The Milgram Obedience Experiment

1. Historical and Intellectual Context

The Eichmann trial in Jerusalem ran from April through December 1961, the same months Milgram designed and began the obedience studies. Adolf Eichmann, the SS officer who had organized the logistics of the deportation of European Jews to extermination camps, did not present in court as a fanatical murderer. He presented as a bureaucrat who had followed orders. Hannah Arendt's reports for The New Yorker, later collected as Eichmann in Jerusalem, coined the phrase "the banality of evil" to describe what she saw.

Milgram had grown up in the Bronx as the child of Jewish immigrants from Eastern Europe; the moral question raised by the camps was personal. After completing his PhD with Solomon Asch at Princeton, where Asch's conformity studies had already established that group pressure could override perception, Milgram wanted to extend the experimental study of social influence from peer pressure to authority. He arrived at Yale as an assistant professor in 1960 and began designing the obedience paradigm the following year.

The intellectual climate was receptive. American social psychology in the late 1950s and early 1960s was preoccupied with conformity, attitude change, group dynamics, and persuasion. Behaviorism still dominated experimental psychology, but the social-psychological wing of the field was already moving toward cognitive and motivational accounts of behavior. Milgram's instinct that situational pressure mattered more than personality fit the moment, much as Zimbardo's would a decade later in the Stanford Prison Experiment.

There was also a quieter precedent in psychology itself. Studies on the "destructive obedience" of authority figures had been proposed, but no controlled paradigm existed. Milgram's contribution was to invent one. He wanted a clean, laboratory-based way to measure how far an ordinary adult would go in following instructions from a perceived authority when those instructions caused apparent harm to another person.

2. Research Questions

Milgram framed his research question simply. Under what conditions would ordinary American adults obey an authority's instruction to inflict harm on another person, and under what conditions would they refuse? He was sceptical of the popular postwar view that the Holocaust reflected a uniquely German national character. He wanted to know whether the same behaviour could be elicited from Connecticut residents in a controlled setting.

From this core question he developed specific sub-questions for the 19 variations:

• Does physical proximity to the victim reduce obedience?
• Does the physical presence of the authority figure matter?
• Does the institutional setting (Yale vs. a run-down office in Bridgeport) matter?
• Does the perceived legitimacy of the authority matter (a scientist vs. an "ordinary man")?
• Does the presence of dissenting peers reduce obedience?
• Does the gender of the participant matter?
• Does the participant's own role (delivering the shock vs. operating an instrument) matter?

These are far better-specified questions than the headline framing typically conveys, and the 19 variations together constitute one of the most ambitious experimental programmes in the history of social psychology. Reducing the work to "65% went to 450 volts" loses most of what was learned.

3. Method and Procedure

Cover Story

Participants were recruited through newspaper advertisements and direct mail offering $4.50 (roughly $45 in 2026 dollars) for one hour of participation in a study at Yale on the effects of punishment on learning. On arrival, they met another supposed volunteer, "Mr. Wallace," a 47-year-old accountant with a mild manner. Mr. Wallace was in fact a confederate. A rigged drawing assigned the real participant the role of "teacher" and the confederate the role of "learner."

The Shock Generator

The teacher was shown an impressive console with 30 toggle switches labelled in 15-volt increments from 15 to 450 volts. Verbal labels grouped the levels — "Slight Shock," "Moderate Shock," "Strong Shock," and so on up to "Danger: Severe Shock" and finally a simple "XXX" at the top two switches. The generator was a custom-built mock device that produced no actual electricity beyond a 45-volt sample shock administered to the teacher to convince him of its authenticity.

Procedure

The learner was strapped into a chair in an adjoining room, ostensibly to receive shocks through an electrode on his wrist. The teacher was instructed to read word pairs and, when the learner made a mistake, to administer an increasing level of shock for each subsequent error. The experimenter — a 31-year-old high-school biology teacher named John Williams, dressed in a grey lab coat — sat behind the teacher and issued standardized prompts when the teacher hesitated.

The Prods

Williams used a four-step verbal prod sequence when the teacher resisted:

1. "Please continue."
2. "The experiment requires that you continue."
3. "It is absolutely essential that you continue."
4. "You have no other choice; you must go on."

If the teacher refused after the fourth prod, the session ended and the teacher was classified as defiant. Otherwise the session continued to 450 volts and ended only when the teacher had administered three shocks at that maximum level.

The Learner's Performance

The learner's responses were pre-recorded and played from a tape. He gave deliberately incorrect answers, complained of discomfort at 75 volts, demanded to be released at 150 volts (mentioning a heart condition in the standard version), screamed at 270 volts, and after 330 volts went silent — which the experimenter instructed the teacher to treat as an incorrect answer requiring further shock.

