Thematic Analysis: Braun and Clarke's Six Phases
The most-used qualitative analysis method, step by step, familiarisation, coding, theme development, and writing up themes that answer your question.
Thematic analysis (TA) identifies patterns of meaning across qualitative data. Braun and Clarke's six-phase approach is the most cited method in social-science dissertations because it is systematic yet flexible. Done well it produces analytic themes. Done poorly it produces a summary of what interviewees said. The difference is in the analysis.
Phase 1: Familiarisation
Immerse yourself in the data, transcribe (or read transcripts repeatedly), noting initial ideas. You cannot analyse data you do not know intimately. Active reading, not passive scanning, is the point.
Phase 2: Generating Initial Codes
Work systematically through the data, tagging features relevant to your question. Codes are concise labels for interesting segments, "fear of judgement", "reliance on peers". Code inclusively at first. You will refine later. Every code should be traceable to a data extract.
Phase 3: Searching for Themes
Cluster related codes into candidate themes, broader patterns that say something meaningful about the question. A theme is not just a topic. It captures something significant about the data in relation to your research question. A thematic map helps visualise how codes group.
Phase 4: Reviewing Themes
Check candidate themes against both the coded extracts and the whole dataset. Do they cohere internally and differ from one another? Split, merge or discard as needed. Some candidate themes collapse. That is the method working.
Phase 5: Defining and Naming Themes
For each theme write a clear definition, its scope and essence, and give it an informative name. "Navigating uncertainty" tells the reader more than "uncertainty". Identify the story each theme tells and how they fit into the overall narrative.
Phase 6: Producing the Report
Write up with vivid, illustrative extracts that evidence the analytic point, not extracts that merely describe. Each theme should make an argument connected to the literature and the research question. Extracts support your analysis. They do not replace it.
Reflexive vs Codebook TA
Braun and Clarke now emphasise reflexive TA, where the researcher's active interpretation is central and coding evolves organically, as opposed to codebook approaches with fixed frameworks and inter-rater reliability. State which you use and why.
Checklist
- Are themes analytic patterns, not topic summaries?
- Is every theme evidenced by traceable extracts?
- Do extracts support an argument rather than just describe?
- Have I stated my TA variant and my role in interpretation?
- Is my reflexive position stated rather than assumed?
- Does each theme answer the research question rather than describe the interview schedule?
The Distinction That Decides Your Mark: Topic Versus Theme
Almost every weak thematic analysis fails at the same point. The writer produces a list of subjects that came up, labels them themes, and illustrates each with a quotation. The result reads like a well-organised summary of the interviews. It is not analysis, because it makes no claim.
A theme makes a claim. It says something about the data that a reader could disagree with. The test is simple: can you finish the sentence "this theme argues that..."? If the answer is "this theme covers what people said about supervision", you have a topic.
| Topic summary (weak) | Analytic theme (strong) |
|---|---|
| Supervision | Supervision is experienced as permission rather than guidance |
| Workload | Workload is described in moral terms, as evidence of commitment |
| Technology | Digital tools are trusted for admin and distrusted for judgement |
| Barriers to help-seeking | Asking for help is negotiated as a threat to competence |
Notice that the strong versions are contestable. Another researcher reading the same transcripts might argue otherwise, and that is precisely what makes them analytic.
From Extract to Code to Theme
The chain from raw data to finished theme should be traceable, and showing part of it in an appendix is one of the easiest ways to demonstrate rigour.
| Stage | Example |
|---|---|
| Extract | "I did not want to email again. He would think I could not manage it on my own." |
| Initial codes | reluctance to contact supervisor, fear of appearing incapable, self-reliance as expectation |
| Candidate theme | Help-seeking carries a competence cost |
| Final theme | Asking for help is negotiated as a threat to competence |
Codes are descriptive and close to the data. Themes are interpretive and sit above it. If your themes look like your codes with capital letters, phases three to five have not really happened yet.
Which Variant of Thematic Analysis Are You Doing?
Braun and Clarke now distinguish sharply between approaches that share the name. Naming yours is not pedantry, because each variant carries different expectations about coding and quality.
| Variant | How coding works | Quality claim rests on |
|---|---|---|
| Reflexive TA | Codes evolve, researcher subjectivity is a resource | Depth of interpretation and reflexivity |
| Codebook TA | Structured codebook, developed early and refined | Consistency of application |
| Coding reliability TA | Fixed codebook, multiple coders | Inter-rater agreement statistics |
A common error is to run reflexive TA and then report a kappa statistic because it looks rigorous. Within reflexive TA, inter-rater reliability is conceptually out of place, since it treats interpretation as something to be standardised rather than developed. Pick a variant, and let its logic run through the whole chapter.
