Asch Conformity Experiment: What the Data Showed

In Solomon Asch's line-judgment experiment, 123 subjects gave the majority's wrong answer on 36.8% of critical trials, against an error rate under 1% alone. Conformity peaked at seven opponents (37.1%) and fell to 31.2% against fifteen. One ally breaking the unanimity cut the effect to a quarter.

Seven students sit in a row and are asked which of three lines matches a fourth. The answer is obvious. Six of them have been told beforehand to give the same wrong answer, out loud, one after another. The seventh has no idea. What Solomon Asch measured in that room is quoted constantly and reported loosely, usually as a single percentage with no mention of which of his experiments it came from. Here is what the 1955 paper actually reports, including the table almost nobody reprints.

What did the Asch conformity experiment actually show?

That 123 people gave a visibly wrong answer on 36.8% of the trials where a unanimous group gave it first, on a task they got right more than 99% of the time alone. Asch recruited subjects from three institutions of higher learning, not including Swarthmore College where he taught, and placed each one in a group of seven to nine people. Everyone else in the room was working from a script. The subject answered second to last, after hearing the majority commit to an answer he could see was false.

That last distinction matters and Asch made it himself. He interviewed every subject afterwards. Some had concluded the group must be right. Others knew the group was wrong and went along anyway, several because they read their own disagreement as a defect they needed to hide. The experiment measures what people say in front of others, which is exactly why it transfers to situations where behaviour is visible.

Does conformity increase with group size?

Up to three people, then it stops, and against fifteen opponents it was lower than against three. Asch varied the opposition from one to fifteen and published eight measurements in the caption of a chart on page 6 of the 1955 article. Of the five pages ranking on this question that our server could read on 21 September 2026, none reprinted that series, and none mentioned the value for fifteen opponents.

The practical reading is the one Asch gave. Three people who agree produce nearly all of the pressure that fifteen do, so the effect is not about being outnumbered. It is about being the only one. That is why crowd size is a poor predictor of how hard it is to hold a position, and it is the finding that makes small groups worth taking seriously, a point the research on social accountability keeps running into.

What made people stop conforming?

One person breaking the unanimity, and nothing else came close. When Asch placed a truthful partner in the group, the majority’s pressure on the subject fell to one fourth of its previous level. In the version where a partner answered correctly for the first six trials, 18 of 27 subjects were completely independent while he was there.

Then Asch ran the version that explains the mechanism. He put in a dissenter who disagreed with the majority and with the subject, picking a line even further from the standard than the majority did. This ally was the least accurate person in the room. Subjects’ errors dropped to 9%, lower than with any correct partner. Asch’s conclusion was that dissent by itself restores independence, regardless of whether the dissenter is right.

The ally who freed people most was the one who was most wrong. What broke the spell was not accuracy, it was the end of unanimity.

The reverse held too, and it held sharply. When the partner switched sides after six trials, errors jumped back to roughly the level of a subject who had never had support at all. Having once stood up to the group bought nothing. When the partner instead simply left the room, announced in advance as having an appointment with the dean, errors also rose but less markedly. Losing an ally to the other side cost more than losing him to the door.

Where does the “75% of people conformed” figure come from?

From a photo caption on page 4 of the 1955 article, and it counts people who conformed at least once, not answers. The caption reads that seventy-five per cent of experimental subjects agree with the majority in varying degrees. The results text states the same fact from the other side: about one quarter of the subjects were completely independent and never agreed with the erroneous judgments.

So the number is real and it is Asch’s. What gets lost is the threshold. A subject who held firm on eleven of twelve critical trials and cracked once is inside that 75%. Quoted as “three quarters of people are conformists” it says something the data does not support, because the same paper reports that the average subject stayed independent on nearly two thirds of critical trials. The two headline figures, 36.8% and 75%, are not in tension. One counts answers, the other counts people, and they need to be labelled.

Does the Asch effect still hold today?

It replicates, but the size of the effect moves with era and culture, so it is not a constant. Bond and Smith pooled 133 line-judgment studies from 17 countries in Psychological Bulletin in 1996 and reported two things worth carrying forward: conformity in United States studies declined over the decades after the 1950s, and countries with more collectivist cultures produced higher conformity than individualist ones. Treat 36.8% as one measurement of one population, not as a human constant.

If you want to read further, the search terms matter more than the experiment’s name. Counts run through the NCBI E-utilities on 21 September 2026, over titles and abstracts indexed in PubMed:

PubMed indexes biomedical literature, so these counts say where a phrase shows up there rather than how much work exists in psychology overall. The useful part is the ranking. The two terms that carry the actual mechanism, normative and informational influence, come from Deutsch and Gerard’s 1955 paper published the same month as Asch’s, and they are the distinction behind most modern work on the types of peer pressure. More numbers from that literature, each traced to its study, are collected in our peer pressure statistics page.

How StellarHabit uses this

The finding we built on is the group-size table, not the headline percentage. Three people who can see what you did produce nearly the whole effect that fifteen do, so a habit tracker does not need an audience to work. StellarHabit shows your habits to a small group you choose, and shows you theirs. Nobody has to send a reminder or ask you anything, which matters because the pressure Asch measured came from visibility alone, not from anyone making a demand. The same mechanism appears in positive and negative peer pressure.

The ally result is the other half. A group that only ever shows wins recreates the unanimous majority, and someone having a bad week reads it as a personal defect to hide, which is what Asch’s interviews found. So progress is a consistency percentage rather than a streak: a missed day moves a number instead of resetting it to zero in public, and the gaps stay visible instead of being erased. Our limits are real. There is no money at stake, so if a financial forfeit is what moves you, a commitment contract fits better, and none of this is built for habits you want to keep private. If you want to try it with people you already know, start with a habit tracker you share with friends.