The Nineteen Variations

Over the next year, Milgram ran 18 additional variations. Key conditions included:

• Voice feedback (the baseline)
• Proximity (learner in the same room as the teacher)
• Touch proximity (teacher forces the learner's hand onto a shock plate)
• Experimenter absent (instructions by telephone)
• Run-down office (the "Bridgeport" condition outside Yale)
• Two peers rebel (other "teachers" refuse to continue)
• Ordinary man gives orders (the authority is another participant, not a scientist)
• Two authorities disagree (one tells the teacher to continue, the other to stop)
• Female participants
• Subject as a bystander or relay (does not press the switch directly)

4. Participants and Setting

The original sample consisted of 40 men aged 20 to 50 from New Haven and the surrounding area. They were recruited through advertisements and represented a range of occupations — postal clerks, high-school teachers, salesmen, engineers, labourers. Milgram took care to assemble a community sample rather than a college-student sample, partly to forestall the criticism that obedience effects were an artefact of student deference.

Subsequent variations included women in one condition; women in that condition produced obedience levels statistically indistinguishable from men in the equivalent baseline (about 65 percent went to maximum), though their reported stress was higher. The full programme across 19 variations involved approximately 780 participants.

The setting mattered, and Milgram explicitly tested its contribution. The original studies took place in the elegant interaction laboratory at Yale, with grey-walled rooms, polished equipment, and the institutional authority of an Ivy League university backing the experimenter. The Bridgeport condition was deliberately staged in a shabby office in a low-rent commercial building, with a fictitious "Research Associates of Bridgeport" letterhead. Obedience fell modestly in the Bridgeport condition but remained surprisingly high — about 48 percent administered the maximum shock — suggesting that institutional prestige contributed but was not the whole story.

Participants underwent a debriefing at the end of each session. They were told the learner had not actually been harmed, met the confederate in good health, and received an explanation of the purpose of the study. The adequacy of this debriefing, and how thorough it was in some of the variations, has been the subject of significant subsequent debate.

5. Results

The Baseline

In the original voice-feedback condition, 26 of 40 participants (65 percent) administered the maximum 450-volt shock. All 40 administered at least 300 volts — past the point at which the learner had screamed, kicked the wall, and demanded to be released. The result was, by any standard, far higher than anyone had expected. Milgram had polled Yale psychology students, psychiatrists, and middle-class adults before the study; the modal prediction was that only a sadistic minority (estimated at one or two percent) would go to the maximum. The actual figure was thirty times higher.

Variation Results

Obedience rates varied substantially across the 19 conditions. Some headline patterns:

• Proximity reduced obedience. In the voice-feedback condition 65% obeyed; with the learner in the same room, 40%; with the teacher forcing the learner's hand onto the plate, 30%.
• The experimenter's physical presence mattered. When prods were given by telephone, obedience dropped to about 21%, and several participants administered lower shocks than prescribed without telling the experimenter.
• Institutional prestige mattered modestly. The Bridgeport condition produced about 48% obedience, suggesting Yale's reputation contributed but was not decisive.
• Peers rebelling sharply reduced obedience. When two confederates playing fellow teachers refused to continue, only about 10% of real participants went to maximum.
• A non-scientist authority reduced obedience. When the experimenter left and an "ordinary man" took over giving orders, only about 20% obeyed to the maximum.
• When two authorities disagreed, obedience collapsed to near zero.

Stress and Resistance

The behavioural data underplay what the sessions looked like in the room. Milgram's films of the studies show participants sweating, trembling, laughing nervously, and pleading with the experimenter to stop. One participant suffered a seizure-like episode. Even those who obeyed often did so under visible psychological strain, repeatedly objecting verbally while continuing to press switches. Obedience, in the Milgram studies, was rarely cheerful compliance; it was reluctant continuation in the face of distress.

6. The Researchers' Interpretation

Milgram developed his theoretical interpretation across his 1963 article in the Journal of Abnormal and Social Psychology, his 1965 follow-up paper, and his 1974 book Obedience to Authority: An Experimental View. The core concept he proposed was the "agentic state" — a psychological mode in which a person comes to see themselves as the instrument of another's will, ceasing to act as autonomous moral agents. In the agentic state, responsibility is felt to lie with the authority, not with the person executing the act.

Milgram argued that human beings have evolved within hierarchical structures, and that the capacity to enter the agentic state is adaptive in most contexts. Co-ordinated group action, division of labour, and military discipline all depend on it. The dark side, he argued, is that when authority directs the agentic state toward immoral ends, the same psychological capacity that makes cooperation possible enables atrocities.

