Purposive sampling (also called purposeful or judgement sampling) is a non-probability sampling method in which the researcher deliberately chooses participants because they have the characteristics, knowledge or experience the study needs. A review by Palinkas and colleagues describes it as ‘widely used in qualitative research for the identification and selection of information-rich cases’.
Because you choose who takes part, a purposive sample gives limited grounds for statistical claims about a wider population. What it can do is give you the people best placed to answer your research question. This guide covers the main types of purposive sampling, with a worked example of each, how it differs from convenience sampling, how two published studies justified their samples, and how to write up your own choice in a dissertation or thesis.
How purposive sampling works
You start from your research question, not from a list of the population. You decide what kind of person (or case, organisation or document) can tell you most about it, set criteria that describe them, and recruit people who meet those criteria.
The UK public health textbook HealthKnowledge sums up the trade-off: judgement sampling is ‘time-and cost-effective to perform’, but it is ‘prone to errors of judgement by the researcher’ and ‘will not necessarily be representative’. Your job in the methodology chapter is to show that your judgement was sound.
Types of purposive sampling
The table follows the objectives Palinkas and colleagues give for each strategy (their Table 1). To compare them on one problem, the examples imagine a master’s dissertation on how hospital pharmacists in England adapted to a new electronic prescribing system. The study is invented for this guide.
| Type | What it’s for | Worked example |
|---|---|---|
| Maximum variation | Finding shared patterns that cut across very different cases | Interview pharmacists from large teaching hospitals and small district hospitals, early-career and senior, in different regions: anything they all report is likely to be a core experience |
| Homogeneous | Describing one subgroup in depth, reducing variation and making group interviews easier | Run a focus group only with pharmacists in their first two years of practice |
| Typical case | Showing what is typical or average to readers unfamiliar with the setting | Choose a hospital whose roll-out followed the standard plan, with no special funding or delays |
| Extreme or deviant case | Learning from unusual cases, to light up both the unusual and the typical | Study the hospital that went live fastest and the one that abandoned its first attempt |
| Critical case | Reasoning that if something holds here, it’s likely to hold elsewhere | Pick the best-resourced hospital: if pharmacists there still struggled with the system, others are likely to have done too |
| Criterion | Selecting all cases that meet a set criterion of importance | Include every pharmacist who has used the system for at least six months |
| Snowball | Finding people with similar characteristics through those who know them | Ask each interviewee to pass your details to colleagues who raised safety concerns about the system |
You can combine types. For example, you might use a criterion (six months’ use of the system) to decide who is eligible, then maximum variation to choose among them.
Which type is most common? There’s no single answer across fields. In the 28 implementation research studies Palinkas and colleagues reviewed, 21 (75%) used some form of criterion sampling and six used some form of maximum variation sampling.
Purposive vs convenience sampling
Both are non-probability methods, and both can involve recruiting whoever responds to an advert. The difference is the reason someone is in the sample.
| Purposive sampling | Convenience sampling | |
|---|---|---|
| Who is selected | People chosen because they have characteristics the study needs (Zrineh and colleagues: ‘the deliberate selection of participants based on specific characteristics’) | People who are available and willing to take part (HealthKnowledge) |
| What decides it | Criteria set from the research question | Access |
| Main risk | Errors of judgement by the researcher (HealthKnowledge) | Results ‘prone to significant bias’ (HealthKnowledge) |
| How you defend it | Explain the logic linking your criteria to the question | Acknowledge it openly and limit your claims |
A recruitment advert doesn’t automatically make a sample convenience-based. In a 2020 paper of research case examples, Campbell and colleagues describe a planned study that recruits new mothers through a flyer at an antenatal clinic, which is partly convenient, but ‘the very specific nature of the criteria for involvement make this design purposive’.
A real UK example: moving from convenience to purposive
A 2023 BMJ Open study of GP well-being during the COVID-19 pandemic, by Jefferson, Heathcote and Bloor, shows the difference in practice:
- The sampling plan came first. The authors sampled NHS GPs purposively across three career stages: early-career (in the final stages of training and the first five years of practice), established, and late-career (including retired GPs who returned to practice during COVID-19). They also sampled for variation in ethnicity, age, gender, contract type and local area characteristics.
