Loading field map
How does language work when it is actually used by real people in real situations? That question has driven a field that is less a single discipline than a series of competing answers, each with its own assumptions about what counts as data, what methods are legitimate, and what the analyst's relationship to the social world should be. Discourse analysis emerged in the mid-twentieth century from a shared dissatisfaction with sentence-level linguistics, but its practitioners quickly split over how to study language beyond the sentence. The result is a landscape of frameworks that have coexisted, borrowed from one another, and sometimes remained in sharp disagreement for decades.
The first systematic attempts to study language use as social action came from two directions that shared an empirical commitment to naturalistic data but drew on very different intellectual traditions.
Ethnography of Communication, developed by Dell Hymes in the 1960s, grew out of anthropological linguistics. Hymes argued that knowing a language meant more than knowing its grammar; it meant knowing how to speak appropriately in a community. He introduced the concept of communicative competence—the ability to use language in socially and culturally appropriate ways—as a direct challenge to Noam Chomsky's notion of linguistic competence, which abstracted away from actual use. Ethnography of Communication treated speech events as the unit of analysis and insisted that the analyst must understand the community's own norms for speaking. Its method was long-term participant observation, and its goal was to describe the full repertoire of ways of speaking in a given community.
Conversation Analysis (CA) emerged at roughly the same time from the work of Harvey Sacks, Emanuel Schegloff, and Gail Jefferson, rooted in Harold Garfinkel's ethnomethodology. Where Ethnography of Communication looked at cultural patterns of speaking, CA focused on the moment-by-moment organization of talk itself. Its central discovery was that ordinary conversation is not chaotic but systematically structured through turn-taking, adjacency pairs, and repair sequences. CA insisted on working with audio or video recordings and detailed transcripts that captured pauses, overlaps, and prosody. The analyst's job was to uncover the methods participants themselves used to produce and understand orderly interaction. CA deliberately avoided invoking external social categories like class or power, arguing that the analyst should stay with what was demonstrably relevant to the participants in the sequence of talk.
These two frameworks shared a commitment to studying language in its natural habitat, but they differed fundamentally in what they took to be the relevant context. For Ethnography of Communication, context was the cultural world the speaker inhabited; for CA, context was built turn by turn within the interaction itself. That tension—between culturally given and locally produced context—would reappear in later frameworks.
While the interactional frameworks focused on spoken language, other researchers turned their attention to written texts and to the ideological forces that shape discourse.
Text Linguistics emerged in the 1970s as an effort to extend linguistic analysis beyond the sentence to the properties of whole texts. Drawing on work by Robert-Alain de Beaugrande, Wolfgang Dressler, and M.A.K. Halliday's systemic functional linguistics, Text Linguistics asked what makes a sequence of sentences cohere as a text. It identified textual features such as cohesion (grammatical links between sentences), coherence (semantic unity), and information structure. Text Linguistics treated texts as autonomous objects that could be analyzed for their internal organization, and it provided tools for describing how texts work that were more systematic than earlier literary or rhetorical approaches. Its limitation, from the perspective of later frameworks, was that it paid little attention to the social conditions under which texts were produced or interpreted.
French Discourse Analysis developed in the same period but from a very different intellectual starting point. Influenced by Michel Foucault's work on discourse as a system of knowledge formation, Louis Althusser's theory of ideology, and Jacques Lacan's psychoanalysis, French Discourse Analysis treated discourse not as a linguistic object but as a site where social identities and power relations are constituted. Its key concept was interdiscourse—the idea that any utterance draws on and responds to prior discourses, so that meaning is never fully present in the text itself but is shaped by the discursive formations that precede it. French Discourse Analysis was less a unified method than a theoretical orientation that insisted on the political nature of discourse. It directly influenced the later development of Critical Discourse Analysis, especially through the work of Michel Pêcheux, whose concept of discursive formation was taken up and transformed by Anglo-American critical linguists.
Text Linguistics and French Discourse Analysis coexisted without much direct engagement. Text Linguistics offered a toolkit for describing textual structure; French Discourse Analysis offered a theory of how discourse produces social reality. The two frameworks answered the question "what is a text?" in incompatible ways: for Text Linguistics, a text was a structured linguistic object; for French Discourse Analysis, a text was a node in a network of ideological forces.
Critical Discourse Analysis (CDA) emerged in the mid-1980s as an explicit synthesis of several earlier threads. Drawing on Halliday's functional linguistics for textual analysis, on French Discourse Analysis for its attention to ideology, and on the Frankfurt School's critical theory for its normative commitments, CDA set out to show how discourse produces, reproduces, and challenges social power relations. Key figures such as Norman Fairclough, Teun van Dijk, and Ruth Wodak argued that discourse analysis could not be neutral: the analyst had a responsibility to expose domination and inequality. CDA introduced concepts such as discursive practice (the processes of text production, distribution, and consumption) and orders of discourse (the configuration of discourses in a social domain). Its methods combined close textual analysis with attention to the broader social context.
