You've coded every transcript. You have a tidy list of codes with supporting quotes. And now you're staring at that list wondering—what do I actually do with these to produce themes?
This is where most qualitative analysis stalls. Not during data collection, not during initial coding, but during the transition from codes to themes. The confusion is so common that Braun and Clarke devoted a full commentary to it, noting that published studies routinely present what are essentially domain summaries—organized topic areas—as though they were themes. The difference matters: domain summaries describe what people talked about; themes tell you something about what it means.
Here is how to make the transition deliberately, in four moves.
What Actually Distinguishes a Theme from a Code?
A code labels a specific idea found in the data—usually one sentence or paragraph. "Participants described avoiding eye contact during supervision" is a code. A theme, by contrast, captures a recurring pattern of shared meaning across the dataset, organised around a central concept. "Invisible hierarchies in supervisory relationships" could be a theme—it's not a topic summary (supervision) and not a single observation (eye contact). It tells you the kind of pattern you found across the data.
The Quirkos blog frames the confusion well: when does a code become a theme? And when was a code really a theme all along? The shorthand is that a good theme has two properties: it recurs across multiple participants or data sources, and it has what Braun and Clarke call a "central organising concept"—a single idea that gives the theme its coherence. If your theme can be summarised as "everything participants said about X," it is probably a domain summary, not a theme. (If you're still working on getting your codebook right before worrying about themes, our codebook sanity check covers that ground first.)
Why Do Researchers Get Stuck Here?
Three reasons consistently surface.
First, most methods training teaches coding well but theme construction poorly. We show students how to highlight passages and assign labels. We rarely walk them through the subsequent step of asking: what shared idea connects these labels? The MAXQDA thematic analysis guide for 2026 notes that Phase 3—generating initial themes—requires a "fundamental shift in analytical thinking, moving from the detailed, segment-level focus of coding to the broader conceptual level where themes reside." That shift is rarely scaffolded.
Second, software tools that are excellent for coding do little to help with theming. Most CAQDAS packages let you drag codes into parent categories. That looks like theme development. It is not. Creating a folder called "Communication" and putting all the communication-related codes inside it gives you a category, not a theme. A theme about communication would tell you something specific—for example, that participants used indirect communication to manage power differentials. The folder analogy is a trap.
Third, researchers conflate "frequent" with "thematic." A code that appears in 40 of 50 transcripts is salient. It is not automatically a theme. Frequency tells you something is widespread; it does not tell you what that thing means in the context of your research question.
How to Move from Codes to Themes in Four Steps
Step 1: Dump the code hierarchy and build a concept map
Take every code you have and write them on physical or digital cards—one code per card. Then arrange them on a table or a whiteboard in whatever clusters seem to belong together. Ignore your code hierarchy for now. Let yourself move cards around, group and regroup, notice when a code keeps wanting to sit in two places at once. That tension is analytically useful.
This works because hierarchies encode who contains whom (parent-child relationships), but themes are about what connects to what (associative relationships). A concept map is better at surfacing associative links than any tree diagram.
Step 2: Ask every cluster: "What is the shared idea here?"
Once you have a cluster of 3-7 codes that seem to co-occur, stop organizing and start interpreting. Write one sentence that states the central organising concept for that cluster. Not a topic label ("Emotions"), but an interpretative claim ("Emotions are treated as professional liabilities that must be managed privately").
If you cannot write that sentence, you do not yet have a theme—you have a collection of potentially related codes. Try again with a different grouping. This is often where people realise their original clusters were based on similarity of topic rather than similarity of meaning.
Step 3: Test the theme against a maximum variation sample
Pull the data extracts for each code in your candidate theme. Read them side by side. Do they actually speak to the same central idea, or are they different things that happen to use similar words? A good theme is not uniform—it captures variation around a shared concept—but the concept must be genuinely shared.
Virginia Braun and Victoria Clarke recommend checking that your themes "work together to tell a story about the data." If you have a theme about "institutional barriers" and another about "individual coping strategies," ask yourself: do these themes speak to each other? Or are they just separate topics you've listed?
Step 4: Name the theme with a claim, not a topic
This is the most visible test. If your theme name could be a section heading in a textbook ("Communication," "Barriers," "Support"), it is a domain summary. If it makes a claim ("When trust breaks, silence fills the gap"), it is a theme.
Claire Moran puts it well: a theme name should tell the reader something specific about what you found, not just where the quote came from. "Coping" tells me the topic. "Coping as a collective performance, not an individual burden" tells me the finding.
Why This Matters for Your Write-Up
When themes are actually themes—when they have a central organising concept and a clear relationship to each other—the writing phase transforms. You are no longer summarising what participants said under topic headings. You are building an argument, each theme a chapter in an analytical story. (Our guide on writing up findings from coded data picks up where this post leaves off.)
A codebook with 80 codes and 12 thematic folders is a management tool. A set of 4-6 well-developed themes, each with a clear central concept, is an analysis. The difference is visible to every examiner, reviewer, and reader. Paideias helps by keeping your codes linked to their source data while you experiment with thematic groupings and test central concepts against the actual evidence—but the analytical work of recognising that shared idea is yours to do.
Frequently Asked Questions
When does a code become a theme?
A code becomes part of a theme when you can articulate the shared meaning that connects it to other codes. A single code, on its own, is rarely a theme—unless that code captures such a rich, recurring pattern that it functions as a central organising concept for multiple data extracts.
How many themes should a study have?
For a typical reflexive TA study, 2-6 themes (with occasional subthemes) is the recommended range. Fewer than two and you may be under-analysing. More than six and the analysis often becomes fragmented—the reader forgets the earlier themes by the time they reach the later ones. The goal is not exhaustive coverage; it is an insightful argument.
What is the difference between a domain summary and a theme?
A domain summary groups everything participants said about a topic. A theme identifies a pattern of shared meaning across the data. "What patients said about waiting times" is a domain summary. "Waiting as a demonstration of institutional power" could be a theme—it makes an interpretive claim that organises the evidence.
Should I use AI to help develop themes?
AI can help you pull relevant excerpts, notice patterns across large datasets, and test thematic groupings—which is what Paideias is designed for. But the conceptual leap from noticing a pattern to naming its central organising concept is fundamentally human. Use the tool to handle the evidence; keep the interpretation for yourself.
Discussion
or sign in to comment with your account