How many people do I need in my qualitative sample?
In this blog, I confront the thorny issue of qualitative sampling. I introduce the concepts of initial, ongoing and final samples and argue we should assess sample fitness rather than sample size.
Interpretive sampling
First, qualitative sampling is interpretive sampling. Qualitative research is designed to interpret or to make sense of other people’s worlds, and to render these meaningful through writing or other forms of communicating. A qualitative sample should include whoever and whatever will enable the researchers to fully explicate the phenomenon and its relevant dimensions, aspects and conditions.
An interpretive sample will be iterative-inductive, by which I mean it constantly moves back and forth between being inductive (open-minded, bottom-up, learning as you go) and deductive (testing out existing and new theories and concepts to see if they help clarify developing understandings). This iterative-inductive approach can be achieved through the techniques of initial, ongoing and final samples.
In qualitative research we usually aim to represent types of people or behaviours, or a range or diversity of experiences or perceptions, as these relate to the issue we are addressing and in the context of certain conditions. However, there is a good chance that we do not know at the outset exactly what dimensions will be meaningful, what diversity of responses will be relevant, or what conditions shape what outcomes.
The initial sample
Qualitative researchers rarely just hang around to see what turns up. You need some idea of how to start, who to access, what to ask and so on. Also, those giving funds, permission, or access (including the participants) will need to know how you intend to start. The most logical and practical way to proceed is to design an initial sample based on what you think you know, using your review of the literature and other expertise.
An interpretive sample must, of course, also link overtly and clearly back to what was described in the design of the research. The initial sample is best designed purposefully to include groups, individuals, or other characteristics that can logically or thematically be taken to be representative of the population you are interested in (parents of young children, pregnant mothers, with disabilities, for example). This sample is representative of the wider population conceptually, not numerically – it reflects certain qualities rather than quantities.
But this initial sample may not remain adequate, because when you start you don’t know everything you need to know!
Ongoing samples
Interpretive sampling continues as the research proceeds, through ongoing assessment of the relevance of data for your eventual findings and, where possible, making adjustments to who and what are included.
Ongoing sampling involves returning to (more or the same) people, to examine emergent ideas, new topics, developing insights, and reviewing what and who is included as analysis proceeds. When I design a project or proposal, I explain to readers why ongoing sampling will be crucial and how it will be practically achieved, perhaps through convenience or snowball sampling, perhaps through longitudinal research or ongoing relationships with participants. Perhaps simply by adding to the sample or the questions as I go along. Sometimes the best you can do is be flexible with what you talk about in interviews, but even that allows you to incorporate relevant details that arise in the moment.
Final samples
The final sample is the people, groups, contexts, and other variables and topics you finally included; they reflect required relevant diversity for specific outputs. It is important that the sample you end up with gives you confidence to say what you want to say, about whom, when.
What is sample fitness?
The best way to address sample adequacy is to look at sample fitness. Qualitative sampling is a matter of quality not quantity. So, if someone asks what size your sample was, it is ok to answer with numbers but also point out that it is the quality of the sample that matters most. We need to encourage the question: does the sample fit?
Initial sample fitness
Initial sample fitness asks: does the initial sample fit with the aims and objectives of the project? Here you will need to estimate size for costings and time but the number or size of the initial sample is far less relevant in terms of fitness than whether you have included the breadth of people (and other variables) to at least begin to address the aims and objectives. An initial sample should fit with:
- the proposed methodology and methods,
- the role of qualitative research in the entire project,
- the location of the work in the literature and prior knowledge,
- relevant practical and design issues, and
- how rich and exploratory the research aims to be
Ongoing sample fitness
Ongoing sample fitness asks how your continued selections and choices, including what methods to use, fit with the interpretivist logic of learning as you go. Objectivity in qualitative research means being responsive and reflexive.
Final sample fitness
Final sample fitness asks: does my final sample fit my conclusions? Do I have a good quality sample with which to talk about this with confidence? Does my sample or selection fit my claims? Final sample fitness is best assessed in relation to different outputs or publications. Small samples do allow you to generalise – it is just that you are generalisng thematically and meaningfully not statistically.
Conclusion
Qualitative sampling is interpretive, balancing deductive and inductive approaches in its efforts to understand and communicate complexity. It is essential to learn from our participants as we go along. Qualitative research is specifically designed flexibly and reflexively to achieve this goal.
Qualitative samples thus need to be drawn and redrawn as initial, ongoing and final samples. The quality of these samples should then be assessed using the concept of fitness.
Initial sample fitness should address fitness for the design stage of a project, ongoing sample fitness should be addressed in relation to the iterative-inductive design of the research, and final sample fitness should be assessed in terms of the claims made in final reports and other dissemination.
I go into all of this in much more depth, with examples, in my new book Qualitative Research Methods for Everyone: An Essential Toolkit (Policy Press). You might also enjoy the book’s accompanying Podcast series. Thanks for reading, and happy sampling! Please feel free to comment.
Author Bio: Karen O'Reilly is Emeritus Professor of Sociology at Loughborough University and Professor II at the Norwegian University of Science and Technology (NTNU). She is widely recognized for her expertise in ethnographic and qualitative research methods, which she has taught for over 30 years. Her research primarily focuses on migration, social theory, and community studies. Some of her most well-known books include: Qualitative Research Methods for Everyone (Policy Press), Ethnographic Methods (Routledge), Key Concepts in Ethnography (Sage), and International Migration and Social Theory (Palgrave).