A small sample size is a group of participants or observations too limited to reliably detect real patterns or differences in a study. There is no single number that makes a sample “small” across all contexts. Whether 20 people counts as small depends on what you’re studying, how large the effect you’re looking for is, and what type of research you’re doing. In practice, though, the consequences of too few data points are consistent: wider margins of error, higher risk of missing real findings, and results that may not apply beyond the group you studied.
Why There Is No Universal Cutoff
You may have heard that 30 is the magic number for a sample to be “large enough.” This idea comes from a statistical principle called the central limit theorem, which says that averages from random samples start to follow a predictable bell-curve pattern as the sample grows. But research from the University of Massachusetts found little documented evidence that 30 actually works as a universal threshold. For some types of data distributions, 30 is plenty. For others, particularly data that is heavily skewed or has extreme outliers, even 300 observations weren’t enough for that bell-curve pattern to kick in.
What actually determines whether your sample is too small comes down to four interconnected factors: the significance threshold you set (how willing you are to accept a false alarm), the statistical power you want (your ability to catch a real effect), the size of the effect you’re trying to detect, and the number of participants. Change any one of these and the required sample size shifts. A study looking for a large, obvious difference between two groups might need only 8 participants per group to have 80% power. The same study looking for a subtle difference could require 788 per group.
How Small Samples Distort Results
The most direct problem with a small sample is low statistical power, which is the probability that your study will detect a real effect when one exists. Most researchers aim for at least 80% power, meaning an 80% chance of finding a true result. When a sample is too small, power drops, and the study becomes far more likely to produce a false negative: concluding that nothing is going on when something actually is. This is known as a Type II error.
Think of it like trying to determine whether a coin is rigged. If you flip it 5 times and get 3 heads, you can’t say much. Flip it 500 times and get 320 heads, and the picture becomes much clearer. With too few flips, you lack the resolution to distinguish a real pattern from random noise.
Small samples also amplify the impact of outliers. One unusual participant in a group of 15 can pull an average dramatically in one direction. That same unusual participant in a group of 1,500 barely registers. This is why small studies sometimes produce dramatic findings that vanish when larger studies try to replicate them.
Margin of Error Gets Worse Fast
In surveys and polls, sample size directly controls the margin of error, which tells you how far your results might be from the true answer. A survey of 1,000 adults typically carries a margin of error around plus or minus 3 percentage points. Drop that sample to 100 and the margin of error balloons dramatically. The relationship isn’t linear, though. Going from 100 to 1,000 participants produces an enormous improvement in precision, but doubling from 1,000 to 2,000 only shaves off about a single percentage point. The biggest gains in accuracy come from moving out of very small sample territory.
This matters in practical terms. If a survey of 100 people finds that 52% support a policy, the true number could easily be anywhere from the low 40s to the low 60s. That’s too imprecise to draw any meaningful conclusion. The same result from 1,000 people narrows the true value to roughly 49% to 55%, which is far more useful for decision-making.
Bias and Generalizability
Small samples are especially vulnerable to sampling bias, where the people in your study don’t reflect the broader population you care about. Every sample carries some risk of this, but larger samples naturally tend to capture more diversity. A small convenience sample, say 20 college students recruited from a psychology class, might skew heavily toward a particular age, income level, or cultural background. Any conclusions drawn from that group may simply not hold true for other populations.
This is the problem of external validity: can you generalize your findings beyond your specific participants? With a small, homogeneous group, the honest answer is often no. Larger and more carefully selected samples spread risk across more people and reduce the chance that any one subgroup’s characteristics dominate the results.
When Small Samples Are Acceptable
Not every study needs hundreds or thousands of participants. In rare disease research, for example, there may only be a few hundred people in the world with a particular condition. The FDA explicitly acknowledges this limitation and exercises what it calls “the broadest flexibility” in evaluating treatments for severely debilitating rare diseases. In these cases, researchers compensate by using more sensitive measurement tools, continuous rather than simple yes/no outcomes, and careful study designs that extract maximum information from each participant.
Pilot studies, which test whether a larger study is feasible, also operate with intentionally small samples. Recommendations for pilot study sizes vary widely, from as few as 24 participants total to as many as 70, depending on the expected effect size and what the pilot is trying to estimate. These aren’t designed to produce definitive answers. They’re designed to generate the preliminary data needed to plan a full-scale study.
Qualitative research follows different rules entirely. Interview-based studies often work with 12 to 15 participants in relatively similar populations, aiming for what researchers call “saturation,” the point where additional interviews stop revealing new themes or insights. In this context, depth of information matters more than breadth of numbers, and a sample of 15 can be entirely appropriate.
How to Tell If a Sample Is Too Small
Rather than memorizing a threshold number, the better question is whether a study was powered adequately for what it set out to find. Four things to look for: Was a power analysis done before the study began? What effect size were the researchers expecting? Did the study achieve at least 80% statistical power? And do the confidence intervals around the results look reasonably narrow?
If a study reports a result that isn’t statistically significant, a small sample is one of the first things to consider. The finding might genuinely be null, or the study might simply not have had enough participants to detect a real but modest effect. These two scenarios look identical in the data, which is exactly why underpowered studies are problematic. They can’t distinguish “no effect” from “not enough information.”
When reading health or science news, pay attention to participant counts. A study of 12 people suggesting a new supplement works should be treated very differently than a study of 1,200 people showing the same thing. The small study isn’t worthless, but it’s a starting point, not a conclusion.