The Milgram experiment results: why most people pressed the highest switch

How the experiment worked

The Milgram experiment is a set of obedience studies run by the psychologist Stanley Milgram at Yale. The first paper, “Behavioral Study of Obedience”, appeared in the Journal of Abnormal and Social Psychology in 1963.

The subjects were 40 men aged 20 to 50 from the New Haven area, recruited by newspaper advertisement and direct mail. They were postal clerks, teachers, salesmen, engineers and labourers, with schooling that ran from unfinished elementary school to doctorates. Each was paid $4.50 and told the money was his simply for coming to the laboratory, whatever happened afterwards.

Each man was told the study was about punishment and memory. A draw decided who would be “teacher” and who would be “learner”. The draw was rigged: the volunteer was always the teacher, and the learner was a member of the research team, seated in the next room.

The shock generator had 30 lever switches running from 15 to 450 volts in 15-volt steps, labelled in groups from “Slight Shock” up to “Danger: Severe Shock”. Every wrong answer meant moving one switch higher. The machine lit up and buzzed, but no shocks were delivered to anyone.

When the teacher hesitated, the experimenter was supposed to work through four fixed prods in order: “Please continue”, “The experiment requires that you continue”, “It is absolutely essential that you continue”, and “You have no other choice, you must go on”.

What it found

The headline Milgram experiment result is this: 26 of the 40 men worked their way to the last switch on the board, the one marked 450 volts. That is 65 percent.

Fourteen stopped. Five of them stopped at 300 volts, the point at which the learner first pounded on the wall. The other nine stopped somewhere between 315 and 375 volts. Nobody refused before 300 volts.

Almost nobody expected that. Before the study, 14 Yale senior psychology majors were asked to predict the outcome; on average they thought about one person in a hundred would go to the end.

Obedience did not look calm. Milgram recorded profuse sweating, trembling and stuttering as typical reactions. Fourteen of the 40 showed definite signs of nervous laughter, and three had what he described as full-blown, uncontrollable seizures.

Replications and criticism

Milgram did not stop at one condition. Gina Perry, who worked through his papers in the Yale archive, counts 24 separate conditions run between 1961 and 1962, with close to a thousand people passing through the laboratory. Only some were written up. Across the variants, the share of participants who reached 450 volts ranged from none of them to almost all of them, which makes the famous 65 percent one point on a wide scale rather than a constant of human nature.

One condition in the archive was never published. Milgram paired people who already knew each other — work mates, best friends, brothers-in-law, a father and son. His own note called it “as powerful a demonstration of disobedience than can be found”. He left no record of why he kept it back.

Perry also listened to the tapes. The experimenter did not stay inside the four prods. He improvised, roaming further and further off script, coaxing and pressing people to carry on. None of that appears in the published account.

The sharpest piece of Milgram experiment criticism came in 2020, when Perry, Augustine Brannigan, Richard Wanner and Henderikus Stam published a reanalysis in Social Psychology Quarterly. They went back to an unpublished 1962 questionnaire study by Milgram's assistant Taketo Murata — 656 post-experiment questionnaires across 23 conditions. Two results stand out. Taken across every condition, a majority of subjects, 56 percent, were defiant at some point rather than obedient. And subjects who fully believed the shocks were real were about two and a half times more likely to defy the experimenter than subjects who doubted the set-up. Believing in the pain pushed people towards refusal, not compliance.

Alexander Haslam and Stephen Reicher have pointed out that the one prod which is a plain order — “You have no other choice, you must go on” — is the one that fails; participants who heard it typically refused. What kept people going was the appeal to the experiment's value. They read this as engaged followership rather than blind obedience: people acting knowingly, not passively.

The nearest thing to a modern replication is Jerry Burger's 2009 study in American Psychologist. Seventy adults took part, 29 men and 41 women, aged 20 to 81. Burger stopped the procedure at 150 volts, on the grounds that 79 percent of Milgram's participants who went past that point went all the way. Seventy percent of his base-condition participants were still willing to continue at 150 volts, against 82.5 percent of Milgram's — a gap that was not statistically significant.

How it is seen today

On Milgram experiment ethics, the verdict is settled. The study would not be approved now. Burger put it plainly: “Current standards for the ethical treatment of participants clearly place Milgram's studies out of bounds.” To get his own partial replication past a review board he used a two-step screening process, told participants at least three times — twice in writing — that they could stop and keep the payment, gave only a mild 15-volt sample shock, and had the learner walk back into the room within seconds of the end.

The debriefing is the hardest part of the record to defend. From the archive, Perry found that three-quarters of subjects left the laboratory without being told the truth at the time. Many received a letter months later. Some learned what had happened much later than that.

So the Milgram experiment explained honestly comes out narrower than the version in circulation. The procedure, the numbers and the distress are all documented in a published paper and can be checked line by line. The conclusion bolted on afterwards — that this shows ordinary people will follow orders blindly — is the part the full record does not carry. Real or Legend files this as confirmed, and that stamp covers the experiment and its data, not the sentence people quote it for.