Thematic analysis is a method for identifying and interpreting patterns of meaning across qualitative data, such as interview transcripts, survey responses, or field notes. It is one of the most widely used approaches in qualitative research, largely because it is not tied to any single theoretical framework and can be applied across disciplines from health sciences to education to business. But that flexibility also makes it easy to misunderstand or misapply, and the method has evolved considerably since its early descriptions, particularly with the distinction between reflexive, codebook, and coding-reliability approaches that researchers now treat as meaningfully different ways of working.
What Thematic Analysis Actually Does
At its core, thematic analysis involves reading through a dataset, assigning short labels (codes) to segments of text that capture something interesting, and then grouping those codes into broader themes that tell a coherent story about the data. The word “theme” here is doing real work: a theme is not just a topic that came up frequently. It is a pattern of shared meaning that captures something important about the data in relation to your research question. That distinction matters more than it might seem, and it is one of the places where researchers most commonly go wrong.
A well-constructed theme is specific and revealing, capturing nuances, shared meanings, and points of divergence in what participants actually said. A poorly constructed one looks like a generic label or topic summary, something like “communication” or “challenges,” that could apply to almost any study. Effective themes go beyond isolated responses to individual interview questions and instead weave together insights from across the dataset into an interconnected analytical story about the phenomenon being studied.1International Journal of Qualitative Methods. Thematic Analysis in Qualitative Research: Common Pitfalls and Practical Insights for Academic Writing
The Spectrum of Approaches
One of the biggest sources of confusion around thematic analysis is that the term actually covers a family of methods, not a single technique. These range from approaches that prioritize coding accuracy and reliability, where multiple coders independently label the same data and then check agreement, to reflexive approaches that treat the researcher’s subjectivity as an unavoidable and even productive part of interpretation.2Counselling and Psychotherapy Research. Can I use TA? Should I use TA? Should I not use TA? Comparing reflexive thematic analysis and other pattern‐based qualitative analytic approaches Knowing which version you are doing, and being transparent about it, is more important than picking the “right” one.
In coding-reliability approaches, the emphasis falls on consistency. Two or more researchers code the same transcripts, and the degree of overlap between their codes is measured statistically. The goal is to demonstrate that anyone applying the same codebook to the same data would arrive at similar results. This works well when the research team needs to show that findings are not idiosyncratic, and it fits naturally into post-positivist frameworks that value replicability.
Reflexive thematic analysis, most closely associated with the work of Virginia Braun and Victoria Clarke, takes the opposite stance on subjectivity. Rather than trying to minimize the researcher’s influence, it treats the researcher as an active participant in knowledge creation. The codes and themes that emerge are understood to be shaped by who is doing the analysis, what they bring to the data, and the questions they are asking. This does not mean “anything goes.” It means the researcher needs to be reflective about the choices they are making and articulate why they made them.
A third family, sometimes called codebook or template analysis, sits in between. The researcher develops a coding template, sometimes partly in advance based on theory or prior research, and then refines it as they work through the data. One study of prehabilitation in gynecologic oncology patients, for example, used inductive thematic template analysis on semi-structured interviews, starting with an exploratory aim and letting the template evolve as patterns emerged from the conversations.3PubMed Central. PRESurgery thoughts – thoughts on prehabilitation in oncologic gynecologic surgery, a qualitative template analysis in older adults and their healthcare professionals
Inductive, Deductive, and Hybrid Coding
Another key decision is whether your coding will be driven by the data (inductive), by existing theory (deductive), or by some combination of both. When you work inductively, you read the data without a predetermined framework and let the codes emerge from what participants actually said. When you work deductively, you bring a set of concepts from existing literature or theory and look for how they appear in the data.
In practice, many researchers blend both strategies. A hybrid approach might start with a set of theory-driven codes drawn from an existing framework but remain open to data-driven codes that capture unexpected patterns. One influential example of this used a hybrid process integrating inductive data-driven codes with deductive theory-driven ones, drawing on social phenomenology to examine the role of performance feedback in self-assessment of nursing practice.4SAGE Journals / International Journal of Qualitative Methods. Demonstrating Rigor Using Thematic Analysis: A Hybrid Approach of Inductive and Deductive Coding and Theme Development The hybrid strategy acknowledges that researchers rarely approach data as blank slates, while still leaving room for the data to push back against prior expectations.
Descriptive Versus Interpretive Codes
Not all codes are doing the same kind of work. Descriptive codes summarize the explicit content of what someone said: if a healthcare worker mentions long shifts, a descriptive code might be “long working hours.” Interpretive codes go deeper, capturing underlying meanings and connections that are not stated outright. An interpretive code applied to the same data might be something like “emotional exhaustion as a barrier to effective care,” linking the surface-level complaint to a broader pattern of burnout affecting clinical performance.5Journal of Medicine, Surgery, and Public Health. Using thematic analysis in qualitative research
The distinction matters because an analysis that stays entirely at the descriptive level tends to produce themes that read more like topic summaries than genuine insights. Moving into interpretive territory is where thematic analysis earns its analytical power, but it also requires more confidence in your reading of the data and more transparency about how you arrived at a particular interpretation. This is one reason reflexive approaches insist so heavily on the researcher documenting their reasoning throughout the process.
