Reflexive Thematic Analysis – The What, Why and How

By Richard Devine, Social Worker for Bath and North East Somerset Council

NOTE: If you are receiving this via e-mail it may be cut short by your e-mail programme and/or the graphics may be distorted. You may wish to click the link and view it in full.

Preface: 

I recently secured a position of British Association of Social Workers In-Practice Researcher at Cambridge University with Dr. Robbie Duschinsky and his team.  As part of this 12-month position, I am involved in undertaking some research with Robbie and Sarah Foster, a collaborator based at Northumbria University. At an initial meeting with Robbie and Sarah, who are, by any standard, incredibly kind and encouraging, I was invited to explore a data set with over 60 interviews with GP’s, Psychologists and Social Workers about social work, risk, and application of theory to practice issues. Robbie suggested that I could undertake a qualitative exploratory study. ‘What on earth is that?!’ I thought to myself, whilst outwardly agreeing with this proposal – as if it was perfectly sensible that someone with no academic background would endeavour to get a paper published. 

A few days later, after I was sent the data set, I opened it up eagerly, excited about delving into the interviews. The first group of interviews I opened, out of the three, was 500 pages. 500 pages! How would I even begin to make sense of that?! I read three interviews, which took over 3 hours, and then I never looked at the data set again for a month. 

Two barriers contributed to this procrastination. Firstly, I felt demoralised by the realisation that reading this data would take far longer than I imagined. Secondly, I had no idea what I was looking for. I had expected some framework, guidance, or instruction on what to look for. After this initially bruising encounter with the data, I put aside a few days to commit to reviewing the data. I managed to read approximately half of all the interviews during this time. Whilst I did so, I highlighted certain sections, especially around the professionals’ response to domestic abuse, primarily because this is an area I am interested in. In addition, I would write notes to track certain thoughts whilst reading the interviews. 

After this process, I e-mailed Sarah and Robbie and shared with them several themes that I had identified. However, I was concerned about the approach I was adopting and wrote the following in my e-mail to them: 

‘When I was reading the transcripts, I realised that there was nothing objective in the approach I was adopting. I was constantly in a process of comparing what I would say, in contrast to what the participant was saying, and the areas or gaps that I identified are based upon certain knowledge I have acquired (either through formal learning or experience)’. 

Wasn’t research supposed to be objective? Deliberately designed to reduce bias and remove the self. 

Both Robbie and Sarah replied to my e-mail. I was relieved to read the following from Sarah,

‘You mentioned that your approach has not been objective and that you have brought your subjectivity into the process.  This is quite normal in quali[tative] research and indeed your subjectivity can be used as a research tool’. 

Amongst many helpful ideas, one suggestion was to read the work of Braun and Clarke.

I can’t say I was overly enamoured at the prospect of reading a book on research methodology. I was very pleasantly surprised, however. Braun and Clarke are extremely engaging, interesting, accessible writers, and would you believe, fun to read. A fun-to-read book on research methodology – now, that’s impressive. They share an infectious passion for qualitative research that pervades their writing.

I read a couple of their papers before reading their most recent book, Thematic Analysis: A Practical Guide (2021). I will outline some key ideas from part 1 of that book here (Note: this is a poor substitute for the actual book, which I can’t recommend highly enough).

Introduction: Research as ‘an adventure, a journey of exploration and discovery’ (p.xxvii)

Braun and Clarke describe reflexive thematic analysis (RTA) as ‘a theoretically flexible method’ (4), for ‘developing, analyzing and interpreting patterns across a qualitative dataset’ (p.4). 

A central component of RTA is that the researcher’s position and contribution is necessary, unavoidable, and an integral ingredient of the process. In other words, the researcher and their subjectivity are tools to consciously and actively utilise. It is not something to remove, reduce, avoid, or minimise, but a valuable resource to be drawn upon. 

