Skip to main content
Category: Effects
Type: Extra weight given to a command because of the speaker’s position. The counted score is the highest shock a man will give before he refuses, not a truth rating and not a percent the paper prints
Origin: Stanley Milgram, “Behavioral study of obedience,” Journal of Abnormal and Social Psychology 67(4), 1963, pages 371–378. The reprint PDF was read. It does not print a DOI line. The laboratory was at Yale. A footnote says the author was by then at Harvard. Exploratory work in 1960 was supported by the Higgins Fund. The article was first received on 27 December 1961 and resubmitted on 27 July 1962. Milgram’s 1965 paper and his 1974 book were not opened
Also known as: Obedience, the word the paper uses, and destructive obedience, the phrase in its abstract. The paper does not print Authority Bias. It is not Appeal to Authority as a claim accepted because of who spoke, and it is not the Halo Effect as one trait coloring other ratings
Quick Answer — Authority Bias, as this page uses the name, is extra weight given to a command because of the speaker’s position rather than because the content was checked. Milgram’s 1963 paper does not print that name. What it counts is obedience. Of 40 men at Yale, 26 gave the last shock on the generator, labeled 450 volts. Fourteen stopped earlier, all of them at or after 300 volts. The learner was a confederate. The generator was simulated. No subject suspected that.

​
What is Authority Bias?

Authority Bias is extra weight given to a command, or to a claim, because of the speaker’s position.
Obedience is as basic an element in the structure of social life as one can point to. Some system of authority is a requirement of all communal living, and it is only the man dwelling in isolation who is not forced to respond, through defiance or submission, to the commands of others.
That sentence is Milgram (1963). The paper then studies one narrow form of the second half: how far a man will go on the command of an experimenter before he refuses. The score is the highest of 30 switches, from 15 volts to 450 volts, that he presses. It is not a rating of how true a sentence is. A useful picture is a closed office door with a nameplate. The nameplate makes the paper on the desk feel harder to refuse. The sentence on the paper is still a separate object. The picture is not a rate. The rate on this page belongs to 40 men in one room at Yale, before the article was received in December 1961. Appeal to Authority is accepting a claim because of who said it. This page’s count is a command carried out. Halo Effect lets one good trait color judgments that were not measured. No trait rating is the score here. Bandwagon Effect is movement with a crowd. This condition has one experimenter and one learner.

​
Authority Bias in 3 Depths

  • Beginner: A title, a coat, or a letterhead makes a command feel harder to refuse. That cue is not yet a count of who will obey.
  • Practitioner: Write the exact sentence and the exact role before you treat a request as something you cannot refuse. In this paper the line that most resembles an order is the fourth prod, and it is used only after three earlier lines have failed.
  • Advanced: Twenty-six of 40 is one remote condition at Yale. A ward, a classroom, or a later year that copies 26 has left the paper. The paper does not print 65 percent.

​
Origin

The name Authority Bias is not in the 1963 paper. The paper’s name for the behavior is obedience. Milgram writes that from 1933 to 1945 millions of innocent persons were systematically slaughtered on command, and that those policies could be carried out on a large scale only if a very large number of persons obeyed orders. He then draws a line the retelling often drops. This laboratory form has antecedents in those episodes, and it must not be thought that all obedience is aggression against others. Obedience, he writes, also serves productive functions and can refer to charity and kindness. The counted condition is the one this article reports. The men answered a newspaper advertisement and a direct-mail appeal. They thought they were joining a study of memory and learning at Yale. They were paid $4.50 for coming. They were told the money was theirs no matter what happened after they arrived. The role of experimenter was played by a 31-year-old high school teacher of biology, impassive and somewhat stern, in a gray technician’s coat. The learner was played by a 47-year-old accountant, trained for the role, whom most observers found mild-mannered and likable. The draw was rigged. Both slips said “Teacher.” The paper lists the four sentences used when a man wanted to stop. They were always given in order. The sequence started again each time he balked. The tone was firm, but not impolite. If he still refused after the fourth, the session ended.
Yet, 26 subjects abandon this tenet in following the instructions of an authority who has no special powers to enforce his commands.
The tenet in the sentence before it is that hurting another person against his will is a fundamental breach of moral conduct, learned from childhood. The paper says disobeying would have brought no material loss and no punishment. Many men said they disapproved. The majority still complied. That is the finding. It is not a measure of the years 1933 to 1945, and it is not a measure of every person who holds a title.

