03 · What You Need to Know
Sample size constrains the claim, not the population's right to be studied
First distinguish a small population from a small sample
A small sample can arise for very different reasons.
Sometimes the target population is large, but researchers recruit only a few participants because of limited time, budget, access, or poor recruitment. In that situation, "the population is difficult to reach" should not automatically excuse an inadequately designed study.
In other cases, the target population itself is genuinely small. No recruitment strategy can produce 500 participants when only 120 eligible people exist in the defined population.
Small sample from a large population
The limitation may arise from recruitment, resources, eligibility rules, or study design and may potentially be addressed.
Inherently small target population
The number of eligible people is itself constrained, so the research design must accommodate that reality.
This distinction should appear explicitly in your justification. Otherwise, an unavoidable population constraint can be confused with a preventable recruitment failure.
There is no single conventional sample-size target
The phrase "required sample size" can sound more universal than it is. Sample-size requirements depend on the research question, design, outcome, statistical model, expected effect or precision, variability, significance level, desired power, allocation, attrition, clustering, and other assumptions.
A target of 300 participants may be inadequate for one analysis and excessive for another. A sample of 30 may be severely underpowered for detecting a modest between-group effect yet entirely appropriate for a different research purpose.
Qualitative research makes the contrast especially clear. Conventional quantitative power calculations are generally not the basis for determining interview sample adequacy. Methodological work instead considers concepts such as saturation or information power, with adequacy depending on the study aim, specificity of the sample, use of theory, quality of dialogue, and analytical strategy.
The correct question is therefore not "Is this sample objectively too small?" It is "Too small for what claim, using what design?"
Statistical power does not decide whether the research question matters
Power calculations help researchers assess the probability of detecting specified effects under a particular statistical design. If the attainable sample provides low power for the planned hypothesis test, that is a genuine methodological limitation.
But the conclusion should be "this design cannot reliably answer this question as currently formulated," not "this population cannot be researched."
That distinction is especially important for rare, highly specialized, or historically excluded populations. If conventional designs were treated as admission tickets to research, populations too small to satisfy standard assumptions could remain permanently absent from the evidence base.
The response should be methodological adaptation, accompanied by appropriately restrained inference.
Start from the maximum feasible evidence, not the sample size you wish existed
When the population is inherently small, estimate realistically how many eligible participants exist, how many can reasonably be approached, and what participation rate is plausible. Then determine what research questions that evidence can support.
This reverses a common workflow. Instead of choosing an elaborate analysis and discovering later that the population cannot support it, researchers allow population size and data structure to inform the design from the beginning.
If nearly the entire target population can be recruited, the study may provide highly relevant descriptive evidence about that defined population even though the absolute number of participants is modest. What remains limited is the precision of estimates, complexity of models, detection of small effects, and generalization beyond the defined population.
Precision may be more informative than chasing statistical significance
With a small population, binary thinking about whether a result reaches a conventional significance threshold can become particularly unhelpful. Small samples often produce wide confidence intervals, meaning the data may be compatible with a broad range of effect sizes.
Researchers should therefore pay close attention to estimates and their uncertainty. An imprecise estimate can still provide information, but its uncertainty must remain visible in the interpretation.
For example, a small study might suggest a potentially meaningful effect while remaining compatible with little or no effect. The correct conclusion is not that the intervention "works" or "does not work" solely because a p-value falls on one side of a threshold. The study has narrowed uncertainty to a particular degree, and that degree should shape the claim.
Do not solve a small-population problem by making a large-population claim
One tempting strategy is to broaden the target population until the sample-size problem disappears. Sometimes that is scientifically legitimate. Sometimes it changes the question.
If your question specifically concerns a small Indigenous community, a rare professional group, a particular disability population, or people with an uncommon condition, combining them with a much larger but substantively different population may produce a statistically convenient dataset while obscuring the very population-specific evidence you intended to obtain.
The relevant question is whether aggregation is conceptually defensible. Similarity should be justified rather than assumed merely because pooling improves power.
