3 Outline of Discourse Transcription
John W. Du Bois
University of California at Santa Barbara
Stephan Schuetze-Coburn
University of California at Los Angeles
Susanna Cumming
University of Colorado at Boulder
Danae Paolino
University of California at Santa Barbara
This article presents a set of basic categories, symbols, and conventions for
discourse transcription. Discourse transcription can be defined as the process
of creating a written representation of a speech event so as to make it
accessible to discourse research. In the following pages, we present in outline
a framework for carrying out such a discourse transcription. (For a discussion
of the principles that underlie the design of transcription systems, see Du Bois,
1991, and Edwards, this volume; for an indepth treatment of the whole
Process of discourse transcription, see Du Bois, Schuetze-Cobum, Cumming,
& paolino, 199 1 .)
For each symbol in this discoucse @anscription system, ,a brief explanation
of usage is given, illustrated with an example drawn from conversational
transcriptions. Where appropriate, we comment on why the discourse f e a m
in question should be attended to.
A word about the examples is in order. All examples cited are drawn from
actual conversations, which have been transcribed and checked by the
autb0rs.l The examples are given in a fairly broad transaiption, which
46 DU BOlS, SCHUETZE-COBURN, CUMMING, AND PAOUNO
includes the most basic transcription information: the words and who they
were spoken by, the division of the stream of speech into turns and intonation
units, the truncation of intonation units and words, intonation contours,
medium and long pauses, laughter, and uncertain hearings or indecipherable
words. In a more detailed, "narrow" transcription, the transcriber would also
include notation of, among other things, accent, tone, prosodic lengtbening,
and breathing and other vocal noises. Such narrow transcription features have
been omitted from most of the transcriptions in this chapter, so as not to
overload the reader with too many new symbols at once; but for those sections
that deal directly with the transcription of these features, the examples will
include them. (Narrowly transcribed versions of most of the examples cited
here are presented in Du Bois et al., 1991.) For brevity's sake, the examples
represent for the most part short stretches of discourse without a great deal of
textual context (cotext). Thus the portions cited are not always whole sentences
or whole interactions, but each line that is cited is whole-that is, each
intonation unit is presented in its entirety-and no omissions have been made
within the stretch of transcription that is cited. Whenever a new notational
convention is being introduced, the symbol is written in boldface letters in the
illustrative examples for that section, in order to highlight the feature in
question.
1. UNITS
One of the most striking, if elusive, features of conversation is its division into
recognizable units at various levels. Any discourse transcription should
indicate at least the most fundamental of these. This section presents symbols
for boundaries between units of various kinds, including the intonation unit and
the word unit, as well as truncated (uncompleted) variants of these units. (The
turn, which is a fundamental unit of conversational discourse, is treated later in
conjunction with the speaker identification label.)
1 . I . Intonation Unit
A carriage return is used to indicate the end of an intonation unit (in effect, the
boundary between two intonation units). Thus, each intonation unit appears on
a separate line.2
2 In tWnerPI, a speaker's intonation unit should not be broken up into two lines; but for the
~~ ktonation unit that is so long that it will not fit on one line, see the section on
P-nuon. Also. it may sometimes be necessary to b d an intonation unit using the "intonation
-tinued'' s~mbgl (described in the section on specialized notations).
Roughly speaking, an intonation unit is a stretch of speech uttered under a
single coherent intonation contow. It tends to be marked by cues such as a
pause and a shift upward in overall pitch level at its beginning, and a
lengthening of its final syllable. For a fuller discussion of intonation units, the
cues that mark them, and the methods for identifying them, see Du Bois et al.
(1991), Chafe (tbis volume), and C~ttenden (1986, pp. 35-45).
(1) ((Aesthetics))
S: That's interesting,
I mean,
th- that you should pair the word aesthetics,
. .. with advertising.
J: Yeah.
1.2. Truncated Intonation U n l
A double hyphen (--) indicates that the speaker breaks off the intonation unit
before completing its projected contour.
This truncation occurs primarily in cases where a speaker utters the initial
portion of a projected intonation unit but abandons it before finishing-that is,
a false start. The double hyphen is not intended to represent the case of a unit
that appears incomplete when measured against the canons of normative
grammar. Intonation units that do not constitute complete clauses are
Commonplace and usually quite normal-and "complete" as intonation units.
