Thematic analysis (TA) is an increasingly popular method for the identification and interpretation of patterns of meaning in qualitative data. There have been various competing uses of the term, some rooted in a quantitative paradigm (e.g., Boyatzis, 1998; Fugard and Potts, 2015).¹ However, as we will document in the following, Braun and Clarke's (2006) TA approach (cf. Braun and Clarke, 2019, 2023b), which emphasizes a qualitative methodology, has been so influential that it is virtually synonymous with TA by now. It involves an iterative engagement with the data across six phases: (1) familiarization with the data, (2) initial coding, (3) searching for themes, (4) reviewing potential themes, (5) defining and naming themes, and (6) producing the report (Braun and Clarke, 2006).

A number of literature reviews have addressed the use of various qualitative data analysis methods in communication research and related fields (e.g., Bradbury-Jones et al., 2017; Jensen et al., 2013; Ngenye and Kreps, 2020; Snelson, 2016). However, we have found only two scoping reviews focusing on TA. First, Bowman et al. (2023) investigated 78 papers published in the Association of Computing Machinery (ACM) CHI conference proceedings on Human Factors in Computing Systems. The authors report on limited documentation of methods and a lack of explicit reflection on methodology (reflexivity) in healthcare human-computer interaction. Second, Braun and Clarke (2023a) reviewed 100 TA papers in five journals from health psychology. The authors lament the predominance of qualitative-positivist research methodologies in this field. In this context, they emphasize TA's potential to promote methodological diversity but worry that in TA practice, “positivism invidiously and invisibly directs research practice” (Braun and Clarke, 2023a, p. 698). They identify areas of problematic practice in TA applications within health psychology, call for greater “qualitative research sensibility” (Braun and Clarke, 2023a, p. 701; see also Braun and Clarke, 2013) and develop a rich set of recommendations for authors, reviewers, and editors to apply adequate and unique quality criteria for TA.

This scoping review documents TA application in a different field and adopts a different perspective. We focus on articles in the research area of media and communication studies, which—particularly in the area of reception studies—is more specialized on qualitative data analysis (Jensen, 2021) than for instance health psychology. Here, TA does not serve to promote qualitative research in predominantly quantitative research areas, but it also complements and possibly replaces other methods of engaging with qualitative data. Thus, deliberate or heedless imposition of positivist-quantitative methodology should be less of a concern. Instead, TA arguably has the potential to bridge traded schisms between different empirical methodologies.

In its current state, TA is in a fascinating position as it strikes pragmatic and constructive compromises between empirical research paradigms that have often been presented as antagonistic and mutually exclusive. On the one hand, TA results are qualitative: compared to quantitative content analysis, for instance, TA preserves and reports relatively much detail and nuance in qualitative data that it categorizes into themes.² Fully developed themes, the final outcome of the analysis, are supported by relevant quotes from the data (Connelly and Peltzer, 2016; Mishra and Dey, 2022). On the other hand, compared to longer established methods of manual qualitative data analysis—such as qualitative content analysis (Vaismoradi et al., 2013), discourse analysis, narrative analysis, or grounded theory—TA puts greater emphasis on systematically documented proceedings (cf. Nowell et al., 2017) akin to quantitative research traditions, and it has succeeded in popularizing a relatively detailed, universally applicable protocol for conducting and reporting on qualitative data analysis.³ Arguably, TA therefore mitigates a common concern regarding manual qualitative data analysis, as it guides researchers to reflect on choices and to discuss them in a reasonably transparent manner (Gough and Madill, 2012), facilitating constructive criticism by others. With the proliferation of unstructured, qualitative data in the context of digitalization (cf. Crowston et al., 2012), there is thus great scope for further TA applications and cross-fertilization with automated natural language processing (NLP), including the use of learning algorithms. To bring any of this to fruition, however, TA applications must convincingly document the data under study and their analysis (transparency) and inform readers on the ifs and buts of the authors' choices and priorities.

With these challenges in mind, using the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) checklist (Tricco et al., 2018) as a guiding principle, we systematically review a “large, complex [and] heterogeneous” body of literature (Peters et al., 2015, p. 141), in order to illustrate the “volume, nature and characteristics” (Arksey and O'Malley, 2005, p. 30) of TA scholarship in top-ranked communication journals. By focusing on the “top-end” of prestigious journal articles, we get cues on “best practice” in TA applications and develop hands-on advice for researchers seeking to develop a state-of-the-art analyses of qualitative data in communication research. However, as Braun and Clarke (2019) have worried, publication success of TA applications has not reliably coincided with good practice according to their general quality criteria. Following our analysis, we then discuss apparent shortfalls and untapped potential even within the top-end of TA applications according to publication success. Our specific research questions are the following:

To be sure, the two aspects of research question 3 entail a shift in perspective on what constitutes “quality” in TA applications. We first elicit lessons to be learned from quality judgements by other parties—journal editors and reviewers. We then also compare the articles that have made it into prestigious communications journals (and thus into our data set) against general quality criteria associated with TA. Arguably, both perspectives will be useful to inform researchers, editors and educators engaging with qualitative data analysis.

