Sampling in Qualitative Research

Overview

Qualitative sampling seeks participants, cases, documents or settings that can illuminate the phenomenon under study. Its logic is usually purposeful and iterative, not probabilistic. The aim is to obtain relevant, varied and sufficiently rich information for interpretation rather than to estimate population parameters from known selection probabilities.

This lecture compares criterion, maximum-variation, homogeneous, critical-case, typical-case, theoretical and snowball sampling. It also explains information power, different meanings of saturation, iterative sample refinement and transparent stopping decisions.

Why this matters

Qualitative studies are sometimes criticised for being “too small” using standards borrowed from surveys, or defended with a vague claim that “saturation was reached”. Both responses miss the methodological question.

A qualitative sample must fit:

• the study aim;
• the methodological approach;
• the specificity of participants’ experience;
• the breadth and complexity of the phenomenon;
• the role of theory;
• the quality of dialogue or observations;
• the intended depth of analysis; and
• the diversity needed to examine variation.

More participants do not automatically produce better interpretation. A very broad, shallow sample may be less informative than a smaller, carefully selected group with rich engagement.

Key concepts

Purposive sampling

Purposive sampling selects cases because they have characteristics, experiences or positions relevant to the research question. Selection criteria should be explicit and defensible.

It does not mean choosing participants who confirm the researcher’s expectations. A strong purposive strategy often seeks contrast, disconfirming experiences and underrepresented perspectives.

Iterative sampling

Qualitative sampling may evolve alongside data collection and analysis. Early interviews can identify experiences, settings or social positions that require further sampling. Changes should be documented as analytic decisions rather than hidden as departures from a fixed numeric plan.

Major purposive strategies

Criterion sampling

All participants meet a predefined experience or characteristic. For example, women must have had gestational diabetes and reached 12 weeks postpartum.

Criterion sampling establishes relevance but may still leave wide variation within the eligible group.

Maximum-variation sampling

Cases are selected across dimensions expected to shape the phenomenon, such as:

• completed versus did not complete testing;
• rural versus urban residence;
• first versus later pregnancy;
• public versus private follow-up;
• different ethnic or language groups;
• varying travel time; or
• different levels of family support.

The purpose is not to reproduce population proportions. It is to explore shared and contrasting patterns across diverse contexts.

Homogeneous sampling

Participants share a narrower experience, allowing detailed exploration within a subgroup. Examples include women who did not complete testing despite receiving a reminder, or nurses responsible for postpartum counselling in district hospitals.

Typical-case sampling

Cases are selected because they reflect an ordinary or common situation. This can make an implementation account accessible but should not be described as statistically representative unless a probability basis exists.

Critical-case and extreme-case sampling

Critical cases are especially consequential for a proposition: “If the pathway fails here despite excellent resources, what does that reveal?” Extreme or deviant cases display unusual success, failure or experience and may expose mechanisms not visible in average cases.

Theoretical sampling

In grounded-theory methodology, sampling decisions are driven by developing theoretical categories. Researchers seek cases that elaborate, compare or challenge emerging concepts. It is not merely purposive recruitment followed by thematic analysis.

Snowball or chain-referral sampling

Participants or community contacts help identify others. This may reach hidden or dispersed groups, but referrals can reproduce network similarity. Initial seeds, referral chains and confidentiality procedures should be documented.

DECISION GUIDE

Choose a qualitative sampling strategy

• Use criterion sampling to ensure everyone has the relevant experience.
• Add maximum variation when contrasting contexts may reveal different mechanisms.
• Use homogeneous sampling for depth within a specific subgroup.
• Use typical cases to describe common processes.
• Use critical or extreme cases to examine revealing successes or failures.
• Use theoretical sampling only when the chosen methodology uses developing theory to guide recruitment.
• Use snowball sampling when direct identification is difficult and network recruitment is ethically appropriate.
• Combine strategies when each has a clear role.

Illustrative purposive sampling matrix

illustrative

Table 1. Initial sample dimensions for postpartum testing interviews

Dimension	Categories sought	Why it may matter
Testing outcome	Completed early, completed late, not completed	Contrasting pathways
Residence	Urban, semi-rural, rural	Travel and service access
Parity	First birth, previous births	Experience and childcare
Follow-up pathway	Public clinic, private clinic, no follow-up	Organisational context
Language	Malay, English, other supported languages	Communication and materials
Reminder exposure	Received, uncertain, did not receive	Intervention reach

Note: This is a planning matrix, not a quota table that must mirror population proportions.