For example, conjunctions (and) and particles (well) frequently appear as
Complete intonation units marked with a comma at the end, which signals
$6 Continuing" intonation (see below)-a kind of incompleteness, if you will,
but one that is distinct in principle from the truncation signaled by double
hyphen. The unit marked with a comma typically constitutes (apparently) all
fbat the speaker projected to say within the current unit, whereas in the unit
marked with a double hyphen the speaker projected to say more witbin the
current unit but abandoned some portion of the projected utterance.
T~ca t ion is thus measured not against normative notions of clause
mmpleteness but against the speaker's presumed projection for tbe current
intonation unit.
(2) ((Ranch))
R: He doesn't have any --
... He doesn't bow what's going on in this world.
48 DU SOIS, SCHUETZE-COBURN, PAOUNO, CUMMlNG
Note that for every intonation unit that is complete, the line should end with
some representation of its intonation contour (see below). If an intonation unit
does not so end, it will in general have the double hyphen that marks
truncation.
1.3. Word
The space character is used to separate words, as in normal orthographic
convention. A space also separates certain other notations, such as those for
laughter, pause, inhalation, and so forth.
(3) ((Aesthetics))
S: Hm.
Hm.
... Okay.
1.4. Truncated Word
A single hyphen (-) indicates where the speaker has truncated a word, leaving
the end of the (projected) word ~nuttered.~
Truncation is often cued overtly via word-final glottal constriction, but not
always-either plienomenon may occur independently of the other. Other
truncation cues may include segment shortening, slight rhythrmc
discontinuities, and so on. (Where it is deemed relevant, the precise
pronunciation of the truncated word can be written using phonetic notation.)
(4) ((Friends))
J: ... You know how they do that,
so you can't s- ha- --
you don't have any balance.
This symbol is not used to mark words that have been pronounced in an
abbreviated fashion as part of an informal speech style. Truncation of a word
is measured not against canons of "nonnal" or "standard" pronunciation but
3. OUTLINE OF DISCOURSE TRANSCRIPTION 49
against the speaker's projected pronunciation for the current word. Only when
a speaker projects pronunciation of a word and then fails to complete that
projected pronunciation is the phenomenon of word truncation involved.
2. SPEAKERS
2.1. Speaker ldentlication and Turn Beginning
To identify the speaker of a given turn in the conversation, a code or a proper
name (written all in capital letters) is inserted at the beginning of the turn.
followed immediately by a colon (:).
(5) ((Door))
A: Now that we have the [side door] fixed
B: m t ' s kind of] --
A: he could.
B: Yeah,
C: Yeah.
D: ... Sure.
Although transcribers often assign prosaic codes such as "A" or "B" to
their speakers, the reader tends to get a more vivid impression of who the
participants are if their utterances are tagged with personal names, which are
always more memorable. The name should in general be a pseudonym, since
in any transcription destined for public presentation, privacy considerations
would ordinarily preclu& use of the speakers' actual names. The choice of
Wries becomes especially important if speakers use names to refer to each
other during the course of a conversation, in which case, the pseudonym in the
speaker identification label should match that in the spoken reference, so that
relevant persons-whether they are spealung or merely spoken about-are
clearly distinguished. If possible, pseudonyms should retain some flavor of the
actual names.
(6) ((Aesthetics))
JEFF: That's all it does.
It doesn't [even] reach a conclusion.
SARAH: [mhml,
JEFF: The conclusion is up to you.
SARAIf: mhm,
JEFF: in going out to --
... to buy the thing.
50 DU 801s. SCHUETZE-COBURN, CUMMING, AND PAOUNO
SARAH: Hm.
Hm.
... Okay.
When it is unclear which of several speakers on a tape is responsible for a
particular utterance or noise, the letter X is used to label the unidentified
speaker.
(7) ((Friends))
X: ((BLOWS WHISTLE))
Note that the stretch of speech between two different speaker labels
constitutes, roughly speaking, the discourse unit known as a turn. The picture
is somewhat complicated, however, by the listener's interjection of
continuative backchannel responses (mhm, yeah, etc.) into a speaker's
extended turn. Although a backchannel response must for clarity's sake bear a
speaker label, as must the two sections of the turn it occurs within, one does
not want to be misled by this practical consideration into overlooking the
essential continuity of the extended turn unit across such fleeting interjections.
