# Discourse Analysis: Read an Exchange Without Guessing Intentions

Learn discourse analysis through a fictional workplace exchange: describe language, compare possible readings, and check the context.

- Canonical URL: https://www.harmate.com/en/blog/discourse-analysis-read-an-exchange-without-guessing-intentions
- Author: Harmate Team
- Published: 2026-10-04
- Updated: 2026-10-04T00:20:25.096036+00:00
- Language: en

## Content

# Discourse Analysis: Read an Exchange Without Guessing Intentions

Discourse analysis examines how an utterance is built and gains meaning in a text or an interaction. It is not a mechanical way to translate a word into a hidden intention. Imagine a fictional exchange between two people preparing a training session: a “yes” answers a question, then a qualification narrows what that agreement covers. The task is to describe turns, roles, pronouns, modality, and context; keep competing readings open; then ask a question that checks what is missing.

## What is discourse analysis?

It is a family of approaches that examines language beyond isolated sentences. Depending on the question, an analyst might study recurring relations between elements in a text, the organization of turns, roles participants take up or assign, pronoun choices, forms of requests, or the circumstances in which something is said. No single checklist fits every corpus and every research question.

In a foundational paper, Zellig Harris describes a method for analyzing connected discourse by examining elements in their relations within the text; he also sets out what this method can reveal about structure without claiming it alone uncovers every particular meaning (originally published as “Discourse Analysis,” *Language*, 1952; French translation: [Harris, “Analyse du discours,” *Langages*](https://www.persee.fr/doc/lgge_0458-726x_1969_num_4_13_2507)). Conversation analysis, meanwhile, draws attention to how roles, context, and meaning can be built through turns rather than fixed entirely before an exchange begins ([Lüger, “Routine conversationnelle et comportement langagier” (in French)](https://www.persee.fr/doc/lsoc_0181-4095_1993_num_63_1_2598)).

These perspectives can help with a workplace conversation, but they do not provide direct access to a person's thoughts. A linguistic form can guide an analysis; it cannot prove why someone chose it.

## A fictional example: “Yes, for the pilot group”

**Known context:** A trainer and a program lead are preparing a session due Thursday. The materials include several items: a core module, a mobile version, and facilitator instructions. They review a fictional message exchange to understand why it did not establish whether everything was ready.

> **Program lead:** “Are the materials ready for Thursday?”  
> **Trainer:** “Yes, the module is good; I sent everything yesterday.”  
> **Program lead:** “Great. Then we can launch as planned.”  
> **Trainer:** “For the pilot group, maybe. The mobile-version instructions haven’t been approved yet.”

The trap is to summarize this as “The trainer is reluctant to commit” or “The program lead does not listen.” Those are possible stories, but the words do not establish them. Start instead with observations and readings that still need checking.

| Observable trace | One possible reading to check | Another compatible reading |
|---|---|---|
| “The materials,” then “the module” | The question and answer may refer to different sets of documents. | The speakers may use “materials” as shorthand for the core module. |
| “I sent everything yesterday” | The answer reports a completed sending action, not necessarily approval by recipients. | The trainer may consider sending everything a sufficient answer. |
| “Great. Then we can…” | The program lead turns “yes” into a conclusion about launch. | The lead may think the conclusion concerns only the module just mentioned. |
| “For the pilot group, maybe” | The agreement is explicitly limited; “maybe” qualifies the proposal. | The trainer may distinguish a pilot from a wider rollout. |
| “haven’t been approved yet” | Approval remains outstanding; the sentence does not say who owns it or when it will happen. | The trainer may be stating a launch condition, not refusing the work. |

We can observe that “yes” may not answer everything the program lead had in mind. From this excerpt alone, we cannot conclude that the trainer wanted to avoid the task, that the lead deliberately ignored a caveat, or that the module is actually ready. Those claims need other evidence or, preferably, direct clarification.

## Four dimensions for reading without overinterpreting

### 1. Turns and their sequence

A response should be read with the turn before it and what it makes possible next. Here, the first answer seems to close the question; “great” and “then” build a consequence. The last turn reopens that conclusion with a condition. The useful unit is not just “Yes”: it is the sequence of question, conclusion, and qualification.

This describes the organization of the exchange, not the participants’ state of mind.

