Discourse analysis in qualitative research: approaches, steps, and when to use it

Discourse analysis is a family of qualitative methods for examining how language is used in context to construct meaning, social identity, and relations of power. Rather than asking what participants think or experience, discourse analysis asks how language works: what social realities does it produce, what positions does it make available, and whose interests does it serve? It is not a single method but a cluster of related approaches, each with different theoretical roots and analytical procedures.

The most widely used variant is critical discourse analysis, which has accumulated over 30,000 citations across Norman Fairclough's foundational work alone, making it one of the most influential methodological frameworks in the social sciences. Despite this reach, discourse analysis remains poorly understood by researchers outside linguistics and political science, and is frequently confused with thematic analysis by those reading methodology chapters in qualitative textbooks.

This guide covers the four main approaches, explains how they differ from each other and from thematic analysis, walks through the practical steps for critical discourse analysis, and addresses the limitations you need to account for when writing up.

What is discourse analysis?

Discourse analysis is a qualitative method that examines how language constructs social reality, positions speakers and subjects, and reflects or reproduces relations of power. Unlike thematic analysis, which identifies patterns in what people say, discourse analysis focuses on how something is said, what that framing makes possible, and what it excludes. It treats language not as a neutral conduit for meaning but as a constitutive social practice.

How is discourse analysis different from thematic analysis and content analysis?

This is the question that causes most confusion for researchers new to the method. The difference is not just procedural but epistemological: the three approaches rest on different assumptions about what language is and what research can discover through it.

	Discourse analysis	Thematic analysis	Content analysis
What it asks	How does language construct meaning and power?	What patterns of meaning exist across the data?	How often do categories of content appear?
Relationship to language	Language is constitutive (it produces social reality)	Language expresses underlying meanings and experiences	Language is a vehicle for content that can be counted
Epistemological stance	Constructionist (typically)	Flexible; can be realist or constructionist	Typically realist or post-positivist
Unit of analysis	Texts, utterances, discursive acts in social context	Coded segments grouped into themes	Coded units counted and categorised
Research output	Analysis of how discourses work and what they produce	Themes describing patterns in the dataset	Frequency counts and comparisons of content categories
Typical sample size	Small (depth of textual analysis over breadth)	Variable, from a few transcripts to hundreds	Variable, well-suited to large corpora
Rigour criteria	Coherence, interpretive plausibility, reflexive depth	Credibility, transferability, reflexivity	Reliability (inter-rater), validity of category system
Difficulty level	High (requires strong theoretical grounding)	Moderate, accessible to beginners	Moderate, accessible but requires consistent coding

For a full comparison of thematic analysis and content analysis, see the dedicated guide. The key distinction between discourse analysis and the others is that discourse analysis is not primarily interested in what people believe or experience. It is interested in the language system itself: the available subject positions, the assumptions embedded in how things are named, and the social consequences of particular ways of speaking.

If your research question is "what do nurses think about patient care?", you want thematic analysis or possibly qualitative content analysis. If your research question is "how does the language of clinical efficiency construct nurses as productive or non-productive subjects?", you want discourse analysis.

What are the 4 main approaches to discourse analysis?

The umbrella term "discourse analysis" covers several distinct traditions, each with its own theoretical heritage and analytical commitments. Choosing between them means understanding what each approach can and cannot do.

Critical discourse analysis

Critical discourse analysis (CDA) was developed principally by Norman Fairclough and Teun van Dijk from the late 1980s onward. Fairclough's three-dimensional model (text, discursive practice, and social practice) provides the most commonly used analytical framework. The core claim is that discourse is both shaped by social structures and helps to reproduce them. CDA attends to how language enacts, legitimates, or challenges relations of dominance.

Van Dijk's version of CDA adds an explicit focus on ideology and cognition: he argues that dominant ideologies are reproduced through everyday text and talk, and that discourse analysis should make these mechanisms visible. His 1993 article "Principles of Critical Discourse Analysis" has accumulated over 4,500 citations according to SciSpace.

CDA is the approach most likely to be taught in social science research methods courses, most widely published in education, politics, and health policy research, and most adaptable to different kinds of data (political speeches, news media, organisational texts, interview transcripts).

Foucauldian discourse analysis

Foucauldian discourse analysis draws on Michel Foucault's account of discourse as a system of knowledge that produces objects of knowledge, positions for subjects, and regimes of truth. A "discourse" in Foucault's sense is not a conversation or a text. It is an historically specific formation that determines what can be said, who can say it, and what counts as knowledge within a given field.

