Multi-Method Qualitative Text and Discourse Analysis: A Methodological Framework

Abstract

The growing interest in combining different approaches to qualitative text and discourse analysis has so far not been met with adapted methodological resources. This article aims to address this gap by developing a methodological framework for combining qualitative text and discourse analysis. First, we introduce four traditions that we identify as four families of methods of text/discourse analysis with different logics: Discourse Analysis, Foucauldian Discourse Analysis, Thematic Analysis, and Qualitative Content Analysis. Second, we review the literature to show how these methods have been combined across disciplines and case studies. Third, we build upon existing literature to unpack the benefits and challenges of multi-method text/discourse analysis, and offer strategies to help navigate the problems that may arise. Overall, this article introduces multi-method qualitative text and discourse analysis (MMQTDA) as a methodological framework to provide guidance and offer solid foundations for an emerging methodological conversation in qualitative text research.

Introduction

Qualitative researchers have developed a wide range of methods of analysis to make sense of textual data, one of the most common forms of data used in qualitative research (Attride-Stirling, 2001; Cho & Trent, 2006; Stenvoll & Svensson, 2011). As a result, qualitative text and discourse analysis (QTDA) has become a thriving methodological space characterized by the diversity of its approaches (Gee & Handford, 2013; Kuckartz, 2014; Schreier, 2012). In parallel, scholars have put forward the benefits of mixed-method research for text analysis on the basis that different traditions complement each other and can benefit the overall research process (Baker & Levon, 2015; Liu, 2019; Zhang, 2012). A growing literature seems to indicate an interest in combining not only methods of text analysis across the qualitative/quantitative divide but also within the qualitative tradition (Darbyshire et al., 2005; Massey et al., 2006). However, this literature has so far been limited to empirical case studies rather than been developing methodological resources. In the absence of such resources, combining methods of QTDA has become a methodological problem that requires clarification: what are the approaches, benefits, and challenges associated with such methodological strategy?

This article aims to address this gap by introducing a methodological framework for multi-method qualitative text and discourse analysis (MMQTDA). Rather than considering qualitative text and discourse analysis methods as a homogeneous field constructed in opposition to its quantitative counterpart, we approach them in their diversity and the complementarity of their tools, logics, and objectives. In this article, we review the existing literature to highlight the benefits and challenges of MMQTDA and illustrate the options available. Overall, this framework aims to support researchers in expanding their methodological imagination and critically exploring potential strategies adapted to their project. It also facilitates the legitimation, adoption, and implementation of MMQTDA. Finally, it offers a starting point for a collective methodological conversation cutting across specific cases and disciplinary traditions.

In doing so, this article expands on the literature focusing on multi-method research design. While the term “multi-method” sometimes qualifies mixed-method projects articulating qualitative or quantitative methods in specific ways (Blatchford, 2005; Hammond, 2005), we approach multi-method research design as research that combines “techniques from different methodological families within a single study” (Seawright, 2016, p. 1) independently of whether the methods are all qualitative or all quantitative (see Kochan, 2016; Palakshappa & Ellen Gordon, 2006). While mixed-method research has now been established as a field of research in its own right (Small, 2011, p. 60), combining methods other than “qualitative + quantitative” has yet to emerge as a collective conversation.

Promoters of multi-method research design have put forward the benefits of such a methodological strategy. As “research traditions exert a powerful influence over the thinking of academic researchers” (Graham, 1999, p. 76), they structure the way we think, which not only influences results but also potentially builds in cookie-cut blinders in our analysis that prevents us from fully grasping the complexity and nuances of the phenomena we study. To address this issue, scholarship has promoted multi-method research design as a way to prevent “methodological complacency” by inviting researchers to go beyond the entrenched domains of inquiries dictated by their methodological traditions (Graham, 1999, p. 76). As such, “well-designed and well-executed multi-method research has inferential advantages over research relying on a single method” (Seawright, 2016, p. 1), for example by “establish[ing] missing links” that connect different phenomena and assess them empirically” (Shibin et al., 2018, p. 908). However, while this literature has successfully demonstrated the rationale for a multi-method research agenda, it has not addressed yet qualitative text and discourse analysis.

Based on the idea that multi-method research design combines different methodological traditions, our first step is to identify the different methodological families within the field of qualitative text and discourse analysis. The first section, therefore, introduces four families that are representative of different ways of analyzing textual data qualitatively: Discourse Analysis (DA), Foucauldian Discourse Analysis (FDA), Thematic Analysis (TA), and Qualitative Content Analysis (QCA). DA, TA, and QCA have been selected because of their wide use, and FDA has been selected to show how methods outside those most familiar can have interesting combinatory advantages. Each family represents both a specific logic of text analysis and comprises different methods that share this methodological logic. In the second section, we review the literature to illustrate four ways of combining these methods: (a) QCA + TA, (b) QCA + DA, (c) DA + TA, and (d) FDA + other methods. In doing so, we provide an overview of how MMQTDA has been applied across diverse research problems, disciplines, and case studies. Finally, based on this literature review and our own experiences of implementing, learning and teaching QTDA, we put forward the main motivations, challenges, and strategies for implementing multi-method qualitative text and discourse analysis.