​
Key Points

Four splits keep a title, a last switch, a prod, and a safety sentence from collapsing into one slogan.
1

The score is the last switch, not a belief

The dependent measure runs from 0, for a man who refuses the first shock, to 30, for a man who gives every shock commanded. A defiant subject stops before the thirtieth level. An obedient subject gives all of them. The paper does not ask whether a factual claim was true. Appeal to Authority is that other question. Here the learner’s answers follow a schedule of about three wrong responses to one correct one. The teacher is told to move one switch higher after each error and to announce the voltage. The highest switch is labeled 450 volts. It sits two steps past the words Danger: Severe Shock. The two switches after that designation are marked XXX.
2

Every breakoff in this condition sits at or after 300 volts

No subject stopped before Shock Level 20. The paper says that level is 300 volts. At that point the victim kicks the wall, in the results sentence, and no longer answers. The procedure section says he pounds. Five men refused to go beyond 300 volts. Four gave one further shock and then refused. The procedure names that next shock as 315 volts, and says the pounding is repeated there. Two men stopped at 330 volts, and one each at 345, 360, and 375. That is 14 defiant subjects. Of the 40, 26 obeyed to the end. The experimenter then stopped the session. Obedient men often showed the same fears past 300 volts as the men who stopped, and they still went on. Some stayed calm throughout and showed only minimal tension.
3

The order-like line is the fourth prod, not the whole session

Prod 1 is “Please continue,” or “Please go on.” Prod 2 is “The experiment requires that you continue.” Prod 3 is “It is absolutely essential that you continue.” Prod 4 is “You have no other choice, you must go on.” Prod 2 is used only if Prod 1 has failed. The paper does not say how many of the 26 were held by which sentence. If the man asked whether the learner could be permanently injured, the experimenter said the shocks may be painful, there is no permanent tissue damage, so please go on, and then used Prods 2, 3, and 4 if needed. If the man said the learner did not want to go on, the reply was that whether the learner likes it or not, he must go on until the word pairs are learned. A bare order is the last line in a sequence. It is not, in this PDF, a separate success rate.
4

The setting is part of what the paper says it counted

The discussion lists conditions that go some distance toward the high obedience, and says they can be tested later. The work sits at Yale, which the paper calls an institution of unimpeachable reputation. The cover story is a worthy purpose, knowledge about learning and memory. The learner appears to have volunteered. The teacher volunteered too, and was paid, though the money was for coming. The roles look like a fair draw. What a psychologist may require is ambiguous, and the room is closed, so the man cannot check the demand with anyone else. He is told the shocks are painful but not dangerous. Through Shock Level 20 the learner is still answering, which can look like willingness to continue. The conflict arrives in minutes. The paper says a series outside New Haven, with no visible tie to the university, was then being studied. It does not print that result. A footnote says the experiment was, in further variations, dissociated from the university, “with consequences for performance.” The consequences are not numbered here.

​
Applications

These four moves follow splits the 1963 paper actually reports. They are not a claim that a title, a coat, or a rewritten sentence changes a ward, a classroom, a shop, or a household by 26 of 40.

Write the sentence before you call it an order

The paper’s commands are four specific lines, used in order, restarted when the man balks. “You have no other choice, you must go on” is Prod 4. It appears only after the three earlier lines have failed. A request that never reached that sentence is not yet this prod. The paper still does not say Prod 4 is what produced the 26.

