A sample in psychology is a smaller group of people (or other units) selected from a larger population to participate in a study. Researchers use samples because studying every single person in a population is almost never practical. Instead, they collect data from this smaller group and draw conclusions that, ideally, apply to the broader population they care about.
That leap from sample to population is the foundation of most psychological research, and it comes with real challenges. How the sample is chosen, how large it is, and who ends up in it all shape whether the findings mean anything beyond the study itself.
Samples vs. Populations
A population is the entire group a researcher wants to learn about. It could be all adults with social anxiety, all children under age five, or all college students in a particular country. The population doesn’t have to be made up of people either. In some studies, the “population” might be therapy sessions, test scores, or brain scans. What matters is that the population is clearly defined, with specific criteria for who or what is included and excluded.
The sample is the subset actually studied. If a psychologist wants to understand how sleep deprivation affects memory in adults, they can’t test every adult on the planet. So they recruit 200 participants, run the experiment, and use statistical methods to generalize the results back to the larger population. Any conclusions drawn from the sample technically apply only to the defined population the sample was properly selected from. This defined group is called the target population, and keeping it clearly in mind prevents researchers from overclaiming what their data actually shows.
Why Sampling Method Matters
Not all samples are created equal. The method used to select participants determines how confidently the findings can be applied to the broader population. Sampling methods fall into two broad categories: probability and non-probability.
Probability Sampling
In probability sampling, every member of the population has a known, nonzero chance of being selected. The simplest version is random sampling, where participants are chosen entirely by chance, like drawing names from a hat. Stratified sampling divides the population into subgroups first (by age, gender, income, or another relevant characteristic) and then randomly samples from each subgroup to ensure they’re all represented. Cluster sampling randomly selects entire groups, such as schools or clinics, and then studies everyone within those groups. These approaches tend to produce samples that closely mirror the population, which makes generalizing the results more defensible.
Non-Probability Sampling
In practice, most psychology studies don’t use true probability sampling. The most common approach is convenience sampling, where researchers recruit whoever is available and accessible. This is fast, inexpensive, and logistically simple, which is why it dominates clinical and social research. The tradeoff is that convenience samples may not reflect the population well, since certain types of people are more likely to show up.
Snowball sampling is another non-probability method, useful when the population is hard to locate. A researcher recruits one participant and asks them to refer others from the same group. This is common in studies of hard-to-reach populations, such as people experiencing homelessness or individuals with rare conditions, where no ready-made list of participants exists. Purposive sampling, where researchers hand-pick participants based on specific characteristics relevant to the research question, is frequently used in qualitative psychology studies.
Representativeness and Generalizability
A sample is considered representative when it reflects the key characteristics of the target population. The common understanding is that a representative sample is one created through random selection, where each person had a known probability of being chosen. But representativeness can also be thought of more broadly: does the sample contain enough variation on the dimensions that matter for the research question?
For example, a study on how stress affects decision-making would ideally include participants across a range of ages, socioeconomic backgrounds, and stress levels. By sampling across these relevant variables, a researcher can test whether the findings hold up across different groups or only apply under certain conditions. This helps map out the real boundaries of the conclusion rather than assuming it applies universally.
Generalizability, the extent to which results apply beyond the specific sample, depends on multiple factors. The characteristics of the participants, how they were recruited, and even the format of data collection (online survey vs. in-person interview) all play a role.
The WEIRD Problem in Psychology
One of the most discussed issues in modern psychology is the overreliance on samples drawn from Western, Educated, Industrialized, Rich, and Democratic societies, known by the acronym WEIRD. An analysis of articles published between 2014 and 2018 in six prominent psychology journals found that both researchers and participants were still predominantly from English-speaking and Western countries, which represent only about 11% of the world’s population.
This is a problem because psychological phenomena, from how people perceive fairness to how they experience emotion, can vary significantly across cultures. When nearly all the data comes from one narrow slice of humanity, findings that get presented as universal truths about “how people think” may actually describe only how certain people in certain societies think. It’s a form of sampling bias baked into the discipline itself, and it remains a major concern despite growing awareness.
Sampling Bias and Its Consequences
Sampling bias occurs when certain members of a population are systematically more or less likely to be included in a study. This can impair a study’s external validity, meaning the results don’t generalize well beyond the sample. Bias can creep in through many routes. Offering a survey in only one format (online only, for example) may exclude people without reliable internet access, people with visual impairments, or older adults less comfortable with technology. The result is a sample skewed toward certain demographics.
Volunteer bias is another common issue. People who choose to participate in research tend to differ from those who don’t. They may be more motivated, more educated, more interested in the topic, or more willing to disclose personal information. When a study’s participants are self-selected, the findings may describe a particular kind of person rather than the population at large. Researchers can partially address this by offering multiple survey formats, using stratified recruitment, or statistically adjusting for known demographic imbalances, but no technique fully eliminates the problem.
How Large Should a Sample Be?
Sample size directly affects a study’s ability to detect real effects. A study with too few participants may miss a genuine relationship between variables simply because there wasn’t enough data to distinguish the signal from random noise. This concept is called statistical power: the probability that a study will correctly identify a real effect when one exists.
The widely accepted standard is that a study should have at least 80% power, meaning an 80% chance of detecting a true effect. The sample size needed to reach that threshold depends on how large the effect is expected to be. A subtle effect (like a small difference in reaction time between two groups) requires far more participants than a dramatic one. Researchers often use software tools like G*Power to calculate the required sample size before they begin collecting data, plugging in the expected effect size, the acceptable error rate (typically 5%), and the desired power level.
Despite decades of recommendations, a review comparing sample sizes in four leading psychology journals across 1955, 1977, 1995, and 2006 found that the push for larger samples had not been meaningfully integrated into core psychological research. Small, sometimes dramatically inadequate sample sizes continued to appear in top-tier journals. This has been a persistent concern, as underpowered studies are more likely to produce false positives and findings that fail to replicate.
Ethics of Selecting Participants
Choosing who ends up in a sample isn’t just a statistical question. It’s an ethical one. The American Psychological Association’s ethical guidelines emphasize three core principles in participant selection. First, respect for persons: participation must be voluntary, with informed consent, and people should be free to withdraw at any time without penalty or stigma. Second, beneficence: the potential benefits of the research should outweigh any risks to participants, and those risks should be minimized. Third, justice: researchers should not target certain groups, such as people on welfare, incarcerated individuals, or members of racial and ethnic minorities, simply because they’re easy to access or in a compromised position. Selection should be driven by the research question, not by convenience alone.
Confidentiality is also essential. Participants’ identities and data must be protected, and individuals should not be publicly identified based on whether they chose to participate or declined. For vulnerable populations, such as children, people with cognitive impairments, or communities affected by trauma, additional protections must be in place before any data collection begins.