CDA's relationship with Conversation Analysis has been one of persistent methodological conflict. CA researchers argue that CDA imports political categories that may not be relevant to the participants themselves, imposing the analyst's ideology on the data. CDA researchers counter that CA's refusal to engage with power and inequality is itself a political choice that tacitly supports the status quo. This disagreement is not merely academic: it reflects fundamentally different views about what counts as evidence and what the purpose of analysis should be. CA aims to describe the mechanisms of interaction; CDA aims to critique and change social arrangements.
The 1990s saw two major expansions that challenged the text-bound and talk-bound focus of earlier work.
Multimodal Discourse Analysis (MDA) argued that meaning is never carried by language alone. Drawing on Halliday's social semiotics and Gunther Kress and Theo van Leeuwen's work on visual design, MDA extended discourse analysis to images, gestures, layout, color, and sound. Its central claim is that all modes of communication are socially shaped resources for making meaning, and that analysts must attend to how modes combine in specific texts and interactions. MDA has been especially influential in the study of media, advertising, and digital communication, where the visual and the verbal are tightly integrated.
Corpus-based Discourse Analysis brought computational methods to the study of large text collections. Using concordancers, frequency lists, and collocation analysis, corpus-based analysts can identify patterns of language use that are invisible to the naked eye. This approach has been particularly productive for Critical Discourse Analysis, where corpus methods can reveal systematic biases in how social groups are represented across large datasets. Corpus-based Discourse Analysis does not replace qualitative analysis but provides a different kind of evidence: it can show that a pattern is widespread rather than anecdotal. Its limitation is that it works best with written texts and struggles with the sequential organization of spoken interaction.
Two later frameworks responded to perceived gaps in the existing landscape.
Mediated Discourse Analysis (MDA, sometimes called nexus analysis) was developed by Ron Scollon and Suzie Wong Scollon to study how discourse is embedded in material practices and technologies. Drawing on Ethnography of Communication's ethnographic sensibility and on activity theory, Mediated Discourse Analysis focuses on the concrete actions people perform with and through discourse, and on the mediational means—technologies, objects, institutional arrangements—that make those actions possible. It has been especially useful for studying digital communication, where the medium itself shapes what can be said and done.
Positive Discourse Analysis (PDA) was proposed by J.R. Martin in the early 2000s as a deliberate counterweight to Critical Discourse Analysis. Martin argued that CDA's relentless focus on domination and inequality had produced a one-sided picture of discourse. PDA set out to analyze discourse that builds community, empowers marginalized groups, and enacts positive social change. Drawing on appraisal theory within systemic functional linguistics, PDA studies how language constructs solidarity, empathy, and hope. It does not reject CDA's political commitments but insists that critique must be balanced by attention to constructive alternatives.
Today, discourse analysis is a pluralistic field. Conversation Analysis remains the dominant framework for studying the sequential organization of talk, especially in institutional settings such as medical consultations and courtrooms. Critical Discourse Analysis continues to be the primary framework for researchers who want to connect language to social power, and it has absorbed insights from Multimodal Discourse Analysis and Corpus-based Discourse Analysis. Multimodal Discourse Analysis has become essential for studying contemporary digital and visual communication. Mediated Discourse Analysis offers a distinctive lens for understanding how discourse is shaped by material and technological infrastructures. Positive Discourse Analysis has carved out a niche in studies of community-building and social movements.
What the leading frameworks agree on is that discourse is a form of social action, not merely a reflection of pre-existing social structures. They agree that analysis must be grounded in empirical data—recordings, texts, corpora—rather than in intuition or introspection. They agree that context matters, though they disagree sharply about what context is and how to delimit it.
What they disagree on is more fundamental. The deepest fault line runs between frameworks that prioritize the internal organization of discourse (Conversation Analysis, Text Linguistics) and those that prioritize its relationship to social power (Critical Discourse Analysis, French Discourse Analysis). A second fault line separates frameworks that treat texts or talk as the primary unit (Conversation Analysis, Text Linguistics, Critical Discourse Analysis) from those that insist on including other semiotic modes (Multimodal Discourse Analysis) or material practices (Mediated Discourse Analysis). A third disagreement concerns the analyst's role: should the analyst aim for neutral description, or is critique an inescapable responsibility?
These disagreements are not signs of a field in crisis. They are the productive tensions that have driven discourse analysis from its earliest days, and they continue to shape its evolution as new forms of communication—social media, algorithmic discourse, multimodal platforms—demand new analytical tools.