Verbatim transcription: rules, examples and when to use it

Verbatim transcription means writing down a recording word for word, exactly as it was spoken, instead of summarising or paraphrasing it. A verbatim transcript of an interview keeps repetitions, false starts and unfinished sentences, and shows who said what. How much else it records, from filler words and pauses to laughter and overlapping talk, depends on the conventions you choose.

What is verbatim transcription?

In qualitative research, a verbatim transcript is the written record you analyse and quote from. It is never a complete copy of the conversation. Tone of voice, timing, gestures and the atmosphere in the room are lost unless you mark them, and every transcriber makes decisions about where a sentence ends or how to spell a half-swallowed word. Blake Poland (1995) pointed out that transcripts are often assumed to be verbatim rather than checked, and argued that reviewing their quality should be a routine part of rigorous research.

There is no single correct format. McLellan, MacQueen and Neidig (2003) note that no transcription format suits every approach, setting or theoretical framework, and set out practical considerations for preparing transcripts systematically. If your handbook simply asks for ‘verbatim transcripts’, ask your supervisor which features to include, and write your rules down before you start so that every interview is transcribed the same way.

True verbatim, intelligent verbatim and edited transcripts

Transcription services usually offer three levels of detail. The names come from the transcription industry rather than from methods textbooks, but students meet them when they compare tools and services.

Methods texts tend to use other words for the same choice. Braun and Clarke (2006) ask for a ‘rigorous and thorough “orthographic” transcript’ for thematic analysis: a verbatim account of all verbal and sometimes nonverbal utterances, such as coughs. Intelligent verbatim leaves some of those utterances out, so if you use it for thematic analysis, agree it with your supervisor first and say so in your methodology. Conversation analysts use far more detailed notation, covered below. In German-speaking countries, many departments refer to Thorsten Dresing and Thorsten Pehl’s rules in Praxisbuch Interview, Transkription & Analyse (9th edition 2024), which distinguish simple from extended transcription.

Whatever the style, never change what someone meant. Be especially careful when participants speak English as a second language: tidying their grammar can put words in their mouths, so keep their wording and correct only your own typing errors.

Verbatim transcription rules, step by step

These are the decisions most transcription keys cover. The symbols vary between conventions; what matters is that you choose one form for each feature and use it in every transcript.

1. 1

   Speaker labels

   Start each turn on a new line with a label, such as INT for the interviewer and P3 or a pseudonym for the participant. Keep your own questions and comments: they are part of the data.
2. 2

   Timestamps

   Add a time at each change of speaker or at regular intervals, for example [00:12:41], so you can find a passage in the recording again. For citing, add line numbers as well.
3. 3

   Pauses

   Mark noticeable pauses. Poland’s suggested conventions, in the revised version of his article in Gubrium and Holstein’s Handbook of Interview Research (2002), use (pause) for a break of two to three seconds and (long pause) for four seconds or more; others write (3) for the number of seconds.
4. 4

   Overlaps and interruptions

   Show where one speaker cuts in on another, for example with a note such as (overlapping) or square brackets around the overlapping words. Mark a word broken off mid-way with a hyphen, as in ‘we were go- going’. Where both voices are unintelligible, services write [crosstalk]. If your key uses square brackets for overlaps, give your own notes, such as inaudible passages, a different mark.
5. 5

   Inaudible passages

   Write [inaudible] with a timestamp, for example [inaudible 00:12:41], so that you or your supervisor can listen again. A best guess goes in brackets with a question mark: [mentor?].
6. 6

   Laughter, other sounds and emphasis

   Note laughter, sighs and coughs where they happen, such as [laughs] or [sighs]. Poland suggests capitals rather than bold or underlining for strong emphasis, because formatting can be lost when files are imported into analysis software.
7. 7

   Anonymisation

   Replace names, places, employers and other identifying details with bracketed descriptions such as [name of university] or with pseudonyms, and use the same replacement every time. Keep the key to the replacements separate from the transcripts.

A pseudonymised transcript is still personal data as long as you or anyone else can link it back to a person, and the recording is never anonymous, because a voice identifies the speaker.

Jefferson notation for conversation analysis

Conversation analysis studies how talk is organised from moment to moment, so it records how things are said as well as what is said. Gail Jefferson’s system, set out in her Glossary of transcript symbols with an introduction (2004), is the standard. It marks features such as:

• (0.6) for a pause timed in tenths of a second, and (.) for a micropause;
• [ and ] where overlapping talk starts and ends, and = for talk that follows with no gap;
• ↑ and ↓ before a marked rise or fall in pitch;
• CAPITALS for talk that is louder than the talk around it;
• >faster< and <slower> talk;
• .hhh for an in-breath and hhh for an out-breath;
• ( ) for talk the transcriber could not hear, and (word) for a doubtful hearing.

