How to Do a Discourse Analysis

How to Do a Discourse Analysis

A toolbox for analysing political texts

Discourse analysis is a useful tool for studying the political meanings that inform written and spoken text. In other posts, I have provided a quick video introduction to the topic, and have discussed the ideas behind discourse theory, the main questions that students and researchers will likely ask as they set up their discourse analysis project, and the things that are worth keeping in mind when working with East Asian language sources. In this post, I offer a handy set of tools for doing a text-based, qualitative discourse analysis. The idea of a discourse toolbox comes from Siegfried Jäger, but I have expanded his approach based on my own experience and the works of other discourse analysts such as Paul Chilton (2004) and Norman Fairclough (1994).

You can go through the whole list of work-steps and tick each item off in turn, which is a good way to practice these methods. However, if you are conducting a specific research project, I would recommend adapting this toolbox to your own needs and tailoring it to fit your concerns. At the end of this post, you will also find a few comments on the limitations of this toolbox plus a list of literature that you can turn to if you want to learn more.

Getting technical: discourse analysis in ten steps

So you have formulated a research question, have collected source material, and are now ready to roll up your sleeves and dig into your sources. But how do you make sure that you have covered all your bases and that you will later be able to make a good case for yourself and your work? Here are ten work steps that will help you conduct a systematic and professional discourse analysis.

1) Establish the context

Before you start chiselling away at your source material, jot down where the material comes from and how it fits into the big picture. You should ask yourself what the social and historical context is in which each of your sources was produced. Write down what language your source is written in, what country and place it is from, who wrote it (and when), and who published it (and when). Also try to have a record of when and how you got your hands on your sources, and to explain where others might find copies. Finally, find out whether your sources are responses to any major event, whether they tie into broader debates, and how they were received at the time of publication.

2) Explore the production process

You have already recorded who wrote and published your sources, but you still need to do a more thorough background check. Try to find additional information on the producer of your source material, as well as their institutional and personal background. For example, if you are analysing news articles, take a look at the kind of newspaper that the articles are from (Jäger 2004: 175): Who are the author and the editorial staff, what is the general political position of the paper, and what is its affiliation with other organizations? Are any of the people who are involved in the production process known for their journalistic style or their political views? Is there any information on the production expenditures and general finances of the paper? Do you know who the general target audience of the paper is? In many cases, media outlets themselves provide some of this information online, for instance in the “about” sections of their websites. In other cases, you will find such information in the secondary academic literature. Don’t hesitate to write the editors an email or call them up: personal interviews can be a great way to explore production backgrounds.

Once you have established the institutional background, take notes on the medium and the genre you are working with. Some scholars go as far to argue that “the medium is the message” (McLuhan 1964/2001), or in other words that the medium in which information is presented is the crucial element that shapes meaning. While I am skeptical of such extreme technological determinism, I do agree that the medium matters: reading an article online is not the same as reading it in a printed newspaper, or in a hardcover collection of essays. Make sure to identify the different media types in which your source appeared, and to also be clear about the version that you yourself are analysing.

For instance, the layout of a newspaper article and its position on the page will be different in a print edition than in an online edition. The latter will also offer comments, links, multi-media content, etc. All of these factors frame the meaning of the actual text and should be considered in an analysis. This may also mean that you should think about the technical quality and readability of your source, for instance by looking at paper quality (or resolution for online sources), type set, etc. You should also take notes on the length of your source (number of pages and/or words) and any additional features of the medium that might contribute to or shape meaning (such as images).

Finally, ask yourself what genre your source belongs to. Are you analysing an editorial comment, and op-ed, a reader’s letter, a commentary, a news item, a report, an interview, or something else? Establishing this background information will later help you assess what genre-specific mechanism your source deploys (or ignores) to get its message across.

3) Prepare your material for analysis

In order to analyse the actual text, it is wise to prepare it in a way that will allow you to work with the source, home in on specific details, and make precise references later. If you are working with a hard copy I would recommend making a number of additional copies of your source material, so that you can write on these versions and mark important features. If you haven’t already, try to digitize your source or get a digital copy. Then add references that others can use to follow your work later: add numbers for lines, headers, paragraphs, figures, or any other features that will help you keep your bearings.

