From Microphone to Manuscript: A Trust-First Routine for Transcribing Dissertation Interviews

Interview transcription can feel like a purely mechanical stage: press play, type, repeat. In practice, it is one of the most ethically sensitive points in a dissertation. A recording captures a person's voice, hesitations and possibly details they did not expect to share. Once those words become a transcript, they are easier to quote, copy and circulate. A little structure at this stage protects participants, strengthens your audit trail and makes later analysis far less fraught.

Begin with the consent you actually have

Before opening the audio file, return to the consent materials you used. Check what participants agreed to: recording, transcription, anonymous quotation, storage duration and who would have access. If your approval covered a professional transcription service, confirm that the arrangement is still within scope. If it did not, do not upload the audio to an external platform simply because it is convenient.

It is also worth checking whether participants were told how extracts might be used. A consent form that permits anonymous quotations does not automatically permit the sharing of full transcripts in an appendix. Where your plans have shifted, speak to your supervisor before proceeding rather than after the fact.

Treat storage as part of the method

Under UK GDPR and the Data Protection Act 2018, interview recordings and transcripts are personal data. Your university's research data policy will set the practical expectations, and your data protection officer or research office can clarify anything ambiguous. In most cases, the sensible baseline is straightforward:

• Keep identifiable audio and transcripts on university-approved, encrypted storage rather than personal cloud accounts.
• Use a clear file-naming convention that does not itself reveal identities, such as a participant code plus interview date.
• Store the key linking codes to names in a separate, access-controlled file.
• Restrict access to named researchers and supervisors, and record who has been given it.
• Note a deletion date that matches your ethics approval and data management plan.

This is not bureaucracy for its own sake. A clean storage routine means you can answer questions about your data with confidence, which matters at a viva or during an audit.

Decide your anonymisation approach before you type

Anonymisation is easier to apply consistently if you settle the rules first. Decide which elements will be replaced: names, workplaces, towns, family relationships, unusual job titles, or identifying combinations of circumstances. Then create a simple substitution list and apply it throughout.

Be alert to indirect identifiers. Replacing a participant's name with 'Participant A' achieves little if the transcript also mentions a distinctive role at a small organisation in a named town. Consider whether generalising those details would preserve meaning without risking recognition. Where a participant has asked to be identifiable, record that explicitly and check that your approval permits it.

It is also wise to distinguish between verbatim and cleaned transcription. A verbatim record captures hesitations, false starts and non-standard grammar; a cleaned version removes them for readability. Neither is inherently more ethical, but your choice should be deliberate and consistent. If quotations will appear in the thesis, decide in advance how you will present spoken language fairly rather than making participants sound less coherent than they were.

Choose tools with your eyes open

Automated transcription services vary considerably in accuracy and in what they do with uploaded audio. Before using one, check whether it is approved by your institution, where data are processed, whether recordings are retained or used for model training, and whether you can delete files afterwards. Free tiers are especially likely to involve unclear retention practices.

Even with an approved tool, keep a human in the loop. Accents, overlapping speech, background noise and subject-specific vocabulary all cause errors. A transcript that mishears a key term can send analysis in the wrong direction, and a misattributed speaker can distort your findings. Treat machine output as a first draft, never as a finished record.

Build a short quality-check routine

A consistent check takes less time than repeated ad hoc repairs. Try this sequence:

1. Listen once without editing. Follow the audio while reading the draft and mark anything that sounds wrong rather than fixing it immediately.
2. Check the anchors. Verify names, numbers, dates, technical terms and anything a participant emphasised.
3. Confirm speaker attribution. Ensure turns are assigned correctly, particularly in group interviews or where voices are similar.
4. Compare a sample against the audio. Re-listen to two or three short passages at normal speed and correct them closely.
5. Apply anonymisation. Replace identifiers consistently and update your substitution list.
6. Record the check. Add a short note to your research log stating who transcribed, who checked, when, and what version is now authoritative.

That final step is easily overlooked but valuable. It creates a modest audit trail showing that the transcript was reviewed rather than assumed correct.

Handle sensitive moments with care

Interviews sometimes include disclosures that were not anticipated. If a participant becomes distressed or reveals something potentially safeguarding-related, pause transcription and consult your supervisor about the appropriate response. Do not quietly remove difficult material simply because it complicates the analysis. Equally, do not quote it prominently without careful thought about dignity and consent.

Where participants have been offered the chance to review their transcript, set a clear deadline and explain what kind of corrections you will accept. Light clarification is usually reasonable; wholesale rewriting of meaning is not. Record any changes and keep the original version in line with your retention arrangements.

A final pass before analysis

Before you begin coding, ask whether the transcript is complete, consistently anonymised, stored correctly and accompanied by a note explaining its provenance. Check that filenames match your schedule and that the master copy is distinguishable from working versions. If a participant has withdrawn, confirm that the relevant files have been handled according to your approved procedure.

None of this need be laborious. A short, repeatable routine does most of the work: check consent, secure the files, apply agreed anonymisation, verify the draft against the audio, and log what you did. Done consistently, it protects the people who trusted you with their words and gives your analysis a firmer foundation.