The foundational text is The Archaeology of Knowledge (1972), in which Foucault sets out his method of analysing discursive formations. Later work, particularly Discipline and Punish (1977) and the three volumes of The History of Sexuality, extended this to show how discourses on medicine, sexuality, and criminality produce the very subjects they appear merely to describe.

In practice, Foucauldian discourse analysis looks for the rules that govern what can be said, identifies the subject positions available within a discourse, and traces how truth claims function to normalise or pathologise. It is particularly powerful for research on professional knowledge (medicine, law, education, social work), on policy texts, and on how categories of identity (race, gender, disability) are produced through institutional language.

Unlike CDA, which often analyses specific texts in close detail, Foucauldian analysis tends to work across broader archives of texts and longer historical timescales.

Conversation analysis

Conversation analysis (CA) has a different intellectual heritage entirely. Developed by Harvey Sacks, Emanuel Schegloff, and Gail Jefferson in the 1960s and 1970s, it begins from an ethnomethodological interest in the methods ordinary people use to produce orderly interaction. The founding paper by Sacks, Schegloff, and Jefferson (1974), "A Simplest Systematics for the Organization of Turn-Taking for Conversation," has over 20,000 citations and remains the most cited article in the journal Language.

CA focuses on the sequential organisation of talk: turn-taking, adjacency pairs (question-answer, invitation-acceptance), repairs, openings and closings, and the moment-by-moment production of mutual intelligibility. It is strictly empirical and grounded in recordings of naturally occurring talk, rather than researcher-generated interview data or written texts.

CA does not typically pursue questions about power or ideology. It is the appropriate tool when your research question is about how people accomplish something through the organisation of talk itself: how doctors manage patient uncertainty in consultations, how job interviews are sequentially structured, how children learn turn-taking in classroom interaction.

Discursive psychology

Discursive psychology (DP) was developed by Jonathan Potter and Margaret Wetherell, whose 1987 book Discourse and Social Psychology: Beyond Attitudes and Behaviour is the foundational text for this approach. DP applies discourse analytic methods to the subject matter of psychology (attitudes, identity, memory, cognition), arguing that these are not private mental states but social and discursive accomplishments.

Where cognitive psychology asks "what attitude does this person hold?", discursive psychology asks "how does this person deploy the language of attitude in this specific interaction, and what work does that do?". DP draws on both CA (for its attention to sequential context and rhetoric in interaction) and to post-structuralist theory (for its constructionist view of psychological categories).

DP is particularly useful for research on identity performance, the discursive construction of mental states, and lay theories of psychological phenomena.

When should you use discourse analysis?

Discourse analysis is the right choice when your research question is genuinely about language and its social functions, not just about capturing people's views. Use it when:

Your research question involves how a phenomenon is discursively constructed, including what language makes possible or excludes
You are studying power relations, ideology, or the reproduction of social inequalities through language
You are interested in subject positions: how language positions people as particular kinds of subjects (patient, expert, deviant, citizen)
Your data are texts with political or institutional weight: policy documents, news media, official reports, professional guidelines, courtroom transcripts
You want to understand how a dominant framing operates and whose interests it serves

Discourse analysis is less well-suited to research questions about individual experience, behaviour, or perception. If you want to understand what it is like to live with a chronic illness, interpretive phenomenological analysis or reflexive thematic analysis are more appropriate. If you want to generate a theory about a social process, grounded theory is the more natural home.

For an overview of how discourse analysis fits within the broader landscape of qualitative research methods, including when to combine methods, see the methods overview guide.

What are the practical steps for critical discourse analysis?

CDA is the most widely used approach and the one most researchers will encounter in social science contexts. The following steps describe the Faircloughian version, which is the most accessible for those new to the method.

Step 1: Define your research question precisely

CDA is not a fishing exercise. You need a focused question that specifies: whose language, in what context, and what aspect of discourse or power you are investigating. Vague questions ("how is healthcare discussed?") produce vague analyses. Strong CDA questions name a discursive object ("how is nursing labour constructed as emotional rather than technical in NHS workforce policy documents from 2010 to 2025?").

Step 2: Select your texts purposively

CDA is almost always purposive rather than representative. You are not trying to generalise to a population; you are trying to understand how a particular discourse works. Select texts that are central to the discursive domain you are studying: canonical policy documents, influential news coverage, key institutional communications. You do not need large samples. Three to twelve texts is typical for a focused CDA study, though this varies by question and scope.

Step 3: Conduct a textual analysis

Analyse the language of your texts closely. At this stage you are attending to features that carry discursive meaning:

Vocabulary choices: what terms are used, what synonyms are avoided, what metaphors are deployed?
Transitivity: who is positioned as agent (doing things) and who as patient (having things done to them)?
Modality: what is presented as certain, probable, possible, or necessary?
Presuppositions: what does the text take for granted without stating?
Nominalisations: what processes are turned into abstract nouns, obscuring agents and causes?
Intertextuality: what other texts or voices does this text draw on, quote, or respond to?