2.2. Speech Overlap
Square brackets are used to indicate the beginning (left bracket) and the ending
(right bracket) of overlap between the utterances of two speakers. One set of
brackets is inserted surrounding the fust speaker's overlapping utterance
portion, and a second set of brackets surrounds the second speaker's
overlapping portion. This notation signals that the two bracketed utterance
portions were uttered at the same time.
For the sake of reading clarity, the second speaker's left bracket is aligned
vertically under the fust speaker's left bracket (by insating as many spaces as
needed). This alignment of space on the page helps to give an iconic sense of
the temporal alignment of the two overlapping utterances. Note that only the
left bracket need be aligned vertically; for reasons of clarity and practicality it
is not advisable to force the right bracket to do so.
...
R: [Because he wanted the stamps],
B: [all those sta~nps],
... Mom let Ted Kemer have.
3. OUTUNE OF DISCOURSE TRANSCRIPTION 51
Wherever several overlaps occur in rapid succession within a short stretch
of speech, distinctive combinations of brackets (e.g., single brackets [ I vs.
double brackets [[ 11) may be needed to make clear what is overlapp'ing with
what. This will be necessary whenever two distinct cases of overlap occur
without at least one line of nonoverlapped text between them, because if only
one kind of bracket were used (e.g., just single brackets) the reader could be
misled to think that the fust and second bracketed portions (marked with single
brackets) were simultaneous with the third and fourth bracketed portions (if
also marked with single rather than double brackets). After one full line of
speech containing no overlaps-when there is no longer danger of confusion-
the use of double brackets can be dropped and single brackets resumed. If
distinctive bracketing is needed again later in the same text, the single and
double brackets should be used in alternation. Whenever no other overlaps
occur nearby, it is best to use just the single brackets alone.
(9) ((Aesthetics))
J: [Yeah].
S: IWhich] colors ... all of the communication,
[[after]] that.
J: [[Yeah]].
When there are many overlaps in very close succession, it may occasionally
be necessary to use more than two kinds of distinctive brackets. For example,
hiple brackets ([[[ I]]) or brackets indexed with numbers ([3 31) can be used to
Create distinctive bracketing?
(10) ((Dinner))
B: Nobody wants [to leave].
A: [They don't] move [[out]].
S: [[Berkeley]] just keeps [3 getting 31 bigger and [4 bigger 41.
B: [3 yeah 31,
14 Yeah 41,
... Well it's amazing to me.
Occasionally, it may be useful to employ distinctively marked (e.g.,
Qubled) brackets even when no other overlaps occur nearby, in order to help
reader follow a complicated conversational exchange.
I
-
4 1 Ibe 1 is avoided because it is easily mistalcen for the lowercase Ietta 1 or even the crpit.l I
" WY t ~ f l c e ~ ; nume& 2 through 9 engenda no such confusion.
52 DU BOIS. SCHUETZE-COBURN, CUMMINQ, AND PAOUNO
(1 1) ((Hypochondria))
G: ... Well,
the worst [[thing I ever had,
K: [[He's a medical miracle]].
G: was brain]] fever,
when I had proposed to her.
K: 8888
. . . From which you haven't recovered.
Often enough, the second speaker in an overlap begins to speak in the
middle of a word being uttered by the fmt speaker. In such cases, it is useful
to keep mck of precisely where the overlap begins, because this may carry
significant information about how the speakers are responding to each other in
"real time" (Schegloff, personal communication). To indicate this, the bracket
is placed within the word at a point corresponding to the overlap. (When a
bracket is written inside a word, no space should be inserted-whether the
bracket is indexed or not-since any space would break up the word and cause
it to appear as two separate words.)
(1 2) ((Hypochondria))
G: ... Then I had,
h,
K: Cytomegalo[virus],
G: [Don't] forget,
cytomegalo [[virus]],
K: [[@I1
D: [[What is that]].