Four Approaches to Qualitative Text and Discourse Analysis

This section introduces the logics of four approaches to QTDA (summarized in Table 1 of the Supplementary material) to reveal possible seams for their combination. Some of these approaches (DA, FDA) come as both an analytical framework and a set of methods. We define analytical framework as the coherent and explicit articulation of concepts previously defined that help researchers inform research design decisions and guide their analysis. Existing theories can be turned into analytical frameworks for the scope of specific projects. As part of a DA or FDA, the concept of discourse represents the core of the analytical framework, backed up by discourse theories the author mobilizes as well as articulated with other concepts that aim to operationalize the project empirically. We define methods as sets of practices that disclose the underpinnings behind the research choices we make and offer overall structure, coherence, and guidance regarding how we produce knowledge. Providing a whole demonstration of how to use QTDA methods falls outside the scope of this article, and we strongly recommend that readers familiarize themselves with how each method can be used individually before embarking on a multi-method strategy.

Discourse Analysis

Since the 1980s, methods under the umbrella term of “discourse analysis” have gained popularity in the humanities and social sciences (Schiffrin et al., 2001; van Dijk, 1993). From discursive psychology to critical discourse analysis (CDA), approaches to DA all mobilize analytical strategies and techniques to understand how discourse—defined here expansively as language in context—produces social configurations and contributes to the (re)production of social and political orders. Rather than a strict rulebook, DA encompasses both an analytical framework—based on the idea that discourse plays a role in society and world politics—and a set of methods of analysis to empirically investigate this idea (Alejandro et al., 2023). In that sense, DA aims to identify linguistic mechanisms and processes in the socio-political context in which discourses are produced and/or received to understand their role and potential effects.

As an analytical framework, DA approaches discourses as social practices constitutive of identities, norms, and perceptions, which comprise both explicit and implicit dimensions. As a set of methods, DA offers specific tools and strategies to investigate these processes empirically. In comparison to other methods of text analysis, DA is unique in the flexibility of its implementation. This flexibility is both an asset as well as a challenge. While it enables researchers to develop strategies specific to each project, it requires them to each time identify the tools best adapted to answer their research question, for example, by reviewing the empirical and methodological literature. As such, DA methodological guidelines oscillate between broad steps to organize the analysis—for example, Fairclough’s (1989) well-established three-dimensional framework—and more “bespoke” approaches (Alejandro et al., 2023), directing researchers to the rich DA toolbox and dictionary type literature already developed (see Baker & Ellece, 2011; Gee, 2014; Taylor, 2013). After mapping the contexts in which the textual material analyzed has been produced and/or received (visual, textual, socio-historical, of utterance) and doing a preliminary analysis focusing on the explicit dimensions of the texts (e.g., what is the text about?), researchers are encouraged to read and re-read the material. This iterative process enables the inductive identification of trends or mechanisms within the material (e.g., some voices seem excluded, some phenomena appear legitimized...) before looking for DA tools that would enable researchers to provide empirical evidence of these mechanisms to an audience. Researchers then systematically analyze the corpus via the tools identified before writing the demonstration mobilizing existing literature about discursive mechanisms and elements of context to support their interpretation. Doing so, DA supports in-depth analysis but with the trade-off that the method is time-intensive which often results in the analysis of relatively small corpora due to feasibility.

Foucauldian Discourse Analysis

Opinions differ as to whether FDA is a method of analysis or a methodology (see Dias & Janjua, 2018; Hook, 2001 for an illustration of the different positions). While we agree this tradition has been less formalized as a method than the others mentioned (see Dias & Janjua, 2018; Kendall & Wickham, 1999 as examples of methodological literature attempting to address this gap), we chose to include it as one of the four QTDA families introduced in this article nonetheless as it offers interesting research design opportunities when used in combination. Despite sharing a similar focus on discourse, we separate FDA from DA as it follows a different logic than the DA family at large.

In line with other DA approaches, FDA comprises an analytical framework based on the concept of discourse—more strictly how it is mobilized in Foucauldian theories—with strategies for empirical operationalization. In contrast with DA methods that largely focus on unpacking specific linguistic mechanisms within discourse, FDA focuses on a discourse as the core of the research design, which Foucault (2002) refers to as a “discursive formation” (p. 34). More particularly, FDA has tended to focus on naturalized discourses based on the premise that some discourses (i.e., ways of speaking about the world) can become so pervasive for certain social groups in specific periods that these groups lose their capacity to decenter themselves cognitively from it, that is, these discourses have become naturalized (Alejandro & Knott, 2022). Through this process, naturalization contributes to producing a socio-discursive world in which people assume the only possible socio-political options are those naturalized by the language they use. In doing so, FDA pursues two main objectives both concerned with the text-context relations: it aims to identify how specific discourses and power relationships have been co-constituted historically and produce specific social and political orders; it also holds a reflexive/critical ambition through its capacity to denaturalize naturalized discourses.

