Transcription is a time-consuming and often tedious task, but it is a nontrivial one with far-reaching consequences. Transcriptions provide the data that are analyzed by researchers interested in language production. However, they are not comprehensive or direct records of what speakers actually say and mean. It is generally acknowledged that setting up a particular recording situation, most often with the observer present, influences what the participants say and how they say it and, consequently, what can be concluded from the data. Choosing a means of recording speech and interactions-for example, audio or video recorder, stationary or remote microphone-further constrains what can become part of the permanent record. Transcription puts the data one step further from what speakers say and do because transcribers must select which aspects of the speech signal and interactional context to record in print. Should the tran-scriber make a written record of every phonetic detail, pause, gesture, change in direction of gaze, and situational feature to approach a complete and accurate representation? The answer is clearly “no,” for reasons
of-among others-readability, affordability, and perhaps even sanity. Ideally, each transcription should be tailored to the purpose of the investigation for which it constitutes the data. In any case, there is no such thing as a neutral transcription. As Ochs (1979) put it in her aptly named chapter, Transcription as Theory, “transcription is a selective process reflecting theoretical goals and definitions” (p. 44).