See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/283600359

Using thematic analysis in psychology, Qualitative

Research in Psychology, 3

Article · January 2006

CITATIONS

1,955READS

2,964

2 authors:

Some of the authors of this publication are also working on these related projects:

Auckland Medical Aid Trust postdoctoral fellowship on Maori and reproduction View project

Collecting Qualitative Date: Beyond the face-to-face interview View project

Virginia Braun

University of Auckland

107 PUBLICATIONS   35,840 CITATIONS   

SEE PROFILE

Victoria Clarke

University of the West of England, Bristol

119 PUBLICATIONS   35,342 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Victoria Clarke on 14 May 2018.

The user has requested enhancement of the downloaded file.

Using thematic analysis inpsychology

Virginia Braun1 and Victoria Clarke2

1University of Auckland and 2University of the West of England

Thematic analysis is a poorly demarcated, rarely acknowledged, yet widelyused qualitative analytic method within psychology. In this paper, weargue that it offers an accessible and theoretically flexible approach toanalysing qualitative data. We outline what thematic analysis is, locating itin relation to other qualitative analytic methods that search for themes orpatterns, and in relation to different epistemological and ontologicalpositions. We then provide clear guidelines to those wanting to startthematic analysis, or conduct it in a more deliberate and rigorous way, andconsider potential pitfalls in conducting thematic analysis. Finally, weoutline the disadvantages and advantages of thematic analysis. Weconclude by advocating thematic analysis as a useful and flexible methodfor qualitative research in and beyond psychology. Qualitative Research inPsychology 2006; 3: 77 �/101

Key words: epistemology; flexibility; patterns; qualitative psychology;thematic analysis

Thematic analysis is a poorly demarcated

and rarely acknowledged, yet widely used

qualitative analytic method (Boyatzis,

1998; Roulston, 2001) within and beyond

psychology. In this paper, we aim to fill

what we, as researchers and teachers in

qualitative psychology, have experienced

as a current gap �/ the absence of a paperwhich adequately outlines the theory, ap-

plication and evaluation of thematic ana-

lysis, and one which does so in a way

accessible to students and those not parti-

cularly familiar with qualitative research.1

That is, we aim to write a paper that will

be useful as both a teaching and research

tool in qualitative psychology. Therefore,

in this paper we discuss theory and

method for thematic analysis, and clarify

Correspondence: Virginia Braun, Department of Psychology, University of Auckland, Private Bag 92019, Auckland,New Zealand.E-mail: [email protected]

# 2006 Edward Arnold (Publishers) Ltd 10.1191/1478088706qp063oa

Qualitative Research in Psychology 2006; 3: 77 �/101www.QualResearchPsych.com

the similarities and differences betweendifferent approaches that share features incommon with a thematic approach.

Qualitative approaches are incrediblydiverse, complex and nuanced (Hollowayand Todres, 2003), and thematic analysisshould be seen as a foundational methodfor qualitative analysis. It is the firstqualitative method of analysis that re-searchers should learn, as it provides coreskills that will be useful for conductingmany other forms of qualitative analysis.Indeed, Holloway and Todres (2003: 347)identify ‘thematizing meanings’ as one of afew shared generic skills across qualitativeanalysis.2 For this reason, Boyatzis (1998)characterizes it, not as a specific method,but as a tool to use across different meth-ods. Similarly, Ryan and Bernard (2000)locate thematic coding as a process per-formed within ‘major’ analytic traditions(such as grounded theory), rather than aspecific approach in its own right. Weargue thematic analysis should be consid-ered a method in its own right.

One of the benefits of thematic analysis isits flexibility. Qualitative analytic methodscan be roughly divided into two camps.Within the first, there are those tied to, orstemming from, a particular theoretical orepistemological position. For some of these�/ such as conversation analysis (CA; eg,Hutchby and Wooffitt, 1998) and interpre-tative phenomenological analysis (IPA; eg,Smith and Osborn, 2003) �/ there is (as yet)relatively limited variability in how themethod is applied, within that framework.In essence, one recipe guides analysis. Forothers of these �/ such as grounded theory(Glaser, 1992; Strauss and Corbin, 1998),discourse analysis (DA; eg, Burman andParker, 1993; Potter and Wetherell, 1987;Willig, 2003) or narrative analysis (Murray,2003; Riessman, 1993) �/ there are different

manifestations of the method, from withinthe broad theoretical framework. Second,there are methods that are essentially in-dependent of theory and epistemology, andcan be applied across a range of theoreticaland epistemological approaches. Althoughoften (implicitly) framed as a realist/experi-ential method (Aronson, 1994; Roulston,2001), thematic analysis is actually firmlyin the second camp, and is compatible withboth essentialist and constructionist para-digms within psychology (we discuss thislater). Through its theoretical freedom, the-matic analysis provides a flexible and use-ful research tool, which can potentiallyprovide a rich and detailed, yet complex,account of data.

Given the advantages of the flexibility ofthematic analysis, it is important that we areclear that we are not trying to limit thisflexibility. However, an absence of clear andconcise guidelines around thematic analysismeans that the ‘anything goes’ critique ofqualitative research (Antaki et al ., 2002) maywell apply in some instances. With thispaper, we hope to strike a balance betweendemarcating thematic analysis clearly �/ ie,explaining what it is, and how to do it �/ andensuring flexibility in relation to how it isused, so that it does not become limited andconstrained, and lose one of its key advan-tages. Indeed, a clear demarcation of thismethod will be useful to ensure that thosewho use thematic analysis can make activechoices about the particular form of analysisthey are engaged in. Therefore, this paperseeks to celebrate the flexibility of themethod and provide a vocabulary and‘recipe’ for people to undertake thematicanalysis in a way that is theoretically andmethodologically sound.3 As we will show,what is important is that as well as apply-ing a method to data, researchers maketheir (epistemological and other) assump-

78 V Braun and V Clarke

tions explicit (Holloway and Todres, 2003).Qualitative psychologists need to be clearabout what they are doing and why, and toinclude the often-omitted ‘how’ they didtheir analysis in their reports (Attride-Stirling, 2001).

In this paper we outline: what thematicanalysis is; a 6-phase guide to performingthematic analysis; potential pitfalls toavoid when doing thematic analysis; whatmakes good thematic analysis; and advan-tages and disadvantages of thematic analy-sis. Throughout, we provide exam-ples from the research literature, and ourown research. By providing examples, weshow the types of research questions andtopics that thematic analysis can be used tostudy.

Before we begin, we need to define a fewof the terms used throughout the paper.Data corpus refers to all data collected fora particular research project, while data setrefers to all the data from the corpus thatare being used for a particular analysis.There are two main ways of choosing thedata set (which approach you take dependson whether you are coming to the datawith a specific question or not �/ see ‘Anumber of decisions’ below). First, the dataset may consist of many, or all, individualdata items within your data corpus. So, forexample, in a project on female genitalcosmetic surgery, Virginia’s data corpusconsists of interviews with surgeons,media items on the topic, and surgeonwebsites. For any particular analysis, herdata set might just be the surgeon inter-views, just the websites (Braun, 2005b), orit might combine surgeon data with somemedia data (eg, Braun, 2005a). Second, thedata set might be identified by a particularanalytic interest in some topic in the data,and the data set then becomes all instancesin the corpus where that topic is referred.

So in Virginia’s example, if she was inter-ested in how ‘sexual pleasure’ was talkedabout, her data set would consist of allinstances across the entire data corpus thathad some relevance to sexual pleasure.These two approaches might sometimesbe combined to produce the data set. Dataitem is used to refer to each individualpiece of data collected, which togethermake up the data set or corpus. A dataitem in this instance would be an indivi-dual surgeon interview, a television docu-mentary, or one particular website. Finally,data extract refers to an individual codedchunk of data, which has been identifiedwithin, and extracted from, a data item.There will be many of these, taken fromthroughout the entire data set, and only aselection of these extracts will feature inthe final analysis.

What is thematic analysis?

Thematic analysis is a method for identify-ing, analysing and reporting patterns(themes) within data. It minimally orga-nizes and describes your data set in (rich)detail. However, frequently if goes furtherthan this, and interprets various aspects ofthe research topic (Boyatzis, 1998). Therange of different possible thematic ana-lyses will further be highlighted in relationto a number of decisions regarding it as amethod (see below).

Thematic analysis is widely used, butthere is no clear agreement about whatthematic analysis is and how you go aboutdoing it (see Attride-Stirling, 2001; Boyat-zis, 1998; Tuckett, 2005, for other exam-ples). It can be seen as a very poorly‘branded’ method, in that it does not appearto exist as a ‘named’ analysis in the sameway that other methods do (eg, narrative

Using thematic analysis in psychology 79

analysis, grounded theory). In this sense, itis often not explicitly claimed as themethod of analysis, when, in actuality, weargue that a lot of analysis is essentiallythematic �/ but is either claimed as some-thing else (such as DA, or even contentanalysis (eg, Meehan et al ., 2000)) or notidentified as any particular method at all �/for example, data were ‘subjected to quali-tative analysis for commonly recurringthemes’ (Braun and Wilkinson, 2003: 30).If we do not know how people went aboutanalysing their data, or what assumptionsinformed their analysis, it is difficult toevaluate their research, and to compareand/or synthesize it with other studies onthat topic, and it can impede other research-ers carrying out related projects in thefuture (Attride-Stirling, 2001). For thesereasons alone, clarity on process and prac-tice of method is vital. We hope that thispaper will lead to more clarity aroundthematic analysis.