Practically, two main strategies are commonly implemented when using FDA as a method: genealogy and problematization. Genealogy is the most formalized method within FDA. Genealogy uncovers how discourses pervasive across institutions, genres, and social groups are legitimized and legitimize social norms and political orders. It takes as a starting point the emergence of a new discourse and other elements such as “institutions, architectural forms, regulatory decisions, laws, administrative measures, scientific statements, philosophical, moral and philanthropic propositions” to trace “the system of relations that can be established between these elements” (Foucault, 1980, pp. 194–228). Genealogy is therefore a multi-method research design in itself as it combines textual and non-textual data. Problematization is a strategy through which researchers identify implicit (problematic) taken-for-granted assumptions within a discourse. It is commonly understood as a necessary step of genealogy, although it is also being used outside of genealogical work. For example, across his works, Foucault (1965, 1979) problematizes the discourse of madness as an illness to be cured by tracing the emergence of psychiatry as a medicalized discourse about madness since the 18th century and the correlated constitution of related architectural forms and judiciary reforms that constitute psychiatry as a discursive institution productive of norms, subjectivities, and power relations.

Overall, FDA requires the use of big textual corpora as it aims to show how a discourse has become pervasive across institutions. Its multi-method design is associated with a clear analytical framework drawing from Foucault’s theorization of the role of discourse in society. It also provides a solid structure for investigating text-context relations, such as connecting textual elements from different sorts (e.g., interview transcripts, legal texts) to macro elements (e.g., neoliberalism and neoliberal institutions).

Thematic Analysis

Different from DA and FDA which come with an analytical framework, TA is a method that formalizes the reading of texts by coding them (i.e., labeling/categorizing segments of text) to form and map themes for interpretation, making our spontaneous approach to understanding textual data more conscious and more rigorous. It aims to identify, analyze, and interpret “patterns of meaning” (Braun & Clarke, 2012, p. 57) through the reading and re-reading of textual data: what is this data about?

The conduct of TA begins with the coding of the textual data. Codes represent “most often a word or a short phrase that symbolically assigns a summative, salient, essence capturing, and/or evocative attribute for a portion of language-based or visual data” (Saldaña, 2015, p. 4). Different types of TA have been developed throughout the years, such as two-cycle coding (Saldaña, 2015), six-stage data coding (Fereday & Muir-Cochrane, 2006), and six-step thematic networks analysis (Attride-Stirling, 2001). Although it can follow different steps, the basic logic of TA, as we described above, is a pattern-finding iterative process (Alejandro, 2020) that goes from identifying codes to grouping these codes into themes that these codes have in common. To take Attride-Stirling’s (2001) demonstration as an example, conducting TA should start with classifying textual data to construct thematic networks, then exploring these constructed thematic networks to create patterns, and finally interpreting the identified patterns. The use of thematic maps has also been recommended to demonstrate the “salient themes” at different levels and illustrate “the relationships between them” (Attride-Stirling, 2001, p. 388).

While thematic networks are most often developed inductively, some adopt a rather deductive approach, for example in the case when an applicable theory has been identified before the coding (Fereday & Muir-Cochrane, 2006). The coding style—what to code for and how to label the code—is also subject to variation depending on the objective of the study, as brilliantly illustrated by Saldaña (2015). Overall, TA does a good job at systematically mapping and organizing the topics within a body of texts, thus providing a holistic view of what is in this body of texts and how to organize it to answer a research question.

Qualitative Content Analysis

Similar to TA, QCA is a method of analysis that does not come with a specific analytical framework. In contrast with TA, it is used to process specific aspects of a bigger corpus. As Schreier (2012, p. 3) comments, QCA “will help you describe your material only in certain respects which you have to specify.” She continues, QCA first “requires you to “translate all those meanings in your material that are of interest to you into the categories of a coding frame,” and second, “it has you classify successive parts of your material according to these categories” (Schreier, 2012, p. 5). The logic of QCA is a logic of reduction, as the construction of the coding frame aims to identify a set of codes—be them simple phrases and words such as for TA, or numbers or letters—that enables the interrogation of only the aspects of the textual material relevant to the research question, problem, theory or empirical priors. QCA thus consists in systematically classifying material (be it newspaper articles, books, adverts, etc.) as instances of the codes of a coding frame; asking whether a code is present or absent in each unit analyzed or in which frequency.

QCA is the more step-by-step method presented in this article: from formulating research questions and hypotheses, to identifying variables and codes, to constructing a coding frame, and to pre-testing and testing the selected corpus under the guidance of the coding frame (Bernard et al., 2017). Due to its reduction attribute, QCA commonly requires a bigger corpus than TA or DA to produce meaningful results; and as Elo and Kyngäs (2008) caution, QCA researchers should consistently keep the research aims and questions in mind, otherwise, they are likely to get lost amid unexpected but exciting unrelated findings resulting from analyzing such a big corpus. In addition to being used as a deductive method, different types of QCA have been developed to serve different purposes. Hsieh and Shannon (2005, pp. 1279, 1281, 1283), for instance, put forward three traditions: “conventional” QCA for “describ[ing] a phenomenon,” “directed” QCA for “validat[ing] or extend[ing] conceptually a theoretical framework or theory,” and “summative” QCA for “understanding the contextual use of the words or content.”