What Separates Thematic Analysis From Content Analysis
Researchers new to qualitative methods often wonder how thematic analysis differs from content analysis, since both involve identifying patterns across text. The two approaches do share significant common ground: both cut across data to search for patterns and themes. The main practical difference is that content analysis allows for, and sometimes emphasizes, quantifying the data. In content analysis, measuring how frequently different categories appear can serve as a rough proxy for significance. Thematic analysis, by contrast, treats frequency as just one possible indicator of importance, and not necessarily the most telling one.6PubMed Central. Content analysis and thematic analysis: Implications for conducting a qualitative descriptive study
A theme that appears in only a handful of interviews can still be analytically significant if it captures something central to the research question. Conversely, a topic that every single participant mentions may not warrant its own theme if it does not add anything beyond what is already obvious. This is a genuine philosophical difference, not just a procedural one, and choosing between the two methods often comes down to whether your research question is better served by counting patterns or by interpreting them.
What Counts as Quality
Because thematic analysis is so flexible, questions about rigor can feel slippery. The markers of quality in a coding-reliability approach (inter-rater agreement, codebook transparency) do not straightforwardly apply to reflexive thematic analysis, and vice versa. This is a persistent source of confusion in peer review, where reviewers sometimes evaluate one type of thematic analysis using criteria designed for another.
A useful way to think about quality across the spectrum is the concept of “knowingness and methodological congruence.” In this view, what matters is not that you followed a particular version of thematic analysis correctly, as if it were a recipe, but that you practiced and reported your analysis in a way that shows you understood what you were doing and why your choices fit together. A researcher who deliberately mixes elements from different approaches and explains the reasoning is in better shape methodologically than one who follows a single approach rigidly but without understanding its assumptions.7PubMed Central. A critical review of the reporting of reflexive thematic analysis in Health Promotion International
This framing puts transparency at the center of quality. Readers of your research should be able to see not just what themes you found but how you arrived at them, what decisions you made along the way, and how your own position as a researcher may have shaped the analysis.
Common Pitfalls
Even experienced researchers stumble over some recurring problems. The most widespread is producing themes that are really just topic labels. “Communication,” “support,” or “barriers” are summaries of what the interview covered, not analytical themes. A good theme has a point of view: it tells the reader something specific about how participants experienced or made sense of the phenomenon. Themes that could be transplanted to almost any qualitative study without editing are a warning sign.
Another common error is treating each interview question as its own unit of analysis, generating themes that map directly onto the interview guide rather than cutting across it. This produces a descriptive summary of the interviews rather than an interpretation. Strong thematic analysis looks for connections between topics, for tensions and contradictions within participants’ accounts, and for meanings that participants may not have stated explicitly but that emerge from the data as a whole.
There is also the issue of what researchers sometimes call “data swamps”: collecting so much data that the analysis becomes unmanageable, or coding at such a fine-grained level that hundreds of codes proliferate without ever coalescing into coherent themes. The remedy is usually to step back and ask what story the data is telling, rather than trying to capture every detail at the code level.
Thematic Analysis in Health Services and Applied Research
Thematic analysis has become especially popular in health services research, where teams often include clinicians, policymakers, and social scientists who may not share a single methodological tradition. Its flexibility is a major advantage here: it can accommodate different disciplinary perspectives without requiring everyone on the team to subscribe to a particular philosophical framework. In health services contexts, thematic analysis has been described as a highly flexible approach that can draw out valuable findings and generate new hypotheses, including when there is little previous research to build on.8BMJ. Practical thematic analysis: a guide for multidisciplinary health services research teams engaging in qualitative analysis
This adaptability also means it is used across a huge range of applied fields: education research, user experience design, market research, organizational studies, social work, and environmental policy. In each of these, the core logic remains the same (code the data, look for patterns, build themes), but the standards of evidence and the expectations around transparency can vary considerably depending on the discipline and the audience.
Software Tools for Managing the Process
Qualitative data analysis software, sometimes called CAQDAS, has become a standard part of many researchers’ workflows. Programs like NVivo, ATLAS.ti, and MAXQDA are fundamentally data management packages: they help you organize transcripts, apply and retrieve codes, search across the dataset, and keep track of your analytical decisions.9PubMed Central. The Implication of Using NVivo Software in Qualitative Data Analysis: Evidence-Based Reflections They do not do the analysis for you, but they make the process of handling large datasets far more efficient than working with printed transcripts and highlighter pens.
