When we make transcription choices we need to keep in mind that any transcription convention embodies a theory of what is relevant in interaction (Ochs, 1979; Psathas & Anderson, 1990). For example, the traditional conversation analytic transcript shows sequential speakership exchange of the form:

speaker A --> speaker B --> speaker A, etc.

with their talk represented in the form:

A: .........
B: ........
A: .........
B: ........
etc.

This way of transcribing emphasizes the mutual relevance of the talk produced by the speakers. Overlaps are indicated by spatial positioning of the transcribed talk or by special symbols. However, as Ochs (1979) points out, sequential layout of talk favors an analytic interest in the sequential organization of interaction, such as speakership exchange, what follows what, interruptions, new starts, etc. If, on the other hand, one were interested in the parallel nature of activities, as Ochs was when investigating the side-by-side play of children, a different form of transcript might be more appropriate, such as a representation of activities in side-by-side columns.³⁷

When more than two or three persons are involved in a conversation the sequential format quickly becomes unwieldy. One solution is to assign a separate "speech line" to each participant, with overlapping talk indicated by placement. For example, Eckert (1990) made a transcript from a group session with six high-school girls, each of whom carried a separate mike. Overlaps here are indicated by spatial positioning of the talk (Appendix F). Alternatively, the six girls could be represented in six parallel columns, in which case overlap becomes more difficult to indicate.

While we have at least some established conventions for verbal transcription, the representation of nonverbal phenomena is in its infancy. The comparatively simple transcription scheme for conversation is inadequate for Interaction Analysis in two ways: first, together with the talk the accompanying non-verbal activities need to be transcribed; and second, since we typically deal with multi-party interactions, multiple activity streams have to be represented. These complexities can be dealt with by two potential solutions: parallel horizontal transcripts and parallel columnar transcripts.

Parallel horizontal transcripts: One way to represent parallel activities is by using multiple horizontal lines that represent talk and nonverbal activities, as in a musical score. In many ways, an "orchestral transcript" where co-participants' actions are represented on parallel horizontal lines is ideal since the relationship of the various lines of activity to each other can be grasped in a glance. For example, Heath (1986) has developed a transcription system which augments detailed verbal transcription with gaze and body movement data such as head nods and hand gestures, posture, and leg and hand movements (Appendix G).

Erickson (1991), in his transcript of a dinner table conversation, shows not only eight different speakers' utterances but also the motions of their hands and arms as they transfer serving dishes across the table and place fork to plate and fork to mouth while eating. This is an extremely complex transcribing scheme that in addition to indicating duration and intensity of movements, also transcribes intonation contours by musical notation that gives pitch and rhythm (Appendix H).

Goodwin and Goodwin (in press) use parallel lines of talk lined up with iconographic representations of the objects at which a worker looks, to show the ways in which environmental resources are used in the course of understanding a question and providing an answer (Appendix I).

Parallel columnar transcripts: Alternatively, multiple analytic streams can be represented in side-by-side columns that include both verbal and nonverbal activities, as Jordan did in representing interaction in obstetric labor rooms (Appendix J). Parallel columns can also be used to track parallel activities in complex, technology-rich work settings where multiple activities happen side-by-side or in classrooms where there may be different small groups working around tables at the same time.

Increasingly, it is becoming important to transcribe the activities of non-human actors in work practices and learning activities. Suchman (1987), in a study of human-machine interaction, juxtaposes human activity with machine activity in a four-column transcript that specifies what parts of these activities are available to human operators and to the machine, respectively (Appendix K). The question of how to transcribe screen events and machine data is, at this time, wide open. Luff and Heath (1991), Frohlich et al. (in press), Whalen (1993) and numerous others have begun to develop notations for transcribing screen events, but no stable conventions have as yet crystallized.

Word processing has helped tremendously in making such complex representation schemes possible. Unfortunately it is not yet simple to maintain automatic temporal alignment of the various activity streams between either columnar or orchestral representations nor are they easy to print out in a form that makes them easy to read. However, given the increasing need to represent multiple data streams, for Interaction-Analytic work as well as for other kinds of Exploratory Sequential Data analysis (ESDA), it is likely that such support will become available in the future.³⁸

In summary, transcription practice at the present time is in flux. We predict that, given the lack of convincing arguments for the benefits of any one particular standard, practitioners will continue to make pragmatic decisions about which transcription convention is best for their particular purposes.