Relatedly, insufficient detail is often gi-ven to reporting the process and detail ofanalysis (Attride-Stirling, 2001). It is notuncommon to read of themes ‘emerging’from the data (although this issue is notlimited to thematic analysis). For example,Singer and Hunter’s (1999: 67) thematicdiscourse analysis of women’s experiencesof early menopause identified that ‘severalthemes emerged’ during the analysis. Rubinand Rubin (1995: 226) claim that analysis isexciting because ‘you discover themes andconcepts embedded throughout your inter-views’. An account of themes ‘emerging’ orbeing ‘discovered’ is a passive account ofthe process of analysis, and it denies theactive role the researcher always plays inidentifying patterns/themes, selectingwhich are of interest, and reporting themto the readers (Taylor and Ussher, 2001).4

The language of ‘themes emerging’:

can be misinterpreted to mean that themes ‘re-side’ in the data, and if we just look hard enoughthey will ‘emerge’ like Venus on the half shell.If themes ‘reside’ anywhere, they reside inour heads from our thinking about our data andcreating links as we understand them. (Ely et al .,1997: 205 �/6)

At this point, it is important to acknowledgeour own theoretical positions and values inrelation to qualitative research. We do notsubscribe to a naı̈ve realist view of qualita-tive research, where the researcher cansimply ‘give voice’ (see Fine, 2002) to theirparticipants. As Fine (2002): 218) argues,even a ‘giving voice’ approach ‘involvescarving out unacknowledged pieces ofnarrative evidence that we select, edit,and deploy to border our arguments’. How-ever, nor do we think there is one idealtheoretical framework for conducting quali-tative research, or indeed one ideal method.What is important is that the theoreticalframework and methods match what theresearcher wants to know, and that theyacknowledge these decisions, and recognizethem as decisions.

Thematic analysis differs from other ana-lytic methods that seek to describe patternsacross qualitative data �/ such as ‘thematic’DA, thematic decomposition analysis, IPAand grounded theory.5 Both IPA andgrounded theory seek patterns in the data,but are theoretically bounded. IPA is at-tached to a phenomenological epistemology(Smith et al ., 1999; Smith and Osborn,2003), which gives experience primacy(Holloway and Todres, 2003), and is aboutunderstanding people’s everyday experi-ence of reality, in great detail, in order togain an understanding of the phenomenonin question (McLeod, 2001). To complicatematters, grounded theory comes in differentversions (Charmaz, 2002). Regardless, thegoal of a grounded theory analysis is togenerate a plausible �/ and useful �/ theory

80 V Braun and V Clarke

of the phenomena that is grounded in thedata (McLeod, 2001). However, in our ex-perience, grounded theory seems increas-ingly to be used in a way that is essentiallygrounded theory ‘lite’ �/ as a set of proce-dures for coding data very much akin tothematic analysis. Such analyses do notappear to fully subscribe to the theoreticalcommitments of a ‘full-fat’ grounded theory,which requires analysis to be directed to-wards theory development (Holloway andTodres, 2003). We argue, therefore, that a‘named and claimed’ thematic analysismeans researchers need not subscribe tothe implicit theoretical commitments ofgrounded theory if they do not wish toproduce a fully worked-up grounded-theoryanalysis.

The term ‘thematic DA’ is used to refer toa wide range of pattern-type analysis ofdata, ranging from thematic analysis withina social constructionist epistemology (ie,where patterns are identified as sociallyproduced, but no discursive analyse isconducted), to forms of analysis verymuch akin to the interpretative repertoireform of DA (Clarke, 2005). Thematic decom-position analysis (eg, Stenner, 1993; Ussherand Mooney-Somers, 2000) is a specificallynamed form of ‘thematic’ DA, which iden-tifies patterns (themes, stories) within data,and theorizes language as constitutive ofmeaning and meaning as social.

These different methods share a searchfor certain themes or patterns across an(entire) data set, rather than within a dataitem, such as an individual interview orinterviews from one person, as in the case ofbiographical or case-study forms of analy-sis, such as narrative analysis (eg, Murray,2003; Riessman, 1993). In this sense, theymore or less overlap with thematic analysis.As thematic analysis does not require thedetailed theoretical and technological

knowledge of approaches, such as groundedtheory and DA, it can offer a more accessibleform of analysis, particularly for those earlyin a qualitative research career.

In contrast to IPA or grounded theory (andother methods like narrative analysis DA orCA), thematic analysis is not wedded to anypre-existing theoretical framework, andtherefore it can be used within differenttheoretical frameworks (although not all),and can be used to do different thingswithin them. Thematic analysis can be anessentialist or realist method, which reportsexperiences, meanings and the reality ofparticipants, or it can be a constructionistmethod, which examines the ways in whichevents, realities, meanings, experiences andso on are the effects of a range of discoursesoperating within society. It can also be a‘contextualist’ method, sitting between thetwo poles of essentialism and construction-ism, and characterized by theories, such ascritical realism (eg, Willig, 1999), whichacknowledge the ways individuals makemeaning of their experience, and, in turn,the ways the broader social context im-pinges on those meanings, while retainingfocus on the material and other limits of‘reality’. Therefore, thematic analysis can bea method that works both to reflect realityand to unpick or unravel the surface of‘reality’. However, it is important that thetheoretical position of a thematic analysis ismade clear, as this is all too often leftunspoken (and is then typically a realistaccount). Any theoretical framework carrieswith it a number of assumptions about thenature of the data, what they represent interms of the ‘the world’, ‘reality’, and soforth. A good thematic analysis will makethis transparent.

A number of decisionsThematic analysis involves a number ofchoices which are often not made explicit

Using thematic analysis in psychology 81

(or are certainly typically not discussed inthe method section of papers), but whichneed explicitly to be considered and dis-cussed. In practice, these questions shouldbe considered before analysis (and some-times even collection) of the data begins,and there needs to be an ongoing reflexivedialogue on the part of the researcher orresearchers with regards to these issues,throughout the analytic process. Themethod section of Taylor and Ussher’s(2001) thematic DA of S&M provides agood example of research which presentsthis process explicitly; the method sectionof Braun and Wilkinson (2003) does not.

What counts as a theme?A theme captures something importantabout the data in relation to the researchquestion, and represents some level ofpatterned response or meaning within thedata set. An important question to addressin terms of coding is: what counts as apattern/theme, or what ‘size’ does a themeneed to be? This is a question of prevalence,in terms both of space within each data itemand of prevalence across the entire data set.Ideally, there will be a number of instancesof the theme across the data set, but moreinstances do not necessarily mean thetheme itself is more crucial. As this isqualitative analysis, there is no hard-and-fast answer to the question of what propor-tion of your data set needs to displayevidence of the theme for it to be considereda theme. It is not the case that if it waspresent in 50% of one’s data items, it wouldbe a theme, but if it was present only in47%, then it would not be a theme. Nor is itthe case that a theme is only something thatmany data items give considerable attentionto, rather than a sentence or two. A thememight be given considerable space in somedata items, and little or none in others, or it

might appear in relatively little of the dataset. So, researcher judgement is necessary todetermine what a theme is. Our initialguidance around this is that you need toretain some flexibility, and rigid rules reallydo not work. (The question of prevalence isrevisited in relation to themes and sub-themes, as the refinement of analysis (seelater) will often result in overall themes,and sub-themes within those.)

Furthermore, the ‘keyness’ of a theme isnot necessarily dependent on quantifiablemeasures �/ but rather on whether it cap-tures something important in relation tothe overall research question. For exam-ple, in Victoria’s research on representa-tions of lesbians and gay parents on 26talk shows (Clarke and Kitzinger, 2004),she identified six ‘key’ themes. These sixthemes were not necessarily the most pre-valent themes across the data set �/ theyappeared in between two and 22 of the 26talk shows �/ but together they captured animportant element of the way in whichlesbians and gay men ‘normalize’ theirfamilies in talk show debates. In this in-stance, her thematic analysis was driven bythis particular analytic question. How she‘measured’ prevalence is relevant, as pre-valence can be determined in a number ofdifferent ways. Prevalence was counted atthe level of the data item (ie, did a themeappear anywhere in each individual talkshow?). Alternatively, it could have beencounted in terms of the number of differentspeakers who articulated the theme, acrossthe entire data set, or each individualoccurrence of the theme across the entiredata set (which raises complex questionsabout where an ‘instance’ begins and endswithin an extended sequence of talk �/ seeRiessman, 1993). Because prevalence wasnot crucial to the analysis presented, Vic-toria chose the most straightforward form,

82 V Braun and V Clarke

but it is important to note there is no right orwrong method for determining prevalence.Part of the flexibility of thematic analysisis that it allows you to determine themes(and prevalence) in a number of ways. Whatis important is that you are consistent inhow you do this within any particularanalysis.

There are various ‘conventions’ for repre-senting prevalence in thematic (and otherqualitative) analysis that does not provide aquantified measure (unlike much contentanalysis, Wilkinson, 2000) �/ for instance:‘the majority of participants’ (Meehan et al .,2000: 372), ‘many participants’ (Taylor andUssher, 2001: 298), or ‘a number ofparticipants’ (Braun et al ., 2003: 249).Such descriptors work rhetorically tosuggest a theme really existed in the data,and to convince us they are reportingtruthfully about the data. But do they tellus much? This is perhaps one area wheremore debate is needed about how and whywe might represent the prevalence ofthemes in the data, and, indeed, whether,if, and why prevalence is particularly im-portant.