To capitalise on this resource, Braun and Clarke invoke the use of the term ‘reflexive’.  Reflexivity involves drawing upon your experiences, pre-existing knowledge, and social position (such as ethnicity, gender, class, etc) and ‘critically interrogating’ (p.5) how these aspects influence and contribute to the research process and potential insights into qualitative data. The invites the researcher to explore, understand, bring forth and make explicit their values, ideas about themselves, the world, and their beliefs. Then consider how, not if, these influence how they interpret and make sense of the research. Reflexive research demands that knowledge is treated as situational – always a consequence of an interaction between the researcher and the data. 

Analysis and interpretation under this conceptualisation can not make simple claims to truth or objectivity, Nevertheless, an analysis can be weak or strong. A reflexive account of the researcher, methodological approach, and situational nature of the research can be unconvincing, superficial (i.e., weak), or compelling and characterised by depth and thoughtfulness (i.e., strong). 

4 Domains of Reflexive Thematic Analysis: 

Braun and Clarke outline 4 domains of reflexive thematic analysis (orientation to data, focus of meaning, qualitative framework, theoretical frameworks) and each one reflects orientations to data. These polarities or dimensions are not mutually exclusive, thus will often overlap. 

Orientation to data 

Less reductive – more reductive: Inductive analysis is an outcome of reviewing the data and identifying the standout themes. It presupposes, in its purest form, that participants’ voices can be accurately captured and fairly represented objectively and without any subjective interference. Braun and Clarke point out that this can’t be fully realised because of what we, as researchers, inevitably and unavoidably bring to the analysis ‘as theoretically embedded and socially positioned researchers’ (p.56). In contrast, deductive analysis begins explicitly with the researcher, in particular, a theoretical or conceptual model that is used as a framework and lens to interpret and extract meaning from the data.  A researcher can bring a theoretical interest into the data set or realise soon after analysing the data that the nature of the themes identified would be best served and illuminated when interpreted through a particular theoretical framework.  

Focus of meaning 

Semantic -Latent: Semantic analysis explores the meaning on a surface level, drawing out themes that are readily and explicitly identified. The researcher remains close to the meaning overtly articulated by the participants, thus tending to produce a more descriptive analysis of the data. Latent analysis, conversely, focuses on exploring the underlying, covert, and often implicit meaning of the data. They tend to be derived from the researcher, or theory-seeking connections and meaning that aren’t self-evident and require some abstracting from the data.   

Qualitative framework 

Experiential – Critical: An experiential analysis focuses on the participants’ voice, drawing out and highlighting their lived experiences and perspectives. It assumes that language is used to accurately communicate meaning. Whereas a critical analysis focuses on a topic or issue and organizing the participants’ contribution around that. 

Theoretical frameworks 

Realist, essentialist – Relativist, constructionist: A realist, essentialist approach endeavors to find the reality and truth encapsulated within the data set. It presupposes that there is an objective reality to be extracted from the data and reported on. A relativist, constructionist approach on the other hand, seeks to examine the meaning the realities expressed within the data set proport. In other words, it attempts to understand the social construction of meaning articulated by the participants. There is no objective reality to be mined from the data because reality is a manifestation of individuals’ and societies sense-making. Therefore, the act and products of this sense-making are what come under scrutiny in the relativist, constructionist approach.

Braun and Clarke note that inductive, semantic, experiential and realist, essentialist approaches tend to group together and deductive, latent, critical and relativist, constructionist group together. However, these dimensions are not dichotomous, and research will often contain elements of multiple dimensions. Recognizing these dimensions however is key, not least to facilitate the researchers’ understanding of the different approaches, but also to make explicit within the methodology which approach has been favored and why.      

The 6 stages of Thematic Analysis: 

Braun and Clarke use the term ‘analytic sensibility’ to refer to the skill of ‘reading and interpreting data to produce insights into your dataset that go beyond the obvious or surface-level content, and to noticing connections between the dataset and existing research, theory and the wider context’ (p.45). The authors contend that this analytic skill needs to be situated within a systematic framework for undertaking qualitative research, which the authors detail in their 6-stage model.   