Among the different traditions introduced in this article, QCA is the closest to a quantitative text analysis approach. Positions vary as to what distinguishes QCA from quantitative content analysis. QCA has been historically developed to address the limits of quantitative content analysis (Kracauer, 1952). We align with Schreier (2012) and others who posit that the distinction between qualitative and quantitative content analysis is a matter of degree (p. 14). We approach content analysis as a spectrum, the qualitative end of it being the most interested in mobilizing contextualization to support its interpretation and acknowledging that codes are themselves interpretative devices at the core of analytical work. Overall, QCA enables researchers to analyze relatively large corpora for a qualitative method by assessing the presence/absence and frequency of relevant elements. Such logic also facilitates producing comparative analysis, for example, to identify the variation of contextual elements related to the sources analyzed such as the socio-demographic characteristics of interviewees or the political leaning of newspapers.

Combining Methods of Qualitative Text and Discourse Analysis: A Review of the Literature

Based on our introduction of the logics and differences among the four families of QTDA, this section makes explicit the rationale and interest of different MMQTDA strategies. Such strategies have been applied in some empirical case studies, however, much like the lack of methodological justification that had traditionally characterized early works in mixed-method research (Small, 2011, p. 71), few of these works justify explicitly why they do so. One exception needs to be noted: studies combining QCA and DA are more likely to have a more transparent methodological rationale. We will start with this strategy and, based on interdisciplinary empirical examples, illustrates how MMQTDA has been implemented through different combinations.

Qualitative Content Analysis + Discourse Analysis

While QCA enables us to study large corpora, DA approaches help us unpack the linguistic mechanisms at play and their potential socio-political effects. QCA provides breadth and DA provides depth, as DA supports a close reading of the text and provides specific tools to evidence the implicit dimensions of meaning-making. These methods, therefore, complement each other well which explains why they are often combined (Herrera & Braumoeller, 2004).

Two main strategies are used to combine QCA and DA. The first one uses DA as an analytical framework and QCA as a method. The results of QCA are interpreted through the lens of discourse. For example, in their article about how the Israeli government used social media to achieve strategic goals such as public diplomacy or propaganda, Heemsbergen and Lindgren (2014) realized the limits of approaching 8,363 tweets and 28 images only via QCA. They pointed out how QCA could not “relate the network affordances of social media or map audience interaction” (Heemsbergen & Lindgren, 2014, p. 583), and thus supplemented QCA with DA to better support the contextualization of the data. Adopting a slightly different approach, Hamid and Jahan (2015) aim to “capture examples of identity representation” through QCA based on “66 letters to the Editor of the Daily Star”, while using CDA to highlight “the production of texts and refer to wider social issues to explain the discourses”; in this way, exploring the mutual discursive construction of language, identity, and social divisions (pp. 81–83).

The second strategy is to use DA as not only an analytical framework but also as a method of analysis alongside QCA. This combination of QCA and DA is the closest to the logic of mixed-method research among the four types of combinations introduced in the article. Starting with a DA-driven research question about how American national identity is generated in different contexts, Bui (2022, p. 9) applied the two research methods in parallel: DA to detect “various features of discourse, including turn-taking between interviewer and participant, categorization, subject positions, rhetorical strategies, and lexical fields”, and QCA to “determine the dominant components invoked in the interview transcripts”; in doing so, she believes this can enable her to not only focus on “the components of interest”, but also to reveal “the inconsistencies or implicit meanings with regards to attitudes” which “cannot be captured in a summary content analysis.”

Among different approaches for MMQTDA, the combination of QCA and DA can be considered relatively well-developed. As early as the 1980s, Achard (1987) provided a brief example to demonstrate the justification of introducing DA to content analysis to investigate economic texts; he highlights that “discourse analysis offers an alternative to forms of content analysis relying too much on intuition without rejecting language phenomena or giving them an excessively marginal place” (p. 31). Methodologists have been trying to demonstrate the rationale and advantages of using QCA and DA in combination. As Hardy et al. (2004) suggest, although the ontology and epistemology of content analysis and DA may conflict, when content analysis becomes more qualitative, the meaning of the text is approached as no longer stable but flexible, making content analysis more compatible with DA. Echoing such an argument, Neuendorf (2004) claims that DA and content analysis provide each other with “clues” that stimulate better research preparation and deeper research perspectives (p. 35). Specifically speaking, for QCA, the addition of DA can bring a critical perspective to investigate meaning in context, while for DA, the addition of QCA can provide a broad dataset to help researchers focus on the temporal and spatial changes in discourse (Feltham-King & Macleod, 2016). Based on this literature, it is thus not surprising that the combination of QCA and DA is often justified more transparently and explicitly than other MMQTDA combinations. Beyond this tradition, other approaches to MMQTDA have been rarely explicitly justified by researchers.

