An Introduction to Discourse Analysis, Second edition, chapter 7

What I read

In this chapter, Gee summarizes the material from the previous sections, using an example of birding, and shows how transcripts play a role in discourse analysis through another example with a child.

First, Gee revisits briefly all of the elements previously discussed, starting with situated meanings and discourse models. A situated meaning is an “image or pattern that we assemble ‘on the spot,’ as we communicate.” The example of coffee being either beans or liquid is used to illustrate the way a word like “mop” or “broom” leads to a new assembled meaning. These specific words also have general meanings, because they are associated with Discourse models (storylines.) Discourse models explain, internally to a group, why words have specific situated meanings within that group. Simulations are ways we play prototypical situations out in our heads. Discourse models are linked to one-another.

Next, Gee introduces reflexivity—the idea that language is a chicken-and-egg in a situation, coming from the situation and simultaneously changing the situation. Situations are where the building tasks converge or activate. These are “discrete situations.” Gee then revisits the seven building tasks:

Significance is making “mountains out of molehills,” or creating importance when looking at a situation.

Activities are the ways things people do are construed in different ways, given the circumstances of the situation.

Identities are the way people take on roles through the things they say.

Relationships are how language is used to indicate how people want to be considered by and with other people.

Politics are how relationships with social or shared goods are communicated.

Connections are how ideas are related to other ideas to create meaning or intent.

Sign systems and knowledge are how people use specialist language to communicate.

Situations are repeated, although never the same, and these embed these building tasks into habits.

Gee describes that language contains clues about the building tasks, which are carried out “all at once and together. Social languages hold these clues, in both use of traditional grammar and sentence structure, and in the second form of grammar—patterns that those in a Discourse will understand and use.

Next, Gee describes how transcripts are used to analyze discourses. Transcripts can be more or less detailed (narrow or broad); small details in a narrow transcript can have important impact on meaning and analysis. Gee uses a transcript analysis example related to a child, being asked about light. He offers a brief view of the syntax used in this transcript analysis, such as the use of stanzas, underlining, capitalized words as emphasis, two periods as a pause, and vertical dots to show drawn out vowels.

The example is used to show how an interviewer “co-constructs” the conversation, and actually contributes to the lack of understanding observed in the child. This is partially because the interviewer doesn’t know or doesn’t recognize the way the child is used to question/answer interactions in school. For example, the researcher tries to move from a simple idea to more abstract principles, but this is not the way the student has been questioned before in school; the student tries to fit their responses into the style of questioning, and Gee concludes that the situation led to “misunderstanding,” which is consequential when that is then assigned to the student rather than the language and diversity.

Next, Gee describes the 26 questions that can be asked during a discourse analysis to arrive at an ideal, or valid, examination of a situation. These questions are grouped under the seven building tasks categories; the questions, summarized, are:

Significance

• What are the situated meanings of the words that are important?
• What situated values are attached to relevant places, people, artifacts, etc?
• What situated meanings are attached to other texts or quotes?
• What Discourse models are being used to connect situated meanings?
• What Discourses are being produced or transformed?

Activities

• What is the activity that is going on?
• What sub-activities compose the activity?
• What actions compose the activity?

Identities

• What roles, and related knowledge, beliefs, and feelings, are in play?
• How are these identities reinforced or changed?
• What Discourses are relevant?

Relationships

• What social relationships are relevant, ignored, or created?
• How are these reinforced or changed?
• How are other texts or quotes used to create relationships?
• What Discourses are relevant?

Politics

• What social goods are relevant or irrelevant?
• How are they connected to the Discourses?

Connections

• What connections are made across utterances?
• What connections are made to things outside of the situation?
• How is intertextuality used to create connections?
• How do connections create coherence?

Signs and knowledge

• What speech, writing, images, and gestures are relevant or irrelevant?
• What knowledge is relevant or irrelevant?
• What “national” language is used?
• What social languages are used, and relevant?
• How is intertextuality used?

