Debunking the Stanford Prison Experiment

Thibault Le Texier, American Psychologist 74(7):823–839, October 2019. Read in full from the PDF. An archival study of the 1971 Stanford Prison Experiment, built from the Zimbardo papers at Stanford Special Collections, the recordings held at Akron, and interviews with 15 of the participants. The spoke's first case study for replication-crisis, and an awkward one, because nothing here failed to replicate. The original never established what it is quoted for.

Seven findings

Le Texier's content analysis produces seven, of which four are new to the record:

1. The design was lifted from a student experiment run three months earlier in one of Zimbardo's classes (Toyon Hall). 11 of the SPE's 17 rules were copied directly, the remaining six adapted to the new space. The class experiment goes unmentioned in the slideshow Zimbardo used for twenty years, in the documentary that replaced it, and in the published papers; the first academic account credits its undergraduate leader only in the acknowledgements.
2. The guards knew what result was wanted. At the orientation on 14 August 1971, which was filmed, Zimbardo told them he had a grant to study "the conditions which lead to mob behavior, violence, loss of identity, feelings of anonymity," and the variables were written on a blackboard behind him: boredom, frustration, fear, arbitrariness, loss of privacy, loss of freedom of action, loss of individuality, powerlessness.
3. The guards received precise instructions on how to treat the prisoners, and were not told they were subjects. They were cast as staff, briefed on the effect to produce, and then their compliance was reported as spontaneous role-adoption.
4. The prisoners could not leave at will, and the harsh conditions were designed by the experimenters rather than improvised by guards.
5. Participants were almost never fully immersed in the situation.
6. Data collection and reporting were incomplete and biased.
7. The conclusions were written in advance, for aims Le Texier calls nonacademic.

The famous details, corrected

The prisoner "breakdown" that ends the story in every textbook is documented from the other side. Prisoner 8612 says in a 1972 tape, and again in a filmed interview in the late 1980s, that he concluded illness would not get him released and so decided to "really act out, really act bizarre, to get out." Le Texier does not claim to know whether he was faking; he notes the testimony is consistent across two decades and that the man wanted to look like he was breaking down. Of the four others released early, two had burst into tears, one had a rash caused by the conditions and by being deprived of his medication, and one was sweeping a corridor when he was told he could go. "Half of them became so psychologically disturbed they had to be released prematurely" is, on this record, an exaggeration.

Why the story survived anyway

The paper's second half is about transmission rather than method, and it is the part this wiki should keep. A majority of textbook authors kept teaching the study without the criticisms that were already published. Fromm's ethical and personality-screening objections date from 1973, two years after the study. Le Texier offers four explanations: Zimbardo actively made the experiment sound more dramatic than it was ("it is really like a Greek drama"), and worked with media on purpose ("I have to give it away to the media and they have to sell it to the public"); the study became a weapon in the situationist-versus-personality argument in social psychology; the Abu Ghraib connection renewed it; and the parable shape — good people, evil place — is more teachable than the result.

What it changes here

replication-crisis has been a claim about a literature with no case attached to it. This is a case, and it points somewhere different from a failed replication. A failed replication says a finding was noise. This says the finding was produced to order and reported as though it had emerged, and that the discipline's error-correction ran forty-six years behind its own critics. The relevant failure was in what got taught, which is a distribution problem, not a statistics one.

It also sharpens the spoke's standing worry about second-hand sourcing (growth edge 2). Every textbook account of the SPE was second-hand, all of them derived from a first-party telling by the person whose grant it was — the same structure as the corpus's Jung pages before psychology-of-the-unconscious and psychological-types, with the added twist that here the primary record contradicts the primary author.

One caution, recorded rather than resolved: Zimbardo published a response in June 2018 to the overlapping claims in Ben Blum's article, and Le Texier says it did not address most of the findings. This wiki holds one side of a live dispute, from the side with the archive.