Qualitative Content Analysis + Thematic Analysis

Despite QCA and TA both relying on codes, these two traditions might have the most complimentary logics among the four traditions introduced. On the one hand, QCA aims at a reduction as it “will help you describe your material only in certain respects which you have to specify” (Schreier, 2012, p. 3). While codes for QCA can be created inductively based on a pilot analysis, it is the tradition of text analysis employed most commonly in a deductive way with the coding frame generated based on prior literature and existing theoretical frameworks. On the other hand, TA provides a holistic view of the material that QCA precisely lacks. This approach is iterative and mostly inductive as it aims to systematically organize the themes within textual materials through different stages of reading/coding and re-reading/re-coding the materials. ¹

To illustrate how the literature has combined QCA and TA, we can refer to Zhao et al. (2022) who designed a QCA based on existing research findings about gay men dating apps, to generalize the characteristics of Chinese gay men’s self-presentation on a question-and-answering web platform; while using TA in a second stage of the research design to explore potential explanations behind the identified strategies, such as privacy considerations, technoculture impacts, and platform comparisons. As Linder and Seitz (2017) highlight, QCA can be used to “descriptively characterise” data and TA can be used to “identify commonalities” (p. 2729). In addition to sequentially combining QCA and TA to enrich the studies, researchers also adopt this strategy to answer different sub-questions. Joseph (2015), for instance, investigates the “the experience of nurses and nurse leaders in hospitals that enable creating and sustaining a climate for innovativeness” by describing the process through which new ideas are generated, the situations that drive their emergence, and the forces that stimulate their acceptance (p. 173). The complementary nature of QCA and TA thus offers possibilities for expanding research dimensions and better understanding research cases.

Overall, researchers combining QCA and TA have put forward two main reasons for this MMQTDA strategy. On the one hand, Gouvias and Alexopoulos (2016, p. 643) highlight its interest in “triangulat[ing] the findings” based on the diverse perspectives that combining QCA and TA can provide. On the other hand, Trautwein and Bosse (2017) showcase its use as a sequential research design, conducting TA first to develop a coding frame for QCA. However, MMQTDA research combining QCA and TA generally lacks explicit methodological justification.

Thematic Analysis + Discourse Analysis

In contrast with MMQDTA strategies combining the reductive power of QCA with in-depth approaches such as TA and DA, combining TA and DA means combining two methods based on the close reading of all the textual materials collected. What could be the benefits of combining these two (time-consuming) approaches? Here, the difference lies within the type of insights the methods can provide. While DA unpacks language use in context and sheds light on the processes through which meaning-making contributes to the social construction of the world, TA focuses on themes: what is said. DA unpacks the implicit dimensions of discourse while TA provides a systematic strategy to organize mainly explicit thematic dimensions of language. In that sense, combining these methods helps to make more rigorous, clear, and transparent, processes that DA researchers have to undertake anyway at some stage of their analysis and that many TA researchers would naturally lean toward. We identify two strategies for combining these approaches.

The first strategy follows one of the logic expressed above that combines QCA with DA: DA is used as an analytical framework, guiding TA’s coding procedure and the interpretation of findings. For example, to identify how language learners construct their identity, Tian and Dumlao (2020) apply Critical Classroom Discourse Analysis as an analytical framework and TA as a method of analysis, to “analyse data and interpret findings inductively and recursively” (p. 1445). Similarly, Liu (2020) examines her corpus of translated news about Belt and Road Initiative Summits in China “using qualitative thematic analysis under the framework offered by CDA [critical discourse analysis]”, through which she demonstrates how different social groups mobilize “different framings of the stories to implant their own interpretation of the events” (p. 399). Another example is Brooks et al. (2019) who produce a “set of themes” for a more efficient DA on the relationship between companion animals and their owners who experience poor mental health (p. 328). Accordingly, using TA and DA in combination can help researchers to more clearly distinguish between two levels of interpretation of TA results, one that we can approach as “text as themes” and the other one as “text as discourse.” For example, in their research on placentophagy, Botelle and Willott (2020) justify their use of DA as a complement to TA by arguing that “themes” could be “analysed at a deeper level” which may be associated with “broader social discourses” (p. 2).

In the second strategy, TA and DA are both used as methods, often sequentially. Namely, TA is used as a first stage of DA, for example, to identify salient topics within a body of texts as a preparatory stage for DA. An example is Marciano’s (2014) research on transgender identity in the online and offline world, in which they “classified texts into thematic categories” as a preparation to “grasp[ing] the general mind-set of the analysed arenas” through DA (pp. 829–830). Notably, this combination formalizes a strategy already present in the DA literature (see for instance van Dijk, 2001). Some authors have even coined the term “thematic discourse analysis” to refer to this combination. For instance, authors like Potter and Wetherell (1987) suggest conducting a thematic level analysis in DA to help the analysts understand more clearly the “common threads and inconsistencies” (Singer & Hunter, 1999, p. 66) of the discourse being analyzed.

Like with other MMQTDA strategies, why and how DA and TA are combined is often not explicitly justified. This raises a particular concern regarding DA and TA, which often follow a similar logic of valuing the depth of the study and iteratively reading and re-reading textual data. Without some clarification of the relationship between the TA and DA, researchers may lose the original strengths of each method while combining them (more details about this risk in the sub-section “The Challenge of Dissolution”). Such a point is illustrated, for example, by articles where it is methodologically unclear what the authors do, both in terms of procedure and objectives.

Foucauldian Discourse Analysis + One of the Other Approaches

As the literature combining FDA with one of the three other approaches is relatively small, we will introduce together in this sub-section the combination of FDA with any of the other methods. FDA is considered difficult to learn and implement on its own, which may explain why its use within MMQTDA is not as common. However, it is notable that FDA combines well with other methods: it aims to identify a discourse across many texts and how this discourse contributes to co-produce non-textual elements, while the other approaches focus on elements within texts (content, themes, and discursive mechanisms).