This is close reading with systematic categories. You are not summarising what the text says; you are documenting how it says it and what that how accomplishes.

Step 4: Analyse discursive practice

The second dimension of Fairclough's framework examines how texts are produced and consumed. Who produced this text, for whom, under what conditions? How is it likely to be interpreted by its intended readers? What other texts does it respond to or draw upon? This step contextualises the textual analysis within the communicative situation.

The third dimension asks: what social relations, ideologies, or power structures does this discourse reflect and reproduce? This is where the critical dimension of CDA is most evident. You are arguing that the textual patterns you identified in step 3 are not accidental but serve social functions: legitimating particular arrangements, naturalising inequalities, positioning certain groups as active and others as passive.

This step requires engagement with relevant social theory and contextual evidence. It is not sufficient to note that particular word choices exist; you need to argue why those choices matter socially.

Step 6: Apply reflexivity throughout

Discourse analysis is particularly demanding in terms of reflexivity. You are making claims about what language does and whose interests it serves. Your own position (your assumptions about power, your theoretical commitments, your relationship to the subject matter) shapes what you see in the data and how you interpret it.

Reflexive practice in CDA means documenting your interpretive choices, being explicit about your theoretical framework, and considering alternative readings of the same texts. Strong CDA does not claim to produce the one true reading of a text; it produces a coherent, evidenced, and theoretically grounded interpretation.

For a detailed treatment of how to practise and write up reflexivity, see the dedicated guide.

What does a worked example of CDA look like?

Consider a study examining how a government's social care policy document constructs the "informal carer": a person who provides unpaid care to an elderly or disabled family member.

A textual analysis might identify:

The term "informal carer" is never unpacked; "informal" obscures that this is skilled, emotionally complex, time-intensive work
Passive constructions ("support is provided by family members") obscure that family members (predominantly women) are making active choices with real opportunity costs
The document uses the metaphor of "resilience" for carers, framing their continued provision of care as a personal capacity rather than a structural arrangement that saves the state substantial expenditure
Modality: carers "can" access support services; the state "will" provide guidance. The state's obligations are stated as commitments; carers' agency is marked as optional

The discursive practice analysis would note that this is an official policy document authored by a government department, designed to be read by local authorities and incorporated into commissioning frameworks, not primarily addressed to carers themselves.

The social practice analysis would connect these textual choices to the broader context of welfare state retrenchment: naming unpaid care as "informal" and framing it as resilience serves to naturalise the reliance on unpaid (predominantly female) labour and depoliticise questions about funding and structural provision.

This is what CDA produces: not a description of what the document says, but a theorised account of what its language does.

What are the key limitations of discourse analysis?

Accounting clearly for limitations is part of producing rigorous discourse analysis.

It does not make claims about individuals. Discourse analysis is about how language works socially, not about what any particular author intended or believed. Critics sometimes challenge discourse analysis by pointing to individual exceptions or arguing about authorial intent. This misunderstands the method; explaining this clearly in your methods section avoids unnecessary reviewer comments.

It cannot be easily replicated. Because discourse analysis is interpretive rather than procedural, two analysts working on the same texts will not produce identical analyses. This is not a flaw but a feature, because interpretation is inherently perspectival. However, it means that rigour is assessed differently than in quantitative work: through coherence of argument, depth of textual evidence, and quality of theoretical engagement, not through inter-rater reliability scores.

It requires strong theoretical grounding. You cannot do CDA without understanding something about social theory and the theoretical framework you are using. Analysts who apply CDA procedurally without engaging with its epistemological commitments often produce superficial analyses that do not go beyond surface description.

It is time-intensive. Close reading of texts takes much longer than people anticipate. A study of three policy documents analysed in genuine CDA depth may produce a richer analysis than a cursory examination of thirty. Scope accordingly.

It carries a political dimension. CDA makes explicit that the analyst has a critical stance: an interest in exposing how language serves power. This is a legitimate scholarly position, but it means reviewers from positivist traditions may challenge the approach. Anticipate this and be prepared to articulate your epistemological commitments clearly in your methodology chapter.

How can software support discourse analysis?

Discourse analysis is primarily a close-reading practice, and no software can replace the interpretive work at its core. However, several aspects of the process benefit from technological support.

