A study becomes famous for one of two reasons.
Either the result is so surprising that nobody expected it, or the result is so tidy that it can be told in a sentence.
The second kind travels much further, because a sentence fits in a lecture slide.
Put ordinary people in a role and the role takes them over, said this one, and that is short enough to survive fifty years of retelling without anybody rereading the paper.
Which is a problem, because the retelling is where the checking should have happened.
It is where it stopped.
What a demand characteristic does to a result
Everyone taking part in a psychology setup is quietly working out what it wants from them.
That is not dishonesty. It is ordinary social behavior, and the cues that give the game away have a name of their own, which is demand characteristics.
If those cues are strong enough, the behavior on the record is not a response to the situation at all. It is a response to what was wanted.
The two look identical from outside the room, and the only way to separate them is a record of what people were actually told beforehand.
For most old work of this vintage no such record exists, because nobody imagined the briefings would matter later.
Here somebody kept them.
Twenty four young men and a corridor
The setting was a converted basement corridor in a university psychology building, fitted out with cells and a yard.
Twenty four male students were screened and assigned at random, eleven as guards, ten as prisoners and two held back in reserve.
The design was intended to run for two weeks with the researcher himself acting as prison superintendent, which is the detail people forget.
Prisoners were arrested at home by real police, brought in, stripped, deloused and given numbers, and the guards worked shifts with dark glasses and batons.
Within days some prisoners had broken down and some guards had become cruel, and the corridor had stopped looking like a university building at all.
A visiting graduate came to run interviews, saw what was happening and told the superintendent plainly that it was terrible to watch.
What the archive turned out to hold
The study began in the middle of August in 1971 and was shut down after six days rather than the planned fourteen.
Decades later a researcher went through the deposited papers and recordings, and published the findings in 2019.
Those materials show guards being actively coached rather than drifting into cruelty on their own. They were told to create a sense of powerlessness and to strip individuality from the prisoners.
One guard who was behaving too gently was told by the warden that every guard was going to be what they called a tough guard.
The most quoted guard of all later said he had a definite plan to force the action so the researchers would have something to work with.
None of that was in the write up, and all of it changes the reading, because coached behavior is not spontaneous behavior.
What the rerun did differently
In 2002 a broadcaster funded a replication with two social psychologists, and the setup was deliberately close to the original.
The difference was the briefing. Guards were given authority and no instructions at all about how to use it, no talk of powerlessness, no expectations set.
Tyranny did not appear. The prisoners formed a group, organized themselves, challenged the guards collectively, and the guards failed to cohere as a unit at all.
That is the opposite outcome from the same physical arrangement, in the way that a tall timber building or a printed school behaves differently once one input changes.
The archival work made the case bluntly, arguing the original guards were told to be tough, which is a documented claim.
What survives the demolition
It would be neat to say the whole thing was worthless and that is not right either.
Something real did happen in that basement. Young men did humiliate other young men within days, and several were hurt badly enough that the study had to stop.
What changed is the explanation. The original story was that roles swallow people automatically, which lets everybody off. The revised story is that people follow leadership when it tells them what the group expects of them.
That second version is less comfortable, because it puts the responsibility back on whoever is doing the telling, and somebody is always doing the telling.
It also survives the archive rather than being demolished by it, which is the test any explanation of this case now has to pass, as the reassessments of it keep concluding.
The prison study did not show that anyone becomes a monster in a uniform.
It showed what instructions do.
