# Verbatim: How to Use Exact Quotes Without Distorting Their Meaning

Learn what a verbatim is, how to separate source, code, theme, and finding, avoid cherry-picking, and protect confidentiality.

- Canonical URL: https://www.harmate.com/en/blog/verbatim-how-to-use-exact-quotes-without-distorting-their-meaning
- Author: Harmate Team
- Published: 2026-09-05
- Updated: 2026-09-05T18:06:57.431006+00:00
- Language: en

## Content

# Verbatim: How to Use Exact Quotes Without Distorting Their Meaning

[Merriam-Webster defines **verbatim**](https://www.merriam-webster.com/dictionary/verbatim) as using the exact words of the original. In a report, study, or article, a verbatim is therefore an excerpt kept as close as possible to the source, with enough context to prevent its meaning from shifting.

The hard part starts after transcription. A quote can give an experience a voice, but it is not automatically a conclusion, a frequency, or evidence of representativeness. To use it responsibly, keep four things distinct: the source, the exact excerpt, the analysis that connects excerpts, and the finding that the evidence finally supports. The method applies to interviews, consultations, questionnaires, and customer feedback.

## Source, verbatim, code, theme, finding: keep five levels apart

“Verbatim” is often used for a whole response, a selected sentence, or a summary. That ambiguity makes reports difficult to audit. A simple chain puts each object back in its place.

| Level | What it is | What it can tell you |
|---|---|---|
| Source | The full response, interview, or document, with an identifier and context | Who spoke, in response to what, when, and under which conditions |
| Verbatim | The exact passage selected and linked back to the source | What the person actually said or wrote |
| Code | A short analytic label applied after reading the material | The issue or action the excerpt helps examine |
| Theme | A meaning pattern built by comparing excerpts and their differences | A structuring idea, not a list of repeated words |
| Finding | A statement answering a decision question, supported by evidence and limits | What the analysis allows someone to decide or explore next |

A code is not hidden in one keyword. A theme is not a word-frequency count. A finding is not a polished restatement of a quote. The thematic-analysis approach described by [Braun and Clarke](https://uwe-repository.worktribe.com/output/1043060) asks analysts to develop patterns of meaning in relation to a question and an explicit interpretation. The route can be inductive, deductive, or mixed, but it should remain inspectable.

## Start with an open question that leaves a trace

An informative verbatim is not manufactured by a question that already suggests its answer. If you want to understand an obstacle, “What is the main problem?” may force a respondent into a category. A situated invitation opens more room: “Tell us about the last time this step slowed you down. What happened, and what did you do next?”

A useful follow-up asks about an episode, decision, consequence, or need. It can request one example while leaving room for nuance, disagreement, or “I don’t know.” [Open questions](https://www.harmate.com/en/blog/open-ended-questions-get-actionable-data-without-bias) help discover unexpected reasons; they do not automatically make responses true, representative, or easy to analyse. An overly broad prompt can also produce short, uneven answers.

Before collecting, decide what will be preserved: original wording, language, date or wave, displayed question, follow-ups, and response status. Preserve the material before reducing it. If translation is needed, keep the source-language version, label the translation, and do not present a translated nuance as the original wording.

## Prepare context before selecting a quote

A sentence removed from an exchange may look clear while actually answering a previous hypothesis, joke, or question. Before extracting it, record at least:

- the question or stimulus that prompted it;
- the contextually relevant population information, without unnecessary identifiers;
- the period, setting, and collection mode;
- omissions, ellipses, brackets, and translations;
- a stable link to the full source and the rights to reuse it.

The [COREQ checklist, introduced by Tong, Sainsbury, and Craig](https://pubmed.ncbi.nlm.nih.gov/17872937/) and listed by the [EQUATOR Network](https://www.equator-network.org/reporting-guidelines/coreq/), contains 32 items for more transparent reporting of interviews and focus groups. The [SRQR by O’Brien and colleagues](https://pubmed.ncbi.nlm.nih.gov/24979285/) contains 21 items for qualitative research reporting more broadly. These are transparency guides, not validity certificates and not automatic quote-selection recipes.

Readers should understand why the excerpt is there: does it illustrate an experience, tension, exception, change, or hypothesis? Without that explanation, the quote becomes decoration with borrowed authority. A report can show the excerpt, its code, the source question, and the finding it contributes to, while preserving a route back to the complete response.

## Select without cherry-picking

The most vivid quote is not always the most representative of the material. A polished sentence may be unusual; a hesitant phrase may contain the strongest clue about a cause. To reduce cherry-picking, write selection criteria before choosing and actively look for cases that challenge the emerging interpretation.

| Selection question | What to check |
|---|---|
| Usefulness | Does the excerpt illuminate the decision question rather than a spectacular detail? |
| Fidelity | Do ellipses, edits, and translations preserve the meaning? |
| Variation | Did I look for a different experience, disagreement, or edge case? |
| Traceability | Can an authorised reviewer find the source and its context? |
| Proportion | Does the space given to this quote match the strength of the evidence? |
| Risk | Could the excerpt expose a person, client, or rare event? |

[Corden and Sainsbury](https://www.york.ac.uk/inst/spru/pubs/pdf/verbusers.pdf), in their study of quotations in qualitative social-research reports, highlight the value of treating participants’ words as part of the argument and reader relationship. That does not mean publishing everything. It means being able to explain what an excerpt contributes and what the rest of the responses prevents you from claiming.