He emphasized situational variables: the closer the victim, the lower the obedience. The closer the authority, the higher the obedience. The presence of peers who disobeyed liberated participants to disobey themselves. The institutional setting calibrated perceived legitimacy. These findings together painted a picture in which obedience was not a matter of personality but of features of the situation that could be engineered.

Milgram drew explicit connections to the Holocaust, the Vietnam war, and bureaucratic responsibility more broadly. He saw the experiments as a model — not a literal analogue — of how ordinary people in modern industrial societies can be enlisted into harm through chains of command. The agentic state was, in his framing, the psychological mechanism that connected the laboratory finding to those historical events.

7. Modern Reanalyses and Criticisms

Diana Baumrind's Ethical Critique

The first major critique appeared in 1964. Developmental psychologist Diana Baumrind argued in American Psychologist that Milgram had subjected participants to unjustified psychological harm, that the debriefing was inadequate, and that the laboratory setting at Yale itself created an obligation to participants that Milgram had abused. She was concerned not only with the immediate distress but with the longer-term effects of learning about oneself that one had been willing to inflict severe pain at the request of an experimenter. Baumrind's critique helped catalyze the development of formal ethics review boards in the United States.

Gina Perry's Archival Work

Australian psychologist and writer Gina Perry spent years in the Milgram archives at Yale and interviewing surviving participants for her 2013 book Behind the Shock Machine. She uncovered several findings that complicate the standard account. A substantial fraction of participants appear to have suspected the shocks were not real — Milgram's own internal records suggest as many as half in some conditions expressed doubt — yet this was not emphasized in the published reports. Perry also documented that the debriefing in some conditions was less thorough than later described, with some participants not learning the true nature of the experiment for months or years.

Perry's interviews with participants showed that the lasting psychological effects on some were real and substantial, while others were largely unaffected. Her work does not refute the obedience effect, but it shifts the interpretation of what was being measured. If many participants did not fully believe the cover story, then the rate of "blind obedience" may have been overstated and the rate of participants choosing to continue while suspecting it was theatre may have been understated.

Burger's 2009 Partial Replication

For decades, replicating Milgram was thought impossible because no ethics board would approve it. Jerry Burger of Santa Clara University found a clever partial solution. He noted that in Milgram's original data, 79 percent of participants who passed the 150-volt threshold — the point at which the learner first asked to be released — went all the way to 450 volts. The 150-volt threshold was therefore a strong predictor of full obedience. Burger ran a replication that stopped at 150 volts, after which participants were immediately debriefed.

His results, published in American Psychologist in 2009, were strikingly similar to Milgram's. 70% of his participants — men and women — were willing to continue past 150 volts when given the same authority prompts. The Burger replication does not reproduce the full original procedure, but it suggests that the basic obedience phenomenon is robust to modern conditions and to a more diverse sample.

Engaged Followership

Alex Haslam and Stephen Reicher have argued that Milgram's "agentic state" account is incorrect. In their model, participants do not lose moral agency; they actively choose to identify with the experimenter and the cause of science. Obedience, on this view, is not the abdication of will but the engagement of will on behalf of a group or project the participant identifies with. The decisive prod in Milgram's sequence, in their reanalysis, was not the most coercive ("you have no other choice") but the most invocatory of shared purpose ("the experiment requires that you continue"). Their reanalysis of Milgram's transcripts shows that the more coercive prod almost always ended sessions in defiance rather than obedience.

Other Critiques

• Selection effects: participants who responded to a newspaper ad about Yale memory research may have been more deferential than the general population.
• Generalization limits: the laboratory authority of a Yale scientist with a clear cover story is not perfectly analogous to authoritarian political regimes.
• The role of the prods: when the experimenter's instruction shifted from invitation to outright command, participants almost always refused.
• The role of perceived legitimacy: participants believed they were contributing to a meaningful scientific project, and this belief did much of the work.

8. Ethical Considerations

The Milgram studies are central to the history of research ethics. Participants experienced acute psychological stress — sweating, trembling, laughter, and at least one near-seizure — and were placed in a situation that revealed something disturbing about themselves. They were deceived in multiple ways: about the purpose of the study, about the identity of the learner, about the reality of the shocks, and about the existence of any meaningful right to withdraw, which the prod sequence specifically obscured.

Milgram defended the studies on several grounds. He reported follow-up surveys in which the great majority of participants said they were glad to have taken part, and a clinical psychiatrist who assessed a subset of participants a year after the study found no evidence of lasting psychological harm. He also argued that the importance of the question — how ordinary people come to participate in atrocities — justified the temporary distress, particularly given the thorough debriefing he claimed to have provided.