Sample size and information power

Information power proposes that fewer participants may be needed when:

• the study aim is narrow;
• participants are highly specific to the aim;
• established theory supports the inquiry;
• dialogue is rich and focused; and
• the analysis is intensive and case-oriented.

More participants may be needed when the aim is broad, experiences are heterogeneous, theory is limited, dialogue is brief or the analysis seeks cross-case comparison across several groups.

A protocol should therefore provide an initial range or expected sample, explain its basis and state how adequacy will be reviewed. A rigid number with no conceptual rationale is weak, but an entirely open-ended plan may be operationally and ethically difficult.

Saturation is not one universal endpoint

Code saturation

Code saturation refers to the point at which additional data yield few new descriptive codes. It may occur relatively early when topics are focused.

Meaning saturation

Meaning saturation concerns the depth, nuance, conditions and variation within codes. It generally requires more data than code saturation.

Theoretical saturation

In grounded theory, theoretical saturation concerns the development and relationships of theoretical categories. It should not be used as a generic phrase for every qualitative design.

Reflexive thematic analysis

In reflexive thematic analysis, themes are generated through interpretive engagement rather than discovered as fixed entities waiting to be exhausted. A mechanistic saturation claim may not fit the epistemology. Sample adequacy can instead be justified through scope, diversity, richness and analytic goals.

Iterative monitoring

The team can maintain a sampling and analysis log containing:

• participant characteristics relevant to the aim;
• recruitment source;
• interview quality and duration;
• emerging concepts;
• areas with thin evidence;
• contradictory or negative cases;
• proposed next participants; and
• rationale for stopping.

Sampling decisions should be discussed within the research team, while recognising that consensus does not remove interpretive judgement.

PROTOCOL CHECKLIST

Qualitative sampling plan to specify

• qualitative approach and study aim;
• population and eligibility criteria;
• initial sampling strategy;
• dimensions of variation or case types sought;
• recruitment channels and gatekeepers;
• expected initial sample or range;
• relationship between data collection and analysis;
• meaning of adequacy or saturation for the approach;
• process and responsibility for reviewing adequacy;
• documentation of sampling adaptations; and
• ethical management of referrals and sensitive identities.

Ethical and governance considerations

Interface with analysis

Qualitative sampling and analysis are linked. Maximum-variation sampling supports comparison across contexts; homogeneous sampling supports intensive within-group interpretation; theoretical sampling supports category development.

Researchers should not transform a purposive qualitative sample into numerical prevalence claims. Counting participants who mention a theme can help describe the dataset but does not estimate the population frequency of that experience.

Mixed-methods studies should state whether qualitative participants are:

• nested within a quantitative cohort;
• selected based on quantitative outcomes;
• recruited independently; or
• connected during integration.

Negative-case samplingMMed extension

Negative or disconfirming cases challenge an emerging explanation. Seeking them can refine the conditions under which a proposed mechanism holds. The goal is not to “disprove” participants’ accounts but to prevent the analysis from becoming a collection of confirming examples. The sampling log should record how such cases were identified and what they changed analytically.

Worked example

WORKED EXAMPLE

Developing an iterative interview sample

Aim: explain how personal, family and service factors shape postpartum testing decisions.

Initial strategy: criterion plus maximum-variation sampling.

The team begins with 18 interviews distributed across testing outcome, residence and follow-up pathway. Early analysis finds rich evidence about transport and childcare but little about women who intended to test yet were uncertain where to go.

Adaptation: Recruit additional women with documented referral ambiguity and several primary-care staff involved in handover.

Later interviews elaborate how discharge messages, appointment ownership and laboratory booking interact. The team stops after 27 women and 6 staff because the selected contrasts are represented, key concepts have depth, and additional interviews mainly refine examples rather than alter the explanatory account.

Reporting: The team states the initial plan, adaptation, final sample and adequacy rationale. It does not claim that 27 is a universal required number.