A rich description of the data set, or adetailed account of one particular aspectIt is important to determine the type ofanalysis you want to do, and the claimsyou want to make, in relation to your dataset. For instance, you might wish to providea rich thematic description of your entiredata set, so that the reader gets a sense of thepredominant or important themes. In thiscase, the themes you identify, code, andanalyse would need to be an accurate reflec-tion of the content of the entire data set. Insuch an analysis, some depth and complex-ity is necessarily lost (particularly if you arewriting a short dissertation or article withstrict word limits), but a rich overall de-

scription is maintained. This might be aparticularly useful method when you areinvestigating an under-researched area, oryou are working with participants whoseviews on the topic are not known.

An alternative use of thematic analysis isto provide a more detailed and nuancedaccount of one particular theme, or group ofthemes, within the data. This might relate toa specific question or area of interest withinthe data (a semantic approach �/ see below),or to a particular ‘latent’ theme (see below)across the whole or majority of the data set.An example of this would be Victoria’s talkshow paper, discussed previously (Clarkeand Kitzinger, 2004), which examined nor-malization in lesbians’ and gay men’s ac-counts of parenting.

Inductive versus theoretical thematicanalysisThemes or patterns within data can beidentified in one of two primary ways inthematic analysis: in an inductive or ‘bot-tom up’ way (eg, Frith and Gleeson, 2004),or in a theoretical or deductive or ‘topdown’ way (eg, Boyatzis, 1998; Hayes,1997). An inductive approach means thethemes identified are strongly linked to thedata themselves (Patton, 1990) (as such, thisform of thematic analysis bears some simi-larity to grounded theory). In this approach,if the data have been collected specificallyfor the research (eg, via interview or focusgroup), the themes identified may bear littlerelation to the specific questions that wereasked of the participants. They would alsonot be driven by the researcher’s theoreticalinterest in the area or topic. Inductiveanalysis is therefore a process of codingthe data without trying to fit it into a pre-existing coding frame, or the researcher’sanalytic preconceptions. In this sense, thisform of thematic analysis is data-driven.

Using thematic analysis in psychology 83

However, it is important to note, as wediscussed earlier, that researchers cannotfree themselves of their theoretical andepistemological commitments, and dataare not coded in an epistemological va-cuum.

In contrast, a ‘theoretical’ thematic analy-sis would tend to be driven by the research-er’s theoretical or analytic interest in thearea, and is thus more explicitly analyst-driven. This form of thematic analysis tendsto provide less a rich description of the dataoverall, and more a detailed analysis ofsome aspect of the data. Additionally, thechoice between inductive and theoreticalmaps onto how and why you are coding thedata. You can either code for a quite specificresearch question (which maps onto themore theoretical approach) or the specificresearch question can evolve through thecoding process (which maps onto the in-ductive approach).

For example, if a researcher was inter-ested in talk about heterosex, and hadcollected interview data, with an inductiveapproach they would read and re-read thedata for any themes related to heterosex,and code diversely, without paying atten-tion to the themes that previous research onthe topic might have identified. For exam-ple, the researcher would not look to theinfluential research of Hollway (1989),identifying discourses of heterosex, andcode just for male sexual drive, have/holdor permissive discourse themes. In contrast,with a theoretical approach, the researchermay well be interested in the way permis-siveness plays out across the data, andfocus on that particular feature in codingthe data. This would then result in anumber of themes around permissiveness,which may include, speak to, or expand onsomething approximating Hollway’s origi-nal theme.

Semantic or latent themesAnother decision revolves around the ‘le-vel’ at which themes are to be identified: ata semantic or explicit level, or at a latent orinterpretative level (Boyatzis, 1998).

6A

thematic analysis typically focuses exclu-sively or primarily on one level. With asemantic approach, the themes are identi-fied within the explicit or surface meaningsof the data, and the analyst is not looking foranything beyond what a participant hassaid or what has been written. Ideally, theanalytic process involves a progression fromdescription , where the data have simplybeen organized to show patterns in seman-tic content, and summarized, to interpreta-tion , where there is an attempt to theorizethe significance of the patterns and theirbroader meanings and implications (Patton,1990), often in relation to previous literature(for an excellent example of this, see Frithand Gleeson, 2004).

In contrast, a thematic analysis at thelatent level goes beyond the semantic con-tent of the data, and starts to identify orexamine the underlying ideas, assumptions,and conceptualizations �/ and ideologies �/that are theorized as shaping or informingthe semantic content of the data. If weimagine our data three-dimensionally asan uneven blob of jelly, the semanticapproach would seek to describe the surfaceof the jelly, its form and meaning, while thelatent approach would seek to identify thefeatures that gave it that particular form andmeaning. Thus, for latent thematic analysis,the development of the themes themselvesinvolves interpretative work, and the ana-lysis that is produced is not just descrip-tion, but is already theorized.

Analysis within this latter tradition tendsto come from a constructionist paradigm(eg, Burr, 1995), and in this form, thematicanalysis overlaps with some forms of ‘DA’

84 V Braun and V Clarke

(which are sometimes specifically referredto as ‘thematic DA’ (eg, Singer and Hunter,1999; Taylor and Ussher, 2001)), wherebroader assumptions, structures and/ormeanings are theorized as underpinningwhat is actually articulated in the data.Increasingly, a number of discourse analystsare also revisiting psycho-analytic modes ofinterpretation (eg, Hollway and Jefferson,2000), and latent thematic analysis wouldalso be compatible with that framework.

Epistemology: essentialist/realist versusconstructionist thematic analysisAs we have argued, thematic analysis can beconducted within both realist/essentialistand constructionist paradigms, althoughthe outcome and focus will be different foreach. The question of epistemology isusually determined when a research projectis being conceptualized, although episte-mology may also raise its head again duringanalysis, when the research focus may shiftto an interest in different aspects of the data.The research epistemology guides what youcan say about your data, and informs howyou theorize meaning. For instance, with anessentialist/realist approach, you can theo-rize motivations, experience, and meaningin a straightforward way, because a simple,largely unidirectional relationship is as-sumed between meaning and experienceand language (language reflects and enablesus to articulate meaning and experience)(Potter and Wetherell, 1987; Widdicombeand Wooffitt, 1995).

In contrast, from a constructionist per-spective, meaning and experience are so-cially produced and reproduced, ratherthan inhering within individuals (Burr,1995). Therefore, thematic analysis con-ducted within a constructionist frameworkcannot and does not seek to focus onmotivation or individual psychologies, but

instead seeks to theorize the socioculturalcontexts, and structural conditions, thatenable the individual accounts that areprovided. Thematic analysis that focuseson ‘latent’ themes tends to be more con-structionist, and it also tends to start tooverlap with thematic DA at this point.However, not all ‘latent’ thematic analysisis constructionist.

The many questions of qualitative researchIt is worth briefly noting that qualitativeresearch involves a series of questions, andthere is a need to be clear about the relation-ship between these different questions.First, there is the overall research questionor questions that drive the project. A re-search question might be very broad (andexploratory), such as ‘how is lesbian andgay parenting constructed?’ or ‘what are themeanings of the vagina?’. Narrower researchquestions might be ‘how and why is lesbianand gay parenting normalized?’ (Clarke andKitzinger, 2004), or ‘what are the discoursesaround vaginal size?’ (see Braun and Kit-zinger, 2001). These narrow questions maybe part of a broader overarching researchquestion, and if so, the analyses they informwould also provide answers to the overallresearch question. Although all projects areguided by research questions, these mayalso be refined as a project progresses.

Second, if data from interviews or focusgroups have been collected, there are thequestions that participants have respondedto. Finally, there are the questions thatguide the coding and analysis of the data.There is no necessary relationship betweenthese three, and indeed, it is often desirablethat there is a disjuncture between them.Some of the worst examples of ‘thematic’analysis we have read have simply usedthe questions put to participants as the‘themes’ identified in the ‘analysis’ �/

Using thematic analysis in psychology 85

although in such instances, no analysis hasreally been done at all!

To sum up, thematic analysis involves thesearching across a data set �/ be that anumber of interviews or focus groups, or arange of texts �/ to find repeated patterns ofmeaning. The exact form and product ofthematic analysis varies, as indicated above,and so it is important that the questionsoutlined above are considered before andduring thematic analyses. Those appro-aches which consider specific aspects, la-tent themes and are constructionist tend tooften cluster together, while those thatconsider meanings across the whole dataset, semantic themes, and are realist, oftencluster together. However, there are no hard-and-fast rules in relation to this, and differ-ent combinations are possible. What isimportant is that the finished product con-tains an account �/ not necessarily thatdetailed �/ of what was done, and why. Sowhat does one actually do? We now providewhat is, we hope, a straightforward step-by-step guide to conducting thematic ana-lysis.