Phase 1: Familiarisation 

This first phase involves two stages. Initially, it involves rigorously engaging with the data. The primary goal here is to immerse yourself in the data and become deeply familiar and intimate with it. There is no short-cut here. It simply involves reading and re-reading the data. This is an especially important phase if you, like me, were not involved in the process of collating the data. The level of familiarity achieved should be such that if the data was suddenly lost, you’d be at a point where you could remember and recall, broadly, the content reasonably well. 

Once familiarity has been achieved, which involves being close to the data, then the research can move onto the next stage which involves critical engagement, and this involves creating some distance with the data. In this stage, the researchers approach moves on from reading and familiarising into engaging critically and reflexively asking questions of yourself (remember, the self is a fundamental element of reflexive TA) and the data. 

Braun and Clarke (p.43, 44) provide some helpful questions to facilitate this process, including: ‘How does the person make sense of whatever it is they are discussing? Why might they be making sense of things in this way (and not in another way)? In what different ways do they make sense of the topic? How ‘common – sense’ or socially normative is this depiction or story? How would I feel if I was in that situation? (Is this different from or similar to how the person feels, and why might that be?) What assumptions do they make in describing the world? What kind of world is ‘revealed’ through their account? Why might I be reacting to the data in this way? What does my interpretation rely on? ‘

This two-staged process can be aided through a third, complementary element of making notes throughout. Braun and Clarke advocate to ‘give yourself dedicated time and space for this’ (p.42) as this phase often takes more time than anticipated. Upon reflection, it could hardly be said that allowing 3 hours to read 500 pages of interviews is ‘dedicated time and space’!’

Phase 2: Coding 

Coding involves methodically reviewing the data and searching for segments that appear interesting, relevant, provoking – in relation to the research question – and then writing brief descriptions (codes) next to them. It is an ‘exploratory’ process (p.64). 

There are two important reasons for coding: insight, and rigour. Coding facilitates insight because it requires purposeful, critical engagement with the data. Rigour, because coding is a systematic interrogation of the data, identifying meaning and patterns across the dataset. Put simply, the researcher is reviewing the depth and breadth of the data. The researcher is mining for ideas and patterns that might otherwise be overlooked in the absence of such scrutiny

Applying insight and rigour facilitates engagement with the data and is intended to yield an intimate understanding of the data and new meanings. It is also an antidote to ‘cherry picking’ whereby the researcher is highly selective about the data, extracting out elements that fit a pre-existing idea or bias. 

As already noted, Reflexive Thematic Analysis encourages the use of the self, and recognition of the use of the self, in all stages of the process and that includes the coding phase. Coding is shaped by the subjectivity and position of the researcher and is in fact, greatly enhanced by this. Braun and Clarke note that coding can often be categorised into the 4 domains of thematic analysis detailed above (less-more reductive, semantic-latent, experiential-critical, Realist, essentialist – Relativist, constructionist). The authors strongly encourage bearing these in mind when coding. Awareness of the researcher’s subjective position and these four domains whilst coding is important in supporting the production of high-quality, cohesive, evidentially sound analysis when engaged in the process of writing up at a later point. 

Practically, coding can be achieved on Microsoft word by adding comments or manually using good old-fashioned, pen and paper. Braun and Clarke prefer the latter, positing that they are more able to engage productively with the material when read as a hard copy compared with electronically. Two rounds of coding are recommended to exploit the utility of the coding process. Reviewing and coding the data in a different order, for example from the end to the beginning, can help the researcher remain freshly engaged with the data. This technique can prevent glazing over which can easily happen when looking at data repeatedly.

To summarise, ‘coding is an organic and evolving process, capturing an interweaving of the knowledge, subjectivity and analytic skill of the researcher, engaging closely and systematically with the dataset’ (p.72).

Coding ‘is about demarcating the variation in the dataset, in order to develop themes robustly based on clusters of pertinent similar meaning’ (p.69).