A Jefferson transcript takes many times longer than a verbatim one, and an automatic draft gives you, at most, the words. Use it if your method is conversation analysis or discursive psychology, and consider transcribing only the extracts you analyse closely in that detail. Linguists in German-speaking countries often use GAT 2 (Selting et al., 2009) instead.

Which transcription style does your method need?

Choose the level of detail from your research question and method, not from what is quickest. As a rule of thumb:

• Reflexive thematic analysis, qualitative content analysis and framework analysis: a verbatim transcript of what everyone in the interview said, including false starts and filler words, which is Braun and Clarke’s minimum for thematic analysis. Some supervisors accept intelligent verbatim, especially for content or framework analysis. Mark laughter or a long hesitation wherever it changes the meaning of a passage.
• Narrative analysis: verbatim, keeping each story whole. Pauses, repetitions and the listener’s responses can matter for how a story is structured and told.
• Discourse analysis: verbatim for critical and Foucauldian approaches; Jefferson notation for discursive psychology. The guide to discourse analysis explains the difference.
• Conversation analysis: Jefferson notation throughout the extracts you analyse.

If you are unsure, transcribe in true verbatim. Removing filler words later is easy; adding them back means listening to the whole recording again.

Invented example: one excerpt in two styles

The excerpt below is invented. It comes from an interview for a master’s dissertation on how international students experience group work. The key picks one form for each feature in the steps above: square brackets mark everything that is not a spoken word, so a pause is written [pause].

The clean version keeps the content and the laughter, which signals that P3 is being ironic about her own decision. It is enough for a content analysis of group work, or for a thematic analysis if your supervisor accepts intelligent verbatim. The true verbatim version also shows the hesitation before ‘okay’, the self-correction in ‘we didn’t really, we didn’t have a plan’ and the ‘er’ and restart before she reports the accusation. For a reading of how P3 manages blame, those details are the data.

Checking an automatic transcript

An automatic transcript is a draft, not a verbatim transcript. It becomes one only after you have checked every line against the recording. Poland’s examples of human transcription errors show what to listen for: a misplaced full stop that changes the meaning, missing words, and words misheard as others, as when ‘communities are the way to go’ was typed as ‘the way to God’. Speech recognition makes the same kinds of mistakes, especially with names, technical terms, accents and people talking at once. It may also merge two speakers or drop short responses such as ‘mm’.

Listen to the whole recording while you read, at reduced speed where the talk is fast. Correct the words, check every change of speaker, then add your conventions: pauses, [inaudible], laughter and overlaps. Anonymise last, and read the finished transcript once more.

Spirelight gives you a draft with speaker labels and a timestamp on each speaker turn. Before you start, you choose ‘Cleaned up’ or ‘Word for word’. For English recordings, Word for word keeps filler words and false starts; for other languages, expect to add them yourself. Neither style applies a transcription convention: the draft has no pause marks, [inaudible] notes, laughter or overlaps, and no Jefferson, GAT 2 or Dresing and Pehl notation. You add those while you listen. The Word export puts the speaker and the time range, such as [4:12 - 4:31], above each turn.

Check your university’s rules before you upload recordings anywhere: a voice identifies the speaker, so a recording is personal data even without names. Spirelight offers no data processing agreement; your university, not you, would sign one. Transcription runs on EU servers, files are stored with Google Firebase in the US, and AI features run on OpenAI in the US, which your ethics application and participant information sheet must state.

Where your university allows it, you can transcribe your interviews in the Word for word style, correct the draft against the audio and export it to Word once your conventions are in place.

Describing your transcription in the dissertation

In the methodology chapter, say which style you used and why it fits your method, which conventions you followed, who transcribed the interviews and how you checked them against the recordings. If you used an automatic draft, say so. Put your transcription key in an appendix or before the transcripts, so that readers can interpret every symbol.

When you quote, refer to the participant and the lines, for example (P3, lines 41–44). Word adds line numbers under Layout, Line Numbers. Some dissertations quote in a lightly cleaned form for readability; if you do, say so, and mark any words you leave out with […]. Ask your supervisor whether full transcripts belong in the appendix, and check every extract against the audio before you submit.