Braun & Clarke Thematic Analysis: The 6-Phase Framework Explained

Braun and Clarke's thematic analysis is a six-phase method for identifying, analyzing, and reporting patterns of meaning across a qualitative dataset. First published in 2006 and substantially updated since (the current version is usually called reflexive thematic analysis), it's one of the most widely used – and, according to recent methodological audits, one of the most widely misapplied – approaches in qualitative research.

The six phases

  1. Familiarize yourself with the data – Read (and reread) the full dataset before you start formal coding. Immersion comes first; jot early observations, but hold off on systematic labeling.
  2. Generate initial codes – Work through the data systematically, tagging anything relevant to your research question with a short label, and keep the extract attached to each code.
  3. Search for themes – Step back from individual codes and start grouping them into candidate themes: broader patterns that make an interpretive claim about the data, not just a bucket of similar quotes.
  4. Review themes – Test each candidate theme against the coded extracts and against the full dataset. Merge overlapping themes, split themes that are really two ideas, and drop ones the data doesn't actually support.
  5. Define and name themes – Write a tight definition of what each theme captures and give it a name specific enough to convey the analytic point, not just a topic label.
  6. Produce the report – Weave theme, evidence, and argument together. Quotes illustrate your interpretive claims; they don't replace the need to make them.

This isn't a strictly linear checklist: Braun and Clarke have been explicit that movement between phases is normal and expected, especially between coding and theme development.

What changed with the "reflexive" version

The original 2006 paper described themes as things that "emerge" from the data. The updated, reflexive approach Braun and Clarke now advocate for explicitly rejects that language: themes don't emerge on their own, they're produced by a researcher who brings a particular theoretical and analytical perspective to the data. That's not a small wording change: it moves the method away from a positivist framing (where two coders should independently reach the same "objective" themes, measured by inter-rater reliability) and toward one where the researcher's interpretive judgment is treated as a resource, not a source of error to be minimized.

Practically, this means reflexive thematic analysis doesn't ask you to calculate inter-rater reliability scores or aim for a fixed, "discoverable" set of themes. It asks you to be transparent and considered about the interpretive choices you made and why.

Where researchers actually go wrong

A 2026 scoping review of 147 papers at CHI (one of the largest HCI conferences, where roughly 11% of accepted papers claim to use reflexive thematic analysis) checked what researchers say they did against Braun and Clarke's own quality criteria. The gap was substantial:

  • Only 32% of papers clearly named all their final themes
  • 63% reported "theme buckets" – topic labels grouping similar content – rather than themes that state a pattern of shared meaning, which is what reflexive TA actually calls for
  • 41% described themes as "emerging" from the data, the exact language the reflexive approach moved away from
  • 29% used a fixed codebook, which sits awkwardly with reflexive TA's more organic, evolving approach to coding
  • Only 20% of papers stated an explicit theoretical or epistemological position, despite reflexive TA requiring one
  • 75% used positivist terminology – "sample," "generalizability," "reliability" – that doesn't fit the method's own underlying assumptions

None of this means thematic analysis is a bad method. It means the method is frequently cited but poorly practiced: often because published examples themselves model the wrong version of it. If you're planning a thematic analysis, the single highest-leverage thing you can do is decide, explicitly, at the start, whether you're doing reflexive TA or a more codebook-driven variant, and then be consistent about it in how you describe your process.

A note on where this fits with AI assistance

Recent empirical work comparing LLM-generated codes to researcher codes on the same interview data found meaningful gaps. In one 2026 study coding 34 interviews, LLMs reached "good agreement" with human coders on only 12–27% of codes, and up to a third of AI-generated codes turned out to be duplicates under different names. That's not an argument against using AI here. It's the same finding across this research every time: AI-generated groupings need a human reviewing, renaming, and merging them, not rubber-stamping them, the same principle reflexive TA itself insists on. The researcher, not an automated process, produces the themes. It's also the design assumption behind Qualibr's approach: AI proposes an initial coding and theme structure, but naming, merging, and the final interpretive call stay with the researcher, every theme traceable back to its source quotes.

Start freeNo card required, try it on your own interviews

FAQ

Is reflexive thematic analysis the same as "thematic analysis"? "Thematic analysis" is an umbrella term covering several distinct approaches. Braun and Clarke's reflexive thematic analysis is one specific, well-documented version of it: the one most people mean when they cite "Braun and Clarke (2006)," even though the method itself has moved on since that original paper.

Do I need to calculate inter-rater reliability for reflexive TA? No. Doing so is actually inconsistent with the method's own logic. Inter-rater reliability assumes there's one correct, objective theme structure two coders should independently converge on. Reflexive TA treats theme development as an interpretive process shaped by the researcher, so it asks for transparency about that process instead.

How is thematic analysis different from axial coding? They come from different methodological traditions. Axial coding is a specific step inside grounded theory's coding paradigm; thematic analysis is a more flexible, standalone approach to identifying patterns across a dataset. See our guide: What Is Axial Coding?

Sources

  • Braun, V. & Clarke, V. (2006). "Using thematic analysis in psychology." Qualitative Research in Psychology.
  • thematicanalysis.ai – third-party guide to the Braun & Clarke framework (consulted for phase wording accuracy)
  • Towards Better Reflexive Thematic Analysis in HCI: A Scoping Review of Practice at CHI (CHI 2026 Proceedings) – 147-paper audit
  • Misra et al. (2026). "Large Language Models in Qualitative Analysis: Comparing Traditional and Researcher-Interpreted Approaches." SAGE.