A research interview is a structured or semi-structured conversation, conducted by a researcher with a participant, in which the participant’s spoken responses are recorded (by audio, video, or note-taking) and treated as data for the study. What separates it from an ordinary conversation is that it is planned in advance around a research question, uses a prepared set of questions or topics (a guide or schedule), and follows a documented protocol for consent, recording, and analysis. Interviews can generate qualitative data (open-ended narrative), quantitative data (fixed-response answers coded numerically), or both, depending on how much the format is structured.
This page is a general orientation to interviews as a method: the three structure types, when interviews are the right choice for a research question, how many participants you need, how interviews compare to surveys, and what ethics and consent require. For the mechanics of building a semi-structured interview guide, see CASRAI’s guide to designing semi-structured interviews. For what happens to the transcripts afterward, see the guide to coding qualitative interview data.
The three interview structure types
Interviews sit on a spectrum from fully scripted to fully open. Where a given interview sits on that spectrum determines what kind of data it produces and how comparable that data is across participants.
| Type | How it works | Data produced | Best suited to |
|---|---|---|---|
| Structured interview | A fixed set of questions, asked in a fixed order and fixed wording, with little or no deviation. Functions much like a spoken questionnaire. | Short, often closed or categorical answers; easy to code and compare across a large sample. | Studies that need standardized, statistically comparable answers from many participants — often layered on top of quantitative survey work. |
| Semi-structured interview | A prepared guide of open-ended questions and topics, but the interviewer can reorder, probe unexpected answers, or drop questions that no longer fit the conversation. | Rich narrative data that is still comparable across participants because everyone is asked about the same core topics. | The large majority of qualitative studies: exploring how people understand, experience, or make sense of something, while keeping enough structure to analyze across cases. |
| Unstructured (narrative) interview | No predetermined question set — the interviewer opens with a broad prompt and follows the participant’s own account, intervening minimally. | Long, participant-led narratives; least comparable across participants, richest for a single case or life history. | Life-history, narrative inquiry, and exploratory work early in a research program, where the researcher does not yet know what the relevant questions are. |
Most published qualitative studies that describe their method as “interviews” are describing the semi-structured form — it is the default because it balances the two competing pressures on interview data: flexibility to follow what the participant actually says, and enough consistency across participants to analyze the data as a set. See CASRAI’s overview of qualitative data collection techniques for how interviews sit alongside focus groups, observation, and document analysis as a family of methods.
When an interview is the right method for a research question
Interviews are not the default choice for every study — they are the right tool for a specific kind of research question. Interviews are generally well suited when the research question asks:
- How people understand, interpret, or make meaning of an experience — for example, how clinicians describe the experience of delivering a difficult diagnosis, or how first-generation students describe navigating a funding application.
- About a process, decision, or experience that cannot be observed directly — internal reasoning, past events, private decisions, or sensitive topics that a participant will only disclose in a private, one-to-one setting.
- For depth on a small number of cases rather than breadth across a large sample — interviews are labor-intensive to conduct, transcribe, and code, which caps the realistic sample size well below what a survey can reach.
- Where the researcher needs to probe and follow up — a survey question that returns an ambiguous or surprising answer cannot be clarified after the fact; an interview question can be probed in the moment.
Interviews are generally a poor fit when the goal is to measure the prevalence, frequency, or statistical distribution of something across a population — that is a survey or structured-instrument question, not an interview question (see the comparison below). Interviews also assume participants are able and willing to articulate their experience verbally; where that assumption doesn’t hold (young children, some cognitive or language barriers, highly sensitive topics), researchers often substitute or supplement interviews with observation, diaries, or other data-collection techniques.
Sampling and saturation: how many interviews is enough?
Interview studies almost never use probability sampling of the kind a survey uses. Instead, most qualitative interview studies use purposive sampling — deliberately selecting participants because they hold the experience, role, or characteristic the research question is about, not because they represent a random cross-section of a population. Related non-probability approaches include snowball sampling (existing participants refer new ones) and criterion sampling (participants must meet a defined set of characteristics). See CASRAI’s overview of probability and non-probability sampling methods for how these approaches differ from the random and stratified sampling used in quantitative work.
Sample size in interview studies is typically justified by saturation rather than by a pre-calculated number: researchers keep interviewing until additional interviews stop producing new themes or codes relevant to the research question, at which point the data set is considered to have reached saturation. In practice this is judged during ongoing analysis rather than fixed in advance, though many funders and ethics committees now expect a proposal to state an anticipated range and the criteria that will be used to judge when saturation has been reached, since “we will interview until saturation” with no stated basis is a common point of reviewer pushback. See CASRAI’s Grounded Theory term for the methodological tradition, associated with Glaser and Strauss, in which theoretical saturation was first formalized as a stopping criterion for data collection.
