Until 31 December 2026 only: try Themera Pro free for 7 days (first subscription). Cancel before and you pay nothing.Try Pro now →
10 min readby Mathias

Interpretative Phenomenological Analysis Steps: How IPA Works, Case by Case

Interpretative phenomenological analysis (IPA) step by step: the double hermeneutic, small homogeneous samples, personal and group experiential themes, typical mistakes, and where AI helps.

IPAResearch methodsPhenomenology

Interpretative phenomenological analysis, usually shortened to IPA, asks how particular people make sense of a significant experience in their lives. It works with small samples, takes one case at a time, and only compares participants at the end. That makes it different from most coding-based methods, and it changes what software can reasonably do for you.

Where IPA comes from

IPA was developed by Jonathan Smith in the mid-1990s; the paper most often cited as its starting point appeared in Psychology & Health in 1996. It began in health psychology and has since spread to clinical, counselling and educational research and beyond. The standard reference is the book by Smith, Flowers and Larkin, "Interpretative Phenomenological Analysis: Theory, Method and Research", first published in 2009, with a second edition in 2022.

The method rests on three commitments. It is phenomenological, because it is concerned with lived experience as the participant describes it. It is hermeneutic, because that experience is only accessible through interpretation. And it is idiographic, because it stays with the particular case before it says anything general.

The double hermeneutic

IPA describes the researcher's position as a double hermeneutic: the participant is trying to make sense of what happened to them, and you are trying to make sense of the participant making sense of it. Your reading is therefore always one step removed, and the method asks you to be open about that rather than hide it.

Sampling and data collection

IPA samples are small and purposive. You look for participants for whom the topic is meaningful and who are reasonably homogeneous with respect to it, so that differences between them can be read against a shared background. The 2009 edition of the book gave a rough guide of three to six participants for undergraduate or master's projects and four to ten interviews for professional doctorates. These figures are orientation, not rules; depth per case matters more than the number of cases.

The typical data source is a one-to-one, semi-structured interview with a short guide and broad, open questions, so that the participant can set the agenda. Interviews are recorded and transcribed verbatim. Focus groups fit less well, because the analysis treats each person as a bounded case.

The steps of IPA

The second edition updated both the wording and some of the terminology. In practice the analysis runs roughly like this:

  1. Reading and re-reading. Immerse yourself in the first transcript, ideally while listening to the recording again.
  2. Exploratory notes. Comment on the text line by line: what is said (descriptive), how it is said (linguistic) and what it might mean (conceptual).
  3. Constructing experiential statements. Condense the transcript and your notes into short statements that capture what seems important about the participant's experience. In older texts this step is called developing emergent themes.
  4. Searching for connections. Group the experiential statements, look at which belong together, which contrast and which carry more weight than others.
  5. Naming personal experiential themes. Organise those clusters into personal experiential themes (PETs) for this one participant, usually in a table that links each theme to passages in the transcript.
  6. Moving to the next case. Repeat the process for each participant, and try to let each case speak on its own terms rather than reading it through the previous one.
  7. Developing group experiential themes. Only now compare across cases. Group experiential themes (GETs) show what participants share and how they differ within a shared theme.

Papers based on the first edition use terms such as superordinate and subordinate themes for the same kind of structure. If your supervisor or your sources use the older wording, say which edition you follow.

How IPA differs from thematic analysis

AspectIPAReflexive thematic analysis
Unit of analysisEach participant as a case, then across casesPatterns across the whole data set
FocusLived experience and sense-makingPatterned meaning; can address many kinds of question
SampleSmall, homogeneousVaries, often larger
TheoryCommitted to phenomenology, hermeneutics, idiographyTheoretically flexible

If your question is about patterns across many participants rather than the texture of individual experience, thematic analysis after Braun and Clarke is often the better fit.

Typical mistakes

Coding the whole data set at once. Pooling all transcripts and coding them together is thematic analysis with an IPA label. The case-by-case structure is what makes the method IPA.

Staying descriptive. A list of what participants said is not yet an analysis. The conceptual notes in step two are where interpretation starts, and the write-up should show it.

Losing the individual in the group themes. A good GET section shows convergence and divergence, with quotes from several participants, not one voice standing in for everyone.

Too large a sample. More interviews do not make IPA more rigorous; they usually make it shallower.

What software can and cannot do here

Be honest with yourself here: the core of IPA is slow, interpretive reading of one person's account, and that work cannot be handed to a tool. An AI-generated code list for a transcript is not the same as your exploratory notes, and it should not replace them.

Where software can help is around that core. Themera transcribes German and English interviews and pseudonymises names before analysis, including inflected German names. You can run a single transcript as its own analysis and use the result as a second reading to compare with your own notes: every code is backed by a verbatim quote, and quotes that are not exactly in the material are rejected automatically, so each suggestion can be traced to the text. The coverage indicator shows passages that received no code, which can point to something you or the tool skipped. You can edit the codebook and re-code, and export results as DOCX or CSV to keep a record next to your PET tables. Data are stored and processed in the EU and not used for training. Which themes matter, and what they mean for this person, remain your decisions. For the general mechanics of coding, see how to code qualitative data. If you use any AI support, state it in your methods section, and check data protection questions with your supervisor or your university's data protection office.

Wording for your methods section

A template with placeholders, not a description of a real study. Adapt it to your procedure, and see writing the methods section for interview studies for the wider structure.

Data were analysed using interpretative phenomenological analysis (Smith, Flowers & Larkin, 2022). [Number] participants were recruited purposively for [shared characteristic] and interviewed using a semi-structured guide. Interviews were recorded, transcribed verbatim and pseudonymised. Each transcript was analysed as a separate case: after repeated reading, exploratory notes were made, experiential statements were constructed and clustered into personal experiential themes. After all cases had been analysed, group experiential themes were developed by comparing personal experiential themes across participants. [If applicable: An AI-assisted tool was used for transcription and as a secondary reading of individual transcripts; all interpretive decisions were made by the researcher.]

In short

IPA suits questions about how a small, well-defined group of people experience something that matters to them. Work one case at a time, move from description to interpretation, build personal themes before group themes, and be transparent about your own role in the reading.

See a sample analysis with codebook, verbatim quotes and the coverage indicator, or try it with your own transcripts: three analyses free, no card needed.

Next step

A codebook and verbatim evidence from your interviews, in minutes

Themera codes interviews and open-ended responses following Kuckartz, Mayring or Braun & Clarke. Every code is tied to a verbatim quote, and anything that could not be coded stays visible.

  • Transcription and pseudonymisation before analysis
  • Edit the codebook and re-run
  • 3 analyses free, no card

Pricing

  • Free€0

    3 analyses, 30 transcription minutes

  • Thesis€69

    once, 90 days, no subscription

  • Solo€39

    per month, cancel any month

Stored and processed in the EU (Mistral AI, Paris). No training on your data, DPA available.

Free, no sign-up

Codebook generator →

Enter your research question, get categories with coding rules and anchor examples.

Free, no sign-up

Cohen's kappa calculator →

Agreement, chance agreement and confidence interval, with an APA-style sentence.