Braun and Clarke updated their thematic analysis framework 3 times (2006, 2019, 2021), and most researchers are still applying the oldest version. The 2019 paper and 2021 book make one fundamental shift: coding is not systematic extraction of meanings from data, but an interpretive act shaped by the researcher's theoretical position. Reflexivity is not a box to tick; it is the method. If you are using their 2006 framework while also calculating inter-rater reliability, you are mixing incompatible paradigms.
Braun and Clarke's 2006 paper "Using Thematic Analysis in Psychology" is one of the most cited papers in all of psychology, with over 100,000 citations on Google Scholar. Ironically, this widespread adoption became part of the problem. The paper was read selectively, stripped of its epistemological context, and applied as a generic coding procedure across research traditions that are fundamentally incompatible with its assumptions. Their 2019 paper "Reflecting on Reflexive Thematic Analysis" and their 2021 SAGE book are, in part, a course correction. This bottom-up, engage-with-everything discipline is exactly what commercial deadlines have historically pushed researchers away from; see the bottom-up rigour problem in commercial research for why that trade-off no longer has to be automatic.
This guide explains what changed, which paper to cite, what reflexivity means in practice, and how to apply the updated framework in ways that will satisfy PhD examiners and peer reviewers. It is written by Henri Schildt, Professor of Strategy at Aalto University, who has published qualitative studies in the Academy of Management Journal and Strategic Management Journal, together with Olli Salo. If you are new to thematic analysis altogether, our accessible introduction is a gentler starting point. And because most researchers now do at least part of their analysis in software, we also cover how AI-assisted tools like Skimle fit a reflexive workflow without compromising it.
The 60-second video below summarises the key points: the four publications, the four things that changed, and the six phases.
What is the difference between reflexive thematic analysis and thematic analysis?
Thematic analysis is a family of methods, and reflexive thematic analysis is one member of it. Braun and Clarke now distinguish three types, and the differences matter because each one rests on different assumptions about what a theme is and how quality is judged. Mixing them, for example by claiming reflexive TA and reporting a kappa score, is the error reviewers most often flag.
| Reflexive TA | Codebook TA | Coding reliability TA | |
|---|---|---|---|
| Examples | Braun and Clarke (2019, 2021) | Framework analysis, template analysis | Boyatzis (1998), Guest et al. (2012) |
| What a theme is | A pattern of shared meaning the researcher constructs | A topic summary or a pattern of meaning, organised by a codebook | A topic summary, often decided early |
| Codebook | None; codes evolve throughout | Yes, developed partly before and partly during coding | Yes, fixed before coding |
| Researcher subjectivity | A resource to use and document | Managed through the shared codebook | A source of bias to control |
| Inter-rater reliability | Not used | Optional, rarely central | Central quality measure |
| Typical use | PhD projects, interpretive and critical studies | Applied and policy research, team projects | Larger teams wanting agreement metrics |
If your project needs a fixed codebook or agreement between coders, codebook or coding reliability TA is a better fit than forcing reflexive TA to do that job. Framework analysis is the most common codebook approach in applied health and policy research.
Which Braun and Clarke paper should you cite? (2006, 2019, 2021, 2022)
Skimle for your work
Skimle for academic researchers
Cite the paper that matches the claim you are making. The 2006 paper is still the standard reference for the six phases, but it does not use the term "reflexive", so a methods section that says "reflexive thematic analysis (Braun & Clarke, 2006)" cites a paper that does not describe the approach by that name.
| Year | Publication | Cite it for |
|---|---|---|
| 2006 | Using thematic analysis in psychology, Qualitative Research in Psychology | The six-phase process and thematic analysis in general |
| 2019 | Reflecting on reflexive thematic analysis, Qualitative Research in Sport, Exercise and Health | Naming your approach as reflexive TA |
| 2021 | One size fits all? What counts as quality practice in (reflexive) thematic analysis?, Qualitative Research in Psychology | Quality criteria, and why you do not use inter-rater reliability or saturation |
| 2021 / 2022 | Thematic Analysis: A Practical Guide (SAGE book) | The fullest account of the method, phase by phase |
| 2022 | Conceptual and design thinking for thematic analysis, Qualitative Psychology | Choosing between reflexive, codebook and coding reliability TA |
| 2022 / 2023 | Toward good practice in thematic analysis, International Journal of Transgender Health | A short checklist of common problems, useful for responding to reviewers |
Is it Braun and Clarke 2021 or 2022? Both appear in reference lists for the same book. SAGE published Thematic Analysis: A Practical Guide in October 2021, and many editions carry a 2022 copyright date, so you will see it cited either way. Use the year printed in the copy you read, and do not confuse the book with the 2021 "One size fits all?" article or the 2022 "Conceptual and design thinking" article, which are separate publications. Similarly, "Toward good practice" was published online in 2022 and appears in the 2023 volume.