The paper proceeds as follows. First, we document the construction of the data set. Second, we map the number TA articles over time, per number of authors, per country, and per topical areas. Third, we report the primary and secondary data sources used in TA articles. Fourth, we identify and discuss the most cited references within the data set. Fifth, we address key features of the most cited articles within the data set. Sixth, we deduce practical advice for scholars considering TA applications. Finally, we take stock of major challenges and desirable areas of advances in TA methods.

There is only a limited number of international scientific databases which allow to obtain robust and reliable bibliographic data for a study of this kind, with Scopus and Web of Science being the major databases in the social sciences, including the field of communication (Singh et al., 2021; Zupic and Cater, 2015). As various authors have noted that Scopus provides more comprehensive and higher quality coverage (Falagas et al., 2008; Mongeon and Paul-Hus, 2016), in this study we exclusively used Scopus to identify relevant academic articles to include in the data set. We used Boolean searches on Scopus and searched for journal articles in the research field of Communication (Scopus code 3315), which feature the expressions “thematic analysis” or “thematic analyses.” The final search was conducted on 6 December 2021. The exact search query used in the advanced search interface on Scopus is “( ( TITLE-ABS-KEY (“thematic analysis” OR “thematic analyses”) ) AND ( PUBYEAR < 2024 ) ) AND ( SUBJTERMS ( 3315 ) ) AND ( LIMIT-TO ( DOCTYPE, “ar”) ).” As indicated in the PRISMA flow chart displayed in Figure 1, after deleting duplicates and removing corrections to published articles, this search yielded 1,038 distinct journal articles.

Scopus makes available much meta-data on each article. However, in order to develop a thorough analysis of key features, we also needed to construct some additional variables by human coding of full text articles. Furthermore, for our purpose of identifying successful practices in TA work, publications in more prestigious journals are of greatest interest. To limit the labor required with minimal adverse effects on the quality of the data set, we excluded articles from less prestigious journals and limited the scope of investigation to the top-65 communication journals according to the 2020 classification of the Scimago Journal and Country Rank (SJR) (Guerrero-Bote and Moya-Anegón, 2012).⁴ Focusing on top-ranked journals and top quality papers only is a common strategy to restrict the size of the data based on journals particularly when a study reports findings of articles and the search on Scopus produces a number of articles that is difficulty to handle.⁵

The data set features 41 iFirst articles that, on 6 December 2021, had not been published in a particular journal volume and issue yet but were listed on Scopus. Hence, the data set includes articles with the final publication years 2022 and 2023. Applying judgement sampling (Maul, 2018), we excluded one article published in 1984 (McAdams and Losoff, 1984) as the article does not feature TA in the contemporary meaning of the term. After these curation steps, the final data set consists of 342 prestigious TA articles, which are classified by Scopus as articles (330) or as reviews (12). We confirmed by reading the abstracts that the 12 reviews included are broader literature review articles with original content, rather than book or policy reviews. All articles were published in English.⁶

In the data set under study, the number of relevant articles per year strongly increased over time (see Figure 2). Since 2013 there has been a rather consistent, approximately exponential growth pattern.

The 342 articles in the data set were published in 46 different journals (see Table 1 and Supplementary material A). Journals that feature most TA articles are Health Communication (55 articles), Journal of Social and Personal Relationships (27 articles), and New Media and Society (23 articles). Among the top-65 communication journals, 19 journals did not feature any articles that fulfilled the search criteria.

TA research is often carried out by small teams (see Tables 1, 2). Overall, the data set features 873 different authors. The mean number of authors per article is 2.66. In the data set, 79 articles were single-authored and 125 had two authors, jointly accounting for 59.65% of all articles. The number of individual authors featuring only once is 843. Thus, there is little indication that individual authors would specialize on TA. What is more, there is no apparent positive effect of author team size on impact as indicated by citation counts (see Table 2). To the contrary, single authored articles are on average cited more frequently (mean 30.82) than multi-authored papers (mean 16.93), and this pattern also holds controlling for article age.

The majority of authors reside in the United States (165 articles), followed by the United Kingdom (65 articles) and Australia (24 articles) (see Table 3). Authors from BRICS countries, such as China, play a much smaller role here than in academic publications overall. Perhaps surprisingly, some large European countries are missing from the top ten countries of origin list. In the case of Germany, this absence can be explained by the widespread use of “qualitative content analysis” in that country, an alternative to TA as featured in Mayring (2010).

Our investigation yields several useful insights and actionable recommendations for researchers engaging with TA. On the one hand, based on common features of the successful TA articles in our data set, we can infer on best practices so far. On the other hand, we discuss apparent shortcomings. Most articles in our data set deviate from the transparency and reflexivity ideals of TA. Most articles in our data set deviate from the transparency and reflexivity ideals of TA (see Herzog et al., 2019; Herzog and Kelly, 2023). In the following we discuss on how to improve current TA practices.

