A sample is an argument, and the strategy's name is the least of it

Every qualitative study begins with a decision that later becomes invisible: who was asked. It is usually made from whoever is to hand. This first of ten units argues that a defensible sample instead follows from the research question and the information content of the case. The nine that follow take each later decision in turn. A strategy's name is worth little unless it names the alternative it rules out.

Availability is not a strategy, and it has a name

Etikan, Musa and Alkassim (2016) compare convenience sampling, which recruits whoever is reachable, with purposive sampling, which selects against criteria derived from the question. Their conclusion is permissive: the two are not ranked. Non-probability sampling earns its place when randomisation is impossible, resources are short and the study does not aim to generalise. Convenience sampling survives only in that last case.

Palinkas and colleagues (2015) give the replacement a shape. Their account sets out the categories of purposeful sampling, from criterion sampling, which takes everyone meeting a stated condition, to maximum variation, snowball and theory-based selection, and reports criterion sampling as the strategy used most commonly in implementation research.

Purposeful selection asks for information, not for a mirror of the population

Patton (2015) states the logic: the power of purposeful sampling lies in selecting information-rich cases for in-depth study. Maximum variation sampling looks for cases that differ, so a shared pattern has to survive the heterogeneity. Extreme case sampling studies the unusual because it illuminates the ordinary.

Gobo (2004) makes the standard explicit. A statistical sample is judged by how well it mirrors a population. A qualitative sample is judged by what it can teach, so Gobo argues that qualitative work has its own logic of social representativeness. Generalisation travels as transferability, through the fit between the setting studied and the reader's own.

The strategy has to name the alternative it rules out

Robinson (2014) turns the argument into a procedure with four parts. Define the sample universe by stating who is included and excluded. Decide the size with the study's epistemological and practical concerns. Choose a strategy, and source the sample. The claim is about transparency as much as technique. How far these concerns are made explicit bears on a study's coherence and trustworthiness. The exclusion criteria matter as much: a study run only in English misses a reader whose first language is another.

Coyne (1997) shows why the naming matters. Purposeful and theoretical sampling are used as synonyms, yet purposeful sampling selects information-rich cases while theoretical sampling selects cases to develop an emerging theory. A name does work only if it tells the reader which alternative the design rejected.

Selection is a judgement about who can answer

Tongco (2007) defends purposive sampling on exactly the ground a random draw would discard. Reliability and competence of the informant govern selection, and a purposive sample can be efficient because of the bias a probability sample treats as a defect.

Fix the problem and the information needed, then define the informant's qualities. Find informants by those qualities rather than by a list. Bah and colleagues (2006) asked a village head to name the healers, then interviewed forty of them across 21 villages.

That warrant must be rebuilt for populations that are not homogeneous. Mohammadi, Konig and Zimmermann (2026) reflect on selecting interviewees from more than 1,200 named experts for the European Research Council, where relevance varies with position. Their answer is a multi-layered, reflexive approach using stratified random sampling to reduce bias and to describe the structure of the expert population itself.

A referral chain is evidence, not only a route

Noy (2008) re-reads snowball sampling through constructivist and feminist hermeneutics and changes the technique's status. The chain of referrals is not merely a route to hard-to-reach people. It generates a social knowledge that is emergent, political and interactional. In her studies of backpacker tourists and marginalised men, the pattern of who referred whom expressed power relations and social capital, so the sampling method is part of the finding.

Heckathorn (2011) supplies the counterweight. He distinguishes snowball sampling as a convenience route from respondent-driven sampling, a method with an explicit weighting model that can yield population estimates under stated assumptions. Noy lets the chain speak about social structure; Heckathorn insists that a population claim needs the arithmetic.

The sample decides the reach of the claim

Suri (2011) examines how purposeful sampling is used in a qualitative synthesis, working through the adaptability of Patton's 16 strategies to combining published studies, and finds the literature says little about how the choice is made. The sampling decisions of the primary studies condition what a synthesis can conclude.

Luborsky and Rubinstein (1995) name the same gap from the other end. Writing in ageing research, they observe that the sampling frame is usually left implicit, and propose qualitative clarity, the qualitative analogue of statistical power, to guide the assessment of a sample. A stated rationale should tell a reader what a qualitative sample could have shown.

Hagaman and Wutich (2017) counted. They analysed a cross-cultural study with 132 respondents in four sites. Where a site's group was relatively homogeneous, 16 or fewer interviews identified common themes. Metathemes cutting across all sites needed 20 to 40 to saturate.

