How do researchers dictate field notes?

Researchers dictate field notes by recording a voice note straight after each session, while the details are fresh. They then transcribe it with a tool their ethics approval allows, and check every quoted passage and name against the audio, not against the transcript. We found no peer-reviewed study that measures voice field notes, so the method rests on what speech and transcription research says about accuracy, correction and privacy. Our guide to dictation for academic writing places this method among the other drafting steps.

Speak the notes straight after the session, while memory is fresh

The case for speaking notes soon after a session is about memory. A voice note captures what you saw and heard in your own words, before the details fade. That is a reasonable case, and it does not depend on any study of voice notes.

Recording is not always the better choice, though. In a 2020 comparative analysis of in-depth interviews, with and without voice recorders, the authors suggest that "in some circumstances not recording is the best approach" ¹. Their study compared interviews rather than field notes, so the point applies by analogy. It is a reminder to decide on purpose whether a session should be recorded at all, and to check that decision against your ethics approval.

Listening back misses more than half of recognition errors

Transcription errors are easy to miss. In an in-lab study, participants missed identifying over 50 percent of speech recognition errors when they listened to the audio output of the recognized text ². In other words, a transcript that sounds right when read back can still be wrong. Check each quoted passage against the original recording, and read the transcript only as a guide to where to listen.

Some transcription tools also invent text. A 2024 study of Whisper, an open speech recognition model, found that roughly 1 percent of its audio transcriptions contained entire hallucinated phrases or sentences that did not exist in the audio ³. An invented sentence in a field note can look as plausible as a real one. Mark any passage that matters to your analysis, and check it against the audio before it enters your notes.

Error rates can differ between speakers, so plan extra checks

Accuracy also depends on who is speaking. In a 2020 study of commercial speech recognition systems, the average word error rate, which is the share of words the system gets wrong, was 0.35 for Black speakers and 0.19 for white speakers ⁴. For field work, this is a reason to plan more checking for speakers whose dialect or speech patterns may differ from the data the tools were trained on, and to say how you checked in your methods section.

Privacy comes before accuracy with participant recordings

Participant audio is governed by your consent forms, your data management plan and your ethics approval, before any tool's settings apply. Apple's own rules show why the tool matters. Apple says dictation sent to its servers is not stored unless you opt in to Improve Siri and Dictation ⁵. Those are Apple's rules for Apple's services, so they do not replace your approval. Check where the audio goes before you record anyone.

What to do after each session

1. Record a voice note straight after the session. On a Mac, Apple says the Voice Memos app can recognize and transcribe the speech in audio recordings ⁶, but check that the service is allowed by your approval before you use it for participant audio.
2. Choose a transcription tool your approval names. Dragon NaturallySpeaking can be used to transcribe recordings made earlier. Sheffield's library describes it as able to dictate into a digital recorder and transcribe it later ⁷, which may suit a team whose university already licenses it.
3. Check every quoted passage and name against the audio. Do this before any quote enters your analysis.
4. Mark anything the transcript invented. Remove it, and note in your records that you did so.

It depends on your ethics approval, your setting and your speakers

If your approval names a tool, use that tool, whatever this page recommends. If you work in noisy places, such as a market or a clinic, expect more errors and check more closely. If your speakers include people whose speech may fall outside the training data of the tools you use, plan more checking, and report it. For drafting the write-up, see our question on how to dictate academic citations. For the broader question of whether dictation suits a thesis, see whether you can dictate a thesis or dissertation.