- The easy route produced plenty of volunteers. Their first recruitment post, on Twitter, was so successful that over 40 GPs offered to take part within 24 hours.
- They didn’t stop there. ‘In order to obtain maximum variation in participant characteristics and reduce the potential for bias’, they also recruited through the Yorkshire and Humber deanery, by snowballing through their networks of clinicians, through a Royal College of General Practitioners group for late-career and recently retired GPs, and by emailing participants in the GP Worklife Survey who had said they were willing to take part in research like this. Potential participants completed a brief survey giving contact details and basic demographic information.
- Result: 40 GPs were interviewed between March and June 2021, and the authors report that data saturation was reached.
The lesson for your own project: a sample made only of the first people to respond is a convenience sample. Setting criteria from your question, recruiting through more than one channel to reach the people your criteria call for, and checking volunteers against those criteria is what makes it purposive.
Two justifications from published studies
Your examiners will want to know why these participants could answer your question. Two short examples of how published authors put it:
- Criterion sampling, child and family health nurses (an Australian case example). In one of Campbell and colleagues’ case examples, the nurses had to have at least two years’ recent experience in the role, because ‘Participants are those who are most likely to have in-depth knowledge and experience of the phenomenon being studied’.
- Career-stage sampling, NHS GPs (UK). Jefferson and colleagues sampled across three career stages and recruited further ‘to obtain maximum variation in participant characteristics and reduce the potential for bias’.
Each names the criterion and links it to the research question in one sentence. That’s the standard to aim for.
How many participants do you need?
Palinkas and colleagues describe the aim in qualitative research as saturation: ‘continuing to sample until no new substantive information is acquired’. In practice, plan a range, recruit in stages, and keep a record of when interviews stop adding new codes or themes, so you can show how you decided to stop.
Your supervisor and your department’s past theses are a good guide to what is expected at your level.
Limitations to acknowledge
- Limited statistical generalisation. Zrineh and colleagues note that purposive sampling ‘limits the ability to make statistical inferences to broader populations’. Don’t report percentages from a purposive sample as if they described everyone.
- Your judgement shapes the sample. Say who you might have missed, for example people who don’t use social media or who declined to take part.
- Self-selection. Volunteers may differ from those who didn’t respond. Screening against your criteria helps, but doesn’t remove it.
Our guide to scope and delimitations explains how to set out what your study does and doesn’t cover.
How to describe purposive sampling in your methodology chapter
Campbell and colleagues advise researchers to ‘clearly, transparently and explicitly describe the logic used in selecting the sample subset’. A template you can adapt:
‘Participants were selected using purposive sampling, specifically [type, for example maximum variation] sampling, because the study aimed to [aim linked to the research question]. Eligibility criteria were [criteria], chosen because [reason]. Within the eligible group, I sought variation in [characteristics]. Participants were recruited through [channels] and screened using [method]. Recruitment continued until [stopping rule, for example no new codes appeared in the final three interviews], giving a sample of [n]. Because the sample was not selected at random, the findings are not intended to be statistically generalisable to [population]; instead they [what the findings offer].’
Your methodology chapter will also describe how you collected the data: see our guide to research instruments. For how qualitative designs fit alongside quantitative and mixed methods, see types of research. If part of your study needs a random sample of a large, spread-out population, our guide to multistage sampling explains one probability approach.
Common mistakes
- Calling any non-random sample purposive. If you recruited whoever replied and set no criteria, it’s a convenience sample. Say so.
- Naming the method without the type. ‘Purposive sampling’ alone tells the examiner little. Say which type and why.
- Criteria that don’t follow from the question. Each criterion should have a reason you can state in one sentence.
- Claiming representativeness. A purposive sample is chosen for information, not to mirror the population.
Next steps
Write your research question at the top of a page, then list the kinds of people who could answer it best and the criteria that would identify them. That list is the start of your sampling strategy. If you’re considering a research degree where you’d design studies like this, you can search current PhD opportunities.
