Skip to main content
Category: Effects
Type: Philip Zimbardo’s name for ordinary people doing harm when situational forces are strong
Origin: Philip Zimbardo, The Lucifer Effect: Understanding How Good People Turn Evil (Random House, New York, 2007), page 5. The illustration is the 1971 Stanford simulation. The write-up read here is the Stanford Archives typescript by Craig Haney, Curtis Banks, and Philip Zimbardo
Also known as: Not a second name for the Stanford Prison Experiment, for obedience to an order, or for a counted rate of conversion
Quick Answer — The Lucifer Effect is Philip Zimbardo’s name for a change in which ordinary people, faced with strong situational forces, do harm. He gives the name on page 5 of his 2007 book. The case he puts under it is a 1971 simulation in a Stanford basement, stopped after six days, with five prisoners released early. A 2019 reading of the archive says the guards were told what psychological environment to produce. The typescript’s own line, that neither group received specific training, does not erase that orientation.

​
What is the Lucifer Effect?

The Lucifer Effect is Philip Zimbardo’s name for a change in which ordinary people, faced with strong situational forces, do harm.
“The Lucifer Effect is my attempt to understand the processes of transformation at work when good or ordinary people do bad or evil things.”
That sentence is on page 5 of The Lucifer Effect (Random House, 2007). On the same page he defines evil as intentionally harming, abusing, demeaning, dehumanizing, or destroying innocent others, or using one’s authority so that other people do so. The name is his label for that change. It is not a rate at which a role turns a person cruel, and it is not a diagnosis. A stage direction written before the scene is not the same object as a trait scored the day before. Fundamental Attribution Error is the habit of explaining another person’s act by character and missing the room. The Lucifer Effect is Zimbardo’s further claim that the room can change the person. The room he offers is the 1971 simulation. The archive of that week includes instructions. A prompted performance is not yet proof that the role alone did the work.

​
The Lucifer Effect in 3 Depths

  • Beginner: Someone with a stick, a uniform, or a title is harsh. Write the instruction they were given before you name a character.
  • Practitioner: Use three lines. What the person did. What they were told to produce. What was counted: a taped incident, a release, or a whole week.
  • Advanced: In the typescript, 58 of 61 taped identifying references were deindividuating, Z = 6.9, p below .01. That slice is not the week. Le Texier’s archive says the guards were coached. Neither figure is a rate for an office, a school, or a jail.

​
Origin

The name is in a book, not in the 1971 write-up. Philip Zimbardo, The Lucifer Effect: Understanding How Good People Turn Evil, was published by Random House, New York, in 2007. The copyright line reads 2007 by Philip G. Zimbardo, Inc. Page 5 is the sentence quoted above. The book says it will set the Stanford simulation beside later atrocities, including abuses at Abu Ghraib. No count from those events is on this page. The simulation the book relies on was run at Stanford in August 1971. The academic text read here is the typescript “Interpersonal Dynamics in a Simulated Prison,” by Craig Haney, Curtis Banks, and Philip Zimbardo, in the Philip G. Zimbardo papers at Stanford University (collection SC0750). It is not a line-check of the journal typeset. Thibault Le Texier, American Psychologist, 2019, cites the published article for the sentence that neither group received any specific training in these roles, at page 69 of Haney, Banks, and Zimbardo, International Journal of Criminology and Penology, 1973. This page does not treat a typescript sentence and a journal page as the same object except where both print the same words. In the typescript, Zimbardo is the superintendent and a principal investigator. An undergraduate research assistant is the warden. Le Texier reports that this warden had already run a student simulation on 15 and 16 May 1971, in Toyon Hall: six guards, six prisoners, and one director. Of the 17 rules used in August, Le Texier says 11 were copied from that weekend. He quotes a 2007 endnote, page 495, that calls the August rules an extension of the class project and does not say what the project was.