Alternative designs can extract different kinds of evidence
A small population does not imply one preferred alternative method. The design should follow the question.
| Research aim |
Possible strategy |
Main limitation to acknowledge |
| Describe the defined population |
Recruit as comprehensively as feasible and report estimates with uncertainty |
Limited precision and potentially limited inference beyond the defined population |
| Understand experiences or processes |
Qualitative interviews, focus groups, observations, or case-oriented designs |
Claims should follow the qualitative methodology rather than mimic statistical population estimates |
| Examine change over time |
Repeated-measures or longitudinal designs when appropriate |
More observations do not create more independent participants, and dependence must be modeled correctly |
| Estimate an intervention effect |
Designs suited to the setting, potentially including repeated measures, multicenter collaboration, or other efficient approaches |
Design assumptions, precision, feasibility, and potential biases remain important |
| Build evidence about a rare phenomenon |
Combine compatible evidence across sites, studies, or data sources where scientifically justified |
Heterogeneity and comparability must be examined rather than hidden by pooling |
Qualitative research is not a fallback for an underpowered quantitative study
Researchers sometimes respond to an unattainable quantitative sample by saying, "Then I will just make it qualitative."
That reverses the logic of research design. Qualitative methods answer different kinds of questions. They can be especially valuable when researchers need depth about experiences, meanings, processes, implementation, or mechanisms, but they should be selected because those questions matter rather than because statistical power is inconvenient.
Qualitative sample adequacy also requires justification. Empirical work on saturation shows that required sample sizes vary with the study's purpose, population heterogeneity, type of saturation sought, and analytical depth. A systematic review found relatively narrow saturation ranges in many studies with homogeneous populations and focused objectives, while more complex studies required larger samples.
Information power offers another way to reason about adequacy: a sample containing highly specific and relevant information may require fewer participants than a broad, heterogeneous sample addressing a diffuse question.
A small population may require a narrower question
When available evidence is limited, the most responsible adaptation is sometimes intellectual rather than statistical.
You may need to ask a narrower question, reduce the number of parameters being estimated, avoid elaborate subgroup comparisons, prioritize a primary outcome, or treat the study as exploratory rather than confirmatory.
This can feel less ambitious. Methodologically, it is often more ambitious in the useful sense: the study attempts to answer a question the available evidence can genuinely support rather than decorating an impossible design with caveats after the analysis.
Small populations can make every participant more identifiable
Research with small populations creates ethical and privacy concerns that are easy to overlook. Even after names are removed, combinations of age, occupation, location, role, experiences, or quotations may allow community members to infer who participated.
Researchers should therefore consider deductive disclosure, especially when reporting qualitative quotations, detailed demographic tables, rare characteristics, or findings from tightly connected communities.
The methodological desire to provide rich contextual detail must sometimes be balanced against the possibility that such detail identifies participants.
Recruiting nearly everyone does not remove all uncertainty
If a target population contains 80 eligible people and 70 participate, researchers may have observed a substantial proportion of that defined population. That can be valuable.
It does not eliminate measurement error, nonresponse bias, missing data, temporal variation, causal ambiguity, or uncertainty about other populations and future members of the population. Nor does it automatically support complex statistical models simply because the sampling fraction is large.
Be precise about what near-complete coverage solves and what it does not.
Feasibility and value should be considered together
Small-population research can impose substantial burdens. When only a few eligible people exist, the same individuals may repeatedly receive invitations from multiple research teams. This can be particularly problematic for populations already subjected to intense academic attention.
The importance of the question should therefore be weighed against participation burden, privacy risk, duplication, and likely informational gain. A study is not justified merely because the population is small and understudied.
The broader principle remains the same as for underrepresentation more generally: absence or scarcity identifies a constraint, but the study still needs a substantive reason to exist.
Watch Out
Do not recruit a small sample for a design that requires much more information and then use the rarity of the population as permission to make the original claims anyway. Rarity explains why evidence is limited; it does not make uncertainty disappear.
