The Stanford Prison Experiment: Great Story, Shaky Science

MayaA young man in a numbered smock is screaming. "I mean, Jesus Christ, I'm burning up inside!" That clip is in pretty much every documentary about the dark side of human nature. Decades later, the man who screamed it said he was faking.

LeoWait, seriously? That's the scream. That's the Stanford Prison Experiment scream.

MayaThat's the one. And it's only the start of how strange this story gets.

LeoHi everyone, I'm Leo.

MayaAnd I'm Maya.

LeoSo here's why we're doing this one now. Netflix just put out a reality show called The New Stanford Prison Experiment. Six episodes, thirty volunteers, unscripted.

MayaThe official website for the original study says it's not affiliated with Zimbardo's estate. Philip Zimbardo, who ran the original, died in October twenty twenty-four.

LeoSo before anyone binges it, let's go back to the original. Set it up for me.

MayaAugust nineteen seventy-one. Zimbardo is a psychology professor at Stanford, and he's working with two graduate students, Craig Haney and Curtis Banks. The money comes from the US Office of Naval Research, because the Navy and Marines were interested in conflict between military guards and prisoners.

LeoHuh, I didn't know the Navy was involved.

MayaMost people don't. They ran a newspaper ad looking for male college students for, quote, a psychological study of prison life. It paid fifteen dollars a day, which is about a hundred and twenty dollars in today's money.

LeoHonestly, as a broke student? I'm in.

MayaSeventy-five men applied. After a questionnaire and an interview, they picked the twenty-four who seemed most stable, most mature, and least involved in antisocial behavior. Mostly white, mostly middle class. Then they were randomly assigned to be guards or prisoners. Random means chance decided, not personality. That's the key to the whole classic interpretation.

LeoHow many on each side?

MayaUsually it's described as nine guards and nine prisoners, plus some alternates. But the official report doesn't agree with itself. In one place it says twenty-two participants, in another twenty-four were selected, and it analyzes ten prisoners and eleven guards. It's a small thing, but it's sloppy.

LeoNoted. Researcher brain is switched on.

MayaOkay, the Sunday after the guards' orientation, real Palo Alto police officers show up at the prisoners' homes with no warning and arrest them for armed robbery and burglary. They're booked, fingerprinted, blindfolded, and driven to the basement of the psychology building.

LeoReal cops? That's wild.

MayaReal cops. The prison is a hallway about thirty-five feet long, with three little cells about six feet by nine feet, three prisoners to a cell. A closet is used for solitary confinement, and they call it the hole. The prisoners get strip-searched and deloused, then they're given a smock with no underwear, a stocking cap, a chain on one ankle, and an ID number.

LeoOof. And the guards?

MayaKhaki uniforms, mirrored sunglasses, wooden batons. They work eight-hour shifts, three guards at a time, and then go home. The prisoners are there around the clock.

LeoThose mirrored sunglasses are doing a lot of work.

MayaThey are. Now, who's in charge? Zimbardo casts himself as the prison Superintendent. An undergrad named David Jaffe plays the Warden. So the man running the study is also a character in it. Remember that, because it matters later.

LeoWhat were the guards actually told to do?

MayaThey were told not to use physical violence. But the orientation was recorded, and Zimbardo told them they could create, quote, boredom, a sense of fear to some degree, and a sense of powerlessness. And that isn't a critic's spin. Zimbardo quotes those lines himself.

LeoHmm. Okay, filing that away.

MayaBefore the results, it's your turn. It was planned to run up to two weeks. What happens?

LeoOkay. Day one, everybody's giggling, it's like a costume party. By day three the guards are getting bossy and kind of mean. Maybe one prisoner quits. And I think they make it, I don't know, ten days before someone calls it off.

MayaSix days. And one prisoner quitting? Try five.

LeoFive?

MayaFive of the roughly ten prisoners were released early. For four of them, the report lists extreme emotional depression, crying, rage, and acute anxiety, starting around day two. The fifth developed a psychosomatic rash, meaning a rash brought on by stress, not by any physical cause. Prisoner eight-six-one-two, Douglas Korpi, the scream guy, was out after about thirty-six hours.

LeoWhat happened on day two?

MayaA rebellion. The prisoners barricaded their cells. The guards sprayed them with fire extinguishers, took away their beds and clothes, and set up a privilege cell.

LeoDivide and conquer. Day two!