Report the sample you obtained, not the sample you planned

Campbell and colleagues (2020) show how a rationale is written. Across three case studies they demonstrate that purposive sampling is more complex than its label suggests, because criteria and purpose shift with design, and conclude that making the approach explicit improves rigour.

The three cases differ in what they can report. The first, a co-led redesign of stroke services in North West Tasmania, let a stakeholder group name the categories. It interviewed half the patients aged over 65 and half aged 65 or under. The second took child and family health nurses with at least two years of recent experience. The third, a study of mental wellbeing in the first year after a first birth, was written before recruitment began, so it reports criteria rather than a sample. The difference between them is the stage of the study rather than the quality of the work.

Contested question: is a qualitative sample judged by representation or by information?

The literature agrees that statistical representativeness is the wrong standard for a qualitative sample, and disagrees about what replaces it. One position holds that the standard is the information a case carries. Patton (2015) states it as the logic of information-rich cases. Gobo (2004) formalises it as social representativeness and transferability. Tongco (2007) locates it in the competence of the informant.

A second holds that the standard is the procedure. A well-argued sample is one whose decisions can be checked. Robinson (2014) sets out four concerns that must be met and made explicit; Mohammadi and colleagues (2026) demonstrate a defensible procedure for an elite population; Campbell and colleagues (2020) tie explicitness to trustworthiness.

A third splits the two camps on referral chains: Noy (2008) treats the chain as knowledge about social structure, while Heckathorn (2011) treats it as a convenience route unless a weighting model is attached. They differ on whether the chain's information can become a population claim without arithmetic.

What would settle it is evidence the literature does not yet contain. A study would pose one question to two samples drawn on the competing logics, fix the frame in advance and record what was obtained. It would then report whether the designs support different conclusions, and at what cost.

The evidence differs in kind. The informational standard leans on a textbook and a collection chapter (Patton, 2015; Gobo, 2004). The procedural standard leans on a guide and three case studies (Robinson, 2014; Campbell et al., 2020). Only Heckathorn (2011) attaches arithmetic, and only for one estimator. While the dispute is open, a designer should report the frame obtained rather than the frame planned, and name the alternative the strategy rules out. The strongest objection is that a procedural standard can be met in full by a sample that answers nothing about the population.

Where the argument is heading

The direction of travel is towards making the sampling frame and the rejected alternative reportable. Mohammadi, Konig and Zimmermann (2026) turn elite sampling into a reflexive, stratified procedure; Noy (2008) pushes the referral chain into the analysis as data.