Inductive and Deductive, Semantic and Latent
Two further choices shape the analysis, and both belong in the methods chapter.
- Inductive coding starts from the data. Deductive coding starts from theory or a prior framework. Most real projects sit somewhere between, and saying where is more honest than claiming pure induction while quietly using your literature review as a code list.
- Semantic coding stays with explicit content, what participants actually said. Latent coding reaches for underlying assumptions and meanings. Latent work is more interpretive and needs more justification, but it is usually where the interesting themes live.
How Thematic Analysis Differs From Its Neighbours
| Method | Aim | Choose it when |
|---|---|---|
| Thematic analysis | Patterns of meaning across a dataset | You want flexibility and a broad view |
| Grounded theory | Build a theory from the data | Theory generation is the goal, and you can sample iteratively |
| IPA | How individuals make sense of lived experience | Small homogeneous sample, experiential focus |
| Content analysis | Systematic, often quantified categorisation | You need frequencies and replicability |
| Discourse analysis | How language constructs meaning and power | The interest is in the talk itself, not what it reports |
Common Mistakes and Fixes
| Mistake | Fix |
|---|---|
| Themes mirror the interview questions | Themes should cut across questions, not restate them |
| Quotation followed by paraphrase | Follow each extract with what it shows, not what it says |
| Counting participants per theme as evidence of importance | Prevalence is not significance in reflexive TA. Argue the meaning |
| Six themes, each with one supporting extract | Fewer, better-evidenced themes almost always score higher |
| No reflexive statement | Say who you are to this topic and how it shaped your reading |
| Claiming saturation without explaining it | Saturation belongs to grounded theory. In reflexive TA, justify sample size differently |
Software
NVivo, ATLAS.ti, MAXQDA and the free option Taguette all help manage coding, but none of them analyses anything. They store and retrieve codes you create. Examiners are unmoved by the software name and interested in the thinking, so mention the tool in one sentence and spend your words on the analysis.
Where Ethical Support Fits
Discussing your emerging codes with a supervisor or a peer is normal qualitative practice, and in reflexive TA a second reader is used to widen interpretation rather than to police it. Getting a methods chapter proofread is fine too.
What must remain yours is the coding, the interpretation and the argument each theme makes. If someone else generated your themes, the analysis is not yours, whatever the transcripts say.
Frequently Asked Questions
How many themes should I have?
For a typical masters dissertation, three or four well-developed themes usually work better than seven thin ones. There is no rule, but depth beats coverage.
How many participants do I need?
It depends on data richness and scope. Braun and Clarke resist fixed numbers, and reflexive TA does not rely on saturation. Justify your sample by reference to the research question and the depth of each account rather than reaching for a threshold.
Do I need a second coder?
Only if you are doing coding reliability TA. In reflexive TA, a second reader can enrich interpretation, but agreement statistics are not the quality criterion.
Can I use thematic analysis on documents or open survey responses?
Yes. TA is method-agnostic about data source. Short open responses give less latent material, so be realistic about how deep the analysis can go.
Should themes appear in the results or the discussion?
Practice varies. Many qualitative dissertations combine them, presenting each theme with its extracts and its link to the literature. Check your department's preference, since some require the split.
What is a thematic map for?
It shows how codes cluster and how themes relate, including subthemes. It is a thinking tool first, and often worth including as a figure because it makes your analytic structure legible at a glance.
Your Next Step Today
Take your current theme names and try to write each as a full sentence making a claim. Any theme that resists the exercise is still a topic, and reworking it is the single highest-value edit available to you at this stage.
Trusted Sources
- Braun V and Clarke V, Thematic Analysis, the authors' own resource site at the University of Auckland, including guidance on reflexive TA and common misapplications. Accessed 11 August 2026.
- Braun V and Clarke V, Using thematic analysis in psychology, Qualitative Research in Psychology 2006, 3(2), 77-101, doi:10.1191/1478088706qp063oa. The original six-phase paper.
- University of Southern California, Qualitative Methods. Accessed 11 August 2026.
- University of Southern California, The Methodology. Accessed 11 August 2026.
Qualitative conventions vary widely by discipline. Where this guide and your department's guidance differ, follow your department.
Related services
Related guides
- Writing the Methodology Chapter: Design, Sampling and Justification
- Grounded Theory Methodology Explained
- Reliability and Validity in Research
- Mixed-Methods Research Design Explained
All academic writing guides or browse our academic writing services.