neur.ro

Milgram experiment

type: effect · origin: study · evidence: supported · themes: how we judge others

Claim assessed: When an experimenter asks them to continue, many ordinary people give what they believe are painful electric shocks to another person.

Published · how we work

In short

In Stanley Milgram’s experiments at Yale, in 1961–1962, ordinary people thought they were taking part in a study on learning. They had to give increasingly strong electric shocks to another participant each time he made a mistake. The shocks were fake and the “learner” was an actor, but the participants didn’t know that. In the best-known version, almost two thirds went all the way to the last switch, 450 volts. The result has been repeated in many countries and in recent replications, stopped at 150 volts for ethical reasons. The rate depends heavily on the situation, though, and the popular explanation, that people blindly obey orders, is contested.

What it says

The experiment. The participant, called the “teacher”, read word pairs to a “learner” in another room and pressed a switch for each mistake, raising the voltage by 15 volts each time. The learner protested, screamed, and at some point stopped answering. When the teacher hesitated, the experimenter used four increasingly firm prods, from “Please continue” to “You have no other choice, you must go on”. Psychiatrists who were given a description of the procedure estimated that almost nobody would go to the end. We haven’t read Milgram’s studies; we describe them from Haslam, Loughnan and Perry (2014) and Doris, Niemi and Machery (2024).

Not one experiment but 23 variants. Haslam, Loughnan and Perry (2014) pooled 21 variants, with 740 participants. Overall, 43.6% went to 450 volts, but the rate ranged from almost nobody to almost everyone:

Percentage who went to 450 volts in some of Milgram's variants: teacher chooses the shock himself, 2.5%; victim is a friend or relative, 15%; teacher holds the victim's hand on the plate, 30%; modest office in Bridgeport, 47.5%; baseline variant, victim heard from another room, 62.5%; participant only reads while someone else presses, 92.5%. How far people obeyed depends on the situation Went all the way to 450 volts Milgram variants, from Haslam et al. (2014) teacher chooses the shock 2.5% victim is a friend or relative 15% holds the victim's hand on the plate 30% modest office, Bridgeport 47.5% baseline, victim in another room 62.5% only reads, someone else presses 92.5% Same authority, very different results.
Haslam, Loughnan and Perry (2014), Table 2; 20 or 40 participants per variant. The baseline is the variant with the victim's voice heard through the wall.

Has it replicated? There is no meta-analysis of the replications. Doris, Niemi and Machery (2024) collected 19 replications and extensions from 10 countries, between 1968 and 2017, with obedience rates between 50% and 94%. Three unpublished US dissertations found lower rates, 30–40%, and an Australian study from 1974 found 16–40% depending on sex. For ethical reasons, recent replications stop at 150 volts, the point after which, in Milgram’s studies, most of those who continued went to the end:

We haven’t read the studies by Burger and Doliński; we describe them from Doris et al. and Grzyb et al.

Did they believe the shocks were real? Critics, including Gina Perry, argue that many participants didn’t believe it, so they didn’t think they were harming anyone. In a questionnaire Milgram sent after the experiments, 56% said they fully believed the learner was getting painful shocks, and another 24% that he probably was; only 2.4% were certain he wasn’t. Perry et al. (2019) showed that those who believed were more often among those who refused. Doris et al. (2024) reanalysed the same data: the difference exists but is small (d ≈ 0.3), and believers and non-believers went, on average, about equally far, roughly two thirds of the way along the panel. The “they didn’t believe it” explanation therefore covers only a small part of the result.

Blind obedience to orders? The popular image is of people carrying out orders. The data show something else. The fourth prod, “You have no other choice, you must go on”, the only one that is a real order, almost always led to refusal, in Milgram’s studies, in Burger’s and in a restaging with actors (Haslam, Reicher and Millard, 2015). Haslam and Reicher propose that people continue when they identify with the experimenter and the scientific purpose of the study, not because they receive orders, and stop when they identify more with the victim. Milgram’s explanation, that people enter an “agentic state” in which they no longer feel responsible, is one of several interpretations.

Example

A new employee is asked by the boss to send customers a message they consider misleading. The boss doesn’t threaten them, but explains why it matters for the company. The employee sends the message, although they don’t like it. If other colleagues had openly refused, they probably would have refused too. The example is an illustration, not a study cited here.

How to apply it

The steps below are a practical way to apply it that we propose, based on the studies cited.

  1. Don’t rely on “I’d never do that”. Almost nobody expected Milgram’s results, neither the psychiatrists nor the participants themselves.
  2. Be the one who refuses first. When others openly refused, very few continued. A single colleague saying “no” can change what the others do.
  3. Get close to those affected. Obedience dropped a lot when the victim was nearby or known. When you make a decision that affects someone, try to see concretely who that person is.
  4. Ask whose purpose it serves. People continued mostly when they believed in the study’s purpose. Check whether the purpose someone presents to you really justifies what you’re being asked to do.

Limits and nuances

Sources

See also: Asch conformity experiments, Bystander effect, Fundamental attribution error, Stanford prison experiment