The sources this unit rests on

• Bah, S., Diallo, D., Dembele, F. and colleagues (2006). Ethnopharmacological survey of plants used for the treatment of schistosomiasis in Niono District, Mali. Journal of Ethnopharmacology, 105(3), 387-399. doi:10.1016/j.jep.2005.11.026 In this primary study, forty healers from 21 villages were interviewed, the sample reached by asking a village head to name them. It reports plant use rather than the sampling decision, so the warrant for the procedure is indirect.
• Campbell, S., Greenwood, M., Prior, S., Shearer, T., Walkem, K., Young, S., Bywaters, D. and Walker, K. (2020). Purposive sampling: complex or simple? Research case examples. Journal of Research in Nursing. doi:10.1177/1744987120927206 Three case studies show that purposive sampling is more complex than its label suggests, because the criteria and purpose shift with the design, and judge the result by four aspects of trustworthiness. Supports the unit's claim that rigour comes from the rationale.
• Coyne, I. T. (1997). Sampling in qualitative research. Purposeful and theoretical sampling; merging or clear boundaries? Journal of Advanced Nursing. doi:10.1046/j.1365-2648.1997.t01-25-00999.x Argues that purposeful and theoretical sampling are routinely used as synonyms although they answer different questions, so conflating them weakens both the rationale and the theory. Gives the unit its clearest case of a strategy's name doing real work.
• Etikan, I., Musa, S. A. and Alkassim, R. S. (2016). Comparison of convenience sampling and purposive sampling. American Journal of Theoretical and Applied Statistics. doi:10.11648/j.ajtas.20160501.11 Compares the two techniques and declines to rank them, concluding that non-probability sampling earns its place when randomisation is impossible, resources are limited and the study does not aim to generalise. Names the default the unit asks students to replace.
• Gobo, G. (2004). Sampling, representativeness and generalizability. In C. Seale, G. Gobo, J. F. Gubrium and D. Silverman (eds), Qualitative Research Practice. London: Sage. https://archive.org/details/qualitativeresea0000unse_i2e8 Argues that qualitative sampling has its own logic of social representativeness, grounded in the social situation rather than the statistical population, and that generalisation works through transferability rather than sampling error. It carries the unit's scepticism about judging a qualitative sample by statistical representativeness.
• Hagaman, A. K. and Wutich, A. (2017). How many interviews are enough to identify metathemes in multisited and cross-cultural research? Field Methods. doi:10.1177/1525822x16640447 Analysed a cross-cultural study with 132 respondents in four sites. Found that 16 or fewer interviews identified common themes from relatively homogeneous sites, while 20 to 40 were needed to saturate metathemes cutting across sites. Shows the required sample is a function of the analytic target.
• Heckathorn, D. D. (2011). Comment: snowball versus respondent-driven sampling. Sociological Methodology. doi:10.1111/j.1467-9531.2011.01244.x Clarifies the confusion between snowball sampling as a convenience route and respondent-driven sampling, a method with an explicit weighting model that can yield population estimates under stated assumptions, and distinguishes seven forms of the estimator. Provides the unit's counterpoint on when a referral chain can support a population claim.
• Luborsky, M. R. and Rubinstein, R. L. (1995). Sampling in qualitative research. Research on Ageing (the journal spells the last word in the American form). doi:10.1177/0164027595171005 Observes that qualitative sampling rationales leave the sampling frame implicit, and proposes qualitative clarity, analogous to statistical power, as a set of principles for assessing a sample. Gives the unit its sharpest formulation of adequacy as an argument a reader can inspect.
• Mohammadi, E., Konig, T. and Zimmermann, B. (2026). Multi-layered sampling strategy for qualitative interviews: methodical reflections on sampling interviews with the European Research Council review experts. International Journal of Social Research Methodology. doi:10.1080/13645579.2025.2453935 Reflects on selecting interviewees from more than 1,200 named experts whose relevance varies with position, proposing a reflexive strategy using stratified random sampling. It argues the strategy also describes the structure of the expert population.
• Noy, C. (2008). Sampling knowledge: the hermeneutics of snowball sampling in qualitative research. International Journal of Social Research Methodology. doi:10.1080/13645570701401305 Re-reads snowball sampling through constructivist and feminist hermeneutics and argues that the chain of referrals generates a social knowledge that is emergent, political and interactional. In studies of backpacker tourists and marginalised men the pattern of referral expressed power relations and social capital.
• Palinkas, L. A., Horwitz, S. M., Green, C. A., Wisdom, J. P., Duan, N. and Hoagwood, K. (2015). Purposeful sampling for qualitative data collection and analysis in mixed method implementation research. Administration and Policy in Mental Health and Mental Health Services Research. doi:10.1007/s10488-013-0528-y Summarises the categories of purposeful sampling and reports criterion sampling as the strategy used most commonly in implementation research. It recommends combining strategies rather than using one template.
• Patton, M. Q. (2015). Qualitative Research and Evaluation Methods: Integrating Theory and Practice (4th ed.). Thousand Oaks, CA: Sage. https://us.sagepub.com/en-us/nam/qualitative-research-evaluation-methods/book232962 States the logic of purposeful sampling as the selection of information-rich cases, and supplies the canonical typology, including maximum variation, extreme or deviant case, typical case, critical and criterion sampling, and snowball or chain sampling, each tied to the question it answers.
• Robinson, O. C. (2014). Sampling in interview-based qualitative research: a theoretical and practical guide. Qualitative Research in Psychology. doi:10.1080/14780887.2013.801543 Sets out a four-point approach: defining the sample universe through inclusion and exclusion criteria, deciding the size against epistemological and practical concerns, choosing a strategy, and sourcing the sample. Argues that how far these concerns are made explicit shapes a study's coherence and trustworthiness.
• Suri, H. (2011). Purposeful sampling in qualitative research synthesis. Qualitative Research Journal. doi:10.3316/qrj1102063 Examines the adaptability of Patton's 16 purposeful sampling strategies to qualitative research synthesis and finds the literature says little about how the choice is made. Shows the sampling decisions of the primary studies condition what a synthesis can conclude.
• Tongco, M. D. C. (2007). Purposive sampling as a tool for informant selection. Ethnobotany Research and Applications. doi:10.17348/era.5.0.147-158 Argues that purposive sampling is most effective for studying a cultural domain through knowledgeable informants, and that the reliability and competence of the informant must be ensured. Claims the method stays robust against random probability sampling, carrying the unit's claim that selection is a judgement about who can answer.