From Codes to Themes in 6 Steps (Braun & Clarke Example with Quotes)
Stuck on how to transform codes into themes? This 2026 guide walks you through changing codes to themes qualitative researchers trust — with a mini transcript, real quotes, and a theme table you can copy.
If you have finished steps coding transcripts and now stare at 60 sticky codes wondering how to transform codes into themes, you are not alone. Almost every masters and undergraduate dissertation gets stuck at exactly this point: changing codes to themes qualitative work requires feels mysterious, subjective, and easy to get wrong.
This guide fixes that. Using Braun & Clarke's reflexive thematic analysis (2006; updated 2022), we show you developing themes from codes as a practical, repeatable process — with a worked mini-transcript, real quotes, and a copy-ready theme table. If you have not coded yet, start with our guide on how to code interview transcripts, then return here to build themes.
Quick answer: codes are labels for small chunks of meaning; themes are patterns of shared meaning organised around a central idea that answers your research question. You get from one to the other by clustering related codes, reviewing them against your data, and defining each theme with a name, definition, and quotes.
What Is the Difference Between Codes and Themes?
Examiners fail this section most often because students treat a code and a theme as the same thing. They are not. A code captures one interesting feature of the data. A theme interprets what several codes mean together.
Braun & Clarke (2022) are explicit: themes do not simply emerge. You actively construct them by asking, "What story do these codes tell together about my research question?" For a deeper conceptual walkthrough, see how to turn codes into themes.
| Feature | Code | Theme |
|---|---|---|
| Size | Small — one data extract | Large — pattern across dataset |
| Example | "Fear of asking questions in seminars" | "Belonging is negotiated through silence and speaking up" |
| Question it answers | What did they say here? | What does this mean for my research question? |
| How many? | 30–100+ in a dissertation | 3–5 main themes (plus subthemes) |
| Test | Could another coder apply it consistently? | Does it have a central organising concept? |
Think of it this way: codes are the bricks, themes are the house. Listing bricks is not the same as describing the building. If your "theme" could be used as an interview question or a topic heading (for example, "Support services"), it is probably still a code or a domain summary — not yet a theme.
A strong theme has a viewpoint. "Support services are hard to access" is a topic. "Seeking help feels like admitting failure" is a theme — it says something interpretive.
What Are Braun & Clarke's 6 Phases in Plain English?
Braun & Clarke's 2006 paper and their 2022 book describe reflexive thematic analysis in six recursive phases. Changing codes to themes qualitative analysis happens mainly in phases 3–5, but you need all six to do it credibly:
| Phase | What you do | Output |
|---|---|---|
| 1. Familiarisation | Read, re-read, listen, note initial ideas | Familiarity + memo notes |
| 2. Coding | Label interesting features systematically | Code list with extracts |
| 3. Generating initial themes | Cluster codes around shared meaning | Candidate themes |
| 4. Reviewing themes | Check themes work against coded extracts + full dataset | Thematic map, merged/split themes |
| 5. Defining & naming | Sharpen central concept, write definitions | Defined themes + theme table |
| 6. Writing up | Weave analytic narrative with quotes | Findings chapter |
Important update from Braun & Clarke (2022): coding can be inductive (data-driven), deductive (theory-driven), or both — but themes in reflexive TA are always interpretive. Do not count frequencies and call the most common code your main theme. Prevalence alone never makes a theme.
How Do You Go From Codes to Themes in 6 Steps?
Below is the hands-on workflow we teach students. We use a running example: a masters study on international students' experience of seminars in the UK. Research question: How do international postgraduate students experience participation in UK seminars?
Step 1: Finish open coding and list your codes with extracts
Before clustering anything, make sure every code has 2–4 example quotes attached. If you coded in NVivo, Word comments, or by hand, export a simple list. Aim for concise, active labels — not whole sentences. For the full technique, see how to code interview transcripts step by step.
Our mini-transcript example (three participants):
"In my country we listen to the teacher. Here they expect you to interrupt, and I feel rude doing that." — Mei (P2)
"I prepare for two hours but still stay silent. I am afraid my accent makes me sound stupid." — Ahmed (P5)
"When the tutor said 'great point' once, I started speaking every week. One sentence changed everything." — Sofia (P7)
Initial codes from these extracts: fear of interrupting; cultural norm of listening; accent anxiety; over-preparation yet silence; tutor validation as trigger; silence as respect vs silence as disengagement.