Doing thematic analysis: a step-by-stepguide

Some of the phases of thematic analysis aresimilar to the phases of other qualitativeresearch, so these stages are not necessarilyall unique to thematic analysis. The processstarts when the analyst begins to notice,and look for, patterns of meaning andissues of potential interest in the data �/this may be during data collection. Theendpoint is the reporting of the contentand meaning of patterns (themes) in thedata, where ‘themes are abstract (and oftenfuzzy) constructs the investigators identify[sic] before, during, and after analysis’

(Ryan and Bernard, 2000: 780). Analysisinvolves a constant moving back and for-ward between the entire data set, the codedextracts of data that you are analysing, andthe analysis of the data that you are produ-cing. Writing is an integral part of analysis,not something that takes place at the end, asit does with statistical analyses. Therefore,writing should begin in phase one, with thejotting down of ideas and potential codingschemes, and continue right through theentire coding/analysis process.

There are different positions regardingwhen you should engage with the literaturerelevant to your analysis �/ with somearguing that early reading can narrow youranalytic field of vision, leading you to focuson some aspects of the data at the expenseof other potentially crucial aspects. Othersargue that engagement with the literaturecan enhance your analysis by sensitizingyou to more subtle features of the data(Tuckett, 2005). Therefore, there is no oneright way to proceed with reading for the-matic analysis, although a more inductiveapproach would be enhanced by not enga-ging with literature in the early stages ofanalysis, whereas a theoretical approachrequires engagement with the literatureprior to analysis.

We provide an outline guide through thesix phases of analysis, and offer examples todemonstrate the process.7 The differentphases are summarized in Table 1. It isimportant to recognize that qualitative ana-lysis guidelines are exactly that �/ they arenot rules, and, following the basic precepts,will need to be applied flexibly to fit theresearch questions and data (Patton, 1990).Moreover, analysis is not a linear process ofsimply moving from one phase to the next.Instead, it is more recursive process, wheremovement is back and forth as needed,throughout the phases. It is also a process

86 V Braun and V Clarke

that develops over time (Ely et al ., 1997),and should not be rushed.

Phase 1: familiarizing yourself with yourdataWhen you engage in analysis, you may havecollected the data yourself, or they may havebeen given to you. If you collected themthrough interactive means, you will come tothe analysis with some prior knowledge ofthe data, and possibly some initial analyticinterests or thoughts. Regardless, it is vitalthat you immerse yourself in the data to theextent that you are familiar with the depthand breadth of the content. Immersionusually involves ‘repeated reading’ of thedata, and reading the data in an active way �/searching for meanings, patterns and so on.It is ideal to read through the entire data setat least once before you begin your coding,as ideas and identification of possible pat-terns will be shaped as you read through.

Whether or not you are aiming for anoverall or detailed analysis, are searchingfor latent or semantic themes, or are data- ortheoretically-driven will inform how thereading proceeds. Regardless, it is impor-tant to be familiar with all aspects of yourdata. At this phase, one of the reasons whyqualitative research tends to use far smallersamples than, for example, questionnaire

research will become apparent �/ the read-ing and re-reading of data is time-consum-ing. It is, therefore, tempting to skip overthis phase, or be selective. We wouldstrongly advise against this, as this phaseprovides the bedrock for the rest of theanalysis.

During this phase, it is a good idea to starttaking notes or marking ideas for codingthat you will then go back to in subsequentphases. Once you have done this, you areready to begin, the more formal codingprocess. In essence, coding continues to bedeveloped and defined throughout the en-tire analysis.

Transcription of verbal dataIf you are working with verbal data, such asinterviews, television programmes or poli-tical speeches, the data will need to betranscribed into written form in order toconduct a thematic analysis. The process oftranscription, while it may seen time-con-suming, frustrating, and at times boring, canbe an excellent way to start familiarizingyourself with the data (Riessman, 1993).Further, some researchers even argueit should be seen as ‘a key phase ofdata analysis within interpretative qualita-tive methodology’ (Bird, 2005: 227), andrecognized as an interpretative act, where

Table 1 Phases of thematic analysis

Phase Description of the process

1. Familiarizing yourselfwith your data:

Transcribing data (if necessary), reading and re-reading the data, noting downinitial ideas.

2. Generating initial codes: Coding interesting features of the data in a systematic fashion across the entiredata set, collating data relevant to each code.

3. Searching for themes: Collating codes into potential themes, gathering all data relevant to eachpotential theme.

4. Reviewing themes: Checking if the themes work in relation to the coded extracts (Level 1) and theentire data set (Level 2), generating a thematic ‘map’ of the analysis.

5. Defining and namingthemes:

Ongoing analysis to refine the specifics of each theme, and the overall story theanalysis tells, generating clear definitions and names for each theme.

6. Producing the report: The final opportunity for analysis. Selection of vivid, compelling extractexamples, final analysis of selected extracts, relating back of the analysis to theresearch question and literature, producing a scholarly report of the analysis.

Using thematic analysis in psychology 87

meanings are created, rather than simply amechanical act of putting spoken sounds onpaper (Lapadat and Lindsay, 1999).

Various conventions exist for transformingspoken texts into written texts (see Edwardsand Lampert, 1993; Lapadat and Lindsay,1999). Some systems of transcription havebeen developed for specific forms of analysis�/ such as the ‘Jefferson’ system for CA (seeAtkinson and Heritage, 1984; Hutchby andWooffitt, 1998). However, thematic analysis,even constructionist thematic analysis, doesnot require the same level of detail in thetranscript as conversation, discourse or evennarrative analysis. As there is no one way toconduct thematic analysis, there is no one setof guidelines to follow when producing atranscript. However, at a minimum it re-quires a rigorous and thorough ‘ortho-graphic’ transcript �/ a ‘verbatim’ account ofall verbal (and sometimes nonverbal �/ eg,coughs) utterances.8 What is important isthat the transcript retains the informationyou need, from the verbal account, and in away which is ‘true’ to its original nature (eg,punctuation added can alter the meaning ofdata �/ for example ‘I hate it, you know. I do’versus ‘I hate it. You know I do’, Poland,2002: 632), and that the transcription con-vention is practically suited to the purpose ofanalysis (Edwards, 1993).

As we have noted, the time spent intranscription is not wasted, as it informsthe early stages of analysis, and you willdevelop a far more thorough understandingof your data through having transcribed it.Furthermore, the close attention needed totranscribe data may facilitate the close read-

ing and interpretative skills needed to ana-lyse the data (Lapadat and Lindsay, 1999). Ifyour data have already been, or will be,transcribed for you, it is important that youspend more time familiarising yourself withthe data, and also check the transcripts backagainst the original audio recordings for‘accuracy’ (as should always be done).

Phase 2: generating initial codesPhase 2 begins when you have read andfamiliarized yourself with the data, and havegenerated an initial list of ideas about whatis in the data and what is interesting aboutthem. This phase then involves the produc-tion of initial codes from the data. Codesidentify a feature of the data (semanticcontent or latent) that appears interestingto the analyst, and refer to ‘the most basicsegment, or element, of the raw data orinformation that can be assessed in a mean-ingful way regarding the phenomenon’(Boyatzis, 1998: 63). See Figure 1 for anexample of codes applied to a short segmentof data. The process of coding is part ofanalysis (Miles and Huberman, 1994), as youare organising your data into meaningfulgroups (Tuckett, 2005). However, yourcoded data differ from the units of analysis(your themes), which are (often) broader.Your themes, which you start to develop inthe next phase, are where the interpretativeanalysis of the data occurs, and in relation towhich arguments about the phenomenonbeing examined are made (Boyatzis, 1998).

Coding will, to some extent, depend onwhether the themes are more ‘data-driven’or ‘theory-driven’ �/ in the former, the

Data extract Coded for

it's too much like hard work I mean how much paper have you got to signto change a flippin’ name no I I mean no I no we we have thought about it((inaudible)) half heartedly and thought no no I jus- I can’t be bothered,it’s too much like hard work. (Kate F07a)

1. Talked about with partner

2. Too much hassle to change name

Figure 1 Data extract, with codes applied (from Clarke et al ., 2006)

88 V Braun and V Clarke

themes will depend on the data, but in thelatter, you might approach the data withspecific questions in mind that you wish tocode around. It will also depend on whetheryou are aiming to code the content of theentire data set, or whether you are coding toidentify particular (and possibly limited)features of the data set. Coding can beperformed either manually or through asoftware programme (see, eg, Kelle, 2004;Seale, 2000, for discussion of softwareprogrammes).

Work systematically through the entiredata set, giving full and equal attention toeach data item, and identify interestingaspects in the data items that may formthe basis of repeated patterns (themes)across the data set. There are a number ofways of actually coding extracts. If codingmanually, you can code your data by writ-ing notes on the texts you are analysing,by using highlighters or coloured pens toindicate potential patterns, or by using‘post-it’ notes to identify segments of data.You may initially identify the codes, andthen match them with data extracts thatdemonstrate that code, but it is important inthis phase to ensure that all actual dataextracts are coded, and then collated to-gether within each code. This may involvecopying extracts of data from individualtranscripts or photocopying extracts ofprinted data, and collating each code to-gether in separate computer files or usingfile cards. If using computer software, youcode by tagging and naming selections oftext within each data item.