The full references with DOIs, in APA 7 format:
For the book: Braun, V., & Clarke, V. (2021). Thematic analysis: A practical guide. SAGE. Journals generally prefer one of the articles above as the primary methods citation, with the book alongside it.
What is the difference between the 2006 and the updated reflexive approach?
The 2006 paper presented thematic analysis as a flexible method that could be used across different epistemological positions, from realist to constructionist. It described a 6-phase process and set a relatively accessible bar for entry. That accessibility was both its strength and its weakness.
By the 2019 paper, Braun and Clarke had observed a recurring pattern: researchers were applying thematic analysis as though it were a mechanical coding procedure, trying to achieve inter-rater reliability, treating saturation as a sample size criterion, and presenting themes as though they were discovered rather than constructed. These practices, they argued, reflect a fundamentally different paradigm (broadly, post-positivist coding reliability TA) rather than the interpretivist approach they had always intended.
Their updated framework makes four major clarifications:
1. Reflexive TA is theoretically positioned, not generic. It sits within a broadly constructionist or critical realist epistemological framework. Researchers who want a more positivist, reliability-focused approach should use codebook TA or coding reliability TA, which are different approaches that Braun and Clarke now distinguish explicitly.
2. Themes are constructed, not found. A theme is not a pattern that exists in the data waiting to be discovered. It is a pattern of meaning that the researcher constructs through their engagement with the data. That constructive role is the point of the method.
3. Researcher subjectivity is a resource, not a bias. The researcher's knowledge, experience, and theoretical commitments actively shape the analysis. This is acknowledged and used, not bracketed or controlled for.
4. Inter-rater reliability is not appropriate and should not be used. Because coding is interpretive and subjective, having two people independently code the same material and then measuring agreement treats coding as though there is a correct answer to be checked. There is not.
What are the 6 phases of reflexive thematic analysis?
The 6-phase structure survives from 2006, but the spirit of each phase is now sharper. The phases are iterative, not linear: you will move between them repeatedly rather than completing each one before starting the next. For a step-by-step walkthrough of all six phases with a worked example, see our dedicated guide.
Phase 1: Familiarisation with the data. Read all your data multiple times. Write initial ideas. This is active engagement, not passive reading, because you are already interpreting. Take notes on what strikes you. For interview studies, how to analyse interview transcripts covers this stage in detail.
Phase 2: Coding. Generate codes that capture something interesting about the data in relation to your research question. Codes are not labels; they are the researcher's interpretive observations. The same data item can carry multiple codes. Keep an analytic memo explaining your coding rationale. Our guide to how to code qualitative data shows worked examples of analytic rather than purely descriptive codes.
Phase 3: Generating initial themes. Group codes into potential themes. At this stage, themes are provisional: working ideas about patterns of shared meaning, not finished products. Create a thematic map.
Phase 4: Reviewing and developing themes. Test your themes against the data. Does each theme have enough data to sustain it? Is the data within a theme coherent? Does it tell a clear story? Merge, split, discard, and rename.
Phase 5: Refining, defining and naming themes. Write a detailed analysis of each theme. What is the essence of this theme? How does it connect to your research question? Name each theme to capture its analytical content, not just its topic.
Phase 6: Writing up. The write-up is part of the analysis, not just a report of it. Weave together data extracts and your interpretation. Quotes do more than illustrate a theme: the theme is argued through them.