Step 2: Cluster related codes by shared meaning (not topic)
Print your codes or put each on a digital sticky. Ask: "Which codes are saying a similar thing about participation?" Group by meaning, not by who said it or where it happened. One useful prompt is: "If these three codes were a chapter title, what argument would they make together?"
Example clusters from our data:
- Cluster A: fear of interrupting + cultural norm of listening + silence as respect → candidate idea about silence having meaning
- Cluster B: accent anxiety + fear of sounding stupid + over-preparation yet silence → candidate idea about language and legitimacy
- Cluster C: tutor validation as trigger + peer encouragement → candidate idea about permission to speak
Discard or park codes that do not relate to your research question in a "miscellaneous" pile — do not force them into a theme.
Coding stuck? Send transcripts — theme table in 48h
Send your transcripts and research question. We return coded extracts, candidate themes, and a clean theme table ready for your findings chapter.
Step 3: Turn clusters into candidate themes with a central concept
Now give each cluster an interpretive spine. A complete candidate theme needs: a working name, a one-sentence central concept, and the codes it contains. Avoid single-word names like "Confidence" — add the argument.
| Candidate theme (working name) | Central concept | Codes inside |
|---|---|---|
| Silence is not absence | Staying quiet can signal respect, caution, or thoughtfulness — not lack of knowledge | fear of interrupting; listening norm; silence as respect |
| Speaking requires feeling legitimate | Students speak when they feel their English and ideas will be accepted | accent anxiety; fear of sounding stupid; over-preparation |
| Permission unlocks participation | A small invite from tutor or peers transforms silence into voice | tutor validation; peer encouragement |
At this stage you should have 4–7 candidate themes. That is normal — you will cut and merge next.
Step 4: Review themes against quotes and the whole dataset
This is Braun & Clarke's Phase 4 and the step students skip. Do two checks. First, read all extracts under each candidate theme — do they cohere? If one quote contradicts the rest, either move it or split the theme. Second, re-read your full transcripts — does the thematic map reflect the whole dataset, or just two vocal participants?
In our example, review showed "Permission unlocks participation" was thin (only two participants) but analytically important because it explained change over time. We kept it as a subtheme under a larger theme rather than deleting it. We also merged two overlapping candidates into one: "Participation must be earned through legitimacy."
Practical test: can you draw your themes as a map on one page with arrows? If two bubbles overlap almost entirely, merge them. If one bubble contains two different stories, split it.
Step 5: Define, name, and write the theme definition
For each surviving theme, write a 40–60 word definition that states: what the theme is about, what it is not about, and how it answers your research question. Then finalise a punchy, analytic name (6–10 words max).
Example final definition: "Negotiating voice in an unfamiliar seminar culture" — This theme captures how international students reinterpret silence from respect into risk, and speak up only when they feel linguistically and socially legitimate. It is about shifting norms, not language level alone, and directly answers how participation is experienced as cultural negotiation.
Good names tell the reader your argument before they read the paragraph. Compare "Language issues" (weak) with "Speaking up means risking legitimacy" (strong).
Step 6: Select quotes and write the analytic narrative
Choose 2–3 vivid quotes per theme that show different facets — not three quotes saying the same thing. Present each quote briefly, then add your interpretation: what does this reveal, why does it matter, and how does it connect to literature? For formatting rules on embedding, trimming, and anonymising quotes, see how to present interview quotes in your dissertation.
"I prepare for two hours but still stay silent. I am afraid my accent makes me sound stupid." — Ahmed (P5, coded: accent anxiety; over-preparation yet silence)
Ahmed's account shows preparation does not resolve the barrier, because the barrier is social legitimacy rather than knowledge. His silence is protective — a strategy to avoid being positioned as less competent.
Follow the pattern claim → quote → interpretation → link for every theme. That single habit lifts most findings chapters from descriptive to analytical.
Can You See a Full Worked Example With Quotes?
Here is how the mini-dataset collapses from raw extracts to final themes:
| Quote (abridged) | Code | Candidate theme | Final theme |
|---|---|---|---|
| "They expect you to interrupt, and I feel rude" — Mei, P2 | cultural norm of listening | Silence is not absence | Negotiating voice in an unfamiliar seminar culture |
| "I am afraid my accent makes me sound stupid" — Ahmed, P5 | accent anxiety | Speaking requires feeling legitimate | Negotiating voice in an unfamiliar seminar culture |
| "When the tutor said 'great point' once, I started speaking every week" — Sofia, P7 | tutor validation as trigger | Permission unlocks participation | Permission transforms silence into voice (subtheme) |
Notice developing themes from codes reduced six codes to one main theme with a subtheme — not by counting, but by interpreting what the codes share: participation is experienced as a negotiation of cultural norms and legitimacy, unlocked by small acts of recognition.