3. OUTLINE OF DISCOURSE TRANSCRIPTION 53
3. TRANSITIONAL CONTINUITY
-I
In speech, important information is carried in the speaker's intonation,
encompassing fluctuations of pitch and other cues. Although a discourse
transcription can never capture a complete representation of the infinite variety
of possible intonation contours, it can nonetheless provide a useful
nwesentation of at least the more critical intonational information by
distinguishing broad classes of contours. It is useful to distinguish here
bemeenfuncrional and phonetic analyses of intonation, each of which has its
place in discmuse transcription. The symbols in this section deal with the
functional of intonation, whereas a set of symbols that address the
phonetic analysis of intonation is introduced in the section on terminal pitch
(see also the section on tone) .5
m e system of categories presented in this section seeks to identify in
general terms one aspect of intonational function, that of marking transitio~l
continuity. When a speaker arrives at the end of an intonation unit, poised to
continue on to the next-or not continue-the intonation contour usually gives
a fairly clear indication of whether the discowse business at hand will be
continued or has finished. This is transitional continuity: the marking of the
degree of continuity that occurs at the transition point between one intonation
unit and the next. The scope of the continuity-the question of what it is that
is being continued or finished-is open-ended: A "final" contour may apply to
the end of a sentence, the end of a turn, or the end of some other discowse
unit. Whereas it may be possible to make finer discriminations in transitional
continuity within the broad class of contours covered by each transitional
continuity symbol, the distinctions among "final," "continuing," and
"appeal" (see below), at least, seem to be basic.
Although the intonation contour classes in this set are defined in terms of
their function, each category is more or less consistently realized by a specific
form: a specific phonetic contour, or a set of contours (in effect, intonational
allomorphs), where each member of the set is determined by its context. The
range of phonetic realizations for a given transitional continuity class differs
somewhat from one language to the next, which is one reason for using
functionally based categories: They help to ensure that similar intonational
functions are written similarly across languages, facilitating comparison even
where phonetic realizations differ. Preliminary observations in a limited
number of languages suggest that, remarkably, all languages are likely to make
intonational distinctions between the transitional continuity classes presented
in this section, though their phonetic realizations may vary.
The symbols used to represent transitional continuity here are drawn from
those employed in written punctuation. Although using commas and periods in
tbat are reminiscent of their function in written language does make it
"ier to remember them, it also means that the transcriber must guard against
into habits of thought associated with written punctuation. In
discourse transcription as presented here, the punctuation symbols comma,
pa and question mark always represent intonation classes and never
-tical or semantic structure per se.
-
sk3 nsearchers who wish to pursue the representation of intonation in discourse further. the wok
Of*-~uhlen (1986). haenden (1986). Crystal (1975), Cutler and Ladd (1983).
(1982* this volume), McLemore (1991). S v ~ v i k (1990). SvMvik and Quirk (1980). and 0 t h ~ ~
be Consulted (see Caper-Kuhlen. 1986. and Cruttenden. 1986. for additional r e f m n ~ ) .
notion of intonation unit, see Chafe (this volume) and Du Bois et d. (1991).
3.1. Final
The period (.I indicates a class of intonation contours whose transitional
continuity is regularly understood as final in a given language. For English
and many other languages, this means primarily (but not exclusively) a fall to a
low pitch at the end of an intonation unit. It is important to recall that, because
this symbol represents an intonational category rather than a syntactic one, it
can appear in places other than the end of a sentence. Conversely, it need not
appear at the end of every (normative) sentence.
( 13) ((Depression))
R: For what.
B: ... They make rope of i t
3.2. Continuing
The comma 6 ) indicates a class of intonation contours whose transitional
continuity is regularly understood as continuing, in a given language. The
contour is often realized in English as a slight rise in pitch at the end of an
intonation unit (beginning from a low or mid level), but it may have other
realizations as well, each of which presumably has slightly different pragmatic
implications. One type of continuing contour is realized by a terminal pitch
that remains level; another, by a terminal pitch that falls slightly, but not low
enough to be considered final.6
(14) ((Ranch))
R: If you think about it,
yeah,
if it rains a lot,
the horse is always wet,
and it's always moist,
it's always on something moist,
... Sure it's going to be softer.
3.3. Appeal
The question mark (?) indicates a class of intonation contours whose
traasitional continuity is regularly understood as an appeal, in a given
language. (For English, this is often realized by a marked high rise in pitch at
6 ~ h e significance of the end point of pitch movement is well known (Coupes-Kuhlen, 1986, pp. 88-
90). As one intonation specialist notes, "The lower the end point, the gcutcr the degn