Phases 2 to 4 are where most of the hours go, and where it is hardest to keep track of why a code moved or a theme split. If you want to try them on your own transcripts, Skimle has a free tier: it suggests candidate codes and themes, links every theme to the excerpts behind it, and leaves every decision to rename, merge or discard with you.
What does reflexivity actually mean in practice?
Recommended reading
AI for better qualitative research - Henri's presentation in New Scholars Webinar Sept 23rd, 2026
Reflexivity is one of those concepts that gets nodded at in methods sections and then disappears. In reflexive TA, it needs to be operational. For the wider concept, including the three types of reflexivity and how to write a positionality statement, see reflexivity in qualitative research.
Keep an analytic journal throughout. Write regular entries documenting your analytical decisions: why you coded something one way rather than another, what surprised you, where you felt uncertain, how your thinking evolved. This journal serves two purposes: it develops your thinking and it provides evidence of your process.
Articulate your positionality before you start coding. Who are you in relation to this topic? What assumptions do you bring from your professional background, lived experience, or theoretical training? Write this down in your methods section. It needs to be relevant rather than confessional.
Document your theoretical framework explicitly. Reflexive TA is epistemologically flexible, but not free-floating. If you are working from a feminist perspective, or a critical realist one, or a social constructionist one, say so. The framework shapes what you see in the data.
Revisit your reflexive account as analysis proceeds. Your initial positionality statement may look different after 50 hours of coding. Note the evolution.
Actively looking for evidence that challenges your reading is part of the same discipline. Henri's New Scholars talk on AI for better qualitative research covers how to do that, and how to stay transparent about your choices, when AI tools are part of the workflow.
How does negative case analysis fit reflexive thematic analysis?
Negative case analysis means deliberately looking for data that does not fit your emerging interpretation: the participant who contradicts a theme, the account that cuts against the pattern. It comes from analytic induction and was popularised as a credibility technique by Lincoln and Guba (1985), where the aim is to revise a hypothesis until it accounts for every case.
Reflexive TA uses the practice, but for a different purpose. Because themes are constructed interpretations rather than hypotheses to be verified, a contradictory case does not "disprove" a theme. It tells you something about the theme's boundaries and texture, and examining it is part of being reflexive about the story you are choosing to tell. In practice:
- During phase 4, ask of every theme: who does this not describe? Go back to the full dataset, not only the coded extracts, and look for participants whose accounts sit uneasily with the theme.
- Decide what the contradiction means. It might show the theme is too broad and should be split, that it holds only for a subgroup, or that participants hold two positions at once. Each is a finding.
- Write the tension into the theme. A theme that says "most participants…, although those in X described…" is richer than one presented as unanimous. Braun and Clarke's quality criteria reward themes with depth and nuance, not themes that appear in every transcript.
- Record the decision in your reflexive journal. Note which cases challenged your reading, how you responded, and whether your own position made you more or less likely to notice them.
- Report it briefly in the methods section. One or two sentences saying that you examined contradictory accounts, and how they shaped the final themes, answers the reviewer question before it is asked.
What to avoid is presenting negative case analysis as proof of validity ("no disconfirming cases were found, confirming the themes"). That framing belongs to a realist logic that reflexive TA does not share.
Hunting for contradictory accounts across 20 or 30 transcripts is slow by hand, which is why it is often skipped. Skimle keeps every theme linked to the excerpts behind it, so you can open a theme, read every passage coded to it, and search the rest of the dataset for accounts that cut against it. See how this fits academic research workflows.
What are 5 common mistakes researchers make when claiming to use reflexive TA?
Braun and Clarke have written extensively about misapplication. These five errors appear frequently in published work:
1. Calculating inter-rater reliability. As noted above, this is incompatible with reflexive TA. If you see an IRR calculation in a paper claiming reflexive TA, the authors have misunderstood the approach (or applied a different one without naming it). Peer reviewers are increasingly catching this.
2. Stopping data collection when themes "saturate". Saturation is a concept from grounded theory's logic of theoretical sampling. Reflexive TA does not use saturation as a stopping rule. Braun and Clarke are explicit about this: sample size decisions should be made in advance based on the research question and practical context.
3. Treating codes as the data rather than interpretations of it. Mechanical coding (applying the same label every time a concept appears) produces a content count, not thematic analysis. Coding should be analytical, asking what this excerpt says about the research question, not just what topic it addresses.