Three limitations of this study are particularly noteworthy. First, we only consider peer-reviewed journal articles. This excludes other relevant work, including seminal TA research published in books and book chapters. Still, we followed this document search strategy and retain this decision to ensure the overall quality of the reviewed documents. Second, the data set analyzed is composed of articles published in the top-ranked journals listed on Scopus. It is known that most international scientific databases, including Scopus, overrepresent scientific documents written in English, which are mostly coming from Western countries exhibiting a bias to the detriment of other languages and non-Western scholarship. Including articles from a larger number of communication journals including those written in Spanish, Portuguese, and other languages may have generated different results. To identify best practice, however, focusing on the most prestigious publications is particularly useful, as it provides a scope on the application and uses of TA in the most influential publications in the academic field. Third, a limited number of journals account for the bulk of articles in our data set. We cannot exclude the possibility that a small number of editors and peer communities have affected time trends, so that the generalizability of aggregate trends remains limited. These limitations should be considered when interpreting the results of our analysis.

According to the current rates of growth identified in this paper, TA is at least bound to become a predominant method of manually analyzing qualitative data. This scoping review illustrates that TA does not only serve to popularize qualitative research in some principally quantitative research areas. It has also come to rapidly complement or possibly replace alternative qualitative data analysis methods in the predominantly qualitative research area of communication.

Nonetheless, our analysis also illustrates considerable challenges. There is some divergence between the TA ideal and TA practice in communications articles. A relative strength of TA in its current state is that it prescribes features of an ideal data analysis process, enabling transparency and reflexivity for specific applications. However, according to our scoping review, explicit reflexivity is not the rule even in highly cited articles. Neither do we find much apparent constructive criticism and general deliberations on how to improve TA.¹⁹ The increasing popularity, ease and broad scope of TA application should not lead to complacency. For instance, as TA is applied across a variety of topics and data sets, tailoring of methods to specific circumstances will be required. In the long run, continued further development is necessary for TA to sustain its momentum as a trend in qualitative data analysis. For that, TA may require a critical mass of ambitious specialists after all.

One way of evolving TA is to anchor it in the methodological literature on manual qualitative data analysis, over and beyond the seminal (Braun and Clarke, 2006). By far the most widely cited source on TA is Braun and Clarke's (2006) classic article. It is cited by over a third of all articles in our data set. No other source comes close. However, as Braun and Clarke have refined and further developed their approach, it can only be recommended to also engage with their later works and other methodological guidance. Further commonly referenced sources found in our data set are displayed in Table 5.

Another way ahead for TA is cultivating linkages with automated NLP, an area of swift development over recent years. Processes akin to TA used to be an essential aspect of the early, “familiarization” and training stages of NLP projects. TA may inform the thousands of applied and academic researchers working on NLP on how to better understand what learning algorithms do, what their limitations are, and how to improve them. As Baden et al. (2022, p. 13) note: “[W]ithout a recognition of the knowledge generated by decades of social scientific, linguistic and other text-based research, computational tool developers are likely to miss or inadequately model important textual properties, and are bound to laboriously reinvent the wheel through trial and error.” By the same token, TA researchers (and any qualitative data analyst, who continues to rely on human reading) may learn from experiences in NLP.

Also, greater transparency in terms of the handling of theory is required in TA applications. Furthermore, all indications are that TA is a versatile tool to analyze any type of qualitative data. However, the potential for analyzing various sets of complementary data on the same or closely related phenomena without having to master different data analysis methods remains largely untapped. Neither do we find attempts at replicating data analyses, which is common in quantitative research. As TA is more systematic than alternative manual data analysis methods, this may well lead to advances in qualitative research, documenting to what extent results hinge on subjective choices.

Finally, TA adopts a middle-ground between the extremes of traditional qualitative and quantitative research. It “works” as a stand-alone method, but also seems well-suited to build bridges and allow for complementary combinations. On the one hand, TA could complement less systematic methods—such as narrative analysis or discourse analysis. TA offers an accessible procedure to more transparently document aspects of data analysis, with relatively little loss of detail and nuance by “forcing” categorizations onto qualitative pieces of evidence. On the other hand, TA could also be combined with quantitative methods—such as (1) quantitative content analysis, (2) statistical analysis of the frequencies of themes subject to categorical or numerical data, for instance regarding interviewees' demographics, or even (3) NLP. On the systematic-algorithmic end, content analysis and NLP are logical TA complements for large data sets, and current AI-assisted thematic coding is a promising avenue requiring further exploration. When engaging in such analyses, TA can serve as a blueprint for preparatory steps for NLP applications, in which authors familiarize themselves with the data before running sophisticated software, or when they “train” algorithms. Combinations along these lines certainly have the potential to inspire further advances in TA.

Overall, TA proponents have established it as one of the workhorses of manual qualitative data analysis, distinct from quantitative content analysis and NLP, but relatively compatible with quantitative research methodology. According to our data, however, the potential for multi-method research and cross-fertilization between TA and the quickly developing field of NLP remains largely untapped. Furthermore, TA practice deviates considerably from the ostensible ideals of transparency and reflexivity, which limits its potential for evolution and adaptation. In practice, even rather successful TA users often seem to take it for granted while applying it. Breaking out of these apparent conventions will be an awarding route for some, more ambitious researchers currently considering TA.