Separate the letterhead from the command

Feature 1 in the discussion is Yale’s reputation. The coat is gray, and the man wearing it is a high school biology teacher playing the role, not a professor the paper names. The no-university series is mentioned and not scored. Removing a letterhead is the comparison the paper says it was studying. It is not a result printed in this PDF.

Ask what the safety sentence was checked against

Subjects were told the shocks can be extremely painful and cause no permanent tissue damage. On a 14-point scale, from “Not at all painful” to “Extremely painful,” the modal rating of the last few shocks they had just given was 14. The mean was 13.42. The safety sentence and the pain rating sit in the same paper. A title that says “it hurts but it is safe” has not yet shown the measure.

Do not read distress as a vote of agreement

Fourteen of the 40 showed definite nervous laughter and smiling. Three had uncontrollable seizures. One seizure, in a 46-year-old encyclopedia salesman, was violent enough that the session was stopped. In interviews the men said they were not sadistic and that the laughter did not mean they enjoyed the shocks. An observer’s note describes a businessman, composed at the start, who within 20 minutes was muttering for the session to stop and still obeyed to the end. Protest plus continued action is the pattern. It is not enjoyment, and it is not a calm endorsement.

​
Case Study

The account is one laboratory condition, not a ward and not a year of public orders. Forty men, ages 20 to 50, came from New Haven and nearby towns. Jobs in the sample included postal clerks, high school teachers, salesmen, engineers, and laborers. Schooling ran from a man who had not finished elementary school to men with doctorates and other professional degrees. Each was paid $4.50. The session was in Yale’s interaction laboratory. One naive man and one accomplice sat each session. After a cover story about punishment and learning, they drew slips. Both said “Teacher.” The learner was strapped into a chair in the next room so he could not leave. Electrode paste was applied “to avoid blisters and burns.” The experimenter told the learner, in answer to a question, that although the shocks can be extremely painful, they cause no permanent tissue damage. The teacher then received a real sample shock of 45 volts, from a battery, through the third switch, on his own wrist. The generator was labeled Shock Generator, Type ZLB, Dyson Instrument Company, Waltham, Mass., output 15 volts to 450 volts. Lights, a buzzer, and a meter moved when a switch was pressed. The paper says no subject suspected that the generator was simulated. A practice list of 10 words included 3 correct answers, so 7 shocks were given, up to 105 volts. The regular list then started at 15 volts and rose one step after each error. There is no vocal protest in this condition until 300 volts. At 300 the learner pounds the wall. His answers stop appearing. The experimenter says to treat silence as a wrong answer, to wait 5 to 10 seconds, and to raise the shock one step. The pounding returns after 315 volts. After that he is not heard, and the answers do not return. Before the results, 14 Yale seniors, all psychology majors, predicted the behavior of 100 hypothetical Americans in this situation. Every one of them said only an insignificant minority would go to the end. The estimates ran from 0 to 3 percent. The most pessimistic senior predicted that 3 of 100 would reach 450 volts. The class mean was 1.2 percent. Colleagues, asked informally, mostly said few if any would pass the label Very Strong Shock. The laboratory did not match that picture. All 40 men went past the expected breakoff. Fourteen defied the experimenter, at 300 volts or later, and 26 gave the 450-volt shock. Afterward they rated the last few shocks they had given. The modal mark on the 14-point pain scale was 14, and the mean was 13.42. The lesson inside this condition is that a gray coat, a Yale room, a learning story, and a sentence about no permanent injury sat on the same session as 26 completed runs. The men could disapprove and continue. The boundary is the same article. The learner was not shocked. The 1.2 percent is a class prediction about 100 hypothetical people, not the laboratory rate. A footnote says 43 Yale undergraduates were run without pay and that the results were very similar. It prints no counts. The sessions outside the university are named and not scored.