In some cases, FDA is used as the analytical framework that helps make sense of and contextualize the results identified via the use of one of the other methods. For example, Cox et al. (2018) use QCA on 250 fire-related articles in a Canadian journal that they interpret via FDA to “identify the systems of meaning (i.e., the order of discourse) and institutionalised relations drawn on in the social construction of recovery as evidenced in the media texts” (p. 472). In Ong’s (2019) article focusing on twenty interviews with older New Zealand-based Filipina migrants, she conducts TA with FDA as an analytical framework to “investigate links between individual narratives and the discourses around carework and aging” (p. 200). In other cases, scholarship demonstrates the interest in combining FDA and other methods side by side. For example, Yamaner (2021) studies the discourses related to the social exclusion of Syrian refugee women in Turkey, using FDA and DA both as analytical methods and methods of analysis (namely combining a genealogy and a toolbox approach to DA inspired by van Dijk’s work on news media). ²

Different advantages of combining the FDA with other methods can be put forward. On the one hand, FDA alone does not offer close reading strategies to systematically analyze texts in the way that methods such as TA and DA can provide. As such, QCA, TA and DA can help structure and empirically strengthen FDA work. Indeed, the identification of a discourse (the core logic of FDA) can be facilitated, for example, by TA, through the close pattern-finding examination of themes present in the material studied. The existence of categories considered as markers of these discourses can be confirmed via QCA, which is especially convenient considering that QCA aims to systematically investigate a specific angle across large corpora. Finally, the empirical demonstration of how the discourse identified co-produces socio-political order (e.g., through processes of naturalization, normalization, legitimation) can analytically gain from the precise demonstration of the discursive mechanisms at play that DA can deliver. On the other hand, FDA offers an analytical framework that helps structure the study of texts to their context, which helps guide the contextualization of results identified with the other methods. Indeed, FDA offers an analytical framework that bridges the micro level of statements and the macro level of institutions and organizations that can help other methods go beyond their study of the textual material.

Finally, it is important to stress that many scholars cite Foucault without conducting FDA, and therefore a QTDA that mentions Foucault is not necessarily multi-method. For example, in the article “Being black, middle class and the object of two gazes,” Canham and Williams (2016) reference Foucault in their theoretical framework as a foundational contribution to what they refer to as “the white gaze” before proceeding to conduct TA. The article is not MMQTDA.

To conclude this section, the above review of these four combining strategies aimed to introduce different ways of doing MMQTDA and illustrate how it has been done with examples from across social sciences. Rather than aiming for exhaustivity, ³ we hope it will provide a starting point for more structured conversations and informed research design decisions.

Benefits, Challenges, and Strategies for Multi-Method Qualitative Text and Discourse Analysis

Based on our review of the literature and our experience in teaching and supervising graduate students using QTDA, we synthesize the benefits of MMQTDA as a research design strategy. We then put forward what we identify as the four main challenges of MMQTDA: the challenges of dissolution, integration, writing up, and high implementation costs. We approach these challenges as the main potential obstacles, difficulties, and resistances commonly encountered when conducting MMQTDA and provide strategies to navigate and overcome them.

Benefits

Multi-method qualitative text and discourse analysis research is motivated by different objectives. Some of these benefits are similar to what methodological literature has put forward regarding mixed-method research overall. In this regard, the first motivation is complementarity understood as the “ability of one type of method to compensate for the weaknesses of the other” (Small, 2011, p. 64), which helps “develop a fuller picture of the research issue” (Hammond, 2005, p. 239). Complementarity is illustrated, for example, by combining the depth of DA with the breath of QCA. A second motivation encountered in both MMQTDA and mixed-method research is confirmation. Also approached as cross-validation or triangulation (Ahmed & Sil, 2012), confirmation deals with the acquisition of additional evidence to increase the validity of results. Namely, “the wish to verify the findings derived from one type of data with those derived from another” (Small, 2011, p. 63) leads to re-analyzing this data through different methods (Cole & Rioufreyt, 2021) and comparing the results. Finally, a third motivation encountered in both MMQTDA and mixed-method research is the opportunity to create further possibilities for research. As Gil-Garcia and Pardo (2006) suggest, in the face of complex social phenomena, the use of a multi-method approach allows us to ask questions from a wider range of perspectives, thus expanding our comprehension of the research case. Especially when expecting to complete a book-length work, such as a Ph.D. thesis, considering the application of a multi-method approach can be enlightening.

In contrast, other types of motivations and benefits appear to be more specific to MMQTDA, and are less commonly mentioned in the broader literature. First, the literature we surveyed illustrates the benefit of combining either DA or FDA as an analytical framework with TA or QCA as a method of text analysis to facilitate the contextualization of the textual elements under scrutiny. Using the concept of discourse provides a theoretical background that can guide researchers in making sense of and interpreting the results found via TA and QCA, methods that provide a useful step-by-step approach but focus on the text without providing a framework articulating the textual results to their context of production or reception. Second, MMQTDA enables scaling up/zooming in the analysis when used in sequential research designs via a two-stage corpus construction. For instance, researchers may want to investigate if the results they identify in a small corpus via DA or TA hold within a bigger sample (using FDA or QCA). Alternatively, researchers may want to bring focus to their FDA or QCA by exploring a sub-sample of their corpus more in-depth via DA or TA.