4) Code your material

When you code data, it means that you are assigning attributes to specific units of analysis, such as paragraphs, sentences, or individual words. Think of how many of us tag online information like pictures, links, or articles. Coding is simply an academic version of this tagging process.

For instance, you might be analysing a presidential speech to see what globalization discourse it draws from. It makes sense to mark all statements in the speech that deal with globalization and its related themes (or discourse strands). Before you start with this process, you need to come up with your coding categories. The first step is to outline a few such categories theoretically: based on the kind of question you are asking, and your knowledge of the subject matter, you will already have a few key themes in mind that you expect to find, for instance “trade”, “migration”, “transportation”, “communication”, and so on. A thorough review of the secondary literature on your topic will likely offer inspiration. Write down your first considerations, and also write down topics that you think might be related to these key themes. These are your starting categories.

You then go over the text to see if it contains any of these themes. Take notes on the ones that are not included, since you may have to delete these categories later. Other categories might be too broad, so try breaking them down into sub-categories. Also, the text may include interesting themes that you did not expect to find, so jot down any such additional discourse strands. At the end of this first review, revise your list of coding categories to reflect your findings. If you are working with several documents, repeat the process for each of them, until you have your final list of coding categories. This is what Mayring (2002: 120) calls evolutionary coding, since your categories evolve from theoretical considerations into a full-fledged operational list based on empirical data.

How the actual coding process works will depend on the tools you use. You can code paper-based sources by highlighting text sections in different colours, or by jotting down specific symbols. If you are working with a computer, you can similarly highlight text sections in a word processor. In either case, the risk is that you will not be able to represent multiple categories adequately, for instance when a statement ties into three or four discourse strands at once. You could mark individual words, but this might not be ideal if you want to see how the discourse works within the larger sentence structure, and how discourse strands overlap.

A real alternative is using other types of software. If you have access to professional research programmes like NVivo, then the software already has built-in coding mechanisms that you can customize and use. There is also open-source software available, for instance the Mac programme TAMS, but I have not tested their functionality. However, even if you only have regular office tools at your disposal, such as Microsoft’s Office or a Mac equivalent, there are at least two ways in which you can code material.

The first is to copy your text into an Excel table. Place the text in one column and use the next column to add the coding categories. You’ll of course have to decide where the line-breaks should be. A sensible approach is to place each sentence of your original text on a new line, but you could also choose smaller units of text.

Another tool that provides coding assistance is Microsoft OneNote 2010, or the Mac equivalent Growly Notes. In OneNote, you can right click anywhere in the text and select “tag” to assign a category to any sentence. You can also customize your tags, create new ones, and easily search and monitor your coding categories and activities. The downside is that you can only tag full sentences, not single words or phrases, but depending on your intentions, this may not be a crucial drawback.

5) Examine the structure of the text

Now that you have prepared your materials and have coded the discourse strands, it is time to look at the structural features of the texts. Are there sections that overwhelmingly deal with one discourse? Are there ways in which different discourse strands overlap in the text? See if you can identify how the argument is structured: does the text go through several issues one by one? Does it first make a counter-factual case, only to then refute that case and make the main argument? You should at this point also consider how the headers and other layout features guide the argument, and what role the introduction and conclusion play in the overall scheme of things.

6) Collect and examine discursive statements

Once you have a good idea of the macro-features of your text, you can zoom in on the individual statements, or discourse fragments. A good way to do this is to collect all statements with a specific code, and to examine what they have to say on the respective discourse strand. This collection of statements will allow you to map out what “truths” the text establishes on each major topic.

7) Identify cultural references

You have already established what the context of your source material is. Now think about how the context informs the argument. Does your material contain references to other sources, or imply knowledge of another subject matter? What meaning does the text attribute to such other sources? Exploring these questions will help you figure out what function intertextuality serves in light of the overall argument.

