Semi-structured interviews for qualitative PhD research

Structured vs semi-structured vs unstructured interviews: which one fits a PhD?

A structured interview is actually a questionnaire which you read aloud (our questionnaire design guide covers that). Many scholars think an unstructured interview will be easy to defend at the DC meeting. Mostly it is harder than they expect. The committee will ask how you will compare one participant with another, and saying “I will see what comes up” will not satisfy them.

That is why most scholars end up choosing semi-structured interviews. If you are not sure interviews belong in your design, sort that first using our guide to research design types.

How do you write an interview guide for qualitative research?

Begin with your research questions. Write each one on top of a separate sheet. Below it, list what a participant must tell you so you can answer that research question. Those points become your main questions.

A good interview guide works in three layers.

• Main questions are open questions with one idea each, in the words people use at home. For 45 minutes to an hour, six to ten questions are usually plenty.
• Probes ask for more detail. Just “Can you give me an example?” or “What happened next?” is enough.
• Follow-ups are sub-questions. You ask them only if the participant doesn’t raise the point.

Keep the easy, factual questions for the start. Sensitive questions go in the middle, after the person trusts you a little. In the end, ask if there was anything they thought you would ask and you didn’t. Many times the most useful material comes out at that point only.

A short illustrative interview guide

Here is part of a guide for an imaginary study on how government school teachers in rural Tamil Nadu use a state teaching app. The research question is “What makes teachers continue or stop using the app?”

1. Tell me about the first time you used the app in a class. Probe: Which class? What were you trying to do?
2. Walk me through the last time you used it. Probe: That day, any problem? Or it went okay? Follow-up (if not raised): Was the network or the device a problem that day?
3. Have there been weeks when you didn’t use it at all? What was going on then?
4. Who else in the school has a view on the app? Probe: How does the headmaster’s view affect what you do?
5. Suppose a new teacher joins your school next week. What will you tell her about the app?

Notice the questions that are not there. The guide does not ask “Do you find the app useful?”, because a teacher will just give a polite yes. It also leaves out “Don’t you think the training was too short?”, since it hands the teacher your own answer.

How many interviews are enough for a PhD study?

There is no correct number. You need a number which you can justify.

A widely cited paper is Guest, Bunce and Johnson’s “How many interviews are enough?”, published in Field Methods in 2006. They did 60 in-depth interviews with women in two West African countries. Saturation means the point after which new interviews give no new themes. In their data it came within the first 12 interviews, and the basic elements were already there by six. Many scholars are quoting “12 interviews” as if it is a rule. But it came from one dataset only. With a mixed sample and a wide topic, 12 may not be enough for you.

Hennink, Kaiser and Marconi (2017, Qualitative Health Research) divided saturation into two kinds. In their 25 interviews, “code saturation”, where you have heard the range of issues, came at nine. “Meaning saturation”, where you understand those issues in depth, took 16 to 24.

Malterud, Siersma and Guassora (2016) suggested information power in its place. So when each participant knows a lot about your exact question, fewer participants will do. It depends on five things, namely the aim of the study, how specific the sample is, whether you use established theory, the quality of the dialogue, and the analysis strategy. Twelve district officials who actually ran a scheme may give you more than forty beneficiaries picked loosely. Braun and Clarke go further for reflexive thematic analysis. As per them, saturation is not a fit for that method at all. Their reason is in our thematic analysis guide.

In the synopsis, give a planned range with your reasoning, for example 15 to 25. In the thesis, explain plainly why you stopped where you did. Recruitment is a sampling decision also, so name your method. Our sampling techniques guide covers purposive, snowball and theoretical sampling.

What should you prepare before the first interview?

Ethics approval and consent

Don’t contact any participant till your ethics committee has given approval. It will usually ask for your interview guide, the participant information sheet and the consent form, so prepare these drafts early. The full process is in our guide to research ethics approval. Give participant these papers in a language they can read. Agreeing to take part and agreeing to be recorded are two separate consents. Some will talk but not want the recorder on. For them, take detailed notes and write them up that same day. For participants who cannot read, ask your ethics committee before going to the field which form of consent it accepts.

Language

Fix the language of each interview in advance. If your thesis is in English and the interviews are in Hindi or Telugu, translate the interview guide and get another person to translate it back to English. If the back-translation comes out different from your English, most probably the question asks something you did not mean. For “stress” or “harassment” many languages have no single matching word. And many participants will just use the English word anyway.

