for those occasions when the transcript is being observed without the accompanying tape. Such descriptions are enclosed within double brackets, for instance, ((smiley voice)).
To summarize this section, then, the transcription system used in CA is designed to produce transcripts that are accurate at the relevant levels of detail (levels of turn-taking and actual speech delivery), while avoiding being technically inaccessible to the majority of readers. The purpose of the transcription conventions, however, is not merely to produce accurate representations of talk, but to focus attention on those features of talk-in-interaction that are analytically significant from the standpoint of CA. We have illustrated that point in brief outline in this section. How the transcription system is used to bring out the finer analytic points in a stretch of talk can be illustrated in more detail by means of the following exercise.
A comparative exercise
A good way of showing up the features of talk that are highlighted by a CA transcript is to compare such a transcript with one produced in a different way. On pp. 86-7 we reproduce two transcriptions of the same stretch of talk, in which two middle-aged, middle-class English women are talking on the telephone. Transcript (A) aims to present the words that were spoken, and some non-verbal activities such as laughter, in a standard orthography which makes the text look like a script for a play. Transcript (B) shows how the CA conventions introduced above are used to transcribe the same stretch of talk.
86
Practices
A Comparison between Two Treatments of the Same Conversation
Transcript (A) [Holt:Xmas 85]
1	L:
2	J:
3	L:
4	J:
5	L:
6	
7	J:
8	L:
9	J:
10	L:
11	J:
12	L:
13	
14	
15	
16	
17	
18	L:
19	J:
20	
21	L:
22	
23	
24	Be
25	J:
26	L:
27	J:
28	L:
29	J:
30	L:
31	J:
Are you not feeling very well? No I'm all right. Yes.
Yes I'm all right.
Oh. You know I... I'm broiling about something.
(Laughs)
What.
Well that sale. At the vicarage. Oh yes.
Your friend and mine was there. Mister 'R'. Oh yes.
And urn, we really didn't have a lot of change that day because we'd been to Bath and we'd been Christmas shopping, but we thought we'd better go along to the sale and do what we could. We hadn't got a lot of ready cash to spend. (Pause)
In any case we thought things were very expensive.
Oh did you?
(Pause)
And we were looking round the stalls and poking about, and he came up to me and he said, 'Oh hello Lesley, still trying to buy something for nothing!' h: (Sharp intake of breath) Ooo Lesley! Ooo! (Laughs) Isn't he... What do you say? Oh isn't he dreadful. Yes.
What an awful man.
Data and Transcription Techniques
87
Transcript (B) [HolCXmas 85]
1   L:   Are you not feeling very rwe.ll,
No I'm all ri:ght
(■) Yes. (0.6)
'Ye:s I'm all right,'
"Oh":.°~.hh Yi-m- You know I-1- I'm broiling about something hhhhehffieh .hhhh
tWha::t.
Well that sa:le. (0.2) At- at (.) the vicarage. (0.6) Oh yeps,
(0.6) '
u (.) ihYour friend 'n mime was the.re (0.2) ~ (hrh hh)
'mMister:, R:, Oh y(h)es, *(hm hm)' (D.4) -
And em: .p we (.) really didn't have a lot'v chamge that (.) day becuz we'd been to Bath 'n we'd been: Christmas shopping, (0.5) but we thought we'd better go along t'th' sale 'n do what we could, (0.2) we hadn't got a lot (.) of s:e- ready cash t'spe:nd. (0.6)
In any^ case we thought th' things were very expensive. Oh did you. (0.9)-
AND uh we were looking round the sta:lls 'n poking about 'n he came up t' me 'n he said Oh: hhello Lesley, (.)"still trying to buy somethingTnotning, .tch! .hhrhaTmhhhhh! Uihoohhhh!
(0.8)
Oor:::   r:L e s 1 e yi 'OO:! lehh heh heh I (0.2)
I:srn   'ti [he
'What' do yjou sa:y. (0.3)
Oh isn't he drea:dful. °eYe::s.' ~ (03)
What'n aw::ful ma:::n
2	J:
3	
4	J:
5	
6	L:
7	
8	J:
9	L.
10	
11	J:
12	L:
13	
14	J:
15	L:
16	
17	L:
18	
19	J:
20	L:
21	J:
22	
23	L:
24	
25	
26	
27	
28	
29	L:
30	
31	J:
32	
33	L:
34	
35	
36	
37	J:
38	
39	J:
40	L:
41	
42	J:
43	L:
44	
45	J:
46	L:
47	
48	J:
88
Practices
Data and Transcription Techniques
89
The first thing that will be noticed about these two renditions is the sheer amount of detail that the CA transcript shows up which is absent from transcript (A). For instance, the CA transcript includes a large number of pauses, overlaps, word stresses and other production features which the first transcript has edited out or 'cleaned up'.