The gap between what the software can do and what researchers actually use it for has been a persistent challenge. Many people learn just enough of a tool to store and code their data, without realizing that the software also supports more advanced analytical strategies like matrix queries, code co-occurrence analysis, and collaborative coding across team members. When analytic strategies are effectively translated into software features, the tools can facilitate not just data management but also deeper analysis and collaboration across research teams.10Sociological Research Online. Bridging the Gap Between Methodology and Qualitative Data Analysis Software: A Practical Guide for Educators and Qualitative Researchers
AI and Large Language Models in Thematic Analysis
The emergence of large language models has sparked real interest in whether thematic analysis can be partially automated. Early experiments show that AI tools can generate thematic structures that align with manual coding to a meaningful degree. One study using an open-source model with 70 billion parameters found that the themes produced by the model achieved moderate to substantial similarity to those produced by human researchers, with similarity coefficients ranging from 0.44 to 0.69 depending on the evaluation method.11PubMed. Inductive thematic analysis of healthcare qualitative interviews using open-source large language models: How does it compare to traditional methods? The model could summarize a full set of interviews in minutes rather than weeks.
That speed is appealing, but the limitations are real. AI-generated codes can be variable across runs, and the models do not explain their decision-making in a way that supports the kind of transparency qualitative research demands. In another study testing AI-powered thematic analysis across three datasets, the model produced consistent thematic structures, but researchers had to intervene repeatedly between steps to refine outputs and maintain methodological integrity.12PubMed Central. Utilizing AI-Powered Thematic Analysis: Methodology, Implementation, and Lessons Learned
Researchers who have used these tools recognize potential gains in efficiency and scalability, but they consistently flag risks around bias, loss of contextual nuance, and reproducibility problems stemming from the rapid evolution of the models themselves. The emerging consensus is that large language models can support interpretive analysis but cannot substitute for it. The need for “prompting literacy,” the skill of crafting instructions that get useful output from a model, is becoming its own methodological competence.13arXiv. LLM-Assisted Thematic Analysis: Opportunities, Limitations, and Recommendations
Working With Visual and Multimodal Data
Thematic analysis was originally developed for text, but researchers increasingly apply it to photographs, videos, social media posts, and combinations of text and images. This creates genuine analytical challenges. Most traditions in qualitative research were built around oral or written data, where words generally have a clear and widely shared meaning. Photographs are different: they carry both a public meaning made up of universally recognizable elements and a private meaning rooted in the photographer’s emotional experience, which can be difficult to express in words.14PubMed Central. Analysing multimodal data that have been collected using photovoice as a research method
Photovoice research, where participants take photographs to document their own experiences and then discuss them with researchers, illustrates the complexity well. The volume of data can vary enormously depending on how many photos participants take, how many group discussions are held, and whether researchers ask for long captions or short ones. There is no established consensus on what the optimal volume looks like, and the answer likely varies by study design. The photographs that were not taken, or that participants chose not to take, can be just as analytically important as the ones that were.
These challenges have prompted researchers to develop adapted frameworks for handling multimodal data, acknowledging that working with participant-generated images requires rethinking methodological aspects of both data generation and analysis.15The Qualitative Report. The Textual-Visual Thematic Analysis: A Framework to Analyze the Conjunction and Interaction of Visual and Textual Data The core logic of thematic analysis still applies: look for patterns of meaning, build codes, develop themes. But the interpretive process is more layered when the data includes images whose meanings are inherently ambiguous.
Ethical Dimensions That Shape the Analysis
Qualitative research in general raises ethical issues that go beyond the standard informed consent form, and thematic analysis is no exception. When you are working closely with people’s words, stories, and sometimes images, the potential for harm does not end when the interview wraps up. Anonymizing transcripts is more complicated than swapping names: a participant’s story may be identifiable from its details even with names removed, especially in small communities or specialized professional settings.
Researchers face ethical challenges at every stage, from designing the study and collecting data to the analysis and final reporting. These include maintaining genuine anonymity and confidentiality, managing the potential impact the researcher has on participants and the impact participants may have on the researcher, and navigating the power dynamics embedded in who gets to decide what the data “means.”16PubMed Central. Ethical challenges of researchers in qualitative studies: the necessity to develop a specific guideline In thematic analysis specifically, the ethical dimension is sharpened by the fact that the researcher is actively interpreting people’s experiences. Representing someone’s account in a theme they would not recognize, or stripping away context that changes the meaning, are not just analytical errors. They are ethical ones.
This is another area where the reflexive tradition’s emphasis on researcher positionality earns its keep. Being transparent about who you are, what assumptions you brought to the data, and how your perspective shaped the themes you constructed is not just a methodological nicety. It is a form of accountability to the people whose lives your analysis is about.