The Milgram experiment: a fingertip hovering over one toggle switch on a worn vintage instrument panel, the rest of the row falling into shadow
Case Studies

The Milgram Experiment: What the Famous 65% Leaves Out

hamza 5 min read

You already know the number. In a Yale basement in the early 1960s, ordinary people were told to give a stranger electric shocks, each one bigger than the last, past where he screamed, then went silent — and about 65% obeyed all the way to the end. That, in a sentence, is the Milgram experiment as most people remember it: proof that if you put a lab coat in the room, most of us will hurt someone on command.

It’s one of the most cited findings in the history of the field. It’s also, once you read Milgram’s own files, far shakier than that single number lets on. Two things the textbooks quietly leave out: most of his subjects didn’t actually obey, and the ones who believed the shocks were real were the likeliest to refuse.

The short version

  • The famous 65% is one condition — 26 of 40 people going to the maximum voltage (Milgram, 1963).
  • Milgram ran roughly two dozen variations. Across all of them, most subjects — 368 of 658, about 56% — defied the experimenter (Perry et al., 2020).
  • His own assistant found that in 18 of 23 variations, subjects who believed the shocks were real gave lower shocks than the ones who suspected a setup.
  • Belief in the pain made people quit. Doubt made them comply. That’s the reverse of the story everyone tells.

The Version You Were Taught

The whole thing was staged, and Milgram built the set carefully. A volunteer turns up believing he’s the “teacher” in a study of memory. The “learner” in the next room is an actor, wired to an electrode. Every wrong answer, the teacher is told to push a lever one notch higher, along a row of switches that ends at 450 volts. The learner protests, then pounds the wall, then falls silent. A man in a grey coat stands at the teacher’s shoulder and, each time he falters, offers one of four flat lines: please continue. In the best-known run, 26 of 40 people went the whole way. Milgram tied the result to the Holocaust, and it lodged in the culture for good.

Then the Archive Opened

Milgram’s papers arrived at Yale in 1993: boxes of notes and data, plus reels of audio from the sessions themselves. The researchers who dug in found the clean story fraying. Those four famous prods weren’t a script the experimenter held to; he improvised, argued, and leaned on people well past the standard lines. Most subjects were never debriefed on the spot — Milgram didn’t run a proper debrief until roughly three-quarters of them had already been through it. And there was data he’d gathered and simply never published.

Did They Even Believe It?

This is the crack running under everything. If some “teachers” suspected the shocks were fake, their obedience was closer to playing along with a bit of theatre. Milgram had the numbers to check. In 1962 he asked his assistant, Taketo Murata, to split subjects by how strongly they’d believed the shocks were real and compare what they did. The finding, buried in a report Milgram never published: in 18 of 23 variations, the people who fully believed they were hurting the learner gave lower shocks than the doubters. Reanalysing that same archive in 2020, Gina Perry and her colleagues stated it without flinching — “the perception of pain inflated subject defiance and that subject skepticism inflated their obedience.” Even among the fully obedient, about one in five didn’t believe the shocks were real.

Most People Said No

Then there’s the headline figure the Milgram experiment is famous for. The 65% belongs to a single baseline. Across the full run of variations, obedience lurched up and down with the staging, and taken together the majority walked away: 368 of 658, about 56%, refused to go to the end. When critics went after the ethics of the study, Milgram defended himself with his belief data — but he laid the table out backwards, claiming three-quarters of subjects had acted certain the shocks were real. The subjects who were most certain were the ones most likely to stop.

What This Doesn’t Prove

Be careful how far you carry it. None of this shows that nobody obeys, or that Milgram cooked the books. The effect of belief on the exact voltage was modest, and some scholars warn that people explaining themselves after the fact tend to flatter themselves. The point is smaller and stranger than a takedown: the result rode on how convincing the performance was, and “most people will obey” quietly drops the disbelief, the improvised pressure, and the majority who quit. The Milgram experiment belongs beside the other stories here where obedience isn’t the simple thing it looks — the loyalty cult leaders manufacture, the machinery of control inside NXIVM. And if you’re sure you’d have pulled your hand back at the first scream, that certainty is worth a second look. It’s the same blind spot that catches the clever.

Frequently Asked Questions

What did the Milgram experiment actually find?

In its best-known condition, 26 of 40 people (about 65%) delivered the maximum shock when told to. But that was one of two dozen variations. Add them all up and most subjects disobeyed — and Milgram’s own data show that believing the shocks were real made people more likely to stop.

Is the 65% obedience figure accurate?

It’s right for one baseline condition and misleading as a headline. Obedience swung wildly across Milgram’s variations, and across all of them about 56% of subjects (368 of 658) refused rather than obeyed.

Did participants know the shocks weren’t real?

Many suspected as much. Milgram’s assistant Taketo Murata linked belief to behaviour and found that in 18 of 23 variations, subjects who fully believed the shocks were real gave lower shocks than sceptical ones — and about a fifth of the fully obedient subjects doubted the shocks entirely.

Sources

The dark side of human behavior. Delivered to your inbox.

You might also like