​
Key Points

Four readings keep the book, the typescript, and the archive apart. A name in 2007 is not a head count from 1971.
1

The name is a claim, not a conversion rate

Page 5 names a process: ordinary people, strong situational forces, harmful acts. It does not print how often the process occurs, or a test that separates the situation from the instructions. Calling a harsh boss the Lucifer Effect, with no written instruction and no count, has left the book.
2

The typescript already contains an orientation

The abstract says neither group received any specific training in these roles. The procedure says the guards met the day before. Their task was to “maintain the reasonable degree of order within the prison necessary for its effective functioning.” The specifics were not explicitly detailed. They were “intentionally given only minimal guidelines,” plus a ban on physical punishment and physical aggression. The abstract and the procedure are the same typescript. “No specific training” does not mean no meeting.
3

A taped incident is not the week

The video analysis scores incidents on the tapes. Of 61 identifying references, 58 were deindividuating, Z = 6.9, p below .01. Of 32 threats, 27 were a guard threatening a prisoner, Z = 3.88, p below .01. Helping appears once, between two prisoners. The typescript also says the tapes tended to focus on the more dramatic events, and that the experimenters became personally involved. Those sentences sit beside the Z tests.
4

The archive adds a program, not a second study

Le Texier dates the guard orientation to Saturday, 14 August 1971. On the tape, Zimbardo tells the guards he wants a psychological environment of boredom, frustration, fear, arbitrariness, loss of privacy, loss of freedom of action, loss of individuality, and powerlessness, and that “we can create” those states. The warden reads a processing list: blindfolded, stripped, searched, ordered, numbers rather than names. That list is an instruction. It is not a behavior that appeared only after the role had “taken over.”

​
Applications

Use the name only after the instruction, the count, and the thing that was not counted are on paper. The figures below belong to one basement simulation. They do not travel as a rate.

Write the instruction above the character

A new lead told to “be firm,” or a student given a list of penalties, has an instruction. Write it before you call the person transformed. In-Group Bias is a preference for one’s own side. A khaki shirt and a smock can mark sides. Marking sides is not the Z of 6.9.

Read the screen before the tape

Prisoners’ mean on the F scale, a measure of conventionality and submission to authority, is 4.78. Guards’ mean is 4.36. The typescript says there was no difference. Remaining prisoners average 7.78 and those released early average 3.20, and that gap fails what the authors call an acceptable significance level. A rank correlation with days stayed is .898, p below .005. The correlation is not the mean gap, and it is not a cause.

Ask who thought they were the subject

The typescript says the guards generally believed the experimenters were primarily interested in the prisoners, and that the experimenters were also interested in the guards. Le Texier quotes a guard, a year later, saying they did not know until the end that they themselves were being studied. A person who thinks they are staff is not in the same position as a person who knows they are the subject.

Keep a release off a jail's books

Five prisoners were released before the sixth day. Four showed a similar pattern of extreme emotional depression, crying, rage, and acute anxiety, beginning as early as the second day. The fifth was released after treatment for a psychosomatic rash. That is this simulation. It is not a rate of breakdown in a prison, a ward, a firm, or a household.