For researchers working across multiple documents (policy archives, interview corpora, news media collections), the practical challenges of managing data, tracking patterns, and organising coded excerpts can be substantial. Tools like Skimle support this through inductive analysis that surfaces recurring linguistic patterns across a document corpus, and through insight and annotation tools that allow you to attach your analytical notes directly to the textual excerpts that prompted them.

Coding qualitative data in a discourse analysis context differs from thematic coding: you are often annotating texts for grammatical and rhetorical features (transitivity, modality, presupposition) rather than semantic themes. A flexible coding approach that accommodates both deductive (applying pre-specified linguistic categories) and inductive (emergent patterns) work is most appropriate.

For researchers combining discourse analysis with other methods, or working on interdisciplinary projects where different team members use different analytical frameworks, Skimle's REFI-QDA export supports interoperability with NVivo and MAXQDA, which some discourse analysis researchers prefer for close textual annotation.

If you work in academic research and want to explore how Skimle fits into a discourse analysis workflow, the academic researchers page covers the practical setup in more detail.

Frequently asked questions

What is the difference between discourse analysis and thematic analysis?

Discourse analysis asks how language constructs social reality, positions subjects, and reproduces or challenges power relations. Thematic analysis asks what patterns of meaning exist across a dataset. Thematic analysis is concerned with what people say and experience; discourse analysis is concerned with how language works as a social practice. They also rest on different epistemological foundations: discourse analysis is typically constructionist, treating language as constitutive of social reality, while thematic analysis can be applied from either a realist or constructionist position. See our full thematic analysis guide for more on TA's approach.

Can discourse analysis be used with interview data?

Yes, but with an important caveat. When interview transcripts are analysed using CDA or discursive psychology, the analytical focus shifts from what the participant is telling you about their experience to how the participant is using language to construct their account, manage their identity, or accomplish particular interactional goals. This is a genuinely different research act from thematic or phenomenological analysis of the same transcript. If you use discourse analysis on interview data, be explicit in your methodology section that you are treating talk as social action, not as a window onto inner states or lived experience.

Is critical discourse analysis qualitative or quantitative?

CDA is a qualitative method. It focuses on depth of textual analysis rather than frequency counts. That said, some researchers combine CDA with corpus linguistics methods that involve quantitative analysis of large text datasets (counting keyword frequencies, identifying collocations, and so on) before conducting qualitative CDA of key extracts. This mixed approach is sometimes called corpus-assisted discourse analysis. Standalone CDA, however, is qualitative throughout.

How do you assess rigour in discourse analysis?

Rigour in discourse analysis is assessed differently from quantitative methods. Key criteria include: coherence of argument (does the interpretation hold together logically?); depth of textual evidence (are claims grounded in specific, detailed examples from the texts?); theoretical transparency (is the analyst's framework stated and the analytical procedures explained?); reflexivity (has the analyst accounted for their own position?); and plausibility (would a reader familiar with the context recognise the analysis as reasonable, even if they might read the text differently?). Inter-rater reliability is not an appropriate criterion for discourse analysis, because the goal is interpretation rather than replication.

What is the difference between Foucauldian discourse analysis and critical discourse analysis?

Both are concerned with discourse, power, and knowledge, but they differ in focus and method. CDA (in Fairclough's version) conducts close linguistic analysis of specific texts and connects textual features to social structures through a three-level framework. Foucauldian discourse analysis works at a higher level of abstraction, examining the broader rules governing what can be said in a field, the subject positions available, and how truth is constituted, often across archives of texts and longer historical periods. CDA is typically more explicitly political and more concerned with the immediate effects of specific texts; Foucauldian analysis is more concerned with the underlying discursive formations that make certain texts possible.

Ready to manage and annotate your discourse analysis corpus without losing track of your textual evidence? Try Skimle for free and work through close textual analysis with full traceability from every analytical observation back to its source excerpt.

Related reading:

About the authors

Henri Schildt is a Professor of Strategy at Aalto University School of Business and co-founder of Skimle. He has published over a dozen peer-reviewed articles using qualitative methods, including work in Academy of Management Journal, Organisation Science, and Strategic Management Journal. His research focuses on organisational strategy, innovation, and qualitative methodology. Google Scholar profile

Olli Salo is a former Partner at McKinsey & Company where he spent 18 years helping clients understand the markets and themselves, develop winning strategies and improve their operating models. He has done over 1000 client interviews and published over 10 articles on McKinsey.com and beyond. LinkedIn profile

Discourse analysis in qualitative research: approaches, steps, and when to use it

Discourse analysis examines how language constructs meaning and power. This guide covers the main approaches, practical steps, and when to choose it over thematic analysis.

What is discourse analysis?

How is discourse analysis different from thematic analysis and content analysis?