8) Identify linguistic and rhetorical mechanisms

The next step in your analysis is likely going to be the most laborious, but also the most enlightening when it comes to exploring how a discourse works in detail. You will need to identify how the various statements function at the level of language. In order to do this, you may have to use additional copies of your text for each work-step, or you may need to create separate coding categories for your digital files. Here are some of the things you should be on the lookout for:

  • Word groups: does the text deploy words that have a common contextual background? For instance, the vocabulary may be drawn directly from military language, or business language, or highly colloquial youth language. Take a closer look at nouns, verbs, and adjectives in your text and see if you find any common features. Such regularities can shed light on the sort of logic that the text implies. For example, talking about a natural disaster in the language of war creates a very different reasoning than talking about the same event in religious terms.
  • Grammar features: check who or what the subjects and objects in the various statements are. Are there any regularities, for instance frequently used pronouns like “we” and “they”? If so, can you identify who the protagonists and antagonists are? A look at adjectives and adverbs might tell you more about judgements that the text passes on these groups. Also, take a closer look at the main and auxiliary verbs that the text uses, and check what tense they appear in. Particularly interesting are active versus passive phrases – does the text delete actors from its arguments by using passive phrases? A statement like “we are under economic pressure” is very different from “X puts us under economic pressure”… particularly if “X” is self-inflicted. Passive phrases and impersonal chains of nouns are a common way to obscure relationships behind the text and shirk responsibility. Make such strategies visible through your analysis.
  • Rhetorical and literary figures: see if you can identify and mark any of the following five elements in your text: allegories, metaphors, similes, idioms, and proverbs. Take a look at how they are deployed in the service of the overall argument. Inviting the reader to entertain certain associations, for instance in the form of an allegory, helps construct certain kinds of categories and relations, which in turn shape the argument. For instance, if I use a simile that equates the state with a parent, and the citizens with children, then I am not only significantly simplifying what is actually a very complex relationship, I am also conjuring up categories and relationships that legitimize certain kinds of politics, for instance strict government intervention in the social sphere. Once you have checked for the five elements listed above, follow up by examining additional rhetorical figures to see how these frame the meaning of specific statements. Things to look for include parallelisms, hyperboles, tri-colons, synecdoches, rhetorical questions, and anaphora, to name only the most common.
  • Direct and indirect speech: does the text include quotes? If so, are they paraphrased or are they cited as direct speech? In either case, you should track down the original phrases to see what their context was, and what function they now play in your source material.
  • Modalities: see if the text includes any statements on what “should” or “could” be. Such phrases may create a sense of urgency, serve as a call to action, or imply hypothetical scenarios.
  • Evidentialities: lastly, are there any phrases in the text that suggest factuality? Sample phrases might include “of course”, “obviously”, or “as everyone knows”. A related question then is what kinds of “facts” the text actually presents in support of its argument. Does the text report factuality, actively demonstrate it, or merely suggested it as self-evident? One of the strongest features of discourse is how it “naturalizes” certain statements as “common sense” or “fact”, even if the statements are actually controversial (and in discourse theory, all statements are controversial). Be on the look-out for such discursive moves.

9) Interpret the data

You now have all the elements of your analysis together, but the most important question still remains: what does it all mean? In your interpretation, you need to tie all of your results together in order to explain that the discourse is about, and how it works. This means combing your knowledge of structural features and individual statements, and then placing those findings into the broader context that you established at the beginning. Throughout this process, keep the following questions in mind: who created the material you are analysing? What is their position on the topic you examined? How do their arguments draw from and in turn contribute to commonly accepted knowledge of the topic at the time and in the place that this argument was made? And maybe most importantly: who might benefit from the discourse that your sources construct?

10) Present your findings

Once you have the answer to your original question, it is time to get your results across to your target audience. If you have conducted a good analysis, then you now have a huge amount of notes from which you can build your presentation, paper, or thesis. Make sure to stress the relevance, and to move through your analysis based on the issues that you want to present. Always ask yourself: what is interesting about my findings, and why should anyone care? A talk or a paper that simply lists one discourse feature after another is tedious to follow, so try to focus on making a compelling case. You can then add evidence from your work as needed, for instance by adding original and translated examples to illustrate your point. For some academic papers, particularly graduation theses, you may want to compile the full account of your data analysis in an appendix or some other separate file so that your assessors can check your work.

