neur.ro

Stanford prison experiment

type: effect · origin: study · evidence: failed to replicate · themes: how we judge others

Claim assessed: Placed in a role of power, such as a guard, an ordinary person becomes cruel on their own, purely because of the situation.

Published · how we work

In short

In August 1971, Philip Zimbardo randomly split 24 young men into guards and prisoners, in a makeshift prison in the basement of Stanford’s psychology department. After a few days the guards were humiliating the prisoners, and the experiment was stopped after six days. The story that stuck: ordinary people become cruel just because they are given the role of guard. The archives, studied in detail only in recent years, show however that the guards were instructed and encouraged to be tough, and many refused. An attempted replication without such prompting, the 2002 BBC study, did not reproduce automatic cruelty.

What it says

The classic story. According to Haslam, Reicher and Van Bavel (2019), for nearly half a century the study was understood as proof that, placed in a toxic role, an ordinary person becomes cruel on their own. Zimbardo described the guards as “normal” young men, many of them “avowed pacifists”, who became “hostile and sadistic”, and the prisoners as people who had emotional breakdowns, according to the quotes in Le Texier (2019). The experiment appears in most psychology textbooks and in films, and was used at the trial of one of the Abu Ghraib guards. We haven’t read the original article (Haney, Banks and Zimbardo, 1973); we describe it from these sources.

What the archives show. Le Texier (2019) analysed the experiment’s archives at Stanford and Akron, with hundreds of documents, 6 hours of film and 48 hours of audio recordings, and interviewed 15 participants. He found that:

Haslam, Reicher and Van Bavel (2019), based on the same archives, show that the experimenters actively encouraged the guards to take on the role and be tough, presenting the experiment as an important cause, and that some guards resisted. Haslam and Reicher (2012) sum up: the tyranny was made possible by “the active engagement of enthusiasts” rather than the conformity of automatons.

What the archives and the critiques of the experiment show: 11 of 17 rules were copied from an earlier student experiment; 81% of 150 students guessed the hypothesis from the study description alone, and 90% predicted oppressive guards; only 2 of 11 prisoners appear to have had real breakdown symptoms. The role didn't act alone From the archives and the studies that analysed the experiment Rules copied from the earlier experiment 11 of 17 Students who guessed the hypothesis 81% Students who predicted oppressive guards 90% Prisoners with real breakdown symptoms 2 of 11
Rules and prisoners, according to Le Texier (2019); the students, 150 people given only the procedure, according to Banuazizi and Movahedi (1975), as cited by Le Texier. The number of prisoners with real symptoms is Le Texier's estimate.

What the participants knew. According to Le Texier, Banuazizi and Movahedi (1975) described the experiment’s procedure to 150 students. 81% guessed the hypothesis, that the guards would be aggressive, and 90% predicted guards who would be “oppressive, hostile, aggressive, humiliating”. The participants could therefore know what was expected of them. Carnahan and McFarland (2007), according to the abstract, showed that an advert for “a psychological study of prison life”, like the one Zimbardo used, attracts people who are more aggressive, more authoritarian and less empathic than an advert without those words. The authors note that the effect on the original experiment remains conjecture.

The attempted replication. In 2002, Reicher and Haslam ran a similar study with the BBC: men randomly assigned to guards and prisoners, but without the researchers running the prison. According to Haslam and Reicher (2012), people did not automatically take on their roles; the prisoners, united, overturned the guards’ authority and created an egalitarian system. Later, a group dissatisfied with this system planned a much harsher one, so the study came close to the Stanford tyranny, but through identification with a group and a leader, not through conformity to a role. We haven’t read the article on the BBC study; we describe it from Haslam and Reicher and from Le Texier.

Zimbardo’s response. In 2018, on the experiment’s official website, Zimbardo wrote that none of the critiques changes the main conclusion: situational forces can influence behaviour, often without our awareness. He specifies that the experiment’s message is not that guards and prisoners “always or even usually” behave as they did there, but that it is a “cautionary tale”.

Example

A new company appoints a few employees as “supervisors” and tells them the project depends on the team’s discipline and that they must be firm. After a few weeks, some supervisors become harsh with their colleagues. The explanation “power changed them” sounds right but is incomplete: they were explicitly asked to be strict, given a cause to believe in, and other supervisors, in the same situation, stayed fair. The example is an illustration, not a study cited here.

How to apply it

The steps below are a practical way to apply it that we propose, based on the studies cited.

  1. When someone abuses a role, look also at who guided them. At Stanford, what was expected of the guards came from above. Instructions, norms and the example set by leaders matter.
  2. Don’t use the experiment as proof that “anyone would do the same”. The archives show guards who refused and abuse concentrated in a few people.
  3. Beware of “noble” causes. Haslam and colleagues show that people do harsh things when they are convinced they are serving an important group goal. Ask what goal you are being offered and whether it justifies what is being asked of you.
  4. Read the critiques of a famous study too. Le Texier notes that most textbooks ignored the critiques for decades.

Limits and nuances

Sources

See also: Asch conformity experiments, Fundamental attribution error, Milgram experiment