Research Design Challenges and Strategies to Overcome Them

The Challenge of Dissolution

Considering that the main rationale for multi-method analysis is to combine different traditions so their specific logic can mutually benefit the overall research design, the first challenge researchers may encounter is to lose the specificity of each tradition in the process of conducting research. This is what we refer to via the metaphor of “dissolution”: the risk of dissolving the specificity of each method when combining them. This may, for instance, occur when researchers are not confident or mature in the practice of one or both approaches and end up inadvertently merging all of them through their implemention. To put it more directly, doing a half-baked DA combined with a half-baked TA (because the researcher might be confused regarding the differences between the two approaches) is more likely to yield less convincing results than the rigorous implementation of any of these methods on their own. The dissolution of the different approaches into a “bad mix” makes them lose their specificity and therefore the added value they could otherwise potentially bring to the table in a successfully implemented multi-method research design.

Based on our experience of seeing junior researchers and students struggle with this challenge, we identify three combinations that raise the most concerns. First, when combining DA and TA, some may find themselves stuck in an unsatisfactory version of thematic DA, not being able to see anything but themes and therefore incapable of demonstrating the discursive mechanisms at play (i.e., identifiable through the use of DA tools such as negative descriptors, hyperboles. . .) while simultaneously not producing a systematic thematic analysis either. Second, when combining QCA and TA, one may be confused by the nature and status of the codes as both methods operate through a coding process. In both approaches, coding refers to the interpretative process through which a segment of a text is summarized, synthesized, and signified by a short phrase or concept as a step in the analytical process. But the way researchers construct and mobilize these codes through the analysis is different between the two methods (see below and in first section of the article). Third, DA and FDA both approach text as discourse. What may raise confusion is that the concept of discourse is often approached differently both ontologically and methodologically between the two approaches. The common countable use of the term “discourse” in FDA (discourses such as “medicine” and “feminist discourse” are made of multiplication of statements) vs the uncountable use of discourse in DA (“language in context”, “language in use” as a dimension of the social world building upon but also different from “language” understood as a system of communication using signs and symbols) might explain the confusion.

To address these challenges, on the one hand, we encourage researchers to put the MMQDTA on pause, go back to the methodological literature about the different methods, and start implementing only one of the methods on their project first. Once the analysis is re-centered on one single method and starts producing meaningful results, then one can reintroduce the combination aspect. On the other hand, we invite researchers to distinguish clearly the differences between different use of “code” and “discourse” before implementing MMQTDA:

• - In TA, the process of constructing codes is mainly inductive, structuring what one would do when reading normally, to subsequently build an analytical pyramid, bottom-up, from codes grouped into themes themselves grouped into global themes. Codes are revised through iterative re-readings of the texts or during the stage of grouping them into themes and may therefore evolve both in scope and labeling. They are not used in a coding frame like in QCA. Indeed, even when the codes of the coding frame are be identified inductively in QCA, these codes will then be used deductively to categorize whether yes or no, or according to which frequency, the unit chosen for analysis (e.g., a newspaper article, a book title, a paragraph) contains the content this code aims to capture.

• - In FDA, the unit at the core of the research design is “a discourse” (or what Foucault refers to as a “discursive formation”) while in DA, the focus of analysis is usually discursive mechanisms within discourse as identifiable by DA tools (e.g., somatization, individualization, blame re-assignment).

Overall, we encourage researchers not to combine methods until they are comfortable using these methods separately to avoid running the risk of inadvertently collapsing them and creating confusion within their analysis.

The Challenge of Integration

This challenge deals with the problems raised when “integrating” different methods of analysis within a coherent research design. By integration we mean the practical ways in which the research has to articulate and make sense of elements of the different methods at different stages of the project (for alternative concepts to “integration,” such as “linking” or “meshing”, see Mason, 2006). When and how to integrate can often make researchers confused and stuck.

To overcome such obstacles, we map below three dimensions of research design that need to be taken into consideration, which we encourage researchers to use to brainstorm decisions related to integration. For each dimension, we make transparent the different research design options available to researchers to help them integrate the methods they combine in a more conscious and informed way.

The first decision to take regarding the integration of the methods deals with the methods’ relation to the data. Researchers can, for instance, analyze (a) the same dataset (e.g., for triangulation), (b) an exploratory sample before constructing the main dataset (e.g., using a pilot analysis to identify patterns in the data that will inform the sampling strategy), (c) the main sample and then a sub-sample (e.g., to bring depth or focus on a certain result), and (d) the main sample and then scale it up (to generalize).

A second decision related to integrating the methods deals with the relation between the analytical tools of each method. These could, for example, (a) be independent of each other (in the case of triangulation), (b) build on each other (e.g., when one uses TA to develop codes for the QCA coding frame), and (c) complement each other with one acting as the analytical framework (DA or FDA) and the other as the method per se understood as operationalizing device (e.g., TA and QCA).