Mind the limitations:


Discourse analysis offers a powerful toolbox for analysing political communication, but it also has its pitfalls. Aside from being very work-intensive, the idea that you only need to follow a certain number of steps to get your results can be misleading. A methodology is always only as good as your question. If your question does not lend itself to this sort of analysis, or if many of the steps I list above do not apply to you, then come up with an approach that suits your project. Don’t be a methodologist: someone who jumps at a set of methods and applies them to everything in a blind fit of activism. Always remain critical of your own work.

This means being mindful of the shortcomings in your approach, so that you do not end up making claims that your material does not support. A common mistake is to claim that a discourse analysis shows what people think or believe (or worse: what entire societies think or believe). Discourse analysis is a form of content analysis. It is not a tool to analyse the impact of media on audience members. No amount of discourse analysis can provide adequate evidence on what goes on in people’s heads.

What we can learn from a discourse analysis is how specific actors construct an argument, and how this argument fits into wider social practices. More importantly, we can demonstrate with confidence what kind of statements actors try to establish as self-evident and true. We can show with precision what rhetorical methods they picked to communicate those truths in ways they thought would be effective, plausible, or even natural. And we can reveal how their statements and the frameworks of meaning they draw from proliferate through communication practices.


Chilton, Paul (2004). Analyzing Political Discourse – Theory and Practice. London: Arnold.

Fairclough, Norman (1995). Critical Discourse Analysis: The Critical Study of Language. Harlow: Pearson Education Limited.

Jäger, Siegfried (2004). Kritische Diskursanalyse. Eine Einführung. (Discourse Analysis. An Introduction). 4th ed., Münster: UNRAST-Verlag.

Mayring, Philipp (2002). Einführung in die Qualitative Sozialforschung – Eine Anleitung zu qualitativem Denken (Introduction to Qualitative Social Science Research – Instruction Manual to Qualitative Thinking). 5th ed., Basel: Beltz Verlag.

McLuhan, Marshall (1964/2001). Understanding Media. New York: Routledge Classics.

Related Posts


  1. Kate

    Thank you!! :)

    • Florian Schneider

      No worries at all. Glad if it helped.

  2. Lauren

    This is gold! I didn’t know how on Earth to start my discourse analysis assignment until I came across this. It has been a life saver. Wish my tutor had taken the time to break it down like you have. Thank you for sharing your knowledge.

    • Florian Schneider

      Thanks Lauren, it’s good to know that this was useful in your studies. Good luck with the assignment!

  3. Patricia E. O'Connor

    Wonderful, cogent, concise description of methodology. My students are thanking you!!

    • Florian Schneider

      Thanks for the kind words, Patricia. I really appreciate it.

  4. whoa!!! thank yuh
    i av learnt alot not jus the discourse analysis.
    ur such a life saver

    • Florian Schneider

      That’s very kind of you Khadijah. Glad I could help.

  5. Elena

    This is the most clear and helpful post about discourse analysis I’ve ever read!
    Thanks a lot for sharing.

  6. Katie

    Yes, excellent. Many thanks.
    Reading all the Fairclough and Foucault in the world doesn’t resolve practical issues like how to cite the analyzed content. Does it belong in works cited? Footnotes? Appendix? I can’t seem to find a natural fit for my research.

  7. Florian Schneider

    This is a good question, Katie. The answer depends on how detailed your analysis of the materials is. I have seen undergraduate studies that cite longer sections in the main body and then list the source in the bibliography like any other materials. In some cases, particularly if the project analyses several texts, it may be good to have two sections in the list of references: one for “primary sources”, one for “secondary sources”. Personally, I like to see the materials that were analysed in an appendix (and then listed in the bibliography alongside the secondary sources). For graduate or post-graduate work, it might even be worthwhile expanding such an appendix to include practical work steps, visualisation of the data, or different rounds of coding on the same document. That way the reader (or examiner) can check the thought process behind the research. Just an idea. Hope this helps!

  8. Putri Nafisah

    How to do a discourse analysis by finding the cohesions and the coherences in the article?