Key advice for this phase is: (a) code for asmany potential themes/patterns as possible(time permitting) �/ you never know whatmight be interesting later; (b) code extractsof data inclusively �/ ie, keep a little of thesurrounding data if relevant, a commoncriticism of coding is that the context is

lost (Bryman, 2001); and (c) remember thatyou can code individual extracts of data inas many different ‘themes’ as they fit into �/so an extract may be uncoded, coded once,or coded many times, as relevant. Note thatno data set is without contradiction, and asatisfactory thematic ‘map’ that you willeventually produce �/ an overall conceptua-lization of the data patterns, and relation-ships between them9 �/ does not have tosmooth out or ignore the tensions andinconsistencies within and across dataitems. It is important to retain accountsthat depart from the dominant story in theanalysis, so do not ignore these in yourcoding.

Phase 3: searching for themesPhase 3 begins when all data have beeninitially coded and collated, and you have along list of the different codes that you haveidentified across the data set. This phase,which re-focuses the analysis at the broaderlevel of themes, rather than codes, involvessorting the different codes into potentialthemes, and collating all the relevant codeddata extracts within the identified themes.Essentially, you are starting to analyse yourcodes and consider how different codesmay combine to form an overarching theme.It may be helpful at this phase to use visualrepresentations to help you sort the differ-ent codes into themes. You might use tables,or mind-maps, or write the name each code(and a brief description) on a separate pieceof paper and play around with organizingthem into theme-piles. A thematic map ofthis early stage can be seen in Figure 2 (theexamples in Figures 2 �/4 come from theanalysis presented in Braun and Wilkinson,2003 of women’s talk about the vagina).This is when you start thinking about therelationship between codes, betweenthemes, and between different levels ofthemes (eg, main overarching themes and

Using thematic analysis in psychology 89

sub-themes within them). Some initial

codes may go on to form main themes,

whereas others may form sub-themes, and

others still may be discarded. At this stage,

you may also have a set of codes that do not

seem to belong anywhere, and it is perfectly

acceptable to create a ‘theme’ called ‘mis-

cellaneous’ to house the codes �/ possiblytemporarily �/ that do not seem to fit intoyour main themes.

You end this phase with a collection of

candidate themes, and sub-themes, and all

extracts of data that have been coded in

relation to them. At this point, you will start

to have a sense of the significance of

individual themes. However, do not aban-

don anything at this stage, as without

looking at all the extracts in detail (the

next phase) it is uncertain whether the

themes hold as they are, or whether some

Figure 2 Initial thematic map, showing five main themes (final analysis presented in Braun andWilkinson, 2003)

Figure 3 Developed thematic map, showing three main themes (final analysis presented in Braun andWilkinson, 2003)

90 V Braun and V Clarke

need to be combined, refined and separated,or discarded.

Phase 4: reviewing themesPhase 4 begins when you have devised a setof candidate themes, and it involves therefinement of those themes. During thisphase, it will become evident that somecandidate themes are not really themes (eg,if there are not enough data to support them,or the data are too diverse), while othersmight collapse into each other (eg, twoapparently separate themes might formone theme). Other themes might needto be broken down into separate themes.Patton’s (1990) for dual criteria judgingcategories �/ internal homogeneity and ex-ternal heterogeneity �/ are worth consider-ing here. Data within themes should coheretogether meaningfully, while there shouldbe clear and identifiable distinctions be-tween themes.

This phase involves two levels of review-ing and refining your themes. Level oneinvolves reviewing at the level of the codeddata extracts. This means you need to readall the collated extracts for each theme, andconsider whether they appear to form acoherent pattern. If your candidate themesdo appear to form a coherent pattern, youthen move on to the second level of thisphase. If your candidate themes do not fit,you will need to consider whether thetheme itself is problematic, or whethersome of the data extracts within it simplydo not fit there �/ in which case, you would

rework your theme, creating a new theme,finding a home for those extracts that do notcurrently work in an already-existing

theme, or discarding them from the analy-sis. Once you are satisfied that your candi-date themes adequately capture thecontours of the coded data �/ once youhave a candidate ‘thematic map’ �/ youare ready to move on to level two of thisphase. The outcome of this refinement

process can be seen in the thematic mappresented in Figure 3.

Level two involves a similar process, butin relation to the entire data set. At thislevel, you consider the validity of indivi-

dual themes in relation to the data set, butalso whether your candidate thematic map‘accurately’ reflects the meanings evident in

the data set as a whole. To some extent,what counts as ‘accurate representation’depends on your theoretical and analytic

approach. However, in this phase you re-read your entire data set for two purposes.The first is, as discussed, to ascertainwhether the themes ‘work’ in relation to

the data set. The second is to code anyadditional data within themes that has beenmissed in earlier coding stages. The need for

re-coding from the data set is to be expectedas coding is an ongoing organic process.

If the thematic map works, then youmoves on to the next phase. However, if

the map does not fit the data set, you needto return to further reviewing and refiningof your coding until you have devised a

Figure 4 Final thematic map, showing final two main themes (see Braun and Wilkinson, 2003).

Using thematic analysis in psychology 91

satisfactory thematic map. In so doing, it ispossible that you will identify potentialnew themes, and you will need to startcoding for these as well, if they are ofinterest and relevent. However, a wordof warning: as coding data and generatingthemes could go on ad infinitum , it isimportant not to get over-enthusiastic withendless re-coding. It is impossible to pro-vide clear guidelines on when to stop, butwhen your refinements are not adding any-thing substantial, stop! If the process ofrecoding is only fine-tuning and makingmore nuanced a coding frame that alreadyworks �/ ie, it fits the data well �/ recognizethis and stop. Consider it as similar toediting written work �/ you could endlesslyedit your sentences and paragraphs, butafter a few editing turns, any further workis usually unnecessary refinement �/ similarto rearranging the hundreds and thousandson an already nicely decorated cake.

At the end of this phase, you should havea fairly good idea of what your differentthemes are, how they fit together, and theoverall story they tell about the data.

Phase 5: defining and naming themesPhase 5 begins when you have a satisfactorythematic map of your data �/ see Figure 4 forthe final refinements of Virginia’s thematicmap. At this point, you then define andfurther refine the themes you will presentfor your analysis, and analyse the datawithin them. By ‘define and refine’, wemean identifying the ‘essence’ of whateach theme is about (as well as the themesoverall), and determining what aspect of thedata each theme captures. It is importantnot to try and get a theme to do too much, orto be too diverse and complex. You do thisby going back to collated data extracts foreach theme, and organizing them into acoherent and internally consistent account,with accompanying narrative. It is vital that

you do not just paraphrase the content ofthe data extracts presented, but identifywhat is of interest about them and why.

For each individual theme, you need toconduct and write a detailed analysis. Aswell as identifying the ‘story’ that eachtheme tells, it is important to considerhow it fits into the broader overall ‘story’that you are telling about your data, inrelation to the research question or ques-tions, to ensure there is not too muchoverlap between themes. So it is necessaryto consider the themes themselves, andeach theme in relation to the others. Aspart of the refinement, you will need toidentify whether or not a theme containsany sub-themes. Sub-themes are essentiallythemes-within-a-theme. They can be usefulfor giving structure to a particularly largeand complex theme, and also for demon-strating the hierarchy of meaning within thedata. For instance, in one of Virginia’sanalyses of women’s talk about the vagina,she identified two overarching themes inwomen’s talk: the vagina as liability, and thevagina as asset (Braun and Wilkinson,2003). Within each theme, three sub-themeswere identified: for liability the sub-themeswere ‘nastiness and dirtiness’, ‘anxieties’and ‘vulnerability’; for asset the sub-themeswere ‘satisfaction’, ‘power’ and ‘pleasure’.However, these eventual final themes andsub-themes resulted from a process of re-finement of initial themes and sub-themes,as shown in Figures 2 �/4.

It is important that by the end of this phaseyou can clearly define what your themes areand what they are not. One test for this is tosee whether you can describe the scope andcontent of each theme in a couple of sen-tences. If not, further refinement of thattheme may be needed. Although you willalready have given your themes workingtitles, this is also the point to start thinking

92 V Braun and V Clarke

about the names you will give them in thefinal analysis. Names need to be concise,punchy, and immediately give the reader asense of what the theme is about.

Phase 6: producing the reportPhase 6 begins when you have a set of fullyworked-out themes, and involves the finalanalysis and write-up of the report. Thetask of the write-up of a thematic anal-ysis, whether it is for publication or for aresearch assignment or dissertation, is totell the complicated story of your data in away which convinces the reader of the meritand validity of your analysis. It is importantthat the analysis (the write-up of it, includ-ing data extracts) provides a concise, coher-ent, logical, non-repetitive and interestingaccount of the story the data tell �/ withinand across themes. Your write-up mustprovide sufficient evidence of the themeswithin the data �/ ie, enough data extracts todemonstrate the prevalence of the theme.Choose particularly vivid examples, or ex-tracts which capture the essence of thepoint you are demonstrating, without un-necessary complexity. The extract should beeasily identifiable as an example of theissue. However, your write-up needs to domore than just provide data. Extracts needto be embedded within an analytic narrativethat compellingly illustrates the story youare telling about your data, and your analy-tic narrative needs to go beyond descriptionof the data, and make an argument inrelation to your research question.

