Conversation analysis is the study of how people use talk to get things done together. It covers how people take turns, pause, interrupt, and repair misunderstandings, not just what they say. There are several types of conversation analysis, but the most useful split is between two: ordinary (or mundane) conversation analysis, which studies everyday talk, and institutional conversation analysis, which studies talk in professional settings like medical consultations and healthcare contact centers.

That distinction matters beyond linguistics. Institutional talk is where business problems actually surface. A patient who calls back three times about the same bill, a benefits explanation that leaves a member more confused, a required disclosure rushed past at the end of a call: each shows up in the structure of the conversation itself. Study that structure closely, across enough conversations, and patterns emerge that no survey or call summary would reveal.

Key Takeaways

• Conversation analysis studies how people use talk, including pauses, turn-taking, and nonverbal cues, to accomplish things together, not just what they say.

• The two broadest types are ordinary conversation analysis (everyday talk) and institutional conversation analysis (talk in professional settings like healthcare, customer service, and call centers).

• Within those, analysts apply different lenses (microanalysis, macroanalysis, and ethnomethodological analysis) depending on whether they're studying a single exchange or broader patterns across many.

• Conversation analysis is related to but distinct from discourse analysis, which looks more broadly at language and its social or cultural context rather than the mechanics of a specific exchange.

• At scale, institutional conversation analysis lets healthcare organizations identify recurring gaps, like confusing explanations, missed empathy cues, and compliance risks, across every interaction instead of a small sample.

What Is Conversation Analysis in Healthcare?

Conversation analysis (CA) is a research method, rooted in sociology and linguistics, that studies social interaction through close examination of recorded talk. Analysts look at words, but also at pauses, overlaps, hesitations, and tone.It pays special attention to the small details people use to signal agreement, discomfort, or confusion.

The field grew out of ethnomethodology in the 1960s and 1970s, through the work of sociologist Harvey Sacks and colleagues Emanuel Schegloff and Gail Jefferson. Their core insight was that ordinary conversation is highly organized. Turn-taking, for example, follows rules speakers rarely notice but almost always follow.

Outside academia, the same close-reading approach helps organizations understand what actually happened in a patient or customer conversation, not just what a survey score or after-call note says happened. In healthcare, where a misunderstood instruction can affect someone's care, that difference matters.

The Main Types of Healthcare Conversation Analysis

Conversation analysis is usually divided by setting: who is talking, and whether their roles are fixed. That gives us two primary types: ordinary and institutional. Within either type, analysts then apply different lenses depending on whether they want to study one exchange closely or patterns across many. Think of the types as what is being studied, and the lenses as how.

Ordinary (Mundane) Conversation Analysis

Ordinary, or mundane, conversation analysis studies everyday, unstructured talk between friends, family members, or strangers. No one has a fixed role or script, so analysts focus on how people coordinate on the fly.

For example, two friends start talking at the same time, both stop, and one says, "Sorry, go ahead." That small moment reveals a shared, unspoken system for managing turns and repairing overlaps. Ordinary conversation is the baseline researchers use to understand how talk works when nothing constrains it.

Institutional Conversation Analysis in Healthcare Settings

Institutional conversation analysis studies talk within professional roles: medical consultations, customer service calls, legal proceedings, and call centers. Participants have defined roles, specific goals, and limits on what each person is expected to say. As a result, institutional talk follows recognizable patterns, and those patterns are exactly what analysts study.

Take a call to a health plan member services line. The member opens by presenting a problem: a claim was denied. The agent verifies identity, asks diagnostic questions, explains the cause, and proposes a resolution. Each participant knows their part.

When that structure breaks down, it's visible in the transcript. An agent starts explaining before the member finishes describing the problem. A call closes before the member confirms they understand the next step. Find the same breakdown across thousands of calls, and it's a systemic issue, not a one-off. That leap from one call to thousands is where institutional conversation analysis becomes an operational tool, which we'll return to below.

Analytical Lenses: Microanalysis, Macroanalysis, and Ethnomethodological Analysis

Analysts can apply any of three lenses to ordinary or institutional talk:

• Microanalysis looks closely at a single exchange: word choice, pauses, and how turns are taken.

• Macroanalysis looks at patterns across many conversations, or at how talk connects to a broader social context.

• Ethnomethodological analysis focuses on the unstated, shared assumptions that make an exchange make sense to the people in it.