    • Florian Schneider

      Dear Putri, I am not sure I have understood your question correctly – do you mean: how can we study the structure of a text? If that is the case, I would try to identify what each paragraph or section does (e.g. does it functions as an introduction, an argument, a counter-argument, an example, a conclusion?) and would try to establish how the author transitions from one section to the next. You could also go into more detail and check what conjunctions or rhetorical tools the text deploys to provide a sense of flow (for instance, if I write: “Discourse influence language. So it also influences politics”, I have linked two separate claims in a way that is by no means self-evident). Some of the literature that I’ve provided in the list of references includes more sophisticated examples than I can provide here, but I hope these brief notes already help.

  9. Ilze

    Many thanks from Latvia!
    I have struggled with discourse analysis for about month now and no one could actually tell me what’s it about and how exactly to do the analysis. This really was a lifesaver for my bachelors degree research. Thank You a lot!

    • Florian Schneider

      Thanks Ilze, I’m glad you found this useful. Good luck with the BA!

  10. Tina Bass

    Well done, Florian. This is very professional and very helpful. I am teaching discourse analysis to undergraduate business students and now I don’t have to create my own video.

    • Florian Schneider

      Hi Tina, thanks for the encouraging words. Hope your students enjoy the introduction and video.

  11. Robyn

    Thank you so much. This is so clear and useful for unpicking political text to illuminate power structures and motivations.

  12. Heidi

    Hi, this article is so useful!
    I am currently conducting discourse analysis of a television travel documentary and was wondering how the stages can be adapted to fit this? Obviously I cannot transcribe and code the hour long programme, so do I therefore transcribe sections which I feel to be most significant and code these?

    • Florian Schneider

      Hi Heidi,
      This is a good question, and depends a bit on the length of the material and what you are trying to achieve. For cases where you are not interested in a shot-by-shot analysis, I would recommend creating a sequence protocol, and then coding those sequences. I recently did this with a colleague of mine to analyse a lengthy Chinese documentary, and it’s a good way to keep track of the content at a macro-level. You can then “zoom in” on specific sequences and examine them in more detail where it’s useful and necessary, for instance shot-by-shot, or transcribing what was said.
      I’ve written a bit more about this here:
      The section on “working with moving images” in particular might be of interest to you.
      Best – FS

  13. leonard makombe

    Thanks a lot for the great work. I am doing a research on how Twitter was used in Zimbabwe during the 2013 elections. I have collected more than 80 000 tweets over 51 days. My question now is: With such a huge dataset, is it possible to do a proper CDA?

    • Florian Schneider

      Hi Leonard,

      This sounds like a fascinating data set, and I do think it is possible to do a discourse analysis on large amounts of text. However, I would only rely on quantitative tools to highlight keyword distribution and check the general thematic structure of the text corpus. I’d always then follow up by looking at representative (or outlying!) examples in more detail for the qualitative part of the analysis.

      By the way, it might also be interesting to see how the people who post these tweets are connected on Twitter, and what kind of networks consequently provide the foundation for the discourse you’re looking at. I can recommend the work by Richard Rogers over at the University of Amsterdam on how to get a handle on such digital methods questions. Oh, and then there’s the very tricky question of reproducing your results without singling out the various posters – if you haven’t read it yet, I can recommend Zimmer’s article on this issue ( Just FYI. :)

      Good luck with this fascinating project! Do let me know what you find.

      Best – F

  14. Nassy

    Hi. i found this article very helpful. Thanks a lot! I’ve got an assignment on media political discourse. Please, i want to know if making use of critical discourse analysis will be an excellent way of analysing a newspaper article.