Pinning down what interpretativeanalysis actually entails

It is difficult to specify exactly what inter-pretative analysis actually entails, particu-larly as the specifics of it will vary from

study to study. As a first step, we recom-mend looking at published examples ofthematic analysis, particularly of the speci-fic version you are planning to use (this ismade somewhat more difficult in that the-matic analysis is often not a named method,but you can find examples, eg, Ellis andKitzinger, 2002; Kitzinger and Willmott,2002; Toerien and Wilkinson, 2004). Inorder to provide a sense of the sorts ofquestions you should be asking of your data,and the sorts of analytic claims you shouldbe seeking to make, we will discuss aparticularly good example of an inductivethematic analysis, which emphasizes un-derstanding men’s experiences in relation tothe broader social context (see Frith andGleeson, 2004).

Frith and Gleeson (2004) aim to ex-plore how men’s feelings about theirbodies influence their clothing practices,and they use data gathered in qualitativequestionnaires from 75 men to answerthis question. They report four themes:practicality of clothing choices; lack ofconcern about appearance; use of cloth-ing to conceal or reveal the body; use ofclothing to fit cultural ideals. Each themeis clearly linked back to the overallresearch question, but each is distinct.They provide a clear sense of the scopeand diversity of each theme, using acombination of analyst narrative andillustrative data extracts. Where relevant,they broaden their analysis out, movingfrom a descriptive to an interpretativelevel (often relating their claims to exist-ing literature). For example, in ‘menvalue practicality’, they make sense ofmen’s accounts in relation to gendernorms and stereotypes, linking the ac-counts individual men provided to theexpectations that men �/ as members ofsociety �/ face. What they do, as analysts,

Using thematic analysis in psychology 93

is relate the patterns of meaning in men’sresponses to an academic analysis of howgender operates. In so doing, they de-monstrate the dual position that analystsneed to take: as both cultural membersand cultural commentators . Their ‘discus-sion’ section makes broader analyticstatements about the overall story thatthe themes tell us about men’s relation-ship with clothing. This story revealsthat men ‘deliberately and strategicallyuse clothing to manipulate their appear-ance to meet cultural ideals of masculi-nity’ (Frith and Gleeson, 2004: 45), in away more traditionally associated withwomen. This analysis makes an impor-tant contribution in that it challengesperceived wisdom about clothing/appear-ance and masculinity.

As this example demonstrates, your ana-lytic claims need to be grounded in, but gobeyond, the ‘surface’ of the data, even for a‘semantic’ level analysis. The sort of ques-tions you need to be asking, towards the endphases of your analysis, include: ‘What doesthis theme mean?’ ‘What are the assump-tions underpinning it?’ ‘What are the im-plications of this theme?’ ‘What conditionsare likely to have given rise to it?’ ‘Whydo people talk about this thing in thisparticular way (as opposed to otherways)?’ and ‘What is the overall story thedifferent themes reveal about the topic?’.These sorts of questions should guide theanalysis once you have a clear sense of yourthematic map.

Potential pitfalls to avoid when doingthematic analysis

Thematic analysis is a relatively straight-forward form of qualitative analysis, which

does not require the same detailed theore-tical and technical knowledge that ap-proaches such as DA or CA do. It isrelatively easy to conduct a good thematicanalysis on qualitative data, even when youare still learning qualitative techniques.However, there are a number of things thatcan result in a poor analysis. In this sectionwe identify these potential pitfalls, in thehope that they can be avoided.

The first of these is a failure to actuallyanalyse the data at all! Thematic analysis isnot just a collection of extracts strungtogether with little or no analytic narrative.Nor is it a selection of extracts with analyticcomment that simply or primarily para-phrases their content. The extracts in the-matic analysis are illustrative of the analyticpoints the researcher makes about the data,and should be used to illustrate/support ananalysis that goes beyond their specificcontent, to make sense of the data, and tellthe reader what it does or might mean �/ asdiscussed above. A second, associated pit-fall is the using of the data collectionquestions (such as from an interview sche-dule) as the ‘themes’ that are reported. Insuch a case, no analytic work has beencarried out to identify themes across theentire data set, or make sense of the pattern-ing of responses.

The third is a weak or unconvincinganalysis, where the themes do not appearto work, where there is too much overlapbetween themes, or where the themes arenot internally coherent and consistent. Allaspects of the theme should cohere around acentral idea or concept. This pitfall hasoccurred if, depending on what the analysisis trying to do, it fails adequately to capturethe majority of the data, or fails to provide arich description/interpretation of one ormore aspects of the data. A weak or un-

94 V Braun and V Clarke

convincing analysis can also stem from a

failure to provide adequate examples from

the data �/ for example, only one or twoextracts for a theme. This point is essen-

tially about the rhetorics of presentation,

and the need for the analysis to be convin-

cing to someone who has not read the entire

data set: ‘The ‘‘analysis’’ of the material. . . is

a deliberate and self-consciously artful crea-

tion by the researcher, and must be con-

structed to persuade the reader of the

plausibility of an argument’ (Foster and

Parker, 1995: 204). In so doing, one avoids

(the appearance of) what Bryman (1988) has

referred to as ‘anecdotalism’ in qualitative

research �/ where one or a few instances of aphenomenon are reified into a pattern or

theme, when it or they are actually idiosyn-

cratic. This is not to say that a few instances

cannot be of interest, or revealing; but it is

important not to misrepresent them as an

overarching theme.The fourth pitfall is a mismatch between

the data and the analytic claims that are

made about it. In such an (unfounded)

analysis, the claims cannot be supported

by the data, or, in the worst case, the data

extracts presented suggest another analysis

or even contradict the claims. The re-

searcher needs to make sure that their

interpretations and analytic points are con-

sistent with the data extracts. A weak

analysis does not appear to consider other

obvious alternative readings of the data, or

fails to consider variation (and even contra-

diction) in the account that is produced. A

pattern in data is rarely, if ever, going to be

100% complete and non-contradicted, so an

analysis which suggests that it is, without a

thorough explanation, is open to suspicion.

It is important to pick compelling examples

to demonstrate the themes, so give this

considerable thought.

The fifth involves a mismatch betweentheory and analytic claims, or between theresearch questions and the form of thematicanalysis used. A good thematic analysisneeds to make sure that the interpretationsof the data are consistent with the theoreticalframework. So, for instance, if you are work-ing within an experiential framework, youwould typically not make claims about thesocial construction of the research topic, andif you were doing constructionist thematicanalysis, you would not treat people’s talkof experience as a transparent window ontheir world. Finally, even a good and inter-esting analysis which fails to spell outits theoretical assumptions, or clarify how itwas undertaken, and for what purpose, islacking crucial information (Holloway andTodres, 2003), and thus fails in one aspect.

What makes good thematic analysis?

One of the criticisms of qualitative researchfrom those outside the field is the percep-tion that ‘anything goes’. For instance, thissentiment is echoed in the first sentence ofLaubschagne’s (2003) abstract: ‘For manyscientists used to doing quantitative studiesthe whole concept of qualitative research isunclear, almost foreign, or ‘‘airy fairy’’ �/ not‘‘real’’ research.’ However, although ‘quali-tative’ research cannot be subjected to thesame criteria as ‘quantitative’ approaches, itdoes provide methods of analysis thatshould be applied rigorously to the data.Furthermore, criteria for conducting goodqualitative research �/ both data collectionand analysis �/ do exist (eg, Elliott et al .,1999; Parker, 2004; Seale, 1999; Silverman,2000; Yardley, 2000). The British Psycholo-gical Society offers relatively succinct on-line guidelines for assessing quality in qua-litative research (see http://www.bps.org.

Using thematic analysis in psychology 95

uk/publications/journals/joop/qualitative-

guidelines.cfm). ‘Criteria’ for assessing qua-

litative research is a not uncontroversial

topic, with concerns raised about rigid

criteria limiting freedom and stifling meth-

odological development (Elliott et al ., 1999;Parker, 2004; Reicher, 2000). Reicher (2000)

takes the critique further, by asking whether

the incredibly diverse range of qualitative

approaches can and should be subject tothe same criteria.

Bracketing these critiques off, the issues

raised in many general qualitative research

assessment criteria can be more or less

applied to thematic forms of analysis. As

thematic analysis is a flexible method, you

also need to be clear and explicit about what

you are doing, and what you say you are

doing needs to match up with what you

actually do. In this sense, the theory and

method need to be applied rigorously, and

‘rigour lies in devising a systematic method

whose assumptions are congruent with theway one conceptualizes the subject matter’(Reicher and Taylor, 2005: 549). A concisechecklist of criteria to consider when deter-mining whether you have generated a goodthematic analysis is provided in Table 2.

So what does thematic analysis offerpsychologists?

We now end this paper with some briefcomments on the advantages and disadvan-tages of thematic analysis. As we haveshown throughout this paper, thematic ana-lysis is not a complex method. Indeed, asyou can see from Table 3, its advantages aremany. However, it is not without somedisadvantages, which we will now brieflyconsider. Many of the disadvantages de-pend more on poorly conducted analysesor inappropriate research questions than on

Table 2 A 15-point checklist of criteria for good thematic analysis

Process No. Criteria

Transcription 1 The data have been transcribed to an appropriate level of detail, and the transcriptshave been checked against the tapes for ‘accuracy’.

Coding 2 Each data item has been given equal attention in the coding process.3 Themes have not been generated from a few vivid examples (an anecdotal approach),

but instead the coding process has been thorough, inclusive and comprehensive.4 All relevant extracts for all each theme have been collated.5 Themes have been checked against each other and back to the original data set.6 Themes are internally coherent, consistent, and distinctive.