What Does a Final Theme Table Look Like?
Examiners love a one-page overview before your narrative. Place this table at the start of your findings chapter. Copy this template:
| Theme (final name) | Definition (40–60 words) | Codes included | Illustrative quote |
|---|---|---|---|
| 1. Negotiating voice in an unfamiliar seminar culture | How students reinterpret silence from respect into risk and speak only when feeling legitimate; about cultural negotiation, not language level alone. | listening norm; fear of interrupting; accent anxiety; over-preparation | "I prepare for two hours but still stay silent..." — Ahmed, P5 |
| 2. Permission transforms silence into voice | Small tutor or peer invitations that convert cautious silence into regular participation; the turning-point mechanism. | tutor validation; peer encouragement | "One sentence changed everything..." — Sofia, P7 |
| 3. [Your third theme] | [What it captures + boundary + link to RQ] | [3–6 codes] | [One vivid quote + ID] |
Keep the table tight: one row per theme, one quote per row. Save your other quotes for the narrative. Label participants consistently (P1, P2) and never use real names without consent.
Coding stuck? Send transcripts — theme table in 48h
Stuck clustering 50+ codes? Send your files today and get a coded extract sheet, thematic map, and polished theme table with definitions — ready to paste into your dissertation.
What Mistakes Turn Codes Into Weak Themes?
Mark these examiner red flags before you submit:
- Renaming a code as a theme. "Lack of confidence" with one code underneath is not a theme. Ask: what do these multiple codes mean together?
- Using topic headings. "Support services," "Challenges," "Benefits" describe aboutness, not meaning. Add your argument.
- Too many themes. Nine themes for eight interviews signals you stopped at clustering. Aim for 3–5 and use subthemes.
- No central concept. If you cannot write a one-sentence "this theme is about..." statement, the theme is not defined yet.
- Quote-dumping. Long block quotes with "as shown above" and no interpretation reads as descriptive. Always follow with so what?
- Ignoring deviant cases. If one participant contradicts your theme, discuss it — examiners reward honesty over neatness.
If you recognise two or more of these in your draft, return to Step 4 and re-review before writing up.
How Do You Know Your Themes Are Finished?
Use this Braun & Clarke–inspired checklist. You are done when you can tick every box:
- Each theme has a clear central organising concept you can state in one sentence.
- Each theme is supported by multiple codes and multiple participants.
- Themes are distinct — no two themes tell the same story.
- Together, themes answer your research question (not just summarise topics).
- Each theme has a sharp name, a written definition, and 2–3 contrasting quotes.
- You can explain in your methodology how you moved from codes to themes (audit trail).
Keep a one-page audit trail: date, decision (for example, "merged X and Y on 12 May because..."), and who was involved. One paragraph on this in your methods chapter dramatically increases trust.
Frequently Asked Questions
Cluster related codes that share an underlying meaning, test each cluster against your quotes and full dataset, then define each theme with a central concept, clear name, and illustrative quotes. Themes interpret what codes mean together in relation to your research question — they are not just frequent codes.
A code labels one feature of the data, such as "fear of asking questions." A theme is a pattern of shared meaning across codes organised around a central idea, such as "Belonging is negotiated through silence." Codes describe; themes interpret.
There is no fixed number, but most dissertation themes draw on 3–8 related codes across several participants. Depth, coherence, and relevance to your research question matter far more than counts. One thin theme with rich explanatory power beats three padded ones.
Rarely. Simply promoting a code to theme status usually produces a topic summary, which examiners penalise. A theme should unite several codes under a new interpretive idea. If only one code supports it, consider making it a subtheme or collecting more data.
Start findings with a theme table (theme, definition, codes, one quote), then devote one section per theme using a claim → quote → interpretation → link structure. In methods, add one paragraph tracing your path from familiarisation through coding, clustering, reviewing, and defining.
Final Thoughts
How to transform codes into themes is a craft, not a click: familiarise, code carefully, cluster by meaning, review against your data, define sharply, and write analytically with quotes. Work through the six steps above on one small batch of transcripts first — once you see developing themes from codes click on a single example, scaling to your full dataset becomes straightforward.
If you are short on time, get a second pair of eyes before submission. Whether you use our codes-to-themes walkthrough, revisit transcript coding basics, or polish how you present interview quotes, the goal is the same: themes with a clear argument, honest audit trail, and quotes that prove it.
