Cognitive Connie
Understanding sampling & ethics
No study can include everyone, and how researchers decide who to include shapes who the findings apply to. Sampling sits at the bridge between the study and the world: a strong sampling method means findings tell us something about people beyond the room where data was collected; a weak one produces results that describe only the specific individuals who happened to participate.
Key figures
Henry Beecher
1904–1976American anaesthesiologist whose 1966 New England Journal of Medicine paper documented 22 cases of unethical research conducted on unsuspecting participants, including vulnerable hospital patients. Beecher's paper was a catalyst for formalising research ethics in medicine and psychology, leading to the creation of institutional review boards and formal ethical approval processes that are now universal requirements.
Diana Baumrind
1927–2018American developmental psychologist who wrote the first major published critique of Milgram's obedience studies on ethical grounds (1964), arguing that the study violated participants' dignity and trust in the research relationship. Baumrind's intervention in the Milgram controversy established the template for ethical critique of psychological research and influenced the formal codes of ethics that emerged from professional psychology bodies in the 1970s.
Stanley Milgram
1933–1984Milgram's obedience studies (1961–1963) remain the most cited case study in research ethics — simultaneously celebrated for their scientific contribution and condemned for their ethical violations. Participants believed they were administering severe electric shocks to another person and showed significant distress. The studies prompted intense debate about deception, psychological harm, and the right to withdraw — debates that directly shaped the ethical guidelines adopted by the APA and BPS in subsequent decades.
Key concepts
Random sampling
Every member of the target population has an equal, independent probability of being selected — achieved via random number generators applied to a complete sampling frame. Tends to produce representative samples and allows generalisation from sample to population. Main limitation: obtaining a complete sampling frame of a large real-world population is typically impossible, and small random samples can still deviate substantially from the population by chance.
Opportunity (convenience) sampling
Recruiting whoever is accessible at the time and place of the study. The most commonly used method in psychology because it is fast and inexpensive. The critical limitation is that the sample reflects whoever happened to be in that place at that time — often over-representing university students, leading to the WEIRD problem (Western, Educated, Industrialised, Rich, Democratic) and limiting generalisability.
Stratified sampling
Divides the population into strata (subgroups based on characteristics relevant to the research question: age, gender, ethnicity, socioeconomic status) and randomly samples proportional numbers from each stratum. Ensures subgroups are represented in proportion to their presence in the population, producing more representative samples than simple random sampling when strata are relevant to the variables studied.
Volunteer (self-selected) sampling
Participants come forward in response to an advertisement or call. Introduces volunteer bias: people who actively volunteer for research tend to differ from the general population — often more curious, more altruistic, or more motivated by personal experience with the topic. Ethical by design: participants have consented before the study begins.
Informed consent
Participants must receive sufficient information about the study's nature, procedures, risks, and their right to withdraw to make a genuinely voluntary decision about participation. Information must be comprehensible, not buried in jargon or given under social pressure. Additional safeguards (parental consent, assent) are required for vulnerable populations including children and individuals with cognitive impairments.
Right to withdraw
Participants may stop at any time during the study without requiring a reason, and without suffering any negative consequence. The right extends retrospectively: after the study ends, participants may request their data be excluded from analysis and destroyed, even when data collection is complete. This is particularly important in studies using deception, where participants only fully understand what they took part in during the debrief.
Deception in research
Some research cannot be conducted without deception because revealing the true purpose would fundamentally alter participants' behaviour. The BPS permits deception only when: the study has genuine scientific value that cannot be achieved otherwise, there is no reasonable alternative, likely harm is minimal, and a thorough debrief is provided. Deception about aspects of participation that participants would have objected to — such as the true risks involved — is not permissible.
Debriefing
A post-study conversation explaining the true purpose of the study, addressing any deception used, checking for distress, providing support if needed, and offering the option to withdraw data. Not a tick-box exercise: if a participant is distressed, the debrief must address this. The BPS requires that participants leave no worse off than before they took part.
Protection from harm
Researchers must take all reasonable steps to protect participants from physical and psychological harm during and after the study. The threshold is that participants should not experience greater risk than they would in everyday life. If a study produces unexpected distress, the researcher must be prepared to stop the study, provide appropriate support, and refer participants to professional help if necessary.
Test your knowledge
Frequently asked questions
Is opportunity sampling always bad?+
Not always. Whether opportunity sampling is problematic depends on the research question. For studies examining basic cognitive or perceptual processes that are likely universal (reaction time, low-level perception, basic memory encoding), a convenience sample of university students may be adequate. For studies of social behaviour, attitudes, development, or psychopathology — where age, culture, education, and socioeconomic status may substantially moderate effects — convenience samples from a single demographic severely limit generalisability. The WEIRD critique (Henrich et al., 2010) showed that the majority of psychology's participant base was drawn from a narrow slice of humanity, and that many "universal" findings did not replicate in other cultures.
When is deception ethically permissible?+
The BPS permits deception when: (1) the research question cannot be validly studied without it — full disclosure would invalidate the results; (2) the potential scientific knowledge gained is significant; (3) there is no foreseeable serious harm; (4) the deception does not concern risks or the fundamental nature of participation in a way that would have affected whether participants consented; (5) ethical committee approval has been obtained; and (6) a thorough debrief will be provided immediately after. The debrief must restore any false beliefs created and give participants the genuine opportunity to withdraw their data.
What makes a sample truly representative?+
A truly representative sample mirrors the target population on all characteristics relevant to the research question. In practice, this is an ideal rarely fully achieved. Stratified random sampling improves representativeness on the stratification variables but cannot control for all relevant characteristics. Random sampling from a complete sampling frame provides the best theoretical guarantee of representativeness, but complete sampling frames are rarely available for real populations. Practically, researchers assess representativeness by comparing their sample's key demographics to population statistics and acknowledging limitations in the discussion section. Replication across different samples and settings is the strongest evidence of generalisability.
Sources
Last reviewed July 2025- 1.
British Psychological Society. (2021). Code of Ethics and Conduct. BPS.
+About this source
The authoritative ethical guidelines for psychological research in the UK, covering consent, deception, debriefing, and protection from harm.
- 2.
American Psychological Association. (2017). Ethical Principles of Psychologists and Code of Conduct. APA.
+About this source
The APA code of ethics covering the principles of research integrity, participant rights, and professional conduct.