What Is Discourse Analysis?
Discourse analysis is a qualitative research method that examines how language, spoken, written, or visual, constructs meaning, shapes social reality, and reflects power relationships within specific contexts. Rather than treating language as a neutral vehicle for transmitting information, discourse analysis starts from the premise that the words people choose, the way they structure arguments, and the topics they include or exclude all carry significance beyond their literal content. The method analyzes texts, conversations, media, institutional documents, and other communicative acts to understand how discourse produces and reproduces social knowledge, identities, and relationships.
Why Discourse Analysis Matters in Research
Discourse analysis reveals what other methods miss: the assumptions embedded in how people talk about a topic, not just what they say about it. Survey data tells you that 70% of customers are satisfied. Discourse analysis of their open-ended comments might reveal that "satisfied" masks resignation, they've lowered expectations rather than genuinely endorsing the experience. In market research, brand research, and policy analysis, understanding how language frames problems and solutions is often more valuable than measuring attitudes on a scale.
How Discourse Analysis Works
Foucauldian Discourse Analysis
Foucauldian discourse analysis (FDA) draws on Michel Foucault's work on power and knowledge. It examines how particular "discourses", broad systems of meaning around a topic, shape what can be said, who can say it, and what counts as truth within a given context.
FDA operates at a macro level. It's less concerned with individual conversations than with the broader patterns of language that circulate through institutions, media, and culture. A Foucauldian analysis of healthcare discourse, for example, might examine how the language of "patient responsibility" shifts accountability from systemic issues to individual behavior.
The analytical process involves identifying discursive constructions (how the object of study is described), locating these constructions within wider discourses, examining what subject positions are made available (who gets to speak with authority), and exploring the practical implications, what becomes possible or impossible within each discursive framework.
FDA is useful when your research question is about power, institutional language, or how framing shapes perception. It's less suited to analyzing conversational mechanics or individual speech patterns.
Conversation Analysis
Conversation analysis (CA) takes the opposite approach, it works at the micro level, examining the sequential organization of talk-in-interaction. Developed by Harvey Sacks, Emanuel Schegloff, and Gail Jefferson in the 1960s and 70s, CA studies how people take turns, repair misunderstandings, open and close interactions, and manage social actions through talk.
CA uses detailed transcription notation that captures pauses (measured in tenths of a second), overlapping speech, emphasis, pitch changes, and inhalation. This level of detail matters because conversational meaning often depends on timing and delivery, not just word choice.
The method is strictly empirical, it avoids speculating about participants' internal states or motivations and focuses on what's demonstrably happening in the interaction. If a participant treats an utterance as a question (by answering it), then it functioned as a question in that interaction, regardless of its grammatical form.
CA is valuable for studying customer service interactions, interview dynamics, medical consultations, sales conversations, and any context where the structure of interaction itself is the research question.
Critical Discourse Analysis
Critical discourse analysis (CDA), most associated with Norman Fairclough and Teun van Dijk, sits between FDA and CA. It examines specific texts in detail (like CA) while connecting those texts to broader social and political structures (like FDA).
Fairclough's three-dimensional model analyzes discourse at three levels:
- Text: the linguistic features of the text itself (vocabulary, grammar, cohesion, text structure)
- Discursive practice: how the text was produced, distributed, and consumed (who wrote it, for whom, through what channels)
- Social practice: how the text relates to broader social, political, and institutional structures
CDA is explicitly political, it aims to reveal how discourse reproduces or challenges inequality. In commercial research, CDA techniques are useful for competitor messaging analysis, brand positioning research, and understanding how media coverage shapes public perception of an industry or product category.
The Analytical Process
Regardless of the specific approach, discourse analysis typically follows these steps:
- Assemble a corpus: select texts that are relevant to your research question. This might be interview transcripts, social media posts, policy documents, advertising copy, or media coverage.
- Read and re-read: familiarize yourself with the data without imposing categories. Note patterns in language use, recurring metaphors, notable absences, and shifts in tone or register.
- Code discursive features: mark up the texts for features relevant to your approach (discursive constructions in FDA, turn-taking patterns in CA, linguistic features in CDA).
- Analyze patterns: identify how these features work together to construct meaning, position speakers, and frame the topic.
- Interpret and contextualize: connect your findings to the broader research question, theoretical framework, and practical implications.
When to Use Discourse Analysis
- You're researching how a brand, product, or organization is talked about across media, reviews, or public forums
- You need to understand the framing effects of specific language choices in marketing, policy, or organizational communication
- You're analyzing interview data where what's not said or how things are said matters as much as the content
- You're studying institutional communication: how organizations construct their identity, authority, or legitimacy through language
- You want to compare how different groups talk about the same topic and identify the assumptions embedded in each perspective
Common Mistakes to Avoid
- Selecting an approach without understanding its assumptions: FDA, CA, and CDA have fundamentally different ontological commitments. Mixing them without acknowledging this produces incoherent analysis.
- Treating discourse analysis as a content summary: listing what people said isn't discourse analysis. The method requires analyzing how language works, not just what it communicates.
- Applying CA transcription standards to data that doesn't warrant it: detailed CA notation is unnecessary if your research question is about discursive constructions rather than conversational mechanics.
- Making claims about speakers' intentions without evidence from the discourse itself. Discourse analysis works with what's in the text, not with speculation about what the speaker meant.
- Analyzing too small a corpus for the claims you want to make. A Foucauldian analysis of "healthcare discourse" based on three interviews isn't credible. The scope of your claims should match the scope of your data.
How Quali-Fi Supports Discourse Analysis
Quali-Fi's Research plan ($1,061/month) provides AI-powered transcription for interview and focus group recordings, preserving the conversational detail that discourse analysis requires. The platform's open-end coding tools let researchers tag discursive features and constructions across multiple data sources, while discussion boards capture naturally occurring written discourse from participants over time, useful data for both FDA and CDA approaches.
Frequently Asked Questions
What's the difference between discourse analysis and thematic analysis?
Thematic analysis identifies patterns of meaning (themes) in data content. Discourse analysis examines how language constructs that meaning, it's concerned with the mechanics and effects of language use, not just its topics. You could run thematic analysis and discourse analysis on the same dataset and get different but complementary findings.
Can discourse analysis use quantitative data?
Some forms incorporate quantitative elements. Corpus linguistics, for example, uses statistical analysis of word frequency and co-occurrence patterns across large text datasets. But the core of discourse analysis is interpretive, numbers support the analysis rather than drive it.
How many texts do you need for discourse analysis?
It depends on the approach and scope of your claims. Conversation analysis can analyze a single interaction in detail. Foucauldian analysis of a broad discourse requires a larger corpus, dozens to hundreds of texts. The key is that your corpus should represent the discursive field you're studying.
Related Topics
- Thematic Analysis
- Content Analysis
- Qualitative vs. Quantitative Research
- Ethnography
- Case Study Research
- Grounded Theory
Ready to go deeper with qualitative analysis? Explore Quali-Fi's Research platform and use AI-powered transcription, coding, and discussion boards to support discourse-level research.