    • Florian Schneider

      Hi Nassy,
      This all depends on the kind of questions you have regarding the newspaper article. If you are trying to find out what it’s position on a specific issue is, and how the author uses language to establish that position, then a discourse analysis might be worth a try. If you are only looking at a single article, though, I’d be careful not to overstate how that piece contributes to broader discourses. That would require either a wider study, or more information on how relevant this particular article is. As with any other subject area, the success of a paper very much hinges on the research question. The selection of methods (for instance: discourse analysis) should follow from that.
      All the best

  15. Mau

    Hi, thanks a lot for this article :) I am actually doing my dissertation on the role of media in environmental protection- using renewable energies…i’m using discourse analysis to analyse newspaper..but i’m a little bit confused..should I separate the analysis of my themes from the grammar part or I mixed both? :s
    Thnks in advance

    • Florian Schneider

      Hi Mau, thanks for the question. I think it depends on the kind of dissertation you are writing, and the level of linguistic detail you plan to go into. If you are working on a research MA or PhD, and have a lot of data, then it might indeed be a good idea to write a chapter that collects and discusses recurring grammatical features in the texts, and to then follow this up with a chapter that discusses what discursive positions are constructed through the language (with examples, of course). To be honest, I myself like it when a thesis tells a story, so I would be tempted to combine these two things: you could structure your thesis according to the different themes you are analysing, and then use the grammar parts as evidence and illustration. In a case like that, you could also provide the more technical details and any primary sources in an appendix, so you can readily reference your analytic work without having to reproduce every minute bit in the main text. So as you see, it’s a matter of preference. I would check with your supervisor to see what makes most sense for your case, and whether your examiners have a preference in this regard. You are, after all, writing for a specific audience… Hope this helps! Best- F

  16. Josh

    Hi Florian,

    Your material is very clear and helpful. Are these methods okay to use for interviews that have been written up at masters level.
    Many Thanks Josh.

    • Florian Schneider

      Hi Josh,
      Discourse analysis is definitely a great way to process interviews – provided you are looking for the (often subconscious) communication choices your interviewees make to get their point across, and if you want to know what kind background knowledge and assumptions informs their views. It’s often quite revealing to see how interviewees tie their arguments together with wider social discourses and the argumentation patterns you’d find there (e.g. the news, academia, work conversations, etc.).
      What I normally do is create a protocol of the interviews (using either my own paraphrasing or rough transcripts), and after coding the meaningful segments I look at specific parts in detail. This can then also include transcribing those parts in a way that marks hesitations, intonations, and other such qualities of the spoken word (Paul Chilton has provided some useful annotation advice for this).
      As you might imagine, this can be a lot of work. So if you are mainly on a “fact finding” mission and are trying to figure out how the topic “works” that your interviewees discuss, then I probably wouldn’t recommend a full discourse analysis based on transcripts: simple protocols might be the better way forward.
      I hope this helps you decide how to approach those materials – good luck with the MA!
      Best – Florian

      • Josh

        Hi Florian,

        Thanks for the replay and advice, this sounds really good. I think I’m going to use some other material as well such as a short film. So mix the discourse analysis with the visual analysis that you have also clearly presented. Could I call this a multimodal discourse analysis? I think the wider context filters through in the interviews and the short films quite well, I also think this is physically impacting on society and possibly playing into Foucault’s ideas about Govenrmentality. Would you recommend analysing some of the physical impacts as well?

        In relation to Paul Chilton is there a link to an example of how to transcribe in a way that marks hesitations, intonations, and other such qualities of the spoken word?

        Many thanks in advance Josh.

        • Florian Schneider

          Hi Josh,

          I am always in favor of including other types of media, and seeing how a discourse works in different “modes”, so this sounds promising. I would be careful to call something a “multimodal” analysis, though: I think the word fits best when you systematically look at how the medium contributes to the discourse. So if you are analysing camera angles, mise-en-scene, editing, etc. in combination with what is said in the film, then the term applies. If you are mainly commenting on the content of the film in relation to your interviews, then I might try to find another word (or point out in a footnote that you are not conducting a full-fledged “multimodal” analysis, and then suggest further reading on that kind of research approach).

          As for “physical” impacts, I find it fascinating to see how discourses crystallize into institutions and then inform such things as buildings, urban planning, use of physical violence, etc. Is that what you have in mind with “physical” impact? That would be the sort of question Foucault indeed looked at. My advice here would be to include such issues if you have good data, and to otherwise note such impacts in the intro/conclusion of your thesis. A risk here is that you might end up doing too many things at once, so be careful that you still narrow down your main analysis enough. This, of course, depends entirely on the kind of thesis you are writing.

