Why the rules existThe history is the argument
Here is the uncomfortable premise of this entire lesson: every clause of modern research ethics is a scar. Nobody sat down in a seminar room and invented informed consent because it seemed polite. Each rule was written after a specific harm, by people trying to make sure that harm could never be called "science" again. So the fastest way to learn the rules is to learn the cases that forced them.
Start in 1946, at the Nuremberg doctors' trial. Twenty-three defendants — most of them physicians — stood accused of experiments performed on concentration-camp prisoners: freezing experiments, deliberate infection with malaria and typhus, mustard-gas exposure, all without any pretense of consent. The judges did something unusual for a criminal verdict: they wrote a code. The Nuremberg Code (1947) lays out ten points, and its very first sentence is the foundation of everything that follows in this lesson: "The voluntary consent of the human subject is absolutely essential." Notice what the Code assumes — that credentialed, educated researchers are capable of atrocity when they convince themselves the knowledge is worth it. The rules exist because good intentions are not a safety mechanism.
It would be comforting to file that under "wartime monsters" and move on. The Tuskegee Syphilis Study (1932–1972) forbids it. The U.S. Public Health Service enrolled roughly 600 Black men in Macon County, Alabama — about 399 with syphilis and 201 without — and studied the untreated course of the disease for forty years. The men were told only that they had "bad blood." They were drawn in with free meals, free exams, and burial insurance; painful diagnostic spinal taps were sold to them as "special free treatment." When penicillin became the standard cure in the late 1940s, it was deliberately withheld, and local doctors were quietly asked not to treat the men (Brandt, 1978). The study ended only when a whistleblower, Peter Buxtun, went to the press in 1972. The fallout — Senate hearings, the National Research Act of 1974, and a national commission — produced the document at the center of this lesson, the Belmont Report. Tuskegee's other legacy is a deep and entirely rational distrust of medical research among many Black Americans, a harm that outlived every participant.
Still, you might think scandals like these are rare, spectacular exceptions. In 1966, the Harvard anesthesiologist Henry Beecher published a short paper in the New England Journal of Medicine that destroyed that comfort. Beecher (1966) collected 22 examples of ethically indefensible studies drawn from mainstream, published research at leading institutions — live cancer cells injected into elderly hospital patients without their knowledge, institutionalized children deliberately infected with hepatitis. His point was not that medicine harbored a few villains. It was that ordinary, ambitious researchers at famous universities would routinely trade patient welfare for publications when nobody made them do otherwise. The problem wasn't monsters; it was incentives without guardrails.
Psychology's own entry in this history is Milgram's obedience research. Milgram (1963) told participants they were in a learning study and ordered them to deliver escalating electric shocks to a screaming "learner" (a confederate; no shocks were real). In the baseline condition, 65% continued to the maximum 450 volts. Participants sweated, trembled, stuttered, and begged to stop — and some left believing they might have seriously harmed or killed a man. The findings are among the most cited in the field, and the study could never be run today in that form: the distress was severe, the deception went to the very heart of what participants thought was happening, and the debriefing did not always promptly undo the harm. Milgram forces the question the IRB now asks of every protocol: is the knowledge worth what it costs the people who provide it?
The obedience studies are real and disturbing, but the tidy "65%" is messier than the textbooks suggest. Re-analysis of the original Yale audio recordings found that among "obedient" participants, essentially all of them broke the study's own required procedure at some point — for example, reading test questions right over the learner's screams, abandoning the "teaching" premise entirely — and the experimenter often failed to correct them, quietly signaling that the science no longer mattered, only the shocks did (Kaposi & Sumeghy, 2026). The lab drifted from "legitimate research" toward unstructured coercion. Cite Milgram for what he showed — situational pressure is frighteningly strong — but hold the precise number loosely.
The three pillarsThe Belmont Report
The National Commission created after Tuskegee spent four years distilling research ethics into something a working scientist could actually use. The result, the Belmont Report (National Commission, 1979), names three principles. Memorizing them is easy; the exam-worthy skill is knowing which concrete protection flows from each.
Respect for Persons
People are autonomous agents with the right to decide what happens to them. The tool: voluntary informed consent. People with diminished autonomy — children, prisoners, people with cognitive impairment — get extra protection, not less. Tuskegee violated this by lying about the diagnosis.
Beneficence
Do no harm; maximize benefit and minimize risk. The tool: a risk–benefit analysis for every protocol — is the knowledge worth the cost to the people providing it? Tuskegee violated this by withholding a known cure.