Finally, one has to figure out the temporal relation between the different methods; whether the methods will be used (a) sequentially (one method used after the other), (b) iteratively (one alongside the other going back and forth and informing each other), or (c) independently from one another (the timing/order does not matter, no specific relation between them) (see the Supplementary Material for a table summarizing the trade-offs regarding these different options).

The Challenge of Writing Up

This challenge might not be the most difficult to overcome, yet based on our experience, it is commonly experienced as problematic by junior researchers engaging MMQTDA. The question relates to how to write the results/analysis section when combining different methods of analysis. Similar to the writing challenges encountered when putting into writing comparative analysis (Sa Vilas Boas, 2012), converting a multi-dimensional research design into linear writing is not self-evident. We suggest three writing strategies for the results/analysis section of a MMQTDA project.

In the first strategy—method by method, the results identified by different methods are presented one after another, with, for example, one sub-section focusing on the result of QCA and another section focusing on the results of TA. This strategy is adapted, for example, when two methods are used as methods of data analysis (rather than DA being used as an analytical framework). It is also useful when the results lead to different outcomes. In that case, a third sub-section or discussion section is required to offer an overarching explanation capable of making sense of the tension existing between the findings yield from the different methods. A second writing strategy—argument after argument—consists in writing the different arguments supporting the answer to the research question one after the other, with evidence from the different methods included as support for each argument. This strategy is particularly useful when the results of the different methods align/support each other or when FDA/DA is used as an analytical framework to bring depth or expand the understanding of the results identified via another method. The third strategy we suggest—question by question—deals with projects where different sub-questions need to be investigated independently to be able to answer the main research question. In that case, the writing of the different sections of the analysis is organized around the sub-questions and how much evidence from each method will be used in each section depends on the relevance of the respective findings.

Beyond the writing of results, we would also like to encourage readers in strengthening the writing of their methodological section when it comes to MMQTDA. First: the what, defining the methods they chose will help the readers navigate the logic of the combination within the possible MMQTDA strategies. Second: the why, making (more) explicit the reasons behind why they chose to combine different methods of QTDA will make their research design strategy more convincing. Third: the how, unpacking the procedure they implemented will make their research process more transparent. So far, literature mobilizing MMQTDA has not reached the quality standards regarding justification and demonstration expected in more traditional—single—analysis methods such as thematic analysis.

The Challenge of High Implementation Costs

The final challenge identified deals with the cost-benefits of using a multi-method research design. Considering MMQTDA has a higher setting up regarding training, and is more difficult and time-consuming to operationalize, is it worth implementing? This challenge is not specific to MMQTDA and might be valid for any research design that is more complex than what the easiest project may look like. Nonetheless, it can be more acutely experienced for MMQTDA than for more established approaches such as mixed-method text analysis because established paradigms might be easier to get published than emerging ones and the potential symbolic capital resulting from their publication might be more readily perceived. For example, the relevance of MMQTDA might be questioned by reviewers and editors potentially not used to such a combination or specialized in only one of the methods and finding the other one superfluous as a result. The cost might also be more clearly perceived because the lack of clearly defined guidelines for quality standards might make the implementation of MMQTDA more stressful and researchers might potentially feel more isolated and doubtful along the way.

Taking into account Robson and McCartan’s (2015) comment that “advocates of multi-strategy designs are evangelical in their zeal” (p. 175), we want to caution readers against the idealization of MMQTDA. When it comes to research design, the more is not always the better, and excellence does not necessarily lay in the multiplication of things to do, especially in times when many researchers experience overwork. When considering conducting MMQTDA, one needs to ask themselves whether this strategy is actually worth it. Balancing feasibility and reasonableness versus the necessity and comparative advantages each method brings for the project. In a word, you want to make sure that MMQDTA increases the quality of your research, rather than multiplies time-consuming steps and complicates the research in a project without a convincing justification. If so, then do not hesitate to convincingly justify your choice of combining methods of QTDA in your methodological section!

Conclusion

Having realized that many researchers, especially junior ones, are confused about how to design and conduct MMQTDA, we aimed to provide researchers with a methodological framework they can use as a brochure to guide their MMQTDA endeavors. We gave a concise introduction to four main approaches to QTDA: DA, FDA, TA, and QCA. We then demonstrated how they have been used in combination through a review of existing studies. Finally, we detailed the main benefits and challenges we identify when it comes to MMQTDA and how these can be solved in practice.

Our review of the empirical literature mobilizing MMQTDA shows a growing interest and acknowledgment of the benefits of this multi-method strategy. However, we believe the methodological standards for MMQTDA still have to catch up with what is currently expected both for the single use of methods of QTDA and mixed-method research. Indeed, the justification of this research design strategy is rarely present, and the procedure through which it has been implemented is often not transparent. We hope that this article can contribute to addressing this situation by facilitating access to the rationale beyond MMQTDA as well as legitimizing this strategy by demonstrating it is already used across different disciplines and case studies.

Supplemental Material

sj-docx-1-qix-10.1177_10778004231184421 – Supplemental material for Multi-Method Qualitative Text and Discourse Analysis: A Methodological Framework

Supplemental material, sj-docx-1-qix-10.1177_10778004231184421 for Multi-Method Qualitative Text and Discourse Analysis: A Methodological Framework by Audrey Alejandro and Longxuan Zhao in Qualitative Inquiry