4. Naming themes as topics rather than analytical claims. "Work-life balance" is a topic. "The impossibility of balance in a culture that rewards overcommitment" is a theme. The theme label should convey what you found, not just what you looked at.
5. Treating the 6 phases as a checklist. Phases are a heuristic for a deeply iterative process. Reporting that you "completed phases 1 through 6" suggests a linear march through a procedure, the opposite of how the method works.
How do you write up reflexive TA for journal submission?
This is where many researchers struggle. The write-up must convincingly show two things: the analytical rigour of your process and the interpretive depth of your themes. Our broader guide on how to write up a thematic analysis covers the mechanics; this section focuses on what reflexive TA specifically requires.
Methods section. Explicitly name the approach as reflexive thematic analysis and cite Braun and Clarke (2019) or the 2021 book alongside the 2006 paper. Describe your positionality. Explain how coding proceeded (iteratively, with memoing). Justify your sample size in terms of depth rather than saturation (how many interviews is enough? covers the reasoning). Do not claim inter-rater reliability.
Results section. Each theme gets a subsection with an analytical label, a few lines explaining the theme's meaning, and 2-4 verbatim quotes presented with contextual attribution. The quotes illustrate the theme but do not constitute it. Your interpretive commentary must do work beyond paraphrase. See our guide on how to write a thematic analysis results section for the structural detail.
Quality criteria. Reflexive TA does not use traditional validity and reliability criteria. Instead, reviewers look for: coherence between epistemological position and method, richness and depth of analysis (not just the number of themes), transparency of the analytical process, and authentic reflexive engagement. Lincoln and Guba's trustworthiness framework (credibility, transferability, dependability, confirmability) is a recognised alternative.
The complete thematic analysis guide covers the full write-up process in more detail.
How does AI-assisted analysis fit with a reflexive approach?
Free webinar
Embracing AI in qualitative research, thoughtfully and responsibly
This is a real methodological question, not just a marketing one. Reflexive TA holds that interpretation is inherently human and researcher-situated. How does AI fit? The answer depends heavily on which tool you use, since they differ in whether the researcher can inspect and override every coding decision. Our comparison of qualitative analysis software sets out where NVivo, ATLAS.ti, MAXQDA and AI-native tools sit on exactly that question. Academic researchers weighing whether AI-assisted analysis can be reconciled with a reflexive stance are asking the right question before adopting a tool, not after.
Henri covers this question at length in his free webinar on embracing AI in qualitative research thoughtfully and responsibly.
AI tools change the mechanics of coding, not its interpretive nature. What AI can do (quickly surface candidate patterns across a large corpus, flag which excerpts might be relevant to an emerging theme, identify where your coding is inconsistent) reduces the cognitive load of the procedural work. What it cannot do is replace your interpretive judgement. How to do thematic analysis with AI walks through this division of labour step by step.
In practice, a reflexive approach to AI-assisted analysis means: treating AI-generated theme suggestions as prompts to your own analysis, not final outputs; being transparent in your methods section about what AI tools did and how you engaged critically with their outputs (our 20-question checklist for AI-assisted analysis is built for exactly this); and maintaining an analytic journal that documents how your interpretation shaped the final themes. For the ethics and transparency questions in more depth, see using AI in qualitative research: a guide for academic researchers.
Skimle's approach to this is to show every insight traced back to the specific excerpts it derives from, so you can interrogate and revise the AI's pattern recognition with your own interpretive lens. The categories view and analyst notes features are designed for exactly this kind of iterative, reflexive engagement. For academic researchers, the predefined categories mode keeps the researcher in control of how themes are structured.
If you are an academic using Skimle, the academic researchers page has more on how the tool supports rigorous qualitative workflows.
Frequently asked questions
Can I use reflexive thematic analysis for a deductive study?
Reflexive TA can incorporate deductive elements (you can start with a theoretical framework that shapes what you code for), but its primary orientation is inductive or abductive. If your study is strongly deductive (testing predefined categories against data), codebook TA or framework analysis may be more appropriate. See our guide on inductive, deductive, and abductive coding for a fuller comparison.