Suggested protocol or reporting wording

Evidence basis and critical appraisal

Information power provides a structured rationale based on study aim, sample specificity, theory, dialogue and analysis [1]. Empirical work distinguishes early code saturation from the greater data needed for meaning and nuance [2]. Broader methodological reviews show that saturation is conceptualised differently across qualitative traditions [3].

For reflexive thematic analysis, saturation may conflict with the view that themes are interpretive products rather than fixed items waiting to be exhausted [4]. Established qualitative-methods texts instead encourage purposeful case selection, reflexivity and methodological congruence [5].

Critical appraisal should examine:

• whether the sampling strategy follows the stated qualitative methodology;
• whether important contrasts were identified prospectively or iteratively;
• how recruitment channels may have limited experience diversity;
• whether data collection was sufficiently rich for the claimed interpretation;
• what evidence supported stopping; and
• whether the report treats the sample as statistically representative.

Sampling across multiple participant groups

Studies may interview patients, family members and health professionals. The sample-size rationale should be considered for each group because their roles and information needs differ. Six clinicians may provide adequate detail about one standardised workflow while six patients may be insufficient to explore heterogeneous access experiences. Combining all interviews into one total can conceal thin evidence within a group.

Researchers should also explain whether groups were sampled independently or relationally. Recruiting a patient’s own clinician creates linked accounts and confidentiality considerations; recruiting separate clinicians may broaden system perspectives but cannot corroborate an individual patient’s episode directly.

Feasibility and diversity are separate considerations

Recruitment may stop because the available time or eligible population is exhausted, but this is a feasibility limit rather than evidence of conceptual adequacy. Researchers should state both: what the data supported and which perspectives may remain thin or absent.

Common mistakes

Knowledge check

What is the purpose of maximum-variation sampling?

Answer and reasoning

To include contrasting cases across dimensions relevant to the phenomenon so that shared and context-specific patterns can be explored. It is not intended to reproduce population proportions.

Why might code saturation occur before meaning saturation?

Answer and reasoning

Researchers may stop encountering new topic labels while still learning new meanings, conditions, mechanisms and variations within the existing codes.

When is theoretical sampling appropriate?

Answer and reasoning

When a grounded-theory approach uses emerging theoretical categories to guide selection of further cases that develop or test those categories.

Can a qualitative sample of 25 participants estimate that 40% of the population experienced a barrier?

Answer and reasoning

Not on the basis of purposive sampling. The sample was selected for relevance and depth, not known population inclusion probabilities.

What should a transparent stopping rationale include?

Answer and reasoning

The study aim, sampling specificity and variation, richness of data, analytical approach, conceptual gaps reviewed, who made the decision and what additional data were no longer expected to change.

Key take-home messages

SUMMARY

Key take-home messages

• Qualitative sampling prioritises conceptual relevance, range and depth.
• Choose the purposive strategy that matches the qualitative aim and methodology.
• Sampling can evolve in response to analysis, but changes must be documented.
• Information power provides a multidimensional adequacy rationale.
• Saturation has different meanings and is not required in identical form for every approach.
• Do not convert purposive samples into population prevalence claims.

References

1. Malterud K, Siersma VD, Guassora AD. Sample size in qualitative interview studies: guided by information power. Qual Health Res. 2016;26(13):1753–1760. https://doi.org/10.1177/1049732315617444
2. Hennink MM, Kaiser BN, Marconi VC. Code saturation versus meaning saturation: how many interviews are enough? Qual Health Res. 2017;27(4):591–608. https://doi.org/10.1177/1049732316665344
3. Saunders B, Sim J, Kingstone T, Baker S, Waterfield J, Bartlam B, et al. Saturation in qualitative research: exploring its conceptualization and operationalization. Qual Quant. 2018;52(4):1893–1907. https://doi.org/10.1007/s11135-017-0574-8
4. Braun V, Clarke V. To saturate or not to saturate? Questioning data saturation as a useful concept for thematic analysis and sample-size rationales. Qual Res Sport Exerc Health. 2021;13(2):201–216. https://doi.org/10.1080/2159676X.2019.1704846
5. Patton MQ. Qualitative research & evaluation methods: integrating theory and practice. 4th ed. Thousand Oaks, CA: SAGE; 2015.