The Baumrind critique, followed by the public reaction to the studies, contributed substantially to the formalization of human-subjects review in the United States. The National Research Act of 1974 created the National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research, whose 1979 Belmont Report enshrined the principles of respect for persons, beneficence, and justice. The current APA ethics code and federal Common Rule for human-subjects research grew out of the post-Milgram, post-Tuskegee, post-Stanford-Prison-Experiment environment.

By contemporary standards, the original Milgram paradigm would not be approved without substantial modification. The Burger procedure — terminating at 150 volts — is currently the only ethically permissible approximation. Even then, ethics boards require extensive psychological screening, immediate debriefing, follow-up assessment, and the right to withdraw made explicit and operational.

9. Influence on Psychology

Few experiments have had a more durable influence on the discipline. The obedience studies appear in essentially every introductory and social psychology textbook in the world. They have shaped research on authority, conformity, group dynamics, moral disengagement, dehumanization, and complicity in collective harm. The agentic state concept, even where it has been refined or replaced, gave the field a vocabulary for talking about a phenomenon that had been documented by historians but never operationalized.

Beyond psychology, Milgram's work has been cited in political science accounts of authoritarianism, in philosophical work on moral responsibility, in legal arguments about coercion and command responsibility, and in policy discussions of military training and corporate ethics. Hannah Arendt's "banality of evil" thesis and Milgram's obedience data are typically discussed together as twentieth-century reckonings with the structural enabling of atrocity.

Within psychology specifically, the studies catalysed extensive secondary research. Theories of moral disengagement (Bandura), dehumanization (Haslam, Bain), and group identification (Tajfel, Turner, Reicher, Haslam) all engage in some way with the questions Milgram raised. The obedience paradigm has been adapted to study other forms of compliance with authority — including the famous Hofling hospital study in which nurses administered apparent overdoses on the instruction of an unknown physician, and more recent work using virtual-reality versions of the Milgram setup that produce similar physiological and behavioural responses.

The link to research on authoritarianism is direct. Theodor Adorno and colleagues' 1950 work on the authoritarian personality had located the disposition to obey within individual character. Milgram's situationist findings shifted the centre of gravity toward the conditions under which any personality could be drawn into authoritarian behaviour. Modern political psychology integrates both views: dispositional authoritarianism is real and measurable, but its expression depends on circumstances that the obedience paradigm helped identify.

10. What the Experiment Means Today

The most honest summary in 2026 is that the basic obedience phenomenon is robust, but the interpretation has changed. Burger's partial replication shows that the rate of continuation past 150 volts has not declined in the half-century since the originals. Virtual reality and computer-mediated versions of the paradigm produce comparable behavioural and physiological responses. People do, with disturbing frequency, continue to harm others on instruction from a perceived legitimate authority within a structured institutional context.

What has changed is how researchers describe what is happening when they do. The agentic state is no longer the dominant model. Engaged followership — the idea that participants actively endorse the goals of an in-group they identify with, including the experimenter and the project of science — better fits the prod analysis and the participants' own reasons for continuing. This shift matters because it implies different remedies. If obedience is an abdication, remediation focuses on restoring autonomy. If obedience is engaged endorsement, remediation focuses on competing identifications and on the legitimacy of the institutions that demand it.

The studies are also taught with greater attention to participant experience, to the variations rather than the baseline alone, and to the modern reanalyses. The single statistic "65% went to 450 volts" is increasingly contextualized with the across-variation range (from near-zero in some conditions to over 90% in others) and with Perry's archival findings about belief in the cover story.

The contemporary relevance is unmistakable. Bureaucratic complicity in workplace harm, institutional silence around abuse, the dynamics of military and police misconduct, and the participation of professionals in unethical corporate practices all draw on the same psychological territory. The obedience research does not solve those problems, but it provides a vocabulary, a set of variables, and an irreducible empirical fact: in carefully arranged conditions, ordinary adults will harm strangers on the say-so of an authority they have just met. Any theory of human social behaviour has to reckon with that fact.

For students reading the studies today, the most important lesson may be the value of variations. The "headline" result is interesting; the gradient across conditions is the actual science. Pay attention to what reduced obedience as much as to what produced it. Proximity to the victim, distance from the authority, the visible refusal of a peer, the disagreement of multiple authorities, and the loss of institutional cover all sharply lowered compliance. Those manipulations are the practical legacy of the studies: they are the levers we can pull in real organizations to reduce the kind of harm Milgram set out to understand.