Phase 3: Generating Initial Themes

Braun and Clarke write ‘a theme captures the patterning of meaning across the dataset’. (p.76). The authors note that they have previously referred to them as ‘shared meaning’, ‘conceptual pattern’, or ‘fully developed themes’ (p.77). Each of these terms, whilst rejected in favour of the term ‘theme’, capture something about the essence of a theme as, ‘a pattern of shared meaning organised around a central concept’ (p.78).  A theme is not to be confused with a topic summary, which is an examination of the different responses or meanings around a specific topic. 

In the initial stages of the process, the codes that have identified a single idea can be clustered together when there is some shared meaning. A theme doesn’t necessarily need to be bounded by a single idea in the way that a code might be, rather it can contain multiple facets, albeit contained by a ‘central organising concept’ (p.80). Themes are to be generated deliberately and thoughtfully by the researcher, drawing upon and integrating their position, knowledge, and intellectual interests to aid this process. To this effect, it isn’t a passive process whereby the themes are self-generated and the researcher simply finds them – it’s a much more rigorous, self-involved, and arguably, exciting process.

During the early stages, the goal is to engage with the data and codes to generate multiple plausible themes that have potential, especially pertaining to the research question. It can be an open, fun, and creative endeavour. Some of the themes may not ultimately be used, so it is advised not to get too attached or fixed on a certain theme. With that said, you don’t want an endless number of themes. Whilst it is possible to generate more themes that you will eventually use, Braun and Clarke recommend that for an 8000-word report two-six themes will suffice.  

A theme doesn’t have to capture everything in the data; however, each theme should have a central organising concept, and preferably be distinctive, bounded and be able to stand out alone. Braun and Clarke provide the following questions to assist in the process, ‘Does this provisional theme capture something meaningful? Is it coherent, with a central idea that meshes the data and codes together? Does it have clear boundaries?’ (p.84). 

In the same way that themes will not simply appear and require the ingenuity of the researcher, the themes will not explain themselves once found. The researcher will need to interpret, elaborate, and explain the relevance of the themes to the research question. It is the researcher’s role to explain the process of theme identification and selection, but also provide a rich, contextualized narrative as to why they matter. 

Phase 4: Developing and Reviewing Themes 

As an extension of phase three, phase four provides an opportunity to review, check and adjust the themes. Reviewing the data set as a whole, and especially the coded data extracts allow the researcher to clarify an understanding of how closely aligned the themes developed in phase three remain to the data. It enables the researcher to check that the themes thus identified have an evidentially and/or compelling story to tell (that is tied to the data). Or ensure the themes aren’t too disconnected from the data set. The latter issue can occur if the researcher, without awareness, projects an idea or a theme that they want to identify from the data. 

This recursive process facilitates a rigour and functions as an insurance policy against weak, fragmented themes that can’t be backed up by the data.  By the end of this phase, you want to have established well-worked, sophisticated, and robust themes that provide a rich, nuanced, and engaging analysis that addresses your research question. This may involve deciding to lose some themes – apparently not uncommon – or adjusting pre-existing themes that take into consideration your re-examination of the data having already undertaken the former three phases. Psychologically, this might be the most challenging stage because it requires letting go of some or all themes altogether. Losing some aspect of a particular theme might be necessary to transform it into a distinctive theme organised around a single central idea. Redevelopment of a theme should be expected and illustrates a sign of effectively and recursively engaging with the data set.

Some useful questions Braun and Clarke (p.98-99) pose to facilitate this phase: ‘Can I identify the boundaries of these theme’, ‘Are there enough meaningful data to evidence this theme?’, Are the multiple articulations around the core idea, and are they nuanced, complex, and diverse?’, ‘Does the theme feel rich?’, ‘Are the data contained within each theme too diverse and wide ranging?’, Does the theme convey something important?’

The authors repeatedly emphasize that the commonality or frequency of an issue does necessarily denote importance nor that it should automatically qualify it as a theme. Instead, the relevance and applicability to the research question is crucial. 