MayaAnd here's the strange part. At a mock parole hearing, five prisoners were asked whether they'd give up their pay to be paroled. Three said yes. Then they went back to their cells.

LeoInstead of just quitting? They could walk out!

MayaThat's the classic reading: they'd stopped seeing it as a study. But there's a wrinkle. Prisoners who asked to leave reportedly weren't let go right away, so walking out may not have felt like an option. Also, roughly ninety percent of what the prisoners talked about among themselves was prison conditions, and only about ten percent was life outside.

LeoAnd the guards turned into monsters. I've heard about a third of them became sadists.

MayaCareful. The report describes three kinds of guard: tough but fair, creatively cruel and harassing, and passive or good. But the version I'm working from gives no counts and no percentages. The one-third figure is Zimbardo's own later description. It's an estimate, not a measurement.

LeoFair. So what did the guards actually do?

MayaCommands were the most common kind of thing guards said, and insults were common too. Guards were never late for a shift, and some stayed late without being paid.

LeoSo how did it end?

MayaChristina Maslach, who had just finished her PhD and was dating Zimbardo, visited on the evening of day five. They married the next year. She saw prisoners being marched to the toilet with bags over their heads and protested strongly. Zimbardo stopped the study the next morning. The official report, to be fair, doesn't give a clear reason for stopping.

LeoSo it took an outsider to say this is messed up.

MayaIn the usual telling, yes. There are a couple more famous stories, and these come from Zimbardo's own accounts, like his book The Lucifer Effect, not from independent sources. Prisoner four-sixteen went on hunger strike and was put in the hole. The other prisoners were told they could give up their blankets to get him out, or keep their blankets and leave him in. Most kept their blankets. Separately, guards had the prisoners chant that prisoner eight-nineteen was a bad prisoner while eight-nineteen sobbed nearby.

LeoThat's bleak.

MayaAnd now the twist. Zimbardo took it to the press and testified to a congressional subcommittee in October nineteen seventy-one, about two months later. That was before any peer review, meaning before other scientists had checked the work. The first refereed paper came out in nineteen seventy-three, in a criminology journal that closed in nineteen seventy-eight.

LeoSo the headline came before the homework.

MayaPretty much. So, did it hold up? Start with the design. There was one run and no control group.

LeoQuick gloss for the people in the back. That's me.

MayaA control group is a comparison group. If you want to know whether the prison setting caused cruelty, you'd compare it to a version without the setting, or without the coaching. Without that, you can't say what caused what. A Stanford colleague, Gordon Bower, reportedly visited and asked what the independent variable was, meaning what are you changing so you can measure its effect? There wasn't one. And the study has never been directly replicated.

LeoNever? It's in every textbook!

MayaIt's a single case study that got treated like a law of nature. In nineteen seventy-five, two researchers, Banuazizi and Movahedi, described the setup to about a hundred and fifty students. Most of them guessed what the study was trying to show, and around ninety percent predicted the guards would be oppressive, hostile, and aggressive. I'll flag that the numbers reach us through a later summary, so they need checking. Their point was demand characteristics.

LeoWhich means people figure out what the experimenter wants and give it to them?

MayaExactly. They were acting out what guards and prisoners are supposed to be like. And that brings us to the most notorious guard, Dave Eshelman, nicknamed John Wayne. He used a Southern drawl, and for decades he's said he was deliberately playing a part, based on the sadistic warden in the movie Cool Hand Luke. He was trying to give the researchers something to study.

LeoOkay, but hold on. Even if it was a role, he chose to be that guy.

MayaTotally fair. He came up with degrading ideas on his own, including making prisoners simulate sex acts. So was he showing us human nature, or doing a performance? Maybe some of both. Then there's Korpi, the scream. He later said he faked his breakdown so he could get out and study for the GRE, the grad school entrance exam. He also said he'd asked to quit and wasn't allowed.

LeoHe faked the most famous meltdown in psychology to go study?

MayaThat's his later account. Zimbardo said he believed the later interview was the lie. And Zimbardo made a fair point: even if Korpi was faking, the researchers had to treat an apparent breakdown as real.

LeoWho dug all this up?

MayaA big one is the researcher Thibault Le Texier. He went through the Stanford archives and interviewed fifteen participants. His book came out in twenty eighteen, and a paper in American Psychologist followed in twenty nineteen. He argued that the guards were briefed and pushed to be tough. He found that many of the rules and schedules came from a prison simulation Jaffe ran as a class project that spring, in a Stanford dorm, which Zimbardo's accounts rarely mention. He also argued that data were collected selectively and that conclusions were announced before the analysis.