Analysis 7 Data have been analysed �/ interpreted, made sense of �/ rather than just paraphrasedor described.

8 Analysis and data match each other �/ the extracts illustrate the analytic claims.9 Analysis tells a convincing and well-organized story about the data and topic.

10 A good balance between analytic narrative and illustrative extracts is provided.Overall 11 Enough time has been allocated to complete all phases of the analysis adequately,

without rushing a phase or giving it a once-over-lightly.Written report 12 The assumptions about, and specific approach to, thematic analysis are clearly

explicated.13 There is a good fit between what you claim you do, and what you show you have

done �/ ie, described method and reported analysis are consistent.14 The language and concepts used in the report are consistent with the epistemological

position of the analysis.15 The researcher is positioned as active in the research process; themes do not just

‘emerge’.

96 V Braun and V Clarke

the method itself. Further, the flexibility ofthe method �/ which allows for a wide rangeof analytic options �/ means that the poten-tial range of things that can be said aboutyour data is broad. While this is an advan-tage, it can also be a disadvantage in that itmakes developing specific guidelines forhigher-phase analysis difficult, and can bepotentially paralysing to the researcher try-ing to decide what aspects of their data tofocus on. Another issue to consider is that athematic analysis has limited interpretativepower beyond mere description if it is notused within an existing theoretical frame-work that anchors the analytic claims thatare made.

Other disadvantages appear when the-matic analysis is considered in relation tosome of the other qualitative analytic meth-ods. For instance, unlike narrative or otherbiographical approaches, you are unable toretain a sense of continuity and contradic-tion through any one individual account,and these contradictions and consistenciesacross individual accounts may be reveal-ing. In contrast to methods similar to DAand CA, a simple thematic analysis does not

allow the researcher to make claims aboutlanguage use, or the fine-grained function-ality of talk.

Finally, it is worth noting that thematicanalysis currently has no particular kudosas an analytic method �/ this, we argue,stems from the very fact that it is poorlydemarcated and claimed, yet widely used.This means that thematic analysis is fre-quently, or appears to be, what is simplycarried out by someone without the knowl-edge or skills to perform a supposedly moresophisticated �/ certainly more kudos-bear-ing �/ ‘branded’ form of analysis likegrounded theory, IPA or DA. We hope thispaper will change this view as, we argue, arigorous thematic approach can produce aninsightful analysis that answers particularresearch questions. What is important ischoosing a method that is appropriate toyour research question, rather than fallingvictim to ‘methodolatry’, where you arecommitted to method rather than topic/content or research questions (Hollowayand Todres, 2003). Indeed, your method ofanalysis should be driven by both yourresearch question and your broader theore-tical assumptions. As we have demon-strated, thematic analysis is a flexibleapproach that can be used across a rangeof epistemologies and research questions.

Notes

1. Boyatzis (1998) provides a much moredetailed account of thematic analysis. However,we do not feel that it is a particularly accessibleaccount for those unfamiliar with qualitativeapproaches. Moreover, his approach differsfrom ours in that, although he acknowledgesthe subjective dimension of qualitative analysis,his approach is ultimately, if often implicitly,located within a positivist empiricist paradigm.

2. Dey’s (1993) account of on ‘qualitative dataanalysis’, which aims to identify shared techni-ques across the diverse range of qualitative

Table 3 Advantages of thematic analysis

Flexibility.Relatively easy and quick method to learn, and do.Accessible to researchers with little or no experience of

qualitative research.Results are generally accessible to educated general

public.Useful method for working within participatory re-

search paradigm, with participants as collaborators.Can usefully summarize key features of a large body of

data, and/or offer a ‘thick description’ of the data set.Can highlight similarities and differences across the

data set.Can generate unanticipated insights.Allows for social as well as psychological interpreta-

tions of data.Can be useful for producing qualitative analyses suited

to informing policy development.

Using thematic analysis in psychology 97

methods, and demonstrate how to do ‘qualitative

analysis’, reinforces this point in that his focus is

largely thematic �/ but not claimed as such.3. Some authors, such as Potter (1997: 147 �/

48) argue that one should not simply provide

‘recipes’ for qualitative methods, such as DA,

because ‘a large part of doing discourse analysis

is a craft skill, more like bike riding or sexing a

chicken than following the recipe for a mild

chicken rogan josh. . . This makes it hard to

describe and learn’. While we do not disagree

that the skills needed for qualitative analyses of

all types need to be learned, others, such as

McLeod (2001), argue that by not discussing the

‘how to’ of analysis, we keep certain methods

mysterious (and thus elitist). Instead, if we want

to make methods democratic and accessible �/and indeed, to make qualitative research of all

forms more understandable to those not trained

in the methods, and arguably thus more popular

�/ we need to provide concrete advice on how toactually do it. We are not questioning the

importance of ‘non-recipe’ forms of training,

but while ‘recipes’ necessarily diminish the

complexity of certain methods, they are impor-

tant for making methods accessible.4. Foster and Parker (1995) suggest one

way to acknowledge the creative and active role

of the analyst is to use the first person when

writing.5. Content analysis is another method that

can be used to identify patterns across qualitative

data, and is sometimes treated as similar to

thematic approaches (eg, Wilkinson, 2000). How-

ever, content analysis tends to focus at a more

micro level, often provides (frequency) counts

(Wilkinson, 2000), and allows for quantitative

analyses of initially qualitative data (Ryan and

Bernard, 2000). Thematic analysis differs from

this in that themes tend not to be quantified

(although sometimes they may be; and Boyatzis

(1998) suggests thematic analysis can be used to

transform qualitative data into a quantitative

form, and subject them to statistical analyses;

and the unit of analysis tends to be more than a

word or phrase, which it typically is in content

analysis.

6. The definition by Boyatzis (1998) of latentand manifest is somewhat narrower than ouridentification of latent and semantic, and heidentifies thematic analysis as incorporatingboth latent and manifest aspects. However, thisresults from the fact that he associates theprocess of interpretation with latent analysis �/whereas we would argue that it should also be animportant element of a semantic approach.

7. We are assuming that you will be workingwith a ‘good quality’ data corpus and data set.We would argue that ‘good data’ are defined by aparticular set of criteria regarding what, why,and how they were collected, and offer rich,detailed and complex accounts of the topic.Good data do not just provide a surface over-view of the topic of interest, or simply reiterate acommonsense account. The challenge for thenovice researcher is to interact with researchparticipants in such a way that they generaterich and complex insights. Producing a goodanalysis of poor quality data is a far moredemanding task for the analyst, although it canpotentially be performed by a skilled and ex-perienced analyst.

8. See Poland (2002) for a discussion of theproblems with the idea of a ‘verbatim’ transcript,and what is left out, and retained, through thisprocess.

9. What we mean by thematic map is similarto, but less detailed than, the ‘codebook’ Ryanand Bernard (2000) refer to, which involves adetailed account of the hierarchical relationshipbetween codes, as well as a description of each,their criteria, exemplars and counter �/ examples,and other such details. Like Boyatzis’s (1998)account of a thematic code, this model is thenapplied to (and revised in relation to) the data.See Figures 2 �/4 for visual representations of athematic maps and its refinement. Another ex-ample of a thematic map �/ this time in tableform �/ can be found in Frith and Gleeson (2004).

References

Antaki, C., Billig, M., Edwards, D. and Potter, J.2002: Discourse analysis means doing ana-lysis: a critique of six analytic shortcomings.

98 V Braun and V Clarke

DAOL Discourse Analysis Online [electronic

version], 1(1).

Aronson, J. 1994: A pragmatic view of thematic

analysis. Qualitative Report 2(1).

Atkinson, J.M. and Heritage, J. 1984: Structures

of social action: studies in conversation

analysis . Cambridge University Press.

Attride-Stirling, J. 2001: Thematic networks: an

analytic tool for qualitative research. Quali-

tative Research 1, 385 �/405.

Bird, C.M. 2005: How I stopped dreading and

learned to love transcription. Qualitative

Inquiry 11, 226 �/48.

Boyatzis, R.E. 1998: Transforming qualitative

information: thematic analysis and code

development . Sage.

Braun, V. 2005a: In search of (better) female

sexual pleasure: female genital ‘cosmetic’

surgery. Sexualities 8, 407 �/24.

Braun, V. 2005b: Selling the perfect vulva. Manu-

script under submission.

Braun, V. and Kitzinger, C. 2001: The perfectible

vagina: size matters. Culture Health & Sexu-

ality 3, 263 �/77.

Braun, V. and Wilkinson, S. 2003: Liability or

asset? Women talk about the vagina. Psycho-

logy of Women Section Review 5, 28 �/42.

Braun, V., Gavey, N. and McPhillips, K. 2003: The

‘fair deal?’ Unpacking accounts of recipro-

city in heterosex. Sexualities 6, 237 �/61.

Bryman, A. 1988: Quantity and quality in social

research . Routledge.

Bryman, A. 2001: Social research methods .

Oxford University Press.

Burman, E. and Parker, I., editors, 1993: Dis-

course analytic research: repertoires and

readings of texts in action . Routledge.

Burr, V. 1995: An introduction to social con-

structionism . Routledge.

Charmaz, K. 2002: Qualitative interviewing and

grounded theory analysis. In Gubrium, J.F.

and Holstein J.A., editors, Handbook of

interview research: context and method .

Sage, 675 �/94.

