Braun and Clarke thematic analysis: six phases, not a protocol

Junu Yang · 2026-09-09

Use Braun and Clarke's six phases as scaffolding for the approach later called reflexive thematic analysis. The phases are not a protocol. Completing them does not make the analysis correct.

This page is for a methods section or a research readout that cites those phases. Citing them is fine. The reading is still yours to defend. The phases do not show that the reading is right.

The 2006 paper outlined the approach. The name came later

The approach later called reflexive thematic analysis was first outlined in 2006, in Using thematic analysis in psychology. That title does not say reflexive thematic analysis. In their 2021 paper, One size fits all? What counts as quality practice in (reflexive) thematic analysis?, Braun and Clarke say they now call the approach reflexive thematic analysis, and that they first outlined it in 2006.

The 2006 paper says its guidelines are not rules. You fit them to the research question and the data. The analysis is recursive, not linear. You move back and forth across the phases. You do not finish a phase and lock it.

The 2006 names, and the names they use now

In the 2006 paper the six phases are:

familiarising yourself with your data
generating initial codes
searching for themes
reviewing themes
defining and naming themes
producing the report

The names have changed. Doing Reflexive TA lists the current names and says they have changed since 2006:

familiarising yourself with the dataset
coding
generating initial themes
developing and reviewing themes
refining, defining and naming themes
writing up

That page says the phases do not prescribe a rigid process. It calls them a guide for a systematic reading, and it says qualitative analysis is not the outcome of following set phases. Analysis is typically recursive, with movement back and forth. With more experience, and with a smaller dataset, some of the phases blur together.

The 2021 paper says the 2006 name "searching for themes" helped people treat themes as things already sitting in the data. The later name, generating initial themes, keeps the researcher in the work of making the first themes.

Reflexive thematic analysis is a different job from codebook or coding-reliability work

The 2021 paper separates three kinds of thematic analysis. They share a name. They do not share a quality check.

Coding-reliability thematic analysis is the cluster concerned with objective, unbiased coding. It uses a codebook and more than one coder. Inter-rater reliability, meaning agreement between coders, is a quality marker. Themes are often set early, sometimes before the analysis.

Codebook thematic analysis uses a structured coding framework. In that 2021 account, agreement between coders and inter-rater reliability are not usually the quality markers. Themes are often developed early, and some versions still revise them or add themes from the data.

Reflexive thematic analysis treats the researcher's subjectivity as part of the analysis. The 2021 paper describes the coding as interpretive. Coding is open, and it is part of how themes develop. It is not, in general, a search for extracts that fit themes you already fixed. A second coder is not required. Inter-rater reliability is a quality marker for coding-reliability thematic analysis. The 2021 paper does not use it as a quality marker for reflexive thematic analysis. It treats the researcher's standpoint as a resource for the reading.

If the methods section cites Braun and Clarke and then offers inter-rater reliability as proof the coding was right, name the method you actually ran. That proof belongs to coding-reliability work.

A theme is a pattern of shared meaning you develop

In the 2021 paper, a theme is a pattern of shared meaning organized around a central idea. The researcher develops it. A topic summary is a different thing. Pricing, privacy, or notifications can be topics in a dataset. They become a theme only when you can say what shared meaning holds the extracts together.

The same paper asks writers to describe theme generation as work they did, and to stop claiming that themes emerged. The 2006 checklist already treats the researcher as active, and it rejects the idea that themes simply appear in the data. The later writing is stricter about what counts as a theme: a pattern of shared meaning, not a bucket of extracts that mention the same subject.

The 2006 paper also says there is no fixed share of the dataset that turns a pattern into a theme. More instances do not make a theme more important. A pattern can matter when it takes up little of the dataset. The researcher judges, and says how. This page does not add a percentage, a theme count, or a quote count.

What to use each phase for

Use the phases to keep a reflexive reading in order. Moving through them does not make the analysis correct. Use the names from Doing Reflexive TA when you are doing the approach they now call reflexive thematic analysis. Keep the 2006 names when you are citing the older paper.

Start by reading the whole dataset, more than once, and noting what strikes you. That is familiarising yourself with your data in the 2006 paper, and familiarising yourself with the dataset on Doing Reflexive TA. The 2006 paper says this reading is the base for the rest, and it advises against skipping it or reading only the parts that look useful. If someone else transcribed the data, spend longer on this reading, and check the transcript against the recording.

Then code the dataset. The 2006 name is generating initial codes. The later name is coding. A code is a short label for a feature of the data that matters to the question. Doing Reflexive TA says to code the whole dataset, in more than one round, and to collate the codes with the extracts. In reflexive thematic analysis those labels can change. Freezing them after one pass is a different method.

Then generate initial themes from the codes and the collated extracts. The 2006 name is searching for themes. The later name is generating initial themes. Use the later name. It says you are making candidate themes. Grouping codes because they share a word leaves you with a topic list. A candidate theme needs a meaning you can state.

Then check the candidates against the coded extracts and the whole dataset. The 2006 name is reviewing themes. The later name is developing and reviewing themes. Doing Reflexive TA says you are asking whether the candidates tell a convincing story of the data that addresses the research question. Themes get developed further, and sometimes split, combined, or discarded. On that page, a theme in their approach is a pattern of shared meaning underpinned by a central concept.

Then work out the scope of each theme and name it. The 2006 name is defining and naming themes. The later name is refining, defining and naming themes. The 2006 paper says that by the end of this phase you should be able to say what each theme is and what it is not. One test they give is whether you can describe the scope and content of each theme in a couple of sentences. Doing Reflexive TA says this is where you decide the story of each theme and choose an informative name. A name that only repeats the topic usually means that story is still missing.

Then write the account. The 2006 name is producing the report. The later name is writing up. Doing Reflexive TA says this weaves the analytic narrative and the data extracts, and sets the analysis against existing literature. The 2006 paper says extracts have to sit inside that narrative, and the narrative has to make an argument about the research question. It says to include enough extracts to show the theme. It does not set a number. This page does not set one either.

Expect to go back. A later phase often returns you to an earlier one. The 2006 paper says writing starts in the first phase, with notes and possible codes, and continues through the analysis. The phases are a sequence you can teach. They are a guide you reuse.

What this page will not tell you

This page will not tell you the analysis is right because the six phases are done. It will not tell you how many interviews to run, how many themes to keep, or how many extracts to print. Those figures are not a Braun and Clarke rule on the pages above.

If the choice is still between this method and grounded theory, that comparison is Grounded Theory vs Thematic Analysis: Which Should You Use and When?. This page does not make that choice. It only says how to use the six phases: as scaffolding, and not as a protocol.