Interviews vs. surveys
Interviews and surveys are often confused because both involve asking people questions, but they serve different purposes and produce different kinds of data.
| Dimension | Interview | Survey / questionnaire |
|---|---|---|
| Typical sample size | Small (commonly single digits to a few dozen) | Large (commonly dozens to thousands) |
| Question delivery | Spoken, one-to-one (or occasionally group), interviewer present | Written, usually self-administered, no interviewer present |
| Flexibility during data collection | High in semi-structured/unstructured forms — can probe and follow up | Fixed — every respondent sees the same instrument, cannot be adjusted mid-collection |
| Typical analysis | Qualitative coding and thematic analysis of transcripts | Statistical analysis of coded responses (frequencies, correlations, etc.) |
| Best for | Depth, meaning, process, sensitive or complex topics | Breadth, prevalence, comparison across a large population |
The two are not mutually exclusive: many studies use both in a mixed-methods design — for example, a survey to establish prevalence across a large sample, followed by interviews with a purposively selected subset to explain the patterns the survey found, or vice versa. See CASRAI’s Research Questionnaire term for how questionnaire design differs from interview-guide design.
Strengths and limitations of interviews as a method
Strengths
- Access to depth and nuance a fixed-response instrument cannot capture — participants can explain reasoning, context, and contradiction in their own words.
- Ability to probe: an ambiguous or unexpectedly interesting answer can be followed up in the moment, which is not possible once a survey has been distributed.
- Well suited to sensitive, complex, or poorly understood topics where a researcher does not yet know what the relevant response categories even are.
- Can surface findings the researcher did not anticipate, since the participant — not a fixed answer list — sets the boundaries of what can be said.
Limitations
- Labor-intensive: scheduling, conducting, transcribing, and coding interviews takes substantially more researcher time per participant than a survey response.
- Small samples limit statistical generalization to a wider population; interview findings are typically presented as analytically or theoretically generalizable, not statistically representative.
- Interviewer effects: the interviewer’s presence, tone, follow-up choices, and even perceived identity can shape what a participant chooses to disclose — a source of bias that has no equivalent in a self-administered survey.
- Social desirability bias: participants may answer in ways they believe are expected or acceptable, particularly on sensitive topics, more so face-to-face than on an anonymous written form.
- Recall and reconstruction: interviews about past events rely on the participant’s memory and retrospective account, which is not the same as a contemporaneous record.
Ethics and consent
Because interviews involve direct, often recorded, interaction with identifiable human participants, they are human subjects research in essentially every institutional and regulatory framework and require review by an institutional review board (IRB) or research ethics committee before data collection begins. At minimum, an interview protocol submitted for ethics review typically needs to address:
- Informed consent — participants must understand what participation involves (topics, time commitment, recording), how their data will be stored and used, and that participation is voluntary and can be withdrawn at any point, including after the interview has taken place. See CASRAI’s Informed Consent term for what a consent process needs to cover.
- Recording and storage — whether audio/video recording is used, how recordings and transcripts are stored and secured, who has access, and how long they are retained.
- De-identification — how names, places, and other identifying details are handled in transcripts and in any published excerpts, particularly important in small purposive samples where participants may be identifiable from their role or context even without a name attached.
- Risk of disclosure — interviews on sensitive topics (health, trauma, illegal activity, workplace conflict) carry a real risk of participant distress or disclosure of information the researcher is obligated to act on (e.g., safeguarding concerns); protocols typically need a plan for handling this before fieldwork starts, not improvised in the moment.
These requirements apply to the interview itself; what happens to the resulting transcripts during coding and analysis is covered separately in the guide to coding qualitative interview data, including de-identification practice at the transcript stage and inter-coder reliability.
Frequently asked questions
What is an interview in research?
A research interview is a planned, recorded conversation between a researcher and a participant, structured around a prepared set of questions or topics tied to a specific research question, and treated as data for that study — distinct from an informal conversation because it follows a documented protocol for consent, recording, and analysis.
What are the main types of research interviews?
Three: structured (fixed questions, fixed order, minimal deviation), semi-structured (a prepared guide followed flexibly, with probing and reordering allowed), and unstructured or narrative (no predetermined question set, participant-led). Semi-structured is the most common form in qualitative research.
Are interviews qualitative or quantitative?
Usually qualitative, but not exclusively. Structured interviews with closed or numerically coded response options can produce quantitative data, and many mixed-methods studies use interviews specifically to generate qualitative data that complements a quantitative survey component.
How many participants do I need for an interview study?
There is no fixed number that applies universally. Most qualitative interview studies justify sample size by saturation — continuing to interview until additional interviews stop producing new themes relevant to the research question — rather than by a pre-calculated target, though proposals are increasingly expected to state an anticipated range and the criteria that will be used to judge saturation.
Should I use interviews or a survey?
Use interviews when the research question is about depth, meaning, or process on a small number of cases, and when you may need to probe an answer to understand it. Use a survey when the research question is about prevalence, frequency, or comparison across a large sample, and a fixed set of response categories is sufficient. Many studies combine both.
For the practical mechanics of building an interview guide, see the guide to designing semi-structured interviews. For how interview transcripts are analyzed once collected, see the guide to coding qualitative interview data. For the broader family of qualitative methods interviews belong to, see CASRAI’s Qualitative Research term.
