Discourse Analysis: Approaches, Steps & Examples

What is discourse analysis?

Discourse analysis treats language as a form of social action. The core premise, shared across its many schools, is that language does not merely describe the world — it actively constructs versions of it. When a newspaper calls a group of people “migrants” rather than “refugees,” or a government department writes that “difficult decisions had to be taken” rather than “we cut the budget,” the linguistic choice is doing interpretive and political work. Discourse analysts study those choices systematically and ask what they accomplish.

The word discourse carries two linked meanings in this field. In the linguistic sense it means language above the level of the sentence — connected text and talk, the way clauses cohere into arguments, narratives and exchanges. In the social-theoretical sense (associated with Michel Foucault) a “discourse” is a whole system of statements, categories and assumptions that makes it possible to talk about a topic in a particular way — “the discourse of mental health,” “neoliberal discourse,” “the discourse of employability.” Most approaches to discourse analysis work somewhere between these two senses, moving from the fine grain of word choice up to the broad frameworks of meaning a culture takes for granted.

Because it is fundamentally interpretive, discourse analysis sits within the qualitative tradition. If you are still deciding between a counting, pattern-finding or interpretive design, our guide to quantitative vs qualitative research sets out where this method belongs.

When to use discourse analysis

Discourse analysis is the right tool when your research question is genuinely about language, meaning and power — not simply about the topics people mention. Choose it when you want to:

• Examine how an issue, group or identity is constructed and represented (for example, how broadsheet headlines frame asylum seekers).
• Trace how power, ideology or expertise is reproduced or challenged through ordinary language.
• Analyse how speakers do things with talk — blame, justify, deny, build solidarity — in interviews, focus groups or recorded interaction.
• Study how institutional texts (policies, prospectuses, clinical letters) position their readers and naturalise particular assumptions.
• Investigate the rhetorical and persuasive structure of a campaign, manifesto or advertisement.

It is not the right tool if you want to quantify how frequently themes occur, generalise to a population, or test a hypothesis. For those aims, content analysis or a quantitative design is more appropriate, and we compare the two methods directly below.

The main approaches to discourse analysis

“Discourse analysis” is an umbrella term covering several distinct traditions, each with its own theory, focus and procedures. Choosing the right one is the single most important methodological decision you will make, because it determines what you look at and what counts as a finding. The four most widely used in dissertations are critical discourse analysis, Foucauldian discourse analysis, conversation analysis and discursive psychology.

1. Critical Discourse Analysis (CDA)

Critical discourse analysis examines how language reproduces or resists social power, dominance and inequality. It is “critical” in the sense that it is explicitly concerned with ideology and is usually on the side of dominated groups. Two figures dominate the field. Norman Fairclough offers the most widely taught analytical framework: his three-dimensional model (shown above) treats every communicative event as three things at once — a text, an instance of discursive practice, and an instance of social practice. Analysis correspondingly has three layers:

• Text (description): close linguistic analysis — vocabulary, grammar, transitivity, modality, cohesion. What words and structures are actually used?
• Discursive practice (interpretation): how the text was produced, circulated and consumed, and which other discourses it draws on (intertextuality). Who made it, for whom, and how is it likely to be read?
• Social practice (explanation): how the text relates to the wider social and ideological order — the relations of power it sustains or contests.

Teun van Dijk brings a complementary, socio-cognitive emphasis: he argues that the link between discourse and society runs through shared social cognition — the mental models, attitudes and ideologies that language activates. His work on the discursive reproduction of racism (the “ideological square” of positive self-presentation and negative other-presentation) is a standard reference for anyone analysing prejudice in text.

2. Foucauldian Discourse Analysis (FDA)

Foucauldian discourse analysis takes the broad, social-theoretical sense of “discourse.” Following Michel Foucault, it treats discourses as historically specific systems of knowledge that produce the very objects and subjects they appear merely to describe — “madness,” “sexuality,” “the delinquent,” “the employable graduate.” FDA is less interested in the fine grain of grammar and more interested in which discourses are available in a culture, what subject positions they make available (who you are allowed to be), and how power and knowledge are bound together. It suits questions about how a category came to be taken for granted and what it permits or forbids.

3. Conversation Analysis (CA)

Conversation analysis, developed from the sociology of Harvey Sacks, Emanuel Schegloff and Gail Jefferson, studies the structural organisation of naturally occurring talk. It works from detailed transcripts and examines turn-taking, adjacency pairs (question–answer, invitation–acceptance), repair, and sequence. CA is rigorously empirical and stays close to what participants demonstrably orient to in the talk itself; it resists importing the analyst’s assumptions about power or society unless the participants themselves make them relevant. It is ideal for institutional interaction — doctor–patient consultations, helpline calls, classroom exchanges.

4. Discursive Psychology (DP)

Discursive psychology, associated with Jonathan Potter and Margaret Wetherell, re-specifies psychological concepts — memory, attitude, emotion, identity — as things people do in talk rather than inner states the talk reports. The analyst asks what an account is designed to accomplish in its immediate setting: how a speaker manages blame, builds a fact as objective, or constructs themselves as reasonable. It combines CA’s attention to interactional detail with a concern for the actions accounts perform.

