Transcription Rules for Qualitative Research

Transcription rules, also called transcription conventions, describe how spoken language is turned into written text. Clear rules make your transcripts consistent, easier to analyze and easier for others to understand. This guide explains the main types of transcription, shows how to mark common features of speech and ends with a simple template you can adapt.

What transcription rules are

Spoken conversation contains much more than words. People pause, laugh, interrupt each other, start sentences again and sometimes speak unclearly. Transcription rules define which of these features you record and how you write them down.

There is no single correct set of rules. The right level of detail depends on your research question and your method of analysis. The University of Bath Library, drawing on the Finnish Social Science Data Archive, describes four levels: summary transcription, basic level transcription, exact transcription and conversation analysis transcription (source). Most theses that use interviews work with something close to the basic or exact level.

Whatever you choose, decide before you begin, apply the rules to every transcript and describe them in your methods chapter. Ask your advisor whether your department expects a particular system.

Verbatim and clean verbatim compared

Verbatim transcription records everything that was said, including filler words, repetitions, false starts and often pauses and nonverbal sounds. It is useful when the way something is said matters, for example in conversation analysis or discourse analysis.

Clean verbatim transcription keeps the full content but leaves out filler words, stutters and obvious slips. The wording and meaning stay the same. It is common in thematic analysis and other approaches that focus on content.

Here is the same statement in both styles.

Verbatim:

R1: So, um, I- I think the, the hardest part was, uh, (2) honestly just finding the time to- to do it. (laughs)

Clean verbatim:

R1: I think the hardest part was honestly just finding the time to do it. (laughs)

Note that clean verbatim does not mean rewriting. You should not correct grammar in a way that changes what the participant meant, and you should not summarize.

Marking pauses, overlaps and inaudible passages

Whichever style you choose, you will need conventions for a few common situations. The following notation is widely used and easy to read, but you can adapt it as long as you apply it consistently.

Speaker labels: Use short codes at the start of each turn, such as "I:" for the interviewer and "R1:" for the first respondent. Start a new paragraph for each change of speaker.
Pauses: Mark short pauses with (.) and longer pauses with the approximate length in seconds, for example (3).
Overlapping speech: Mark the start of the overlap with // in both speakers' lines, or describe it, for example [both speak at once].
Interruptions and broken-off words: Use a hyphen, for example "I was go-".
Inaudible passages: Use [inaudible] together with a timestamp, for example [inaudible 00:14:32], so you can find the passage later.
Uncertain words: Put your best guess in parentheses with a question mark, for example (workshop?).
Nonverbal sounds: Describe them in parentheses, for example (laughs), (sighs) or (phone rings).
Emphasis: If emphasis matters for your analysis, mark it consistently, for example with CAPITAL LETTERS. Avoid bold or colored text, because formatting can be lost when you import transcripts into analysis software.
Anonymized details: Use square brackets, for example [Company A] or [City 1].

A template you can adapt

Copy the following template into a document, adjust it to your project and keep it open while you transcribe.

Transcription rules for [project name]

Style: clean verbatim / verbatim (choose one).
Speaker labels: I = interviewer; R1, R2 = respondents.
New paragraph for every change of speaker.
Timestamps: at every change of speaker / at the end of each paragraph / every few minutes (choose one).
Filler words (um, uh): included / omitted.
Pauses: (.) for a short pause; (3) for a pause of about three seconds.
Overlaps: // at the start of overlapping speech.
Broken-off words: hyphen, for example "go-".
Inaudible: [inaudible hh:mm:ss]. Uncertain: (word?).
Nonverbal: (laughs), (sighs), (phone rings).
Dialect and colloquial speech: kept / lightly adapted to standard spelling.
Anonymization: [Person A], [City 1], [Company B]; key stored separately.
Header for each transcript: interview code, date, length, interviewer, transcriber.

Applying the rules to automatic transcripts

If you use automatic transcription to create a first draft, your rules still apply. Automatic transcripts can contain errors and do not follow your conventions on their own. Listen to the recording, correct the text, check the speaker labels and add pauses, overlaps and other markings according to your rules.

How Stillscript can help