Justice
Fairness in who bears the burden of research and who reaps its benefits. One group must not be exploited so another can profit — the burdened should be among the beneficiaries. Tuskegee violated this by loading all the risk onto poor Black men while the knowledge served everyone else.
Belmont is not a poster on a wall. Each principle became specific machinery that you will personally encounter the first time you submit a study for review:
| Belmont principle | What it becomes in IRB practice |
|---|---|
| Respect for persons | The informed-consent process and its required elements; children give assent while parents give permission; special federal protections for prisoners and other vulnerable groups. |
| Beneficence | The written risk–benefit justification in every protocol; the "minimal risk" standard that decides review level; confidentiality and data-security plans; the obligation to halt a study that starts causing unexpected harm. |
| Justice | Fair participant selection — you cannot recruit the poor, the captive, or the convenient simply because they are easy to get; the population that carries the risk should be one the findings can help. |
Who signs offThe IRB and its three review tiers
The Institutional Review Board (IRB) is the committee that turns Belmont into decisions, protocol by protocol. Under the federal Common Rule, every institution taking federal research money must have one, with at least five members from varied backgrounds — including at least one non-scientist and at least one community member unaffiliated with the institution, precisely so the committee never becomes researchers rubber-stamping researchers. Its job is prospective: no data collection begins until the protocol is approved.
Not every study gets the same scrutiny. Proposals are sorted into three tiers by risk. Exempt review covers the lowest-risk categories — say, an anonymous survey on a non-sensitive topic — but here is the detail students miss: exempt is a determination the IRB makes, not a box you check yourself. You still submit. Expedited review covers studies with no more than minimal risk — risk no greater than everyday life — and is handled by the chair or a subcommittee rather than the full board. Full board review is for everything that raises the stakes: more than minimal risk, vulnerable populations, sensitive topics, or deception that could matter.
| Tier | Risk profile | Who reviews | Typical example |
|---|---|---|---|
| Exempt | Lowest-risk federal categories | IRB staff/chair make the determination | Anonymous survey of study habits |
| Expedited | No more than minimal risk | Chair or a designated subcommittee | Reaction-time task with a consent gate and no sensitive data |
| Full board | More than minimal risk; vulnerable groups; sensitive topics; consequential deception | The entire convened committee | Trauma interviews; studies with children or prisoners |
Sit in on an IRB meeting and you'll discover the debates are rarely about grand philosophy. They are relentlessly practical: is the consent form written at a reading level this population can actually understand, or is it eight pages of legal armor protecting the university? Is $50 fair compensation for an hour, or quiet pressure on broke undergraduates? Where exactly will the data live, who holds the key, and what happens to it when the researcher leaves? Whose phone number rings if a participant gets upset at 9 p.m.? Two honest caveats come with this: IRBs are made of humans, so a protocol approved at one university is sometimes rejected at another — and a committee's approval is a floor, not a moral guarantee. The ethics of your study remain yours.
The core transactionInformed consent, element by element
Informed consent is where respect for persons becomes an actual conversation. The APA Ethics Code (American Psychological Association, 2017, Standard 8.02) specifies what that conversation must cover, and the list is worth knowing cold: the purpose of the research, its expected duration and procedures; the right to decline or withdraw at any time, and the consequences (there should be none worth fearing) of doing so; any foreseeable risks or discomfort; any prospective benefits; the limits of confidentiality; any incentives for participation; and whom to contact with questions about the study or your rights. All of it in language the participant actually understands — a consent form written like a software license fails the standard even if every element is technically present.
One more distinction with teeth: written documentation of consent is generally required when risk climbs above minimal, but consent itself is required in some form for nearly everything. Online studies handle this with a consent page that participants must actively click through before the first trial can load — the gate comes first, always, before a single datum is collected.
Students constantly collapse these two. Informed consent is a process: communicating purpose, risks, protections, and rights so a person can freely decide, and staying open to their questions throughout the study. The consent form is merely the paperwork that documents the process happened. A signed form obtained through mumbling, rushing, or jargon is documentation without consent — the legal shell of the thing, missing the thing itself. And remember whom the process protects: the participant, not the experimenter.
The slippery oneDeception and the debt of debriefing
If consent is essential, how is deception ever permitted? Because some phenomena evaporate under full disclosure. Announce "this is a study of conformity" and nobody conforms; announce "we're measuring prejudice" and everybody performs virtue. Deception buys naturalistic behavior, and it comes in two forms: commission (actively telling participants something false — a cover story) and omission (withholding a key detail until the end). The Ethics Code permits it only when the study has genuine value, no non-deceptive alternative exists, and — the bright line of Standard 8.07 — never about anything that would affect willingness to participate: physical pain, serious emotional distress, foreseeable harm. You may disguise the purpose of a task; you may not disguise its costs.
