Interview Transcription: Verbatim vs Intelligent Verbatim, Rules

Why transcription matters

Recorded speech cannot be coded systematically; text can. Transcription creates the data you will analyze, quote and archive. It is never a neutral copy: you decide what to keep (pauses, laughter, dialect) and what to leave out. Brinkmann and Kvale stress that transcripts are translations from an oral to a written mode, so the rules you choose should fit your research question and analysis method.

Transcribing yourself is also a good way to get to know your data. Many researchers note their first analytic ideas while transcribing.

Transcription styles compared

Choose the style before you start and apply it consistently.

A set of transcription rules

Write down your rules and put them in the appendix. A typical rule set for intelligent verbatim:

• Speaker labels: I for interviewer, P1, P2 … for participants; new paragraph at each change of speaker.
• Fillers (um, uh, er) and stutters are omitted unless they carry meaning.
• Emphasis is not marked unless clearly relevant; laughter and other non-verbal events in round brackets: (laughs), (sighs).
• Pauses longer than about three seconds marked as (pause).
• Unclear words: (unclear) or (inaudible 00:12:34) with a timestamp.
• Dialect and colloquial language are kept, not standardized, unless meaning becomes unclear.
• Timestamps at each speaker change or every few minutes, to find passages in the audio.
• Identifying information replaced with descriptive placeholders in square brackets: [name of university], [colleague 1].

Pseudonymization and anonymization

Interview transcripts almost always contain identifying information: names, places, employers, rare events. Under the UK GDPR and EU GDPR, pseudonymized data (names replaced, but a key exists) is still personal data; only truly anonymized data, where people can no longer be identified by any reasonably likely means, falls outside data protection law. In qualitative research, full anonymization is often hard because of rich context.

• Replace names with codes (P1, P2) or realistic pseudonyms, consistently across all transcripts.
• Generalize details: “a large hospital in Manchester” → “[large urban hospital]”.
• Store the key linking codes and real names separately and securely, and delete it when no longer needed.
• Check quotes before publication: could colleagues recognize the person from the combination of details?
• Follow the retention and deletion periods in your ethics approval and consent form.

Time and tools

Manual transcription typically takes 4–8 hours per hour of audio, depending on audio quality, the number of speakers and the level of detail. For ten one-hour interviews, that is 40–80 hours, a substantial part of your dissertation timeline.

Automated speech recognition can produce a first draft in minutes, but you must correct it line by line against the recording, especially for names, technical terms, accents and overlapping speech. Before uploading recordings to any service, check that it complies with your ethics approval and data protection rules: where is the data processed, is it used for training, and is there a data processing agreement? Many universities provide approved tools; use those. A foot pedal or keyboard shortcuts for pausing and rewinding help with manual correction.

Quality check

Before analysis, listen to each recording again while reading the transcript, at least for a sample of passages. Check speaker labels, unclear passages and consistent pseudonyms. In your methods chapter, state the transcription style, the rules, who transcribed, whether software was used and how the transcripts were checked. This supports the dependability of your study.

Next step: analysis

Once your transcripts are clean and pseudonymized, you can start coding. At /codieren you can import transcripts, build a coding frame and mark passages directly. The appropriate transcription level depends on your analysis method, and some departments prescribe specific rules; your university’s own guidelines take precedence. Upload them to Professor Digital and the rules are tailored to them.