How many themes should a reflexive thematic analysis produce?
There is no correct number. Most published reflexive TA studies identify 3-6 themes, but this depends entirely on the dataset, the research question, and the analytical depth you are aiming for. More themes does not mean better analysis. If anything, the pressure is to consolidate: if you have 12 themes, most of them probably belong together under a more analytically ambitious heading.
Do I need to use member checking in reflexive TA?
Member checking (returning interpretations to participants for validation) is sometimes associated with qualitative rigour, but it is not required or expected in reflexive TA. In fact, returning your constructionist interpretations to participants, who may not share your analytical framework, can create confusion rather than validation. Braun and Clarke's quality criteria for reflexive TA do not include member checking. Transparency of process, reflexive engagement, and analytical depth are the relevant criteria. If you want participants involved, Braun and Clarke (2023) suggest "member reflections" instead: inviting participants to reflect on the analysis to add insight and generate further data, rather than to confirm that you got it right.
How do I reference reflexive thematic analysis correctly?
Cite the 2006 paper for the six phases and the 2019 paper, the 2021 "One size fits all?" article or the book for the reflexive framing. Citing only 2006 while calling your method reflexive thematic analysis is the most common referencing error. The citation table above lists every reference with its DOI.
What are the limitations of reflexive thematic analysis?
Reflexive TA is not the right method for every project, and reviewers increasingly expect authors to say so explicitly.
Inter-rater reliability does not apply. This is the most frequent source of confusion. Braun and Clarke are explicit that calculating inter-coder agreement (Cohen's kappa and similar) is conceptually incoherent in reflexive TA: the method treats the researcher's interpretation as a resource rather than a source of bias, so two coders converging on identical codes demonstrates training, not validity. If a reviewer asks for a kappa score, the appropriate response is to explain the epistemological mismatch, not to compute one. Use a codebook or coding reliability variant instead if your project requires agreement metrics.
It does not produce generalisable findings. Reflexive TA generates situated interpretations, not population estimates.
It is difficult to do well at speed. The method depends on prolonged engagement with the data. Analyses produced in a few days typically show it, usually as themes that are really just topic summaries.
It requires an epistemological position you can defend. Reflexive TA sits within a broadly constructionist frame. Pairing it with a realist framing of your findings produces the incoherence reviewers most often flag.
Ready to try reflexive thematic analysis on your own data? Create a free Skimle account and run phases 2 to 4 on a few transcripts. Every theme stays linked to the excerpts behind it, and every naming, merging and discarding decision stays yours.
Related reading:
- Guide for PhD students doing qualitative analysis of interviews
- Reflexive thematic analysis: the six phases with a worked example
- How to write a thematic analysis results section
- Reflexivity in qualitative research: types, examples and positionality statements
- Inductive, deductive, and abductive coding: when to use each
- The complete guide to thematic analysis
About the authors
Henri Schildt is a Professor of Strategy at Aalto University School of Business and co-founder of Skimle. He has published over a dozen peer-reviewed articles using qualitative methods, including work in Academy of Management Journal, Organisation Science, and Strategic Management Journal. His research focuses on organisational strategy, innovation, and qualitative methodology. Google Scholar profile
Olli Salo is a former Partner at McKinsey & Company where he spent 18 years helping clients understand the markets and themselves, develop winning strategies and improve their operating models. He has done over 1000 client interviews and published over 10 articles on McKinsey.com and beyond. LinkedIn profile
Sources
- Using thematic analysis in psychology - Braun & Clarke (2006), Qualitative Research in Psychology
- Reflecting on reflexive thematic analysis - Braun & Clarke (2019), Qualitative Research in Sport, Exercise and Health
- Thematic Analysis: A Practical Guide - Braun & Clarke (2021), SAGE
- One size fits all? What counts as quality practice in (reflexive) thematic analysis? - Braun & Clarke (2021), Qualitative Research in Psychology
- Conceptual and design thinking for thematic analysis - Braun & Clarke (2022), Qualitative Psychology
- Toward good practice in thematic analysis - Braun & Clarke (2023), International Journal of Transgender Health, PMC
- Lincoln, Y. S., & Guba, E. G. (1985). Naturalistic inquiry. SAGE.