“Critical discourse analysis (CDA) is a type of discourse analytical research that primarily studies the way social power abuse, dominance, and inequality are enacted, reproduced, and resisted by text and talk in the social and political context.” (Source: van Dijk, 2001)

The steps in a discourse analysis

Although the schools differ, a well-run discourse analysis follows a recognisable sequence. The following six steps work for a critical or interpretive project; conversation analysts would add a formal transcription convention (the Jefferson system) at the data stage.

1. Frame the research question and choose an approach. Decide whether you are asking about power and representation (CDA), the conditions of a discourse (FDA), interactional structure (CA) or the actions of accounts (DP). The approach dictates everything that follows.
2. Gather and bound your material. Select a focused, justified corpus — a defined set of headlines, a policy document, a set of interview transcripts. Keep it small enough to analyse in genuine depth; discourse analysis trades breadth for interpretive richness.
3. Establish the social and textual context. Note who produced the text, for whom, when, through which channel, and what debate it enters. Meaning is context-dependent, so this groundwork is part of the analysis, not a preamble.
4. Code at the linguistic level. Work line by line, annotating the features your approach makes relevant — lexical choice, modality, agency and passivisation, pronouns, metaphor, presupposition, and how turns are organised.
5. Interpret patterns into discourses. Move from individual features to the broader meanings, framings and subject positions they construct. Ask what each pattern does and whose interests it serves.
6. Build and evidence the argument. Write up your reading as a sustained, transparent argument anchored in specific extracts, acknowledging the analyst’s role and considering alternative readings.

Because your corpus is itself a kind of data, it is worth planning collection carefully; our guide to methods of data collection covers how to assemble and document textual and interview material defensibly.

Key concepts and the linguistic toolkit

Discourse analysts share a vocabulary for the textual features that carry meaning. These are the tools you apply at the coding stage.

• Framing: how a text sets up an issue so that some interpretations seem natural and others unthinkable — calling a fee a “contribution,” or a cut an “efficiency saving.”
• Lexical choice: the selection of one word over near-synonyms (“migrant”/“refugee,” “regime”/“government,” “collateral damage”/“civilian deaths”), each carrying different evaluative loading.
• Modality: the degree of certainty, obligation or commitment a speaker signals — the difference between “this may help,” “this will help” and “this must happen.” High modality projects authority; low modality hedges.
• Agency and passivisation: who is shown as acting and who is acted upon. The passive voice and nominalisation can delete the agent altogether — “mistakes were made” hides who made them — which is one of the most consequential moves a text can make.
• Presupposition: assumptions smuggled in as given — “when will the council finally fix this?” presupposes a failure not yet established.
• Pronouns and positioning: the in-group “we” and out-group “they” that construct belonging and distance.

A full worked example: analysing a single headline

The best way to see discourse analysis in action is to take one short text and work outward from word to interpretation. Below, a single newspaper-style headline is analysed end to end using Fairclough’s three layers.

Discourse analysis vs content analysis

Students often confuse these two text-based methods, but they answer different questions. Content analysis is typically systematic and (in its quantitative form) counts the frequency of pre-defined categories so results can be summarised and compared. Discourse analysis is interpretive: it asks how meaning is constructed and what language does, and it treats context and the analyst’s argued reading as central rather than as bias to be eliminated. The same headline corpus could be content-analysed (“how often does ‘flood’ appear across 200 headlines?”) and discourse-analysed (“what does the flood metaphor do?”).

If you want a counting-based design instead, see our guides to content analysis and thematic analysis; for a broader treatment of analysing texts as data, the quick guide to textual analysis places discourse analysis alongside its neighbours.

Strengths and limitations

Discourse analysis offers interpretive depth that quantitative methods cannot, but it asks for transparency and reflexivity in return.

• Reveals the hidden work of language — power, ideology and assumption that a frequency count would miss entirely.
• Takes context seriously, producing rich, situated understanding of how meaning is made.
• Flexible across data types — written texts, talk, images and multimodal material.

Its limitations are equally real and should be acknowledged openly:

• Findings are interpretive and not generalisable to a population in the statistical sense.
• The analyst’s own position shapes the reading, so reflexivity is essential rather than optional.
• It is labour-intensive, which forces a small corpus and limits breadth.
• Quality criteria differ from positivist reliability and validity — you demonstrate rigour through transparency, coherence and grounding in the data, not through inter-coder agreement.

Common mistakes to avoid

• Doing a summary, not an analysis — paraphrasing what a text says instead of examining how it says it.
• Reading politics into the data without anchoring every claim in a specific linguistic feature of a specific extract.
• Mixing approaches incoherently — borrowing CA’s transcript symbols but ignoring its commitment to participant orientation, for example.
• An over-large corpus that forces shallow treatment and defeats the method’s purpose.
• Ignoring reflexivity and presenting an interpretation as the single objective truth of the text.

How to do discourse analysis well

Strong discourse analysis is disciplined as well as creative. Commit clearly to one approach and apply it consistently. Keep your corpus small and your reading deep. Quote enough of each extract that the reader can see the feature you are analysing and judge your interpretation. Make the analytical move explicit every time — name the linguistic feature, then state what it does, then connect it to the wider discourse. Be reflexive about your own standpoint, and consider rival readings rather than asserting the only one. Done this way, discourse analysis produces some of the most insightful, original findings available to a qualitative researcher.

Related methodology guides

• Narrative Analysis
• Qualitative Data Analysis

Frequently Asked Questions