          As for the transcription advice, I couldn’t find Chilton’s notations online, but I have reproduced some of them in the figure in this blogpost: As you’ll see, the full list of notations is on page 206 of his book “Analyzing Political Discourse”.

          Let me know how your analysis proceeds!

          Best – Florian

  17. Michele

    Hi Florian,
    Firstly, this post on discourse analysis is incredibly helpful so cheers for that!
    I’m embarking on my MA dissertation and am doing a critical discourse analysis. I’ve focused on one online newspaper and its coverage of immigration, but am feeling overwhelmed by the amount of data generated. I was wondering if you could recommend how many articles to study, as CDA is so intense, would 2 or 3 be ok?

    • Florian Schneider

      Hi Michele, I agree that a full discourse analysis of a large number of texts is almost impossible for anything smaller than a research MA or PhD thesis. There are three ways you can still make a contribution in an MA thesis, but without overwhelming yourself. The first is to consciously phase out certain analytic aspects, for instance by choosing to not explore all linguistic features of the texts in detail. In that case, you’ll have to find a good justification for your choices, and should probably point out at the end what follow-up research would now be necessary as a next step. The second option is to chose materials that are particularly representative. If you have evidence that a particular newspaper article kicked off a huge debate, or that a specific policy document is of paramount importance (e.g. a state-of-the-union address, etc.), then you may not need more materials – you should, however, then point out what limitations this particular “window” into the discourse has. Thirdly, you could take a classic hermeneutical approach by starting with one text, qualitatively mapping out the discourse and its features there, and then moving on to a second text, a third text, and so on, until you are no longer finding any major new discursive features. I believe Jäger recommends such an approach. If you narrow your topic down well enough at the outset, you may indeed be in a position to justify using only a handful of texts rather than a large corpus.

      • Michele

        Thanks Florian, this is really helpful. I’ve decided to do a general analysis of headlines from one month of the newspaper I’m using, focusing on elements of structural feminism and critical discourse analysis, and then shall do a more detailed breakdown of 3 of the most relevant articles. Do you think that would be okay, as long as my limitations are explicit?

        • Florian Schneider

          This sounds very cool – I personally like approaches that take a bird’s-eye view first (through headlines in newspapers, structure of TV series seasons, etc.) and then pick a representative sample for detailed analysis, based on that initial work. This could work nicely. Let me know how it shapes up! Best – F

  18. D CHEN

    It’s a good article to teach people how to conduct one discourse analysis. I learn much from it when I designed my research.
    I just have one question, is it suit for analysing the official documents like rules, laws, regulations etc.?
    I think it is good for assessing news, but I’m not sure if it can be applied to some official papers.

    • Florian Schneider

      Discourse analysis can definitely be used on policy documents. It is indeed easier to analyse news articles, since they are often rather explicit about their “discursive position”, but legal texts also appeal to certain categories, draw from assumptions, and establish self-evident truths. The important thing to keep in mind is that a legal document is a specific genre, and that different genre conventions consequently apply. I know Fairclough has looked at official documents, and you’re likely to also find such studies in the established journals as well (e.g. Discourse & Society), so I’d recommend taking a look at such examples for inspiration.

  19. Louise

    Hi Florian, this is a great article, and is one of the first I read that really explains the process clearly. I’m looking for some advice, however, as I’m writing about improving the nuclear Non-Proliferation Treaty and I’ve decided to conduct a discourse analysis on the Treaty, the IAEA statute and political speechs made by America as well as non-Western orgs such as the NAM or the Arab League on the subject, to compare discourse between the two sides and how to bridge the gap. Is this too ambitious? How many documents do you recommend for a decent analysis, and of such variety in genre?

    • Florian Schneider

      Dear Louise,
      This does indeed sound quite ambitious – I assume you are writing an MA thesis? You may want to take a look at my comment above, on Michelle’s project. She had similar concerns about narrowing down her material, and I suggested three different options to her on how to handle that challenge. In your case, you may have to make a choice: you could use one of the two sets of texts (official IAEA documents vs. speeches) as background and the other to do a detailed analysis. For instance, I think doing a discourse analysis on the actual Treaty and the IAEA statutes makes good sense, and should be doable at the MA level. On the other hand, if you want to cover speeches, I would probably recommend taking a quantitative approach first, for instance using WordSmith ( or some similar tool. You could then “zoom in” on specific features of the discourse and discuss those in more detail. Otherwise you might end up with a lot of speeches, particularly on a topic such as this one, that you may not be able to assess in detail at a qualitative level. Unless of course you are doing a PhD, in which case this sounds like the kind of work that would make a good doctoral thesis.
      I hope this helps!