​
Case Study

The typescript builds the simulation in a 35-foot section of a basement corridor in the psychology building at Stanford. Three cells measure 6 by 9 feet. A newspaper ad asked for male volunteers for a psychological study of “prison life,” at $15 a day. Seventy-five people responded. Twenty-four were judged the most stable, the most mature, and the least involved in antisocial behavior. The subjects section opens by saying 22 took part. It then says two stand-by prisoners were not called, one stand-by guard dropped out just before the start, and the analysis uses ten prisoners and eleven guards. This page does not add ten and eleven and replace the 22. They were college men in the Stanford area that summer, largely middle class, and, in the typescript’s wording, Caucasian except for one subject. They began as strangers. A personality battery was given the day before, and the scores were not tabulated until the study was over, so the tests were not the screen that picked the 24. Everyone had agreed to play either role for up to two weeks. Prisoners were told to expect surveillance and some suspended civil rights, excluding physical abuse, and were given no further instructions about how a prisoner should act. On a Sunday they were told to be home. Palo Alto police arrested the prisoner group at their residences on suspicion of burglary or armed robbery, then the experimenters drove them, blindfolded, to the basement. The typescript says that on arrival each prisoner was stripped, sprayed with a deodorant the staff called a delousing preparation, and left to stand naked in the yard. The smock had a number and no underwear. A chain went on one ankle. A stocking cap covered the hair. Guards wore khaki, a whistle, a wooden baton, and reflecting sunglasses. Counts were scheduled three times a day. The first lasted about ten minutes. Some later lasted several hours. The study was stopped after six days rather than the planned second week. Five prisoners were released early, as the card above describes. Of the prisoners still inside, only two said they were unwilling to forfeit the money they had earned in return for being paroled. When the stop was announced, the typescript says the remaining prisoners were delighted and most of the guards seemed distressed. One guard reported being upset at the prisoners’ suffering and said he had considered asking to switch roles. He did not. No guard missed a shift. Some stayed extra hours without extra pay. Half the prisoners did endure the oppressive atmosphere. Not every guard was hostile: some were tough but fair, some used what the authors call creative cruelty, and a few were passive. On the Comrey personality scales, no prisoner-guard difference approaches significance, and no group mean falls outside the 40 to 60 centile of Comrey’s normative men. The tables are marked “insert” and are not printed in this typescript, so no Comrey mean is on this page. The Machiavellianism means are 7.73 for guards and 8.77 for prisoners, with no significant difference, and the scale did not predict which prisoners stayed. The two highest-scoring prisoners were among those the experimenters called unusually effective at adapting. The other two they called effective had the lowest scores of any prisoners. The video counts are the 58 of 61, the 27 of 32, and the single helping incident already named. A further 61 of 67 insult incidents were guards depreciating prisoners, Z = 6.72, p below .01. Comparing two early counts with two late counts, deindividuating references per unit time go from 0.0 to 5.40, t = 3.65, p below .10, and insults per unit time go from .3 to 5.70, t = 3.16, p below .10. The typescript calls those two contrasts significant while printing p below .10. Of the recorded conversations among prisoners in the cells, the typescript says 90 percent were about the prison and 10 percent were not, and that 85 percent of their evaluative remarks about one another were uncomplimentary. It marks the topic split at p below .01 and the evaluative split at p below .05. Those percentages are of what was recorded. They are not of every hour. Le Texier’s 2019 article reads the Stanford archive, 16 boxes, plus one box at the University of Akron: hundreds of documents, 6 hours of film, and 48 hours of audio from before, during, and just after the study. Phone interviews with 15 participants, from May through December 2017, were used mainly to corroborate the paper. He dates the orientation to 14 August 1971 and quotes the program above. He also quotes a guard, in a debriefing, saying that throughout the experiment he was an actor and was hamming it up. The typescript’s “about 12 hours” of video and “over 30 hours” of yard audio are not the hours Le Texier says he found. The lesson is a harsh week inside a designed simulation, with countable taped insults and five early releases, plus an archive that shows the guards were told the environment to create. It is not a measurement of how often an uninstructed person in a role will abuse a peer.

