Researchers: How to Anonymize Interview Transcripts

A stack of interview transcripts is a promise kept to every person who spoke with you. Anonymizing them means removing or disguising anything that could point back to a real name, a workplace, or a town, while keeping the words useful for research. Do it early, do it consistently, and keep a written record of every change.

When Should You Plan Anonymisation?

Anonymisation works best as a plan, not a cleanup job done at the end. Before you record a single interview, decide what participants will be told. Will their transcript be shared with other researchers, archived, or quoted in a thesis? The UK Data Service says anonymisation is most effective when it is planned early and applied consistently throughout a project, not added afterward. Write this plan into your consent form, in plain language, so participants know what anonymised will actually mean for their own words.

How Do You Pseudonymize Names and Log the Changes?

Give every participant one fake name, like Jane Example, and use it everywhere that person appears, in every transcript and on every page. Then write down what you changed. An anonymisation log is a separate file listing the original word, its replacement, and where it appeared. Without it, you cannot check your own work later, and you cannot answer a participant who asks what happened to their words.

• Create the log as a spreadsheet before you start reading, with columns for the original text, the replacement, and the category, such as name, place, employer, or date.
• Replace a name the same way everywhere, even if a participant spelled it differently across interviews.
• Store the log apart from the anonymised transcripts, on a different drive or folder, since together they undo the whole exercise.
• Decide who can open the log and for how long, then delete it once you no longer need to reverse anything.

Building this log by hand across dozens of transcripts is slow work. For a transcript saved as a Word file, ONYRI Sanitize can handle the mechanical part: Token mode, on the Pro plan, gives each detected name one consistent label, like [NAME1], repeated everywhere it appears, and custom rules add a specific place or employer name. It keeps no anonymisation log of its own, so you still need to record your own reasoning.

What Indirect Details Can Still Identify Someone?

Removing a name is rarely enough. The UK Data Service warns that combinations of ordinary details, a rare job, a small town, an exact date, can point to one person even with no name attached. This is sometimes called jigsaw identification. No single detail gives someone away, but three or four together do.

How Should You Handle Quotes in Papers and Theses?

A single published quote can undo pages of careful anonymisation if it is too specific. Before you quote someone, check the sentence on its own. Could a colleague, a neighbor, or a manager recognize the speaker from it, even without the pseudonym attached? If a quote names a rare project or an exact date, generalize it the same way you generalized the transcript.

• Read each quote in isolation, outside the transcript, and ask who could still recognize the speaker.
• Combine two very similar quotes from different people into one line only if your consent form allowed it.
• Leave a rare or distinctive turn of phrase out of a quote if it would let colleagues recognize the speaker's voice.
• If your consent form promised participants a review of quotes, send them before submission, not after.

Should You Keep the Original Audio Recording?

Anonymising the transcript does not anonymise the recording. A voice carries an accent, a speech pattern, and sometimes background sound that a transcript never captures. Decide in the consent form whether you will keep the audio after transcription, for how long, and who can listen to it. If you promised participants an anonymised transcript, treat the original recording as identifying data with its own storage rules.

What Should You Check Before You Archive or Share?

• State in the consent form how anonymised transcripts will be used, shared, or archived.
• Give every person one pseudonym and log every change in a file kept apart from the transcripts.
• Generalize any job, place, age, or date that could single someone out on its own.
• Reread each transcript and each quote as a stranger would, looking for combinations that still identify someone.
• Decide the audio recording's fate, keep, destroy, or restrict, separately from the transcript's fate.