Deception also creates a debt, repaid at debriefing: the researcher reveals the deception, explains why it was necessary, undoes any false beliefs or discomfort, and restores an honest relationship before the participant walks out the door. Good debriefing does two jobs — dehoaxing (correcting the false information) and desensitizing (addressing any feelings the study stirred up). And there is a self-interested reason to use deception sparingly: deceived participants talk. A subject pool that expects trickery arrives suspicious, and suspicious participants guess at hypotheses and act on their guesses — a contamination problem you will meet again in Lesson 12 as demand characteristics.
Free choice, actually freeCoercion and undue influence
Consent must be voluntary, and voluntariness dies quietly. Coercion is a threat — explicit or structural — for refusing; undue influence is an offer too large or a relationship too unequal for refusal to be realistic. The two populations where this bites hardest are the two psychology uses most. Students: a professor who requires participation in their own study holds grading power over the "volunteers," which is why subject pools must always offer an equivalent non-research alternative (a short paper, attending a talk) for the same credit. Prisoners: for much of the twentieth century, American prisons were a favorite site for medical and drug testing — captive, cheap, and compliant — and Belmont's justice principle was written with them squarely in mind. Federal rules now sharply restrict prison research, on the logic that an environment of total control makes truly free consent nearly impossible. Payment follows the same principle: compensation for time is fine; an amount that makes people accept risks they would otherwise refuse is pressure wearing a gift's clothing.
The other committeeAnimal research: the IACUC and the 3 Rs
Human research goes to the IRB; animal research goes to the Institutional Animal Care and Use Committee (IACUC), which reviews housing, procedures, pain management, and whether the study is justified at all — and which, like the IRB, includes an outside community member and, typically, a veterinarian. The field's guiding framework is older than Belmont: the 3 Rs of Russell and Burch (1959). Replacement — use non-animal alternatives (simulations, tissue cultures, or a "lower" species) wherever possible. Reduction — design studies to use the fewest animals that can answer the question, which makes good statistics an ethical duty, not just a technical one. Refinement — modify procedures to minimize pain and distress. One honest caveat about coverage: the federal Animal Welfare Act's oversight famously excludes purpose-bred rats, mice, and birds — the overwhelming majority of lab animals — though institutions taking Public Health Service funds must apply protections to all vertebrates, and invertebrates remain largely outside the rules entirely.
When scientists cheatFabrication, falsification, plagiarism
Everything so far concerns how researchers treat participants. Research misconduct is about how they treat the truth, and the federal definition names three sins. Fabrication: making up data or results. Falsification: manipulating materials, procedures, or data so the record no longer honestly represents what happened — trimming inconvenient participants, doctoring images, quietly changing values. Plagiarism: presenting someone else's words or ideas as your own. Note what the definition protects: not participants directly, but the scientific record — and everyone downstream who trusts it, including patients, policymakers, and the students who build careers on it.
The modern cautionary tale is Diederik Stapel, a celebrated Dutch social psychologist who, for years, simply invented his datasets. He would design a study with collaborators, claim to run it at schools he had special access to, and return with clean, publication-ready numbers — for research that had never happened. Weeks before the fraud surfaced in 2011, he was in the international news claiming data showed that merely thinking about meat made people more selfish. Three university committees ultimately documented fraud across dozens of papers (Levelt Committee et al., 2012), and his retraction count reached 58. Two details deserve your attention. First, the data themselves raised flags — results too clean, effects too regular — which is why a methods course teaches you to read numbers skeptically. Second, the collateral damage: graduate students who did nothing wrong watched dissertations built on his "data" turn to ash. Fraud is not a victimless shortcut. Keep misconduct distinct, though, from the questionable research practices — flexible analyses, selective reporting — that turned out to be far more common than outright fraud; that reckoning is Lesson 15's story.
The 1971 SPE — remembered as proof that good people turn cruel in a bad situation — sits uncomfortably close to this section. Archival recordings and interviews show the "guards" were explicitly coached and pressured toward cruelty; the famous "prisoner breakdown" was later described by the participant himself as acting; and the study was run by its own "prison superintendent," steering events toward the expected result, without ever being peer-reviewed as an experiment (Le Texier, 2019). Whatever you call it — staged demonstration, self-fulfilling theater — its real lesson is how an evidence-free result can harden into fifty years of textbook "fact." Treat it as a case study in why the scientific record needs guarding, not as evidence about human nature.