Gee then describes validity, which has “continually vexed so-called ‘qualitative research.’” Validity is based on four elements: the convergence of answers to the above 26 questions towards something mutually compatible, agreement with the answers from other researchers or people in the Discourse, how much the analysis can be applied to other related things, and how much it is tied to “linguistic structure.” It is “highly improbable that [these elements] will converge unless there is good reason to trust the analysis”—this is validity. Discourse analysis is the argument that these elements are valid in supporting a hypothesis.

Gee concludes the chapter by providing a brief overview of how to do a discourse analysis. He recommends selecting a piece of data that “both interests you and that you believe will speak to or illuminate an important issue or question… with an eye to the features you think will be most important for the issue or question in which you are interested.” A researcher then should think about the situated meanings of the content, and which Discourses are at play. Then, after reflection, they should ask the 26 questions above, looking to where answers converge on a same point or theme. Then, a researcher can expand this approach to other parts of the data, or to new data.

What I learned and what I think

This is very much aligned to a designerly way of thinking and engaging with participants during research: dive deep into individual sections or parts of a transcript, and interrogate it from a number of perspectives. We almost always use perspectives related to the subject of the design problem, so it’s a lens of education when we’re working with an education client, finance for finance, etc.

That’s one core difference here, I think. We’re looking at a problem with a problem-finding and problem-framing goal, which is different than an understanding or hypothesis-validating goal. The design lens is shaped also by the agenda of our client, their stated roadmaps and business goals, the way we’ve interpreted their appetite for risk and their pre-conceived ideas of where the research should head (and how much we will capitulate to that, given that this is work-for-hire.) In a way, we’re doing discourse analysis during our interactions with the client about the research that then does a form of similar interpretation.

In contrast, the approach here is on the way towards validity, which is a logical inevitability. That’s just not happening in design, maybe because of lack of rigor, or maybe because it’s just a completely different output lens. I like the way validity is described, because it doesn’t abdicate that the results are relevant and valuable, but does shift the focus of validity definition away from repeatability or some sort of sample-to-population predictability. I do wonder about the small amount of snark about “so-called qualitative research,” which is probably a shot at thematic analysis in academic research, but maybe could be shot also at designers playing analyzers (same critique from ethnographers.)

Another difference, although maybe we just haven’t gotten there yet, is the lack of exploration across participants. The analysis has been focused on one transcript of one person. I’m fast-forwarding to when I actually go do this, and right now, I see that my interpretation cross-participant is going to probably use this type of philosophy with a more reflexive thematic analysis. I don’t know if that “counts.” I’m not sure it matters, except that it will change and probably blow up the validity definition of inevitability.

The seven building tasks make sense, as they are “just” ways of poking at a piece of data from a variety of angles, all related to the Discourse. It’s a way of forcing that rigor, in the pieces that Gee has established as relevant. I really don’t understand or like the name “building tasks.” They seem like lenses or filters, at least for the analysis, and aspects of Discourse for the speaker. They really don’t seem like tasks, because a task is very goal-focused and overt. However, “aspects” doesn’t really do much either.

The ”just think of the 26 things” is driving me crazy, in terms of teaching and learning. It’s literally impossible to keep a list of 26 things in the brain in any depth at once, and I can’t believe he means to actually use these as a checklist (in fact, he goes out of his way to say not to do that.) After doing this over and over, I can see internalizing and then intuiting those 26 things, but right now, it’s not just intimidating, it’s not useful. But, in looking at them, they are very consistent within each of the seven tasks, so I think I can manage in the generalities.

I’m going to continue to critique and not like the “5 of this, 2 of this” approach. We’re at 7 building tasks, 26 things to keep in mind for an ideal analysis, 4 elements of validity, 3 discourse model types, 2 grammars, and 4 tools of inquiry. The building tasks seem the most pragmatic for actually doing the analysis, so I’ll focus on those for now.