VI
Mark học AIThe Asch Conformity Experiments · Solomon Asch, Swarthmore College
When the whole group says something plainly wrong, do we dare say what we see? Asch's answer: most of the time, yes, but group pressure is strong enough to pull about a third of answers toward the wrong one.
Participants thought they were taking a vision test. In fact everyone else in the room worked for Asch.
10 things often said about the experiment, checked against current research. Click each row for the reasoning.
Asch (1955) reported 36.8% with 123 people. The controlled replication by Franzen and Mader (2023) with about 200 students in Bern found 33%, very close to the original.
In the variant with an "ally" giving the right answer, errors fell to about a quarter (Asch, 1955). Even when the dissenter was also wrong, just in a different way, conformity fell to about 9% (Asch, 1956). Later studies found the same direction.
Asch (1955): almost no effect with 1 person, 13.6% with 2, 31.8% with 3, little increase beyond. Bond and Smith's (1996) meta-analysis also found larger majorities exert more pressure.
When participants wrote their answer instead of saying it aloud, conformity fell clearly (Asch, 1956). Deutsch and Gerard (1955) also found anonymity reduces the "fear of standing out".
Bond and Smith (1996) pooled 133 studies from 17 countries: collectivist countries showed more conformity than individualist ones. This is a correlation between countries, not an experiment, but it held across three different culture measures.
Bond and Smith (1996) found conformity in US studies declining since the 1950s. But the studies differ in participants and methods, so causes are hard to separate. The 2023 Swiss replication still found a rate close to Asch's.
Perrin and Spencer (1980) found only 1 conforming answer in 396 trials among UK engineering students. But that is a special group (trained in precise measurement), and other replications still find the effect. Better: conformity depends on people and context.
Bond and Smith (1996) found a small difference; Mori and Arai (2010) in Japan found women went along on 3.44 of 12 trials on average, close to Asch's 4.41, while men were not swayed. Sex differences depend heavily on culture, measures and experimenter; many researchers still dispute them.
In Asch's data about 2 in 3 answers stayed correct, about a quarter of people never went along, and only about 5% went along every time. Griggs (2015) reviewed 20 textbooks: 14 gave the conformity rate, only 1 said most answers were independent.
Asch himself was interested in independence and struck by how many held their ground. Friend, Rafferty and Bramel (1990) called the "conformity study" retelling a misinterpretation. Hodges and Geyer (2006) argue that those who went along were sometimes weighing truthfulness against respect for others, not just afraid.
Group pressure is real and replicable. But people are not sheep: most answers stay independent, and a single ally changes the picture.
Asch's results appeared in early-1950s America, when worries about conformity and political suspicion ran high. The story "people follow the crowd even when they can see clearly" fit that worry, so it was retold again and again.
In textbooks, "3 in 4 people conformed at least once" is easier to remember than "2 in 3 answers stayed independent". Griggs (2015) shows that over time the independence almost disappeared from the retelling.
The 2023 replication still found about 33%. So the effect has not vanished. The right lesson lies in between: group pressure is strong enough to spoil a meeting, but weak enough to be broken by one person who speaks up.
Work skill: daring to disagree in meetings, and if you run the meeting, knowing how to invite dissent.
Every number in the article, the Facebook carousel and the captions comes from the studies below.