11.2 — Obedience, and the Power of the Situation

A volunteer is told he is helping with a study of memory. He reads word pairs to a learner in another room and, each time the learner answers wrongly, presses a switch delivering an electric shock. The switches rise in fifteen-volt steps to 450 volts, labelled from Slight Shock to Danger: Severe Shock to a final XXX.

At 150 volts the learner asks to stop. At 300 he screams and refuses to answer. Beyond that, silence.

There is no shock. The learner is an actor. The question is how far the volunteer will go.

Psychiatrists asked to predict beforehand said almost nobody would reach the end.

In Stanley Milgram's baseline condition, about 65 per cent went to 450 volts.

What the study actually showed

The number is famous and the conditions are what matter.

Milgram ran many variations, and obedience moved enormously.

When the experimenter gave instructions by telephone rather than in person, obedience dropped sharply.

When the participant had to physically place the learner's hand on a plate, it dropped further.

When two other apparent participants refused and walked out, obedience collapsed to a small fraction.

When the study was moved from Yale to a run-down office in another town, it fell.

And when the experimenter's authority was ambiguous, or two authorities disagreed, it fell to near zero.

So the real finding is not that people obey. It is that obedience is highly sensitive to specific features of the situation — physical distance from the victim, physical presence of the authority, the perceived legitimacy of the authority, and whether anybody else has refused.

The single most powerful variable was the presence of another person refusing, exactly as in 11.1.

The criticisms, which are serious

This study is quoted constantly and the problems with it are rarely mentioned.

Ethically it would not be approved today. Participants were placed under severe distress, and the debriefing has been argued to have been less thorough than Milgram described.

Archival work has complicated the picture. Examination of the original recordings and records — notably by the psychologist Gina Perry — found that the experimenter frequently went beyond the four scripted prods, improvising and pressuring participants; that many participants expressed doubt that the shocks were real; and that the reported obedience rates varied substantially across the many unpublished variations.

And the interpretation has shifted. Milgram framed it as people surrendering their agency to authority. An alternative account, supported by analysis of what participants actually said, is that they were persuaded to identify with the scientific project — obeying not blindly but because they had come to see themselves as part of something worthwhile. Under this reading it is less about submission and more about engaged followership, which is arguably more disturbing rather than less.

What survives. Ordinary people, in the right situation, will do things they would have said they never would. The specific numbers should be held loosely and the direction is not in doubt — conceptual replications in modified, ethically acceptable forms have found broadly similar patterns.

The Stanford prison study, which was not an experiment

You have heard that ordinary students, randomly assigned to be guards, became sadistic within days — supposedly showing that roles corrupt.

The account that has emerged from the archives is different, and the correct lesson is nearly the opposite of the popular one.

There was no control condition and no experimental manipulation. It was a simulation, not an experiment.

The guards were coached. Recordings released later show the research team instructing guards to be tough and to create an atmosphere of powerlessness. One guard's much-quoted brutality was, by his own later account, a deliberate performance — he had decided to give the researchers what they wanted.

And the most famous breakdown was reported by the participant himself, years later, to have been faked in order to be released.

Zimbardo disputed aspects of these criticisms. The methodological objections — no control, an experimenter who was also the prison superintendent, demand characteristics throughout — are not seriously disputed.

And a later study, run for the BBC with independent oversight, found no such drift. The guards there did not spontaneously become tyrannical, and the group that eventually became authoritarian was not the one the original account predicts.

The corrected lesson: people do not automatically become brutal when given a role. They become brutal when an authority they identify with tells them that brutality serves a worthwhile purpose — which is a more specific claim, better supported, and considerably more useful.

The bystander effect, corrected

The third story in this cluster, and it also needs updating.

The original claim: the more witnesses present, the less likely anyone is to help, because responsibility is diffused. The framing case was the 1964 murder of Kitty Genovese, reported at the time as having been witnessed by 38 people who did nothing.

The Genovese reporting was substantially wrong. Later investigation found far fewer people had a clear view or understood what was happening, several did act, and at least one called the police.

The laboratory effect is real, and meta-analyses confirm that intervention rates fall as group size increases.

But field research changed the picture significantly. A study using CCTV footage of real public conflicts across three countries found that someone intervened in the large majority of cases — around nine in ten — and that more bystanders present meant a higher chance that at least one person acted.

How to hold both. Diffusion of responsibility is a real mechanism that reduces any individual's likelihood of acting. In real emergencies, with enough people present, somebody usually does. The practical implication is unchanged: if you need help, name a specific person. You, in the blue shirt, call an ambulance. That removes the diffusion entirely.

The general lesson, which is the point of the page

Situations explain more behaviour than character does, and people consistently underestimate this — which is the fundamental attribution error from 5.8, in its most consequential form.

Two implications, in opposite directions, and you need both.

Be less confident that you would have behaved well. Almost everybody believes they would have refused, and the base rates say most would not have. Believing you are the exception is precisely what makes you vulnerable, because it means you are not watching the situation.

And be more willing to change situations rather than people. If ordinary people do bad things in certain arrangements, the leverage is in the arrangement — the reporting line, the incentive, the anonymity, the distance from consequences, the absence of anyone who can say no.

What makes it easier to refuse

Directly from the variations, and each is actionable.

One other person refusing. The strongest factor in both Milgram and Asch. Which means being the first refuser is the highest-value thing you can do in a room — not because you will win, but because you make refusal available to everybody else.

Proximity to the consequences. Distance from the person harmed makes harm easier. Anything that makes the affected person concrete and visible increases resistance, which is why decisions that harm people are so much easier to take at a distance.

Ambiguity of authority. Obedience requires a clear legitimate authority. Asking a plain question — "who is actually authorising this?" — degrades that surprisingly effectively.

Early refusal. Every step taken makes the next one easier to take (8.7's evidence mechanism, running the wrong way). It is much easier to refuse at 75 volts than at 300, because by 300 you have already established what kind of person you are being.

And an exit that exists. People comply partly because leaving feels unavailable. Knowing in advance what your exit is — what you would say, where you would go — makes it usable.

What to do with this

Stop assuming you are the exception. The useful version of this material is not learning that other people are weak. It is knowing which situational features make anybody more likely to go along, so you can notice them.

Learn the four warning features: an authority you respect, a gradual escalation, distance from the person affected, and nobody else objecting. When three of the four are present, you are in the situation.

Be the first refuser. It is the single most useful thing on this page and it works on everybody else in the room.

Name the specific person when you need help. You, in the blue shirt.

And decide your line in advance. Not in general — specifically. What would you not do, and what would you say. A line decided beforehand is available under pressure; a line worked out during is not (5.9).

Next: 11.3 moves from what one person does under pressure to what happens to the thinking of a group as a whole.