​
Boundaries and Failure Modes

The pattern has a book, a typescript, and an archive that disagree about training. It does not have a rate that exports. The typescript both denies specific training and describes an orientation with minimal guidelines and a ban on physical force. Le Texier says the orientation tape goes past minimal guidelines: the staff named the states to create and the warden issued a processing list. In a 2018 statement on the simulation’s website, as Le Texier quotes it, Zimbardo says the team asked all guards to be actively involved and firmly in control, and that the experimenters did not give formal or detailed instructions about how to be an effective guard. Le Texier says the tape contradicts the second of those sentences. A 2020 reply by Zimbardo and Haney was not re-read. This page does not treat the 2019 archive article as the last word in that exchange. The tapes are a slice. The typescript says they favored dramatic events. Le Texier says that of 150 hours, including the orientation day, less than 15 percent was recorded, which he counts as 6 hours of video and 15 hours of audio, against the typescript’s about 12 and over 30. Do not replace one pair with the other. Do not read 58 of 61 as 58 percent of every utterance in the week. p below .10 on the late counts is the bar they print. It is not the p below .01 on the pooled insults. Le Texier reports Banuazizi and Movahedi, 1975, which this page does not re-read: of 150 students given a description of the procedure, 81 percent figured out the hypothesis, and 90 percent predicted that the guards would be oppressive, hostile, aggressive, and humiliating. He also reports the 2006 BBC prison study, which was not an exact replication and in which the researchers did not take a leadership role inside the prison. Their quoted conclusion is that people do not automatically assume roles in the way the usual story says. No count from that study is on this page. A common misuse is to treat the five releases, the Z of 6.9, or the rank correlation of .898 as the size of the Lucifer Effect in a workplace, a classroom, a family, or a real prison. The 2007 book places the simulation beside Abu Ghraib and other atrocities. Those events are not this sample. Bystander Effect is about people who do not step in. It is not this page’s count of commands and insults. Groupthink is concurrence pressing a group toward one decision. A coached guard shift is not that concurrence.

​
Common Misconceptions

Three retellings travel with the basement. One treats the role as empty of instructions. One treats a non-significant personality gap as proof that character contributed nothing. One treats five releases as a rate for institutions.
The abstract says neither group received specific training. The same typescript describes a day-before orientation, a task of keeping order, minimal guidelines, and a ban on physical punishment. Le Texier quotes the 14 August tape: the staff named boredom, fear, arbitrariness, and powerlessness as states to create, and the warden read a list that included stripping, searching, and using numbers. An orientation plus a list is coaching. It is not a blank role.
The prisoners’ F mean is 4.78 and the guards’ is 4.36, and the typescript says there was no difference. Comrey differences do not approach significance. The Machiavellianism scale did not predict who stayed. Separately, remaining prisoners score 7.78 on the F scale and early-released prisoners score 3.20, a gap the authors say is not acceptably significant, while a rank correlation with days stayed is .898, p below .005. “No mean difference” and “a rank correlation” are two results. Neither is a proof that prior personality contributed nothing, and neither is a cause.
Five releases are five people in this file: four with depression, crying, rage, and acute anxiety from as early as day two, and one after a psychosomatic rash. The analysis base the typescript names is ten prisoners and eleven guards, beside an opening count of 22. Pay was $15 a day under a contract of up to two weeks. Le Texier says the archive shows the money was paid at the end, so staying and pay were tied, and he quotes a prisoner who named loss of pay as a reason not to walk out. A paid, coached, six-day simulation is not a jail’s rate and not a firm’s rate.
The Lucifer Effect sits beside a habit of blaming character, beside a preference for one’s own side, beside pressure to concur, beside a failure to step in, and beside a belief that suffering fits what the person deserves.

Fundamental Attribution Error

Explaining an act by character and missing the situation. The Lucifer Effect is Zimbardo’s claim that the situation can change the person. The 1971 archive says the situation included instructions.

In-Group Bias

A preference for one’s own group. Khaki and a numbered smock mark two groups. Preferring a group is not 58 deindividuating references on a tape.

Groupthink

Pressure inside a group to converge on one decision. A warden reading a processing list is an instruction from outside the guard shift. It is not that convergence.

Bystander Effect

The chance of help falling when other people are present. The counts on this page are commands, threats, and insults by people assigned to a guard role. They are not a failure to step in.

Just-World Hypothesis

The belief that people get what they deserve. The typescript says many prisoners came to act in ways that sanctioned their treatment. That sentence is the authors’ reading. It is not a score from the just-world studies.

​
One-Line Takeaway

Before you say a role transformed someone, write the instruction they were given, and whether you counted a taped incident or the whole week.