VI
Mark học AIObedience to Authority · Stanley Milgram, Yale University
How far will an ordinary person go in harming someone, just because a man in a lab coat says "please continue"? Today's answer: further than we think, but it depends heavily on the situation, and it may not be "obedience" at all.
Participants were told it was a study of "the effect of punishment on memory". In fact, what was being measured was them.
Wording from the 1963 paper. The surprising finding (Burger, Girgis and Manning, 2011): the more direct the order, the more likely people were to stop. What kept them going was the reason "for the experiment".
If many went to the end because they guessed nobody was being hurt, then 65% measures something other than "willingness to harm".
10 things often said about the experiment, checked against current research. Click each row for the reasoning.
14 Yale students predicted on average that 1.2% would reach 450V; in fact 26 of 40 did (Milgram, 1963). Partial replications also find most people continue past 150V: 70% in the US (Burger, 2009), 90% reaching the top button in Poland (Doliński et al., 2017).
Within Milgram's own data, the share reaching 450V ranges from 2.5% to 92.5% depending on the condition (Haslam, Loughnan and Perry, 2014). An absent order-giver, a nearby victim and refusing peers all lower it.
The condition with 40 women also gave 65% (Milgram, 1974, compiled in Haslam et al., 2014). Burger (2009) found no significant difference between men and women.
65% is one condition. The average across 21 conditions is 43.6% of 740 people (Haslam, Loughnan and Perry, 2014). In 12 of the 21 conditions, most participants stopped before 450V.
A sizeable share had doubts. A reanalysis of 656 of Milgram's questionnaires found that those who believed were about 2.5 times more likely to stop (Perry, Brannigan, Wanner and Stam, 2020). So the "obedience" rate may be inflated.
The bare command ("no other choice") actually made people stop (Burger, Girgis and Manning, 2011). Haslam and Reicher's reading: people continue when they identify with the experimenter's goal. A well-grounded explanation, not yet the final word.
Burger (2009): 70% continued past 150V, against 82.5% in Milgram's comparable condition, a difference that was not statistically significant. But every replication stops at 150V for ethical reasons, so we cannot know how far people would go.
Milgram (1964) reported that 84% said they were glad to have taken part. But the questionnaire was sent months later, and Perry (2013) shows many were not properly debriefed straight after their session.
The tapes in the Yale archive show the experimenter sometimes strayed from the four prods, repeating and adding to them (Perry, 2013). The conditions are therefore not fully comparable.
Milgram himself drew the link to Nazi-era crimes. Haslam and Reicher (2012) argue that real perpetrators were usually active and believed in what they did, rather than passively "following orders". A one-hour lab session cannot explain all of history.
The phenomenon is real and unsettling: ordinary people can go a long way one small step at a time. But 65% is not "human nature", and "blind obedience" may not be the right explanation.
The 1963 paper itself describes many participants under obvious strain: trembling, sweating, nervous laughter. In 1964 Diana Baumrind wrote in American Psychologist that inflicting such stress was unacceptable, and that the deception could leave participants distrustful.
Milgram replied the same year, citing his questionnaire: 84% said they were glad to have taken part. Gina Perry (2013) later listened to the tapes and interviewed former participants, showing the debriefing was less complete than Milgram described.
These debates fed into modern rules: ethics board approval in advance, the right to withdraw at any time, full debriefing right after the session. That is why Burger (2009) had to stop at 150V, screen out vulnerable people, and remind participants several times that they could withdraw.
The experiment does not only show that people comply easily. It also shows which situations help people stop. Use those situations.
Every number in the article, the Facebook carousel and the captions comes from the studies below.