THE STANFORD PRISON EXPERIMENT

How a six-day simulation became a powerful-and disputed-story about human nature

Realistic editorial reconstruction: the researcher, guard and prisoner occupied different positions inside the same designed institution.

A young man is arrested at home, searched, issued a number, and led into a windowless basement. Another student puts on a khaki uniform, carries a baton and watches from behind mirrored glasses. Both know they volunteered for research. Yet within days, some prisoners are distressed, some guards are humiliating, and the psychologist directing the study is acting as the prison superintendent.

This is the familiar opening of the Stanford Prison Experiment. The usual ending is even more dramatic: ordinary people became cruel or submissive simply because they were assigned a role. That conclusion made the study a cultural symbol of situational power. It is also much stronger than the evidence permits.

The simulation documented real suffering and troubling conduct. But later archival work, methodological criticism and a different prison study show that behaviour was shaped by leadership, expectations, identification, resistance and the design of the institution-not by role labels operating automatically. Understanding that distinction makes the study more useful, not less.

The question behind the experiment

In August 1971, Philip Zimbardo and colleagues created a simulated prison in the basement of Stanford University’s psychology building. They wanted to examine the psychological effects of becoming a prisoner or prison guard. Their wider concern was whether destructive prison behaviour reflected the personalities of the people involved or the social environment they inhabited.

The study was planned to last two weeks. It ended after six days. The early termination, the distress shown by prisoners and the increasingly degrading behaviour of some guards became the foundation of its fame.

Who took part?

The original report describes 75 men who answered a newspaper advertisement seeking male college students for a psychological study of “prison life.” After interviews and psychological screening, 24 were selected as healthy, psychologically stable participants. They were paid $15 per day.

The men were randomly assigned to prisoner or guard roles. Eighteen began the simulation-nine prisoners and nine guards, with each guard working one of three shifts-while other selected volunteers were kept in reserve. Random assignment reduced pre-existing differences between the role groups, but it did not make the entire project a clean test of role effects.

How the simulated prison was created

Realistic editorial reconstruction: volunteers receive role assignments and institutional expectations in a 1970s research room.

Prisoners were unexpectedly arrested by local police, booked and brought blindfolded to the mock prison. They wore loose smocks, stocking caps and identification numbers; a light chain around one ankle served as a reminder of confinement. Guards wore khaki uniforms, carried wooden batons and used mirrored sunglasses to reduce eye contact. Physical violence was prohibited.

The basement environment included small cells, a corridor called “the yard,” surveillance and a solitary-confinement closet. The research team controlled meals, sleep, movement, information and the rules. Zimbardo served simultaneously as principal investigator and prison superintendent; graduate student David Jaffe acted as warden. Those dual roles would become central to later criticism.

What happened during the six days?

The first day was relatively uneventful. On the second, prisoners rebelled: they blocked cell doors and removed identifying numbers. Guards responded with fire extinguishers, stripped prisoners, removed beds, isolated leaders and created a “privilege cell.” Counts became longer and more punitive. Some guards imposed degrading exercises and arbitrary rules, especially during night shifts.

Several prisoners displayed acute distress and were released early. One developed disorganised thinking and uncontrollable crying after roughly 36 hours. Others showed anxiety, rage, withdrawal or psychosomatic symptoms. The original researchers interpreted these reactions as evidence that the prison had become psychologically real.

The simulation ended on the sixth day after Christina Maslach, a psychologist invited to observe, objected strongly to the treatment she witnessed. Her intervention mattered, but the study’s ending also exposed a deeper problem: the investigators had become so absorbed in running the prison that independent ethical judgement was weakened.

The classic conclusion: situations can overpower character

Zimbardo and colleagues argued that the prison environment encouraged guards to dominate and prisoners to become passive, dependent and helpless. This message challenged comfortable explanations that abusive institutions are created only by unusually cruel individuals. Systems distribute power, define normal behaviour, conceal responsibility and make some actions easier than others.

That insight remains important. Yet the stronger popular claim-that assigned roles alone inevitably transformed ordinary men-does not follow from this study. The simulation contained active leadership, communicated goals, uneven participation, and extensive researcher involvement.

The experimenter was not outside the system

Guards were not simply handed uniforms and left to invent the meaning of their role. The orientation conveyed that the researchers wanted a powerful prison environment and that prisoners should experience surveillance, loss of control and powerlessness. Archival analyses also describe interventions encouraging guards to behave more firmly when their conduct seemed insufficiently “tough.”

This does not mean every abusive act was ordered or scripted. It means the behaviour cannot be attributed to role assignment alone. Participants were responding inside an institution whose leaders had goals, expectations and authority. The researcher was part of the causal environment being interpreted.

Not every guard became cruel

Guard behaviour varied considerably. Some acted harshly, some enforced rules without escalating humiliation, and some were comparatively friendly or performed small favours. Prisoners also resisted, negotiated, complied selectively and tried to leave. Variation is not a footnote: it contradicts the idea that the role produced one automatic response.

A more defensible interpretation asks why particular participants identified with a harsh model of guarding, why others resisted it and how leaders and group norms made certain versions of the role appear legitimate.

Demand characteristics and recruitment bias

Participants try to understand what a study is about. In 1975, Ali Banuazizi and Siamak Movahedi showed descriptions of the prison procedure to students and found that many could infer the expected hypothesis and predict hostile guard behaviour. This raised the possibility of demand characteristics: people may enact what they believe the researchers expect, especially when the situation is ambiguous.