Conversation Analysis vs. Discourse Analysis

Conversation analysis and discourse analysis are often confused, but they ask different questions. Conversation analysis focuses tightly on the mechanics of a specific spoken exchange: turns, pauses, overlaps, and repairs. Discourse analysis looks more broadly at language, text, and talk in their wider social, cultural, or ideological context. It isn't limited to spoken conversation and might examine a policy document, a news article, or an ad campaign.

Consider a patient who says, "I guess I'll just pay it, then," after a billing explanation:

• A conversation analyst would note the two-second pause before she spoke, the hedge "I guess," and that the agent moved to close the call without checking her understanding.

• A discourse analyst might ask how healthcare billing language positions patients as responsible for navigating a system they didn't design.

Both are valid. For improving specific interactions, conversation analysis is the more direct tool.

How to Conduct a Conversation Analysis

The six steps below follow one running example: a health plan wants to understand why members keep calling back about prior authorization status.

1. Select the conversations. Choose calls tied to the question. Here, that's recorded calls where a member asked about a pending prior authorization and called again within a week.

2. Transcribe them. Capture the words and the details that matter: pauses, overlaps, and shifts in tone. A long silence after "it's still under review" is data.

3. Identify the sequence of actions. Map what each turn does. The member asks for status. The agent gives it. The member asks how long it will take. The agent says it depends, and the call closes.

4. Identify patterns. Compare across calls. Agents consistently answer the status question but rarely say what happens next or when the member will hear back.

5. Analyze. Interpret why the pattern matters. Without a clear next step, the member has no reason to stop calling. The breakdown is in the sequence, not the agent's tone.

6. Write a report. Summarize findings and recommendations, such as a standard next-step statement and a proactive update when an authorization runs past its expected window.

This process works well for research, where a handful of calls can answer a focused question. It breaks down at operational scale. A contact center handling thousands of patient or member interactions each week can't transcribe and code them by hand, so most teams analyze a small sample and hope it's representative. The patterns that matter most, like the repeat-contact loop above, are often the ones a sample misses.

Features of Conversation Analysis in Healthcare

Five features do most of the work in any conversation analysis. Each has a direct parallel in healthcare calls.

• Turn-taking: How speakers trade the floor. An agent who talks over a patient describing symptoms may miss the one detail that matters.

• Adjacency pairs: Paired actions where the first sets up an expected second, like question and answer or offer and acceptance. When an agent asks, "Would you like me to schedule a callback?" and gets silence or "I guess," that hesitant response is worth following up on.

• Repair sequences: How speakers catch and fix misunderstandings. A caller asks, "So I owe $40?" and the agent clarifies, "Your copay is $40, but this visit also applied to your deductible," before the caller hangs up confused.

• Sequence organization: How actions are ordered across the whole call. An agent who closes before confirming next steps has a sequence problem, even if every individual answer was correct.

• Nonverbal cues: On the phone, these are pauses, sighs, pace, and tone. A long pause after a cost estimate often signals confusion or worry the caller doesn't put into words.

How Authenticx Applies Conversation Analysis at Scale

Everything above describes institutional conversation analysis done by hand. Authenticx applies the same approach across 100% of healthcare interactions, using AI to handle the transcription, coding, and pattern-finding a manual analyst can only do for a sample.

The line from method to platform is direct. The features a researcher codes manually, such as repair sequences, turn-taking breakdowns, and tonal cues like hesitation or frustration, are the signals Authenticx surfaces automatically across every call, chat, and email. The prior authorization pattern from the example above doesn't depend on someone pulling the right ten calls. It shows up because every call is in the data.

AI alone can miss the nuance qualitative analysis depends on, so Authenticx uses on-shore, human-reviewed AI: human reviewers layered on top of automated analysis so findings hold up.

A few distinctions are worth keeping clear:

• Conversation analysis is the research method this page describes.

• Conversational analytics is the AI-powered application of it: software that measures and surfaces patterns across large volumes of interactions.

• Expert-led analysis is an option for organizations that want the work done for them rather than running it in-house, through Authenticx's Targeted Conversation Analysis service.

Day-to-day, KBo, Authenticx's AI conversation guide, helps teams work with these findings, so insights reach the people who can act on them instead of staying in a report.

See how Authenticx applies conversation analysis at scale: across 100% of interactions, not a sample.

Schedule a Demo