The frontierEthics in the data age
The Belmont Report was written for clinics and laboratories — settings where a researcher and a participant could look each other in the eye. It never imagined a study run silently on hundreds of thousands of people through their news feeds. Yet that study happened. In 2014, researchers at Facebook and Cornell manipulated the emotional tone of 689,003 users' feeds for a week to test whether moods spread through networks — they do, a little (Kramer et al., 2014). No one was asked; no one could decline; "consent" was a clause buried in the terms of service. The backlash was global, and PNAS took the unusual step of publishing an Editorial Expression of Concern about the paper's consent and opt-out procedures.
The Facebook study defined the modern problem, and it has only grown. Platform-scale consent is broken: nobody reads the terms of service, a product team's A/B test can silently enroll millions, and "the platform agreed" is not the same as "the person agreed." Public is not the same as consented: scraping people's posts, photos, or profiles for research uses data they made visible for one purpose to serve another they never contemplated — and de-anonymization is far easier than it looks, since a few "anonymous" data points can often be linked back to a name. The Cambridge Analytica scandal showed the dark endpoint: personality data harvested from millions under the cover of a research app, repurposed for political targeting. The Belmont principles still govern this terrain in spirit — autonomy, harm, fairness — but the machinery of consent hasn't caught up, and your generation of researchers inherits the retrofit.
So how do you evaluate the ethics of any study — historic, modern, or algorithmic — without memorizing a new rulebook each time? Bioethicists distilled the requirements into a portable checklist (Emanuel et al., 2000), and it doubles as your interrogation kit:
Value: does the question matter enough to justify using people to answer it? Validity: is the design rigorous — because a study too flawed to yield real knowledge wastes its participants' risk by definition? Fair selection: who bears the burden, and were they chosen for scientific reasons or convenience? Risk–benefit: what does participation cost, and is it minimized? Independent review: did anyone outside the research team sign off? Consent: did participants agree — knowingly, freely, and in language they understood? Respect throughout: could they withdraw, was their privacy protected, were they debriefed? Run Tuskegee through this list and it fails all seven. Run the Facebook study through and watch which ones buckle.
Which Belmont principle is violated? Tap a scenario to reveal it.
SourcesCited in APA 7
American Psychological Association. (2017). Ethical principles of psychologists and code of conduct (2002, amended effective June 1, 2010, and January 1, 2017). https://www.apa.org/ethics/code/
Beecher, H. K. (1966). Ethics and clinical research. New England Journal of Medicine, 274(24), 1354–1360. https://doi.org/10.1056/NEJM196606162742405
Brandt, A. M. (1978). Racism and research: The case of the Tuskegee Syphilis Study. Hastings Center Report, 8(6), 21–29. https://doi.org/10.2307/3561468
Emanuel, E. J., Wendler, D., & Grady, C. (2000). What makes clinical research ethical? JAMA, 283(20), 2701–2711. https://doi.org/10.1001/jama.283.20.2701
Kaposi, D., & Sumeghy, D. (2026). From legitimate to illegitimate violence: Violations of the experimenter's instructions in Stanley Milgram's "obedience to authority" studies. Political Psychology, 47(2), Article e70112. https://doi.org/10.1111/pops.70112
Kramer, A. D. I., Guillory, J. E., & Hancock, J. T. (2014). Experimental evidence of massive-scale emotional contagion through social networks. Proceedings of the National Academy of Sciences, 111(24), 8788–8790. https://doi.org/10.1073/pnas.1320040111
Le Texier, T. (2019). Debunking the Stanford Prison Experiment. American Psychologist, 74(7), 823–839. https://doi.org/10.1037/amp0000401
Levelt Committee, Noort Committee, & Drenth Committee. (2012). Flawed science: The fraudulent research practices of social psychologist Diederik Stapel. Tilburg University, University of Amsterdam, & University of Groningen.
Middlemist, R. D., Knowles, E. S., & Matter, C. F. (1976). Personal space invasions in the lavatory: Suggestive evidence for arousal. Journal of Personality and Social Psychology, 33(5), 541–546. https://doi.org/10.1037/0022-3514.33.5.541
Milgram, S. (1963). Behavioral study of obedience. Journal of Abnormal and Social Psychology, 67(4), 371–378. https://doi.org/10.1037/h0040525
National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research. (1979). The Belmont report: Ethical principles and guidelines for the protection of human subjects of research. U.S. Department of Health, Education, and Welfare.
Russell, W. M. S., & Burch, R. L. (1959). The principles of humane experimental technique. Methuen.