Recruitment may matter too. In 2007, Thomas Carnahan and Sam McFarland compared volunteers attracted by an advertisement for a “psychological study of prison life” with volunteers responding to a generic psychology-study advertisement. The prison-study volunteers scored higher on aggression, authoritarianism, narcissism and social dominance, and lower on empathy and altruism. This does not prove that Stanford’s final participants had those exact traits, but it warns that self-selection can shape who enters such research.

What later research changed

The BBC Prison Study, conducted by Stephen Reicher and S. Alexander Haslam and published in 2006, was not a direct replication. It used different safeguards, theory and procedures. Its results nevertheless challenged the idea that people naturally conform to unequal roles. Guards did not automatically develop a shared identity or sustain control. Prisoners gradually formed a cohesive group, challenged the hierarchy and helped bring the guard regime to an end.

Reicher and Haslam argued that tyranny depends on social identification, leadership and the active construction of group norms. People do not lose agency when they enter roles; they interpret roles and may embrace, reshape or reject them. Harm becomes more likely when leaders define domination as legitimate and followers identify with that project.

What the Stanford study does, and does not show

• It shows that a highly controlled, unequal institution can generate distress, conflict and degrading conduct.

• It shows how difficult it can be for researchers and participants to retain perspective once they identify with an institutional mission.

• It does not isolate “the role” as a single causal variable: expectations, leadership, surveillance, rules and researcher interventions changed at the same time.

• It had no adequate control condition, no standardised behavioural outcome and a very small, self-selected sample of men.

• It does not show that most people inevitably become sadistic when given authority, or passive when deprived of it.

• It cannot support precise universal claims about prisons, workplaces, schools, hospitals or humanity as a whole.

Why the study is so difficult to treat as an experiment

Randomly assigning volunteers to guard and prisoner roles is an experimental feature. But causal inference requires more than random assignment. The intervention must be sufficiently defined, competing influences controlled and outcomes measured consistently. In Stanford, the researchers also managed the institution, improvised responses to events and sometimes encouraged a preferred form of guard behaviour.

The study is therefore better understood as a dramatic, historically important simulation and case study with experimental elements-not as decisive evidence that role assignment caused the reported behaviour. Thibault Le Texier’s archival reassessment concluded that the standard textbook narrative rests on selective reporting, researcher direction, and claims the design could not establish. Zimbardo and Craig Haney later disputed aspects of that critique and continued to emphasise the power of dehumanising environments. The controversy should be reported, not edited away.

Ethics: why this study would not be approved in the same form

The participants consented to a prison simulation, but meaningful consent also requires understandable risks and a genuine ability to withdraw. Prisoners who asked to leave were at times treated as though they needed “parole,” blurring the difference between a volunteer and an inmate. The surprise arrests, sleep disruption, humiliation, loss of privacy and intense distress created risks that were not adequately contained.

Zimbardo’s superintendent role also produced a conflict between protecting participants and preserving the institution. Contemporary review would demand independent safety monitoring, explicit withdrawal procedures, pre-defined stopping rules, proportional risk, protected privacy, rapid clinical support and a research team whose scientific judgement was not fused with operational authority.

Why the experiment still matters

Modern application: a junior clinician raises a concern while senior colleagues stop and listen.

The study’s useful lesson is not that people are puppets of roles. It is that systems shape what people notice, what they believe is expected, which actions receive approval and how easily responsibility can be displaced. Leadership can normalise harm-but leadership can also authorise dissent, care and restraint.

In hospitals, schools, companies and public institutions, safer systems create independent oversight, protect people who speak up, rotate concentrated authority, monitor outcomes rather than appearances and make stopping rules explicit. Teams should ask not only “Who behaved badly?” but also “What goal, norm, incentive or silence made this behaviour seem acceptable?”

Individual courage matters. Maslach’s objection helped puncture the shared reality of the simulated prison. But organisations should not rely on one unusually courageous outsider. Ethical resistance must be designed into the institution.

The takeaway

The Stanford Prison Experiment deserves a place in psychology’s history, but not as a simple demonstration that a uniform turns ordinary people into tyrants. Real distress occurred. Some guards behaved abusively. Researchers also helped construct the norms they later attributed to the situation, and many participants did not conform in the same way.

The stronger modern lesson is more demanding: harmful systems are made through leadership, expectations, identification, unequal power and failures of oversight. Because they are made, they can also be interrupted.

A role offers a script. It does not remove the person, the leader-or the responsibility to refuse it.

References and further reading

Haney, C., Banks, W. C., & Zimbardo, P. G. (1973). Interpersonal dynamics in a simulated prison. International Journal of Criminology and Penology, 1, 69–97.

Banuazizi, A., & Movahedi, S. (1975). Interpersonal dynamics in a simulated prison: A methodological analysis. American Psychologist, 30(2), 152–160. https://doi.org/10.1037/h0076835

Carnahan, T., & McFarland, S. (2007). Revisiting the Stanford Prison Experiment: Could participant self-selection have led to the cruelty? Personality and Social Psychology Bulletin, 33(5), 603–614. https://doi.org/10.1177/0146167206292689

Reicher, S., & Haslam, S. A. (2006). Rethinking the psychology of tyranny: The BBC Prison Study. British Journal of Social Psychology, 45(1), 1–40. https://doi.org/10.1348/014466605X48998

Le Texier, T. (2019). Debunking the Stanford Prison Experiment. American Psychologist, 74(7), 823–839. https://doi.org/10.1037/amp0000401

Zimbardo, P. G., & Haney, C. (2020). Continuing to acknowledge the power of dehumanizing environments: Comment on Haslam et al. (2019) and Le Texier (2019). American Psychologist, 75(3), 400–402. https://doi.org/10.1037/amp0000576

American Psychological Association. (2017, amended 2025). Ethical Principles of Psychologists and Code of Conduct. https://www.apa.org/ethics/code