### 2. Pronouns and references

“I sent everything” says who sent something; it does not necessarily specify what “everything” includes. “Then we can launch” leaves open who decides: who is “we,” and what does “launch” mean here? Pronouns and phrases point to a reference question, not proof that someone is being excluded or shifting responsibility.

### 3. Modality: certainty, possibility, obligation

“Maybe” makes a statement less categorical; “haven’t been approved yet” presents an action as unfinished. The lead’s “we can” is also modal: Eddy Roulet’s study of the French modals *pouvoir* (“can”) and *devoir* (“must”) shows that they participate in how permission or requests are performed and take their value within interaction ([Roulet, “Modalité et illocution”](https://www.persee.fr/doc/comm_0588-8018_1980_num_32_1_1486), in French). “Then we can launch” may therefore state a possibility, propose a decision, or implicitly ask for agreement. The words alone do not reveal confidence, motivation, or the reason for the delay. In a work setting, “maybe” may keep an option open, recall a condition, or request a decision. Look at the next turn or ask what the speaker meant to leave open.

### 4. Roles and context

The speakers may have different responsibilities: one may decide whether to launch; another may prepare materials or obtain approval. “Ready?” sounds simple, but “ready” might mean written, tested, approved, accessible, or ready to facilitate. Record what is established—date, channel, documents, explicitly known roles—and what the excerpt does not say.

## Turn an analysis into a clarification question

A useful analysis produces a next question, not a verdict about a person. For this exchange, one could ask: “When you say the module is good, which part have you checked? What still needs approval before Thursday, and who can do that?” The question returns to ambiguous terms, distinguishes the materials, and makes the pending decision visible.

Avoid: “Are you unsure about your work?” or “Why haven’t you finished it?” Those formulations insert an explanation and push the other person to confirm, deny, or defend it. A neutral follow-up is not magic: it can still be misunderstood, and the other person is not required to share more than they choose.

For your own data, separate the source material, the observation, and the interpretation. The [verbatim guide](/en/blog/verbatim-how-to-use-exact-quotes-without-distorting-their-meaning) covers exact quotation and preserving context. [Open-ended questions](/en/blog/open-ended-questions-get-actionable-data-without-bias) can collect accounts; without an analysis and checking method, they do not establish a motive or a general fact.

## A six-move reading method

1. **Set boundaries.** Are you studying a conversation, interview, report, or set of texts? Define the question; “analyze the discourse” without an object is too broad.
2. **Gather necessary context.** Note the situation, known roles, date, or channel that matters. Do not fill gaps with intuition.
3. **Describe before interpreting.** Record turns, repetitions, pronouns, negation, modality, requests, and time expressions relevant to the question.
4. **Tie each reading to a trace.** Separate an observation about the text from an explanation of intent.
5. **Form competing readings.** Ask what could distinguish them; a reading that cannot be questioned can turn into a flattering story.
6. **Return to the person or context if possible.** Ask for clarification, or state explicitly that the available material does not settle the question.

This is not a complete school of discourse analysis, and it cannot replace a specialized method when a project requires one. It helps reduce the gap between what an exchange says, what an analyst notices, and what they can reasonably conclude.

## What discourse analysis cannot prove

One excerpt may not represent a whole relationship or team. A reply may be ironic, unfinished, constrained by time, or answer only part of the preceding turn. Transcripts often lose pauses, gestures, intonation, and the conditions under which people spoke. The more context is missing, the more modest the conclusion should be.

Do not use a word list to classify a person as “resistant,” “committed,” or “unmotivated.” The same words can perform different functions depending on the participants, subject, and moment. Nor should analysis become hidden surveillance: explain why exchanges are collected, limit the scope, protect people, and avoid publishing recognizable quotations without an appropriate basis.

## When to connect the method to response collection

A training or advisory project may use collected or imported responses to retain context, surface themes supported by verbatims, and prepare a report. That does not replace analyzing a complete dialogue when turns matter, and it does not reveal respondents' private motives. In practice, before accepting a theme, reread a few of its supporting verbatims with this article’s questions: what is the person answering, what restriction or modality appears, and what context is missing? The practitioner decides which readings are defensible, which contradictions remain open, and what question to ask next.

Discourse analysis becomes more careful when it slows interpretation down: describe forms, show how they relate to context, preserve competing readings, and acknowledge what is missing. The goal is not to guess what someone “really meant,” but to explain what the traces allow us to understand—and what they do not yet settle.