Pilot interviews

Do two or three pilot interviews with people who are like your participants but will not be in the final sample. In the pilot recordings, look for one-line answers and questions people took wrongly. Also note where you yourself talked too much. Each pilot show you something to fix, and almost every interview guide changes after the pilots. Writing about those changes in the thesis counts in your favour.

How do you conduct a semi-structured interview well?

Most of interviewing is listening. Play back your recording. If you hear yourself explaining your study, completing their sentences or agreeing loudly, then you are shaping the data.

• Before switching on the recorder, spend a few minutes on tea and small talk. Then confirm consent on the recording itself.
• Probes should not lean to any side. “Tell me more about that” is fine. Saying “That must have been frustrating” puts your words in their mouth.
• If the person goes quiet, let them be. Mostly the real point comes after that pause.
• Ask about one actual event (“the last time…”) instead of their general opinion, since people find it harder to polish a story.

Officials, elites and power

With a Block Development Officer or a hospital superintendent you may get twenty minutes, a gatekeeper in the room, and the official line. So carry a short guide with only the questions you can’t drop. If you know the scheme in detail they can’t talk around you. Keep asking for examples. With contract workers, village women or students from your own department, power goes the opposite way. They may say what they think you want to hear, more so if you reached them through their employer. Tell them clearly there are no right answers and nothing they say goes to their boss.

Phone and video interviews

Most ethics committees allow phone and video interviews. They tend to be shorter and give thinner data, and they work best with people you have met before. Also check that where the video platform keeps recordings matches your ethics approval.

How should you record, store and transcribe interviews?

A separate voice recorder is better. If you only have your phone, keep it in flight mode. The same day, copy the files off the device into encrypted or university storage. Never put them in a WhatsApp chat. The key linking codes (P01, P02) to names should stay in a separate protected file. India’s Digital Personal Data Protection Act, 2023 may apply to the personal data of participants, so ask your ethics committee or research cell what your university expects.

Do the anonymising while you transcribe. In a small place, even “the only woman medical officer in the block” tells everyone who she is, same as her name.

Verbatim or clean transcription?

In verbatim you type every “ah” and “matlab” also, and the repetitions, false starts, laughter and pauses. A clean (intelligent verbatim) transcript removes fillers and false starts, but the words and the meaning stay. For thematic analysis clean is usually enough. Discourse or conversation analysis needs full verbatim. Write down rules for pauses, inaudible words and overlapping speech, and use them in every transcript.

Code-switching and translation

Many participants switch between English and their own language in middle of a sentence. Transcribe in the language they spoke, either in its own script or in one consistent romanisation. Full translation of every transcript is not a must. Only the quoted extracts will also do. Either way, keep the original next to every translated quote in your files.

How long it takes, and automatic tools

One hour of clear speech usually takes several hours to transcribe by hand. Talk that keeps jumping between two languages, or a noisy recording, eats up even more hours. Plan the timeline with that. For a rough draft you can try automatic transcription. How correct it comes out on Indian languages and mixed speech varies a lot. So sit with the audio and check each line. One more check before uploading. Where does the tool keep the recordings, and does your consent form allow that?

What should you report about interviews in your methodology chapter?

Examiners use this section to decide if they can trust your findings chapters. At the least, cover these points.

• The reason semi-structured interviews suit your research questions.
• Who the participants were and how you found them (sampling method, gatekeepers). Also how many you approached and how many agreed.
• How you decided that the number of interviews was enough.
• Where each interview took place (office, home, phone, video), who else was there, and how long it went on.
• The language of each interview, who did the translation and transcription, and how you checked the translation.
• How you took consent, the ethics approval reference, and the way you anonymised the data.
• What changed after the pilot. Keep the final interview guide in an appendix.
• Your own position. That means your relationship with participants and anything about you (gender, caste, institution, job) which may have shaped their answers.

If you want to check yourself more fully, go through COREQ, the Consolidated criteria for reporting qualitative research. Tong, Sainsbury and Craig published it in the International Journal for Quality in Health Care in 2007. It has 32 items in three domains, namely research team and reflexivity, study design, and data analysis and reporting. COREQ was made for health research, but it works as a self-check in other fields also. This section is one part of that chapter, and our guide on writing the research methodology chapter covers the rest.