However, we have emphasized that the transcription system is not just aimed at accuracy of detail. Like all transcription systems, it is designed to highlight analytically relevant features of talk-in-interaction. We can illustrate this point by focusing on a particular section of the interchange transcribed in these two extracts. It will be clear that what happens during this interchange is that L tells a story to J about having been insulted by someone referred to as 'Mister R' (this pseudonym is used on the tape itself, not just in transcription) at the local vicarage sale. The story is introduced at the beginning of the extract as something that L is 'broiling about'; just what she is broiling about, it turns out, is Mister R's comment that L, by rummaging about in the stalls at this charity sale, is 'trying to buy something for nothing' (that is, pick up a bargain).
What we are interested in is what happens immediately after L has delivered the punchline of her story: that is, the sequence which begins in line 24 of transcript (A), and line 36 of transcript (B). In each of these transcripts, what happens here is represented radically differently. In Transcript (A), we find the following:
22 L:   and he came up to me and he said, 'Oh hello Lesley,
23 still trying to buy something for nothing!'
24 Both: (Sharp intake of breath)
25 J:   Ooo Lesley!
26 L:   Ooo! (Laughs)
27 J:   Isn't he...
28 L:   What do you say?
29 J:   Oh isn't he dreadful.
While transcript (B) gives us the following:
34 L:   about 'n he came up t' me 'n he said Oh: hhello
35 Lesley, (.) still trying to buy something f'nofFiing,
36 .tch! .hhihaEHhhhhh!
37 J: Lhhoohhhh!
38 (0.8)
39 J:   Oor: : :t: L e s 1 e y ]
40 L:       l00:.'lehh"h"eh herJ
41
42
43
44
45
J: L:
J:
(0.2)
I:srn   'ti rhe
iWhatl do y'ou sa:y. (0.3)
Oh isn't he drea:dful.
There are two immediately noticeable differences. The first is that what is discursively represented in transcript (A) (line 24) as: 'Both: (Sharp intake of breath)' is represented typographically in transcript (B):
36 L:   .tch! ,hh [hahhhhhhh!
37 J: lhhoohhhh!
The second difference is that transcript (B) shows up how the reactions to the story do not occur simply as a series of utterances following one after another, as suggested by transcript (A). Rather, they occur in three couples, separated by short pauses. Moreover, each of the coupled reactions occurs in overlap with its partner.
What is the relevance of this? First of all, we have mentioned that CA transcripts have moved from representing laughter (and other non-lexical phenomena) discursively to developing distinctive typographic representations. What occurs immediately after L's story in this interchange is not laughter but something similarly non-lexical: emphasized, open-mouthed in-breaths which are conventionalized expressions of moral indignation (and so are quite well fitted to the character of the punchline).
However, it is not clear that on reading transcript (B) alone (that is, without the tape) we would be able to interpret these vocalizations in the way they were produced. This appears to be one advantage that transcript (A) has over the CA transcript. In describing what happens as a 'sharp intake of breath', it informs us in a way that transcript (B)'s line of h's does not. But this only serves to emphasize the importance, for CA, of the original data tape, as we stressed earlier on in this chapter. Without the tape it is difficult to 'hear' what is going on at this point in transcript (B). With the tape, however, that problem disappears. So transcript (A) only has an advantage over transcript (B) on the assumption that the analyst does not have access to the tape, which is not the case in CA research.
At the same time, transcript (B) allows us to analyse the
90
Practices
interactional production of this stretch of talk in ways which are not possible with transcript (A). The fact that we can see the set of coupled reactions each being done in overlap brings out the collaborative nature of these expressions of indignation. That is, it is not just that both speakers react with sharp intakes of breath, followed by indignant 'Ooo!'s, but they do so almost simultaneously. By this means, the story recipient, J, displays for the storyteller the fact that she understands and empathizes with L's sense of being insulted by Mister R. She does this by vocalizing, on two closely timed occasions, precisely the indignant reaction which L's punchline requires.
The fact that their reactions are not exactly simultaneous is itself important in this respect. Looking closer still, we find that on the first occasion, L, as the storyteller, is momentarily the first to embark on a response:
36-* L:   .tch! ■hhrhahhhhhhh!
37 J: Ihhoohhhh!
On the next occasion, however, J takes over the first starter role:
38 (0.8)
39-» J: Oor: : :r: L e s I e yi 40    L:       lQO:!lehhlieh hehi
The point is that not only are these reactions closely coordinated; there is even more delicate interactional work going on which is revealed in the timing. Remember that L had set the story up as something she is 'broiling about'. One question that J faces as the story recipient is what form of response would be appropriate to Mister R's comment. This issue is resolved when L herself embarks on a response by manifesting the feeling she had at the time. Having heard the beginning of this sharp intake of breath, J is able to join in (line 37), in a way treating L's response as a 'cue' to show how the story should be treated. After the pause in line 38, J is able to take over the leading response role without waiting for a further cue from the teller. This in turn can then be treated as a cue by L, with which she can coordinate a matched response.
