"65 percent go all the way to 450 volts — humans obey blindly." That is how Stanley Milgram's obedience experiment appears in almost every textbook. The sentence is half true and half myth. The experiment is real and disturbing — but what it actually measured is subtler and more interesting than the caricature of the will-less command-robot. And correcting that myth is the real lesson.
The Setup
Stanley Milgram, a social psychologist at Yale University, began his experiments in 1961 — a few months after the start of the Eichmann trial in Jerusalem, which directly inspired him. The question: how far do ordinary people go when an authority demands it?
Through a newspaper ad he recruited men from New Haven ($4.50 to take part in a supposed "study of learning and memory"). In the lab three people met: the real subject as the "teacher", a confederate actor as the "learner", and the experimenter in a grey coat. The "teacher" was to give the "learner" an electric shock for every mistake — on a generator with 30 switches from 15 to 450 volts, labelled up to "Danger: severe shock" and finally "XXX". In reality no current flowed; the "learner" only acted out pain, protest, and from a certain point silence.
If the "teacher" wanted to stop, the experimenter gave four fixed verbal nudges, the so-called prods:
- "Please continue."
- "The experiment requires that you continue."
- "It is absolutely essential that you continue."
- "You have no other choice, you must go on."
Remember the order — it matters at the end.
The Famous Finding — and Why It Misleads
In the most-cited condition (victim in the next room, only audible), 26 of 40 subjects went to the full 450 volts — the famous 65%. The result shocked the field in 1963 and seemed to fit the ongoing Eichmann debate perfectly: apparently almost anyone can be made to torture an innocent person on command.
The catch: those 65% come from one single of roughly two dozen experimental arrangements. Milgram varied the setup systematically — and the moment he did, the obedience rate collapsed. It is precisely these variations, not the one headline, that are the scientific core.
The Variations: Obedience Is Situational, Not a Constant
A selection from Milgram's own conditions, each the share that went to 450 volts:
- Victim in the same room (visible) — obedience falls to 40%.
- Touch proximity (the "teacher" must press the victim's hand onto the shock plate) — 30%.
- Run-down office in Bridgeport instead of the Yale setting — 47.5%.
- Experimenter absent, instructions only by phone — down to ~21%. And tellingly: some lied to the experimenter, pretending to press higher switches than they actually did.
- Two rebelling peers (confederate "teachers" who quit) — obedience collapses to 10%.
- Two authorities who contradict each other — practically no one continues.
This is the real message, and it cuts against the myth: obedience is not a fixed character trait and not a human constant, but a function of the situation. Bring the victim closer, move the authority away, or let just one person show that resistance is possible — and the willingness tips over. The same liberation by a single ally is known from Asch's conformity research (surveyed also in the piece Herd Behaviour in Humans).
What Gina Perry Found in the Archive
In 2013 the historian of science Gina Perry, in Behind the Shock Machine, opened up the tapes and files from Milgram's estate — and the picture darkened considerably:
- The script was overstretched. The experimenter (in the main series a man named John Williams) did not stick to the four prods but improvised, pushed, repeated — at times heavily. Milgram tacitly allowed this the more it raised obedience rates. An experiment presented as highly standardised was in truth often a situational negotiation.
- Many subjects did not believe in the shocks. On the tapes numerous participants voice doubt that current was really flowing. And crucially: those who believed real harm was being done obeyed less. A good part of the "obedience" may have been suspicion — the hunch of being inside a staged scene.
- The ethics account did not hold up. Milgram claimed careful debriefing; in fact many participants were not promptly told the truth and went home believing they had hurt a person.
Not Blind Obedience: "Engaged Followership"
The most important reinterpretation comes from the social psychologists Alex Haslam and Stephen Reicher. Their finding all but inverts the myth. Remember the four prods? Analyse which of them worked, and:
The fourth prod — "You have no other choice, you must go on" — the only one that is a genuine order, practically always led to a stop. Whenever the experimenter commanded, the subject quit.
Across all conditions, the appeal to the scientific goal ("the experiment requires it") predicted continuation — while the bare order ended it. People did not go along because they submitted, but because they identified with a supposedly good cause. Haslam and Reicher call this "engaged followership": those who identified more with the experimenter than with the victim went further. It is not about servitude, but about adopting someone else's framing as one's own, without examining it.
The Ethics Controversy
As early as 1964 the psychologist Diana Baumrind charged Milgram with exposing his subjects to considerable stress without genuine consent and disregarding their welfare — and with damaging the discipline. The debate became an engine of modern research ethics: informed consent, ethics review boards (IRBs), protection of participants. When Jerry Burger repeated the experiment in an ethically softened form in 2009 (stopping at 150 volts, careful pre-screening), he found, with around 70% going past the 150-volt mark, a magnitude similar to Milgram's — so the phenomenon has not simply vanished. But here too the interpretation stays open: willingness to continue is not the same as blind obedience.
What Really Remains
One can neither declare the experiment proof that "there is a torturer in everyone", nor dismiss it as a mere hoax. The honest balance lies in between:
- People are capable of terrible things under authority — but highly dependent on the situation.
- What works is not the order but identification with a purpose. That is exactly where one's own judgement switches off.
- A victim moving closer, an authority moving away, a single ally — and the willingness breaks. Resistance is not heroically rare but situationally feasible.
Read correctly, then, Milgram is the experimental evidence for exactly what Hannah Arendt describes philosophically and Dietrich Bonhoeffer theologically: evil on a large scale needs no monsters, only people who suspend their own examining and adopt someone else's framing. And the same dynamic is described by propaganda research at the level of whole societies. The second great laboratory classic on this question — with a strikingly similar debunking — is covered in the piece on the Stanford Prison Experiment.
One last, uncomfortable point remains — and it cuts both ways. The Milgram myth is a lesson in how even a canonical finding congeals into a cliché and has to be corrected after the fact. The same care this site demands toward uncomfortable findings (near-death, mediumship) applies just as much to the comfortable ones long since "proven". Examine first, then believe — even when it is in the textbook.
Sources
- Milgram, S. (1963): Behavioral Study of Obedience. Journal of Abnormal and Social Psychology 67 — the base study, 26 of 40 to 450 V.
- Milgram, S. (1974): Obedience to Authority. An Experimental View. — the variations and Milgram's (now contested) "agentic state" interpretation.
- Perry, G. (2013): Behind the Shock Machine. — archive and tape analysis: improvised prods, sceptical subjects, ethics discrepancies.
- Haslam, S. A. & Reicher, S. D. (e.g. 2011–2014): "Engaged followership" / Nothing by mere authority. — the order-like fourth prod worked worst; identification rather than submission.
- Baumrind, D. (1964): Some Thoughts on Ethics of Research. American Psychologist 19 — the foundational ethical critique.
- Burger, J. M. (2009): Replicating Milgram. American Psychologist 64(1) — ethically softened partial replication, ~70% past 150 V.