      • Louise

        Thanks Florian, I’m looking into WordSmith now. After reading your response for Michele as suggested, would you recommend Fairclough over, say, Foucauldian Discourse Analysis, for such a project? What would you say are the benefits of CDA?

        • Florian Schneider

          Hi Louise.
          I’m afraid it’s not all that easy to draw a clear line between different approaches to discourse, such as Foucauldian analysis, CDA (e.g. Fairclough), political discourse analysis (e.g. Chilton), or discourse-historical analysis (e.g. Wodak). They often overlap and draw from each other, and many of the distinctions are subtle theoretical differences (for instance how “constructivist” the respective author is) rather than completely different methodological approaches. For a good introduction of how Fairclough aligns himself with Foucault’s aims, I can recommend this short text:
          To answer your question, I think you could make a distinction at the methodological level between studying 1) primarily and in great detail the linguistic features of a discourse, 2) the socio-historical context of the discourse (and its development over time), and 3) the strategic communication choices and social practices of different actors at a particular point in time (e.g. framing, self-other representations, etc). Most discourse analysts will look at all three, and if you want to read a good article that covers all of these angles for the Scottish case, I can recommend this piece by my colleague Johnny Unger for inspiration:
          For an MA thesis I think it would be fair to emphasise one of these levels of analysis, as long as you also acknowledge the others. You could, for instance, provide the socio-historical context in your introduction and could then explore how different actors frame the issue, building in examples from the language as you go along. Your limitations/future research section in the conclusion could then point out how more detailed linguistic analysis and historical tracing of the discourse can shed light on additional questions you have raised in your thesis. Just a thought.

  20. Jalton

    Hi Florian. Thank you very much for putting this website up. I am currently writing a proposal for a PhD dissertation on energy policy formulation and have been wondering about a specific kind of discourse analysis — argumentative discourse analysis (M. Hajer). There seems to be a dearth of resources about it, especially as a method. I surmise that it emphasizes certain dimensions of discourse compared to the “conventional” discourse analysis which generally explores/examines text (linguistic), the rhetorical component, as well as context (socio/historical). I was wondering if you were kind enough to offer suggestions or tips (perhaps even some general “red flags”). Many thanks and more power to you.

    • Florian Schneider

      Hi Jalton,
      Sorry for keeping you waiting – you caught me during the Easter break.
      Thanks for pointing out Hajer’s work, which I think has a lot in common with the issues I’ve discussed here. At the risk of doing his work injustice, it seems his argumentative discourse analysis (ADA) is very much interested in the structure of texts and conversations and in the rhetorical and argumentative strategies that people deploy. For instance, he’d be interested in classic argumentative fallacies such as appeals to authority or begging the question, which I agree are very useful when examining arguments. I particularly like the fact that he places a strong emphasis on how people perform their role in social interactions, which is something my colleagues and I are also interested in. In that sense, I don’t think ADA stands in opposition to other forms of discourse analysis – it simply draws attention to specific aspects of communication and would probably fit very nicely into the “toolbox” I’ve put together above.
      There are of course also differences, for instance in the way Hajer writes about “discourse coalitions” when talking about groups that share similar discursive positions – a context in which I would probably use a network approach – but these distinctions are rather subtle. I would have to talk to him and his colleagues to see where we potentially disagree. My guess would be that I place a tad more emphasis on agency whereas he might be a bit more interested in structures. At any rate, something I find highly valuable is his definition of “dominant” discourse, which you’ll find here: (under “influence of discourse”).
      Not sure whether I’ve helped or muddied the waters further… let me know what your PhD research uncovers, and what sort of approach you ended up adopting. The project sounds fascinating.
      Best – F

Leave a Comment