Phase 5: Refining, defining, and naming themes

A key aid in refining and defining your theme is to create an abstract for your theme. The development of an abstract is a test for your theme (a kind of shock test). In the abstract of a theme, you should be able to explain the central organising principle, convey the uniqueness of the theme and how it will contribute to the overall analysis. Even at this late stage, there is an opportunity to revise, or even more severely, let certain themes go altogether. Writing the abstract will require the researcher to succinctly capture the gist, and this process will facilitate rigour and identify any further need for refinement. 

Another important task is to name the theme. Braun and Clarke encourage researchers to be creative and playful during this process and identify a theme name that is ‘informative, concise and catchy’ (p.111). An effective theme name has two important functions. Firstly, it provides the reader an indication, and hopefully an enticing indication, as to the unique, informative analysis your paper will contain. Secondly, a poorly developed theme name that fails to capture the essence of the theme, misrepresents the analysis but also indicates that the theme is, in fact, a topic summary! This is not the desired outcome of Reflexive TA. Relatedly, Braun and Clarke advocate avoiding one-word theme names because whilst they can capture a topic, it is typically insufficient to capture the richness and full meaning encapsulated in a theme. 

Phase 6: Writing up: ‘Arriving home and telling a story about your adventure’ (p.118)

‘We want to write in a way that invites our readers to stay, and ultimately rewards them for staying. A recognition of the gift of their time and attention’ (p.118).

Writing up as storytelling: Braun and Clarke propose that an analysis should intrigue and engage the reader whilst simultaneously convincing them of the validity and robustness of the arguments and themes the researcher has identified. In some sense, you are describing your journey of exploration and discovery, however, the process of writing this is also explorative and thus is the final part of your adventurous journey. There still remains an opportunity to refine and reconceptualise some aspects of your journey, including the themes. 

‘Establishing the gap model’ versus ‘making the argument model’: Braun and Clarke identify two approaches to writing up the analysis, ‘establishing the gap model’ and ‘making the argument model’ (p.120). The former is predicated on the notion that gaps in knowledge can be identified through an absence of research in a particular area or the current research on a topic is lacking in some way, and importantly, these gaps need to be identified and filled. The research agenda is therefore positioned to resolve this gap. Braun and Clarke argue that this approach ‘effectively reproduces a positivist-empiricist idea of research as truth-questing’ and isn’t congruent with a qualitative approach which is localised, contextualised and endeavours to contribute to ‘a rich tapestry of understanding that we and others are collectively working on, in different places, spaces and times’ (p. 120). For this reason, the authors advocate for the ‘making an argument model’ as this places the rationale for the research in the context of pre-existing knowledge and theory. Instead of showing that you have found a piece of truth that the rest of humanity has yet to discover, you take a more modest, yet equally valuable approach, of providing a rich, contextualized, compelling representation of a certain issue. For this reason, in most but not all instances, a literature review doesn’t need to be undertaken until after the analysis of the data has been completed. You’ve yet to uncover what you will discover and therefore you don’t know what knowledge or theory will be relevant.  

Methodology: This section involves writing about what you did (descriptive) and why you choose the approach (analytical). This includes documenting how you integrated Reflexive Thematic Analysis principles and phases as well as how the use of the self, influenced the process. Braun and Clarke caution against simply detailing the six phases because this denotes a linear recipe that can be applied to research divorced from any researcher influence and input.  Taking ownership for how you, in your unique position, have integrated these general phases illustrates transparently the methodological tools used to interact with the data and how this has contributed to the findings. In other words, the researcher should clearly be visible and for this reason, first-person language is recommended.  

Writing about themes: The aim is to produce a story about each theme that should have a distinct element, yet connections between themes can be made such that a thread runs throughout. 

To begin with, a brief summary of each theme is suggested. 

Thereafter, each theme is explored drawing upon extracts from the data. Braun and Clarke advise a 50/50 ratio of data with analysis. Including data extracts serve two important functions. Firstly, it provides an evidential basis for your analytical claims and secondly, allows the reader to evaluate the validity of the claims made based on the source data. 