Clarke, V. 2005: ‘We’re all very liberal in our

views’: Students’ talk about lesbian and gay

parenting. Lesbian and Gay Psychology Re-

view 6, 2 �/15.

Clarke, V. and Kitzinger, C. 2004: Lesbian and gay

parents on talk shows: resistance or collu-

sion in heterosexism. Qualitative Research

in Psychology 1, 195 �/217.

Clarke, V., Burns, M. and Burgoyne, C. 2006:

‘Who would take whose name?’ An explora-

tory study of naming practices in same-sex

relationships. Manuscript under submission.

Dey, I. 1993: Qualitative data analysis: a user-

friendly guide for social scientists . Routledge.

Edwards, J.A. 1993: Principles and contrasting

systems of discourse transcription. In Ed-

wards, J.A. and Lampert, M.D., editors,

Talking data: transcription and coding in

discourse research. Lawrence Erlbaum As-

sociates, 3 �/31.

Edwards, J.A. and Lampert, M.D., editors, 1993:

Talking data: transcription and coding in

discourse research . Lawrence Erlbaum As-

sociates.

Elliott, R., Fischer, C.T. and Rennie, D.L. 1999:

Evolving guidelines for publication of qua-

litative research studies in psychology and

related fields. British Journal of Clinical

Psychology 38, 215 �/29.

Ellis, S.J. and Kitzinger, C. 2002: Denying equal-

ity: an analysis of arguments against low-

ering the age of consent for sex between

men. Journal of Community and Applied

Social Psychology 12, 167 �/80.

Ely, M., Vinz, R., Downing, M. and Anzul, M.

1997: On writing qualitative research: living

by words . Routledge/Falmer.

Fine, M. 2002: Disruptive voices: the possibilities

for feminist research . University of Michi-

gan Press.

Foster, J.J. and Parker, I. 1995: Carrying out

investigations in psychology: methods and

statistics . BPS Books.

Frith, H. and Gleeson, K. 2004: Clothing and

embodiment: men managing body image and

Using thematic analysis in psychology 99

appearance. Psychology of Men and Mascu-

linity 5, 40 �/48.

Glaser, B. 1992: Basics of grounded theory

analysis . Sociology Press.

Hayes, N. 1997: Theory-led thematic analysis:

social identification in small companies. In

Hayes, N., editor, Doing qualitative analysis

in psychology. Psychology Press.

Holloway, I. and Todres, L. 2003: The status of

method: flexibility, consistency and coher-

ence. Qualitative Research 3, 345 �/57.

Hollway, W. 1989: Subjectivity and method in

psychology: gender, meaning and science .

Sage.

Hollway, W. and Jefferson, T. 2000: Doing quali-

tative research differently: free association,

narrative and the interview method . Sage.

Hutchby, I. and Wooffitt, R. 1998: Conversation

analysis: principles, practices and applica-

tions . Polity Press.

Kelle, U. 2004: Computer-assisted analysis of

qualitative data. In Flick, U., von Kardorff,

E. and Steinke, I., editors, A companion to

qualitative research . Sage, 276 �/83.

Kitzinger, C. and Willmott, J. 2002: ‘The thief of

womanhood’: women’s experience of poly-

cystic ovarian syndrome. Social Science &

Medicine 54, 349 �/61.

Lapadat, J.C. and Lindsay, A.C. 1999: Transcrip-

tion in research and practice: from standar-

dization of technique to interpretive posi-

tionings. Qualitative Inquiry 5, 64 �/86.

Laubschagne, A. 2003: Qualitative research �/airy fairy or fundamental? Qualitative Re-

port [electronic version], 8(1).

McLeod, J. 2001: Qualitative research in counsel-

ling and psychotherapy. Sage.

Meehan, T., Vermeer, C. and Windsor, C. 2000:

Patients’ perceptions of seclusion: a qualita-

tive investigation. Journal of Advanced Nur-

sing 31, 370 �/77.

Miles, M.B. and Huberman, A.M. 1994: Qualita-

tive data analysis: an expanded sourcebook ,

second edition. Sage.

Murray, M. 2003: Narrative psychology. In Smith,J.A., editor, Qualitative psychology: a practi-cal guide to research methods . Sage, 111 �/31.

Parker, I. 2004: Criteria for qualitative research inpsychology. Qualitative Research in Psy-chology 1, 95 �/106.

Patton, M.Q. 1990: Qualitative evaluation andresearch methods , second edition. Sage.

Poland, B.D. 2002: Transcription quality. InGubrium, J.F. and Holstein, J.A., editors,Handbook of interview research: contextand method. Sage, 629 �/49.

Potter, J. 1997: Discourse analysis as a way ofanalysing naturally occurring talk. In Silver-man, D., editor, Qualitative research: theory,method and practice . Sage, 144 �/60.

Potter, J. and Wetherell, M. 1987: Discourse andsocial psychology: beyond attitudes andbehaviour. Sage.

Reicher, S. 2000: Against methodolatry: somecomments on Elliott, Fischer, and Rennie.British Journal of Clinical Psychology 39, 1 �/6.

Reicher, S. and Taylor, S. 2005: Similarities anddifferences between traditions. Psychologist18, 547 �/49.

Riessman, C.K. 1993: Narrative analysis . Sage.

Roulston, K. 2001: Data analysis and ‘theorizingas ideology’. Qualitative Research 1, 279 �/302.

Rubin, H.J. and Rubin, I.S. 1995: Qualitativeinterviewing: the art of hearing data . Sage.

Ryan, G.W. and Bernard, H.R. 2000: Data man-agement and analysis methods. In Denzin,N.K. and Lincoln, Y.S., editors, Handbook ofqualitative research , second edition. Sage,769 �/802.

Seale, C. 1999: The quality of qualitative re-search . Sage.

Seale, C. 2000: Using computers to analysequalitative data. In Silverman, D., Doingqualitative research: a practical handbook .Sage, 155 �/74.

Silverman, D., editor 2000: Doing qualitativeresearch: a practical handbook . Sage.

100 V Braun and V Clarke

Singer, D. and Hunter, M. 1999: The experienceof premature menopause: a thematic dis-course analysis. Journal of Reproductiveand Infant Psychology 17, 63 �/81.

Smith, J.A. and Osborn, M. 2003: Interpretativephenomenological analysis. In Smith, J.A.,editor, Qualitative psychology: a practicalguide to methods . Sage.

Smith, J.A., Jarman, M. and Osborn, M. 1999:Doing interpretative phenomenological ana-lysis. In Murray, M. and Chamberlain,M.M.K., editors, Qualitative health psychol-ogy: theories and methods . Sage.

Stenner, P. 1993: Discoursing jealousy. InBurman, E. and Parker, I., editors, Dis-course analytic research: repertoires andreadings of texts in action. Routledge, 94 �/132.

Strauss, A. and Corbin, J. 1998: Basics of quali-tative research: techniques and proceduresfor developing grounded theory. Sage.

Taylor, G.W. and Ussher, J.M. 2001: Making senseof S&M: a discourse analytic account. Sex-ualities 4, 293 �/314.

Toerien, M. and Wilkinson, S. 2004: Exploringthe depilation norm: a qualitative question-naire study of women’s body hair removal.Qualitative Research in Psychology 1, 69 �/92.

Tuckett, A.G. 2005: Applying thematic analysis

theory to practice: a researcher’s experience.

Contemporary Nurse 19, 75 �/87.

Ussher, J.M. and Mooney-Somers, J. 2000: Nego-

tiating desire and sexual subjectivity: narra-

tives of young lesbian avengers. Sexualities

3, 183 �/200.

Widdicombe, S. and Wooffitt, R. 1995: The

language of youth subcultures: social iden-

tity in action . Harvester Wheatsheaf.

Wilkinson, S. 2000: Women with breast cancer

talking causes: comparing content, biogra-

phical and discursive analyses. Feminism &

Psychology 10, 431 �/60.

Willig, C. 1999: Beyond appearances: a critical

realist approach to social constructionism.

In Nightingale, D.J. and Cromby J., editors,

Social constructionist psychology: a critical

analysis of theory and practice . Open Uni-

versity Press, 37 �/51.

Willig, C. 2003: Discourse analysis. In Smith,

J.A., editor, Qualitative psychology: a prac-

tical guide to research methods . Sage, 159 �/83.

Yardley, L. 2000: Dilemmas in qualitative health

research. Psychology and Health 15, 215 �/28.

About the authorsVIRGINIA BRAUN is a senior lecturer in the Department of Psychology at the Universityof Auckland, where she teaches, supervises and conducts qualitative research. Herresearch interests are primarily focused around women’s health, gendered bodies, andsex and sexuality, and the intersections between these areas. She is currently working onprojects related to ‘sex in long-term relationships’, ‘female genital cosmetic surgery’ and‘the social context of STI transmission’.VICTORIA CLARKE is a senior lecture in social psychology at the University of the Westof England. She has published a number of papers on lesbian and gay parenting, andco-edited two special issues of Feminism and Psychology on Marriage (with Sara-JaneFinlay and Sue Wilkinson). She is currently conducting ESRC-funded research on same-sex relationships (with Carol Burgoyne and Maree Burns) and co-editing (with ElizabethPeel) a book on LGBTQ psychology (Out in psychology, Wiley).

Using thematic analysis in psychology 101

View publication statsView publication stats