A [2025 methodological synthesis](https://pmc.ncbi.nlm.nih.gov/articles/PMC12362350/) identifies the risk of cherry-picking and recommends considering a quotation’s function, practical constraints, ethics, readability, and inclusion. A selection log is often more honest than memory: source, provisional theme, reason for inclusion, divergent case found, sensitivity, published version, and verification link. If three excerpts say similar things, choosing the clearest one does not justify “most participants.” That phrase needs a denominator, a coding rule, and documented coverage. Otherwise write “this excerpt illustrates,” or “several participants described,” only when the latter is genuinely supported.

## Protect confidentiality: removing a name is not enough

A verbatim can identify someone through their story. A rare job, date, place, client name, or combination of details may enable re-identification in a small group. Consent to quote does not remove the need to consider risks to other people mentioned in the response.

[CNIL distinguishes anonymisation from pseudonymisation](https://www.cnil.fr/fr/technologies/lanonymisation-de-donnees-personnelles): anonymisation aims to make identification practically impossible and irreversible; pseudonymisation replaces identifiers while attribution remains possible with additional information. A pseudonymised quote is therefore still personal data. [Kaiser’s article on confidentiality in qualitative research](https://pmc.ncbi.nlm.nih.gov/articles/PMC2805454/) describes deductive disclosure as a practical issue: combining contextual clues can reveal an identity even after a name has been removed.

Before publication, minimise details that are not needed, generalise a location or replace a sensitive example, and log the transformation. Brackets should show an editorial intervention; they should not improve style or make a story more persuasive. When risk remains high, do not publish the excerpt, seek specific consent, or use a controlled paraphrase. Keeping the full source in a restricted workspace does not mean making it public.

## What a verbatim shows—and what it does not

The scope of a quote depends on the design that produced it. It can show how a person experiences a situation, the vocabulary they use, a sequence of events, or a reason they give. It can support a hypothesis to investigate. It cannot measure the size of a phenomenon by itself.

| A verbatim can help show | It cannot establish on its own |
|---|---|
| How a step is experienced or described | How many people share that experience |
| Why one person gives a reason | That the reason is the main cause |
| That a formulation or tension exists in the material | That an entire population thinks this way |
| An edge case or important divergence | That a result is statistically representative |

The number of visible quotes is not a respondent count. Repeating a quote does not create additional observations. To use frequency language, count responses using an announced rule, retain ambiguous cases, and state the analysed base. For exploratory work, access to disagreement and limits may be more useful than a ranked list of themes.

## A concrete example: understanding an onboarding delay

Imagine, for teaching purposes, an internal consultation of twelve new hires about their first week. The open question is: “Describe the moment when your arrival was hardest, and what would have helped.” One full response contains several episodes. The team selects this excerpt:

> “I had access on the first morning, but nobody knew who was meant to show me the project. I waited three days before I dared to ask.”

The source records the question, the relevant week, and a pseudonymised identifier. A provisional code could be “unassigned onboarding ownership.” A theme might become “start-up blocked by implicit coordination,” if other responses and different cases support it. A cautious finding is: “This excerpt shows how the absence of an explicit hand-off delayed entry into the project.”

What it cannot support by itself: “The main onboarding problem is missing ownership,” or “new hires wait three days.” Those claims require reviewing all twelve responses, documenting the coding, looking for opposite experiences, and supplying the matching denominator. Another person may say: “I was productive on day one because my buddy sent the context before I arrived.” That divergence does not invalidate the theme; it identifies a condition that can prevent the problem.

The verbatim preserves one person’s account of causality without turning it into a general truth. The code supports comparison, the theme organises interpretation, and the finding remains proportional to the evidence.

## A checklist before publishing a quote

1. Is the complete source retained, findable, and accessible to authorised reviewers?
2. Is the published text exact, with edits, translations, and omissions marked?
3. Are the question, period, and necessary context visible?
4. Is the reason for including the excerpt recorded during analysis?
5. Were divergent or negative cases actively sought and retained?
6. Could the details re-identify the person or a third party?
7. Do consent and usage rights cover this form of publication?
8. Does any frequency or majority statement rely on a real denominator?
9. Are code, theme, and finding distinct from the quote?
10. Can a reader verify the limit of what the excerpt supports?

For a related workflow, see [how to read 200 open responses without fooling yourself](https://www.harmate.com/en/blog/analyzing-200-open-responses-without-bias-a-method-for-actionable-decisions) and [how to create a word cloud without distorting responses](https://www.harmate.com/en/blog/word-cloud-how-to-create-one-without-distorting-your-responses).

## FAQ

### Can a verbatim be corrected?

You can document a transcription issue, remove an identifying detail, or translate, but each intervention should remain visible and preserve meaning. For a substantial correction, use a paraphrase and label it clearly as such.

### How many quotes should a report contain?

There is no universal number. Include enough to show the reasoning, relevant variation, and limits—not enough to create an illusion of precision. The decision question and the diversity of the material should guide the selection.

### Is a quote evidence?

Yes, as a qualitative piece of evidence linked to a source; no, as a frequency measure or proof of representativeness. State what it supports and what it does not.

### Are verbatim and quotation the same thing?

A quotation can be an exact reproduction, but it can also mean a cited passage or reference more generally. “Verbatim” emphasises fidelity to spoken or written words. In every case, preserve the source and context.

### Can AI choose the best verbatims?

AI can speed up retrieval or suggest groupings. It should not be the sole judge of fidelity, sensitivity, or relevance. Check every excerpt against its source, document exclusions, and keep uncertainty visible.

Open collection preserves material that prepared categories might miss. A tool such as [Harmate](https://www.harmate.com/en/product/open-questions) can help connect responses, themes, excerpts, and findings in a reviewable reading. The decision remains tied to sources, divergences, and limits: a quote opens the reasoning; it does not close it.