Braun and Clarke provide several tips for selecting extracts: Select rich and engaging extracts. Select extracts from across the data set and don’t rely on one participant in the data set. Use a range of quotations for each theme – aim for data that shows the depth and breadth of the central organising concept. Select extracts that clearly and concisely illustrate the analytic claims. Remember though, the extracts don’t speak for themselves and will require the researcher to interpret the data effectively and convincingly. Avoid paraphrasing the data. Avoid repeating extracts and edit out any unnecessary material. Some data will require contextualising, otherwise it might be unclear.

The authors make a distinction between ‘illustrative’ and ‘analytical’ (p.105) ways of reporting on the data. When using data illustratively, the researcher provided a thorough, rich, and nuanced account of a theme and the data could essentially be inserted afterward, to further exemplify the argument being made. To this effect, the analysis would be coherent even without the insertion of the data. Using data analytically involves an explicit integration of the data with the analysis, with specific elements of the data being commented upon the data. In contrast to an illustrative approach, it would be difficult to make sense of the analysis without the data.  

Conclusion: ‘Conclusions are the ultimate so what of your story’ (p.146). A conclusion is an opportunity to explain what can be taken away from this research. In characteristically helpful fashion, Braun and Clarke provide six domains that can be used to draw conclusions from: 

1. Conclusions about the data and the analysis – what do we now know that we didn’t before and what implications could be derived from this?
2. Conclusions related to existing literature and knowledge – how does this research contribute, add something new, or differ from pre-existing knowledge?
3. Conclusions about methodology – what does the method used reveal about the utility of the method to examine the issues in the study and what effect did the methodology have on finding meaning from the data? 
4. Conclusions about theory – in what way does the analysis support or contradict theory on this explored issue?
5. Conclusions about practice – what are the implications, if any, for how the findings should or could inform and alter practice? 
6. Conclusions for wider society – what do the findings mean, more broadly, for narratives, social structures, intervention and/or policy? 

This list isn’t exhaustive and there may be other domains from which to conceptualise the conclusion. It is likely that more than one of these areas will be used in the conclusion.  

Braun and Clarke emphasize that the conclusion does not necessitate a conclusive and complete analysis. The intention of RTA is to contribute to the pre-existing body of knowledge by providing rich, contextualised, creative, and thus original research undertaken in the spirit of curiosity and reflexivity.  To this effect, generalisability is made cautiously with acknowledgment of the ways in which the researcher, the research design, the participants, and the context shaped the findings. 

A couple of final points. Firstly, allow time for editing. Every stage, including editing, takes longer than anticipated. To do justice and to honour the participants and your own contribution to this research allow this final, important stage sufficient time. Secondly, if permitted, share reflections from first person perspective about the research choices made and the learning derived from those choices and the process of producing research. Not only is this in keeping with the ethos of RTA it will allow other researchers to learn from your experiences. 

Concluding thoughts: 

Braun and Clarke conceptualize their approach, and qualitative research in general, as an adventure. I found this to be an extremely useful and engaging analogy to describe the process. Consequently, I was able to reframe a daunting, and slightly overwhelming process as an adventurous one that will undoubtedly be challenging, but attainable and rewarding. As pointed out by the authors, ‘you need to be psychologically ready for a rich, unexpected, sometimes frustrating, but ultimately achievable adventure’ (p.76).

Braun and Clarke’s book also partially disabled my inclination to be caught up with achieving an outcome – getting to an endpoint – which inevitably results in dissatisfaction with the present (all that there ever is). The journey analogy has reoriented my focus on the process – the journey, the adventure, and all that it has to offer. A psychologically healthier position, surely.

On a final note, I am fearful that in summarising Braun and Clarke’s book, I have stripped away the sense of humour and fun that imbues their writing. I can’t offer much in the way of compensation, so in the most unoriginal fashion, I googled jokes about qualitative research. In the spirit of the author’s intelligent wit, I’ll end here with this joke I found: 

‘I have a joke about qualitative research, but everyone is asking me to provide statistical evidence’.

(Source: https://twitter.com/drrobertsakic/status/1287036927158505473?lang=en)

By Richard Devine (12.11.21)