LeoIs there a tape? Please tell me there's a tape.

MayaThere's a tape. It became public around twenty eighteen. Jaffe, the Warden, is pressing a reluctant guard, and he says, "We really want you to get active and involved." Haslam, one of the researchers behind a later study we'll get to, called it a smoking gun.

LeoSo much for the guards turning cruel all on their own.

MayaAnd this is my favorite bit of analysis. Haslam, along with Reicher and Van Bavel, went through that same conversation. The Warden used words like we and us fifty-seven times, roughly once every thirty words. He's building a team with a mission. Their reading is that the cruelty came from identity leadership. Toughness was presented as serving a worthy cause. The roles alone didn't do it.

LeoLike a manager saying "we're a family" right before asking you to work the weekend.

MayaKind of, yes. Then there's who signed up. In twenty oh seven, Carnahan and McFarland ran the ad again, once with the words "prison life" and once without. The people who answered the prison version, about thirty of them against sixty-one for the other ad, scored higher on aggressiveness, authoritarianism, and narcissism, and lower on empathy.

LeoSo the ad may have attracted a certain type.

MayaMaybe. It's a small sample, and Haney and Zimbardo disputed it. A related study by Bartels found that people who read Zimbardo's orientation wording expected more hostility, including from themselves. But that measured expectations, not behavior.

LeoHas anyone actually tried to rerun it?

MayaThe closest attempt is the BBC Prison Study in two thousand two, run by Reicher and Haslam. Fifteen men, five guards and ten prisoners, eight days, and this time the guards were not briefed to be tough. The guards didn't really identify with the role and didn't want to use their authority. The prisoners overpowered them and set up an equal-shares commune. When that fell apart, some participants pushed for a tyrannical regime.

LeoThat's a whole political science course in a week.

MayaIt also had its own catch. It was made for TV, and the cameras may have kept people on their best behavior. Their conclusion was that people don't automatically take on assigned roles. It depends on whether they identify with the group and on how they're led.

LeoDid Zimbardo ever push back?

MayaHe did, and we should give his side. He said guards were explicitly told not to use force. He said Jaffe's push was mild compared with the pressure inside real institutions. And he pointed out that some guards stayed consistently good, so not everyone was following a script. Le Texier wrote a reply called "The SPE Remains Debunked." American Psychologist rejected it, and it's only available as a preprint, meaning a version that hasn't been peer reviewed.

LeoSo should the study be retracted?

MayaA Retraction Watch guest post in twenty twenty-five raised that, and there's a funny problem: the original journal doesn't exist anymore, so there's no one to retract it. The author argues for correcting the textbooks instead. And ethically, it's become a standard example in the history of stricter review boards for research on people.

LeoOkay, so what do I take into real life?

MayaThe defensible version is still unsettling. A setting, the way experimenters frame things, and leadership produced real distress and some real abuse within a few days. Uniforms and anonymity can make cruelty easier. Most of all, so can leaders who present harshness as serving a good cause.

LeoSo if you run anything, a team, a classroom, a study, your expectations are steering people even when you don't say them out loud.

MayaYes. And notice who volunteers for power, and what they're told is expected of them. Also, don't play in your own experiment. Zimbardo himself said he got caught up in being the Superintendent, and that Maslach had to snap him out of it.

LeoAnd what doesn't it show?

MayaIt doesn't show that anyone in a uniform turns into a sadist. Most guards didn't, and the BBC guards wouldn't. It doesn't carry any statistical weight: one run, no control group, roughly twenty self-selected white male students. It doesn't prove Abu Ghraib was just the situation. Zimbardo testified as a defense expert in one of those cases, which was advocacy, not a test of the theory. And it doesn't let individuals off the hook. Even Zimbardo said understanding doesn't excuse our liability.

LeoSo when the Netflix show comes on, the question isn't whether these people will turn evil. It's who's telling them what the show wants.

MayaThat's the episode. Here's my one-liner: the Stanford Prison Experiment says less about what roles do to people and more about what leaders ask of them.

LeoLove it. Thanks for listening, everybody. I'm Leo.

MayaAnd I'm Maya. Stay curious, and keep asking where the control group is.