Thematic analysis is a method for finding, organising, and reporting patterns of meaning, called themes, across qualitative data such as interview transcripts, focus groups, and open survey responses. It is the most widely used qualitative approach in dissertations because it is flexible and not tied to one theoretical school. Done properly it is a disciplined, auditable process that moves from raw text to codes to themes you can defend, not a quick summary of what people said.
Why a code is not a theme
The single mistake that weakens most qualitative chapters is treating codes and themes as the same thing. A code is a short label attached to a segment of data that captures what it is about, for example fear of failing the viva. A theme is a broader pattern of shared meaning that runs across many codes and speaks to your research question, for example assessment as a source of identity threat. A list of codes, however tidy, is not a set of themes. The analytic work, and the part examiners scrutinise, is moving from many codes to a smaller number of themes that each carry one clear, central idea. This is the same discipline of matching method to question that governs how to choose a statistical test on the quantitative side.
The six phases of reflexive thematic analysis
The most cited framework, from Braun and Clarke, sets out six phases. They are not a rigid recipe to rush through but a recursive process you move back and forth across as your understanding deepens.
- Familiarisation: read and reread your transcripts, noting first impressions before you code anything.
- Generating initial codes: work systematically through the data, tagging every feature that is interesting or relevant, and keep a short definition for each code.
- Searching for themes: cluster related codes into candidate themes, asking what each potential theme is really about.
- Reviewing themes: check candidate themes against the coded extracts and the whole dataset, then merge, split, or discard them so each one genuinely holds.
- Defining and naming themes: write a clear definition and a concise name for every theme so a reader grasps it at once.
- Writing up: produce the analytic narrative, weaving in verbatim quotations as evidence for each theme.
Notice that themes are constructed through this work, not lying in the data waiting to be found. The quality of your analysis rests on the judgement you apply across these phases and on your ability to show that judgement, which is why an audit trail matters as much here as a clean analysis script does when you write the results chapter of a dissertation.
Inductive or deductive, semantic or latent
Before you code, settle two choices and state them, because they shape the whole analysis and an assessor will expect a deliberate decision. First, your coding direction: an inductive analysis builds themes from the data itself, while a deductive analysis reads the data through an existing theory or framework. Second, your level of meaning: a semantic reading stays close to what participants said explicitly, while a latent reading interprets the underlying assumptions beneath the surface. Many dissertations sensibly blend these, but you should know and declare where you sit.
The three types of thematic analysis
Thematic analysis is now usually described as a family of three approaches that differ in how themes are developed and whether agreement between coders is measured.
| Approach | How themes are developed | Best suited to |
|---|---|---|
| Coding reliability | Codebook applied by several coders, agreement measured | Teams wanting demonstrable consistency |
| Codebook | Structured codebook, developed and refined as you go | Applied or policy work with some predefined structure |
| Reflexive | Themes developed through the researcher's active interpretation | Most qualitative dissertations exploring meaning |
Reflexive thematic analysis is the most common choice in doctoral work. It treats your own position as part of the analysis rather than a source of bias to be eliminated, so a short statement of reflexivity belongs in your write-up.
Showing rigour, and where software fits
Qualitative work is judged on whether a reader can trust how you reached your themes, so build the evidence of rigour as you go: keep an audit trail of decisions, a codebook with definitions and examples, a theme map showing how codes were grouped, and quotations tied to their sources. Tools such as NVivo, MaxQDA, or ATLAS.ti organise codes and extracts and make large datasets manageable, but they do not create themes or judge meaning. The interpretive work, and the credibility of the result, rests with you. The same caution applies to generative artificial intelligence: it can help organise or suggest, but a human makes every coding and theming decision, and no tool should be presented as having produced your findings on its own.
Thematic analysis is one of several approaches, so it is worth seeing it beside the alternatives in our overview of qualitative data analysis methods. And if your study pairs interviews with a survey, you can analyse open responses thematically while handling the rating items as described in analysing Likert-scale survey data, keeping the two strands of a mixed-methods thesis speaking to each other.