When we transcribe conversations, we take two things into account and try to join them: On the one hand, we have to be as accurate as possible and transcribe everything that could be important. On the other hand, our transcriptions also have to be readable. Our choices are affected by the conversation analytical research tradition, which samtalegrammatik.dk is also born from. One of the founders of this tradition, Gail Jefferson, has invented a transcription system that strives for accuracy in some areas, and is less accurate in others.
When we investigate Danish talk in interaction, we don’t just look at the transcriptions. We also listen and look at the original data (for more, go to Data). But the transcriptions still affect what we will notice and therefore they have an impact on our investigations and results. We will therefore make a short summary of the principles behind Jefferson’s (and our) system. You can see the actual transcription system under Transcription system, which is to be found in the navigation column to the left.
The principles are:
- Nothing that can be heard is ignored. Sentences that are restarted, incomplete utterances, clearing of the throat, coughing, laughing, and other sounds which from a traditional view are not linguistic, are included in the transcription. The reason for this is that these non-written phenomena have shown to have many functions in talk in interaction.
- Much accuracy is important in the course of time of the conversation. Transcription signs are invented to precisely indicate pauses, overlapping talk, elongations, and changes in speed.
- Intonation, stress, and specific voice qualities are being depicted. Therefore, indications of (different degrees of) stress, course of intonation (the melody), creaky voice, smiley voice, and so on are indicated. These indications are temporarily not as detailed as they could be. As an example, movement in intonation is normally only indicated at the end of an utterance. This is because the transcription still has to be readable. In some investigations, we zoom in on sound features like these and they are in this context shown with greater accuracy (e.g. by showing pictures of the psychical sound made in a phonetics program).
- Phonetic and phonemic writing is not used. Instead, we depict the words that are being said with a modified version of normal orthography. The modification is specific to words which are pronounced in distinctively different ways in our data. As an example, “hvad” (what) can in the spoken language be pronounced in clearly different ways, as hva [væ], hvad [væð] og hvar [vɑ]. These distinctions are used in our own corpus, AULing (for more, go to Data) and have been used now for a longer period of time. This has revealed that is it likely that “hvad” is not just one word, but three different words in the spoken language. The reason that we don’t use phonetic or phonemic writing in transcriptions is that it would be too time consuming to produce and it would be incomprehensible to other people. In the treatment of specific phenomena we sometimes used phonetic writing, and when we do, it is a rough version of IPA (the International Phonetic Alphabet). This is the sort of writing we also see in Den Danske Ordbog (the Danish dictionary).
- Since we don’t use phonetic writing, it also means that we don’t systematically indicate differences in pronunciation that comes from people speaking different variants of Danish (for more, go to Variation in spoken language).
Further reading
Jefferson (2004) is Gail Jefferson’s own review of the transcription system she has invented.
Steensig (2005) goes through the conversation analytic transcription system and compares it to other systems.