By these means the two women display for one another that they are 'with' each other on this tale. But there is evidence to suggest that this does not just happen by accident. They are, it
Data and Transcription Techniques
91
seems, actively coordinating their actions by cueing each other, so that their talk is brought off as closely matched both in timing and in content. This sort of matching is likely to continue only for a short time without becoming a joke (interestingly, notice that L breaks into laughter as early as the second 'round': line 40). Consequently, after the next short pause in line 41, they both go off on separate paths:
41 (0.2)
42- » J:    I:sm   'ti rhe
43- > L:      What) do yiou sa:y.
44 (0.3)
45 J:   Oh isn't he drea:dful.
These fine-grained observations about the interactional production of this sequence are only made possible because of the CA transcript's focus on features of detail such as overlap. At the same time, we only notice the overlapping production of the in-breaths following the punchline because the CA transcript favours the typographic representation of sounds over their discursive representation. In short, it is clear that the more complex CA transcript gives us access to the interactional management of talk at a much deeper analytic level than the 'cleaned up' version.
In the way that transcripts are laid out on the page, as well as in the kinds of phenomena that are represented, we find a reflection of CA's distinctive perspective on talk-in-interaction. It has not been our aim in this chapter to discuss the relative merits of the various transcription systems that currently exist, nor to compare their different theoretical underpinnings. But we have emphasized that different research interests will require different selections from the vast range of features of talk-in-interaction that can possibly be transcribed. CA has a specific set of research interests, and the transcription system is extremely well fitted to the associated requirements.
We have also stressed the intrinsic relationship between the processes of transcription and the analysis of data. The development of the transcription conventions themselves testifies to that relationship. The system was not just invented in the abstract, but evolved as analysts sought to understand new features of interactions recorded on tape, leading to the development of new means of transcribing these features.
Overall, this suggests that the CA transcription system is not a
92
Practices
finished object, but one that may develop and evolve as new analytic themes themselves emerge. Transcripts, too, develop and evolve. They are not intended as 'objective' representations of social reality, as Graddol, Cheshire and Swann (1994:181) somewhat critically propose. Transcripts are necessarily impressionistic: they represent the analyst/transcriber's hearing of what is on the tape. And of course, that hearing may alter. Repeated listening to tapes almost always throws up phenomena which were simply missed the first time round. So as a piece of data is subjected to closer and closer analysis, the transcript itself evolves as part of that analytic process. This again illustrates the close connection between data, transcription and analysis.
A final point is that transcripts play a key role in the claim of CA to be a rigorous empirical discipline. An important aspect of this is that analyses produced by one researcher do not amount merely to idiosyncratic and untestable assertions about what is going on in a stretch of talk. Rather, the analysis is projected into a public arena in which it can, if necessary, be challenged and even altered. This is made possible not only by the fact that publications in CA routinely contain examples of data transcripts (as, indeed, we do throughout this book); but also by the fact that conversation analysts' transcripts (and, ideally, the data they are transcriptions of) are made publicly available to anyone who requests them in order to test the accuracy of the analysis, or to reanalyse the data. By this means, transcripts are central to guaranteeing the cumulative and publicly verifiable nature of conversation analytic research.
CHAPTER 4
Analysing Phenomena I
Building a Collection
Having introduced the transcription procedures of conversation analysis, we begin in this chapter to look at some of the basic techniques with which researchers approach the analysis of their data. Conversation analysts place great emphasis on building 'collections' of instances of a particular conversational phenomenon. The aim is to produce analyses of patterns in the sequential organization of talk-in-interaction. Analysing patterns in this way enables the analyst to make robust claims about the 'strategic' uses of conversational sequences: the ways in which culturally available resources may be methodically used to accomplish mutually recognizable interactional tasks.
In this chapter we look in detail at the principal analytic techniques which are used in building collections. Using case studies of three pieces of CA research, ranging in time between an early 1968 study and a more recent 1992 paper, we focus on techniques for the identification, description and analysis of singular conversational devices, which can be shown to function in robustly patterned ways within talk-in-interaction. In different ways, these case studies allow us to illustrate the steps by which an analysis of such patterned phenomena can be built up.
However, it is important to emphasize that while this chapter is concerned with analytic procedures and techniques, we do not aim to provide a set of guidelines for doing CA of a 'manual' type. The reason for this is simple. Although as chapter 3 showed, CA has adopted a relatively strict and systematic style of data transcription, the techniques with which researchers approach the actual analysis of data rely as much on what Schenkein (1978) described as the 'conversation analytic mentality' as on any formal rules of research method. Like most forms