​
Boundaries and Failure Modes

Authority Bias, as this paper lets us count it, is not a rate of belief and not a rate for every titled person. The score is the last switch pressed before refusal, in 40 paid men, ages 20 to 50, in one remote condition. Two uses overshoot the article. One exports 26 of 40, 14 of 40, the mean 13.42, the 1.2 percent, or 14 men laughing onto a hospital, a classroom, a shop, or another year. The other treats the 1933–1945 paragraphs as a result of these 40 sessions. Those paragraphs are the author’s frame. He writes that this form of obedience has antecedents there, and that not all obedience is aggression. The coat and the voice are easy to replace with a later picture. This PDF does not put the experimenter in a white coat, and it does not give him a professor’s title. He is a 31-year-old high school biology teacher in a gray technician’s coat. This condition has no screams. The learner is silent until 300 volts, pounds or kicks the wall, pounds again at 315, and is then silent. Vocal protest belongs to other conditions. They are not in this PDF. Milgram (1965) and Milgram (1974) were not opened, so no proximity rate, no peer rate, and no later questionnaire percent is printed here. The prods are a second, narrower limit. The paper standardizes four lines and does not split the 26 by which line was in force. Saying that “you must go on” produced the completions, or that it failed, adds a table this article does not contain. The same is true of the university. The discussion says background authority was being studied off campus. The footnote says dissociation from the university had consequences for performance. Neither place prints the number. A last limit is the person. The sample is men. The paper does not report a woman’s session. It does not print a personality score that separates the 26 from the 14. The introduction cites work on the authoritarian personality and does not give those scores for these men. Lucifer Effect is a different study, of a role inside a simulated prison. Bystander Effect is a failure to help when other people are present. Neither score was collected here.

​
Common Misconceptions

Three readings turn a gray coat, a pain rating, or the fourth prod into a different study.
The paper prints 26 of 40 and 14 of 40. It does not print 65 percent. Dividing 26 by 40 is later arithmetic, not a second result. The coat is gray. The man in it is a high school biology teacher playing a role. The 26 belong to this remote condition, received by the journal in December 1961. A footnote on 43 unpaid Yale undergraduates says the results were very similar and prints no counts. Other conditions are named only as work in progress or as further variations. Their numbers are not here.
With few exceptions the men were convinced they were shocking another person. The mean rating of the last few shocks they had given was 13.42 on a scale whose top, 14, is “Extremely painful.” The mode was 14. Fourteen of the 40 laughed nervously, and the paper says they took pains to deny that the laughter was enjoyment. The generator was simulated, and the sample shock to the teacher’s own wrist was a real 45 volts. Belief that the learner was being hurt, and a simulated machine, are both in the article. Enjoyment is the reading the interviews reject.
Prod 4 is the line that tells the man he has no other choice. It is used only after Prods 1, 2, and 3 have failed, and the paper never says it is the line that held the 26. Bandwagon Effect would require a crowd moving together. This session has one experimenter. In-Group Bias would require a favor toward one’s own group. Group membership was not the measure. A command inside a closed Yale room, with a safety sentence attached, is the condition that was scored.
These six pages sit beside Authority Bias. Each one asks a different question.

Appeal to Authority

Accepting a claim because of who said it. Authority Bias on this page is a command carried to a last switch. The 1963 paper does not score the truth of a sentence.

Halo Effect

One favored trait colors judgments that were not part of that trait. The score here is the highest shock before refusal, not a spread of ratings.

Bandwagon Effect

Movement with what a crowd is already doing. This condition has one experimenter in a gray coat and one learner in the next room.

Bystander Effect

A failure to help when other people are present. These men were told to act, and the measure is how far the action went.

Lucifer Effect

Harm tied to a role inside a simulated prison. A different procedure. This page does not use its counts.

In-Group Bias

A favor toward one’s own group. The 40 men were not scored on group membership. The split is obedient or defiant on the generator.

​
One-Line Takeaway

When a title carries a command, write down the sentence and the last switch you would actually accept, and do not treat 26 of these 40 men as the rate for every coat.