Social Proof: Solomon Asch and the Length of a Line

The task was so easy it was insulting. On the wall, a card with a single vertical line. Beside it, a card with three lines of obviously different lengths. Which of the three matches the first? A child could do it, and in control conditions people got it right more than ninety-nine percent of the time.

Solomon Asch's arrangement, run at Swarthmore from 1951, put seven or eight people in the room and asked them to answer aloud, in order. All but one were confederates. For the first couple of rounds everyone gave the obvious answer. Then, on a prearranged trial, the first confederate gave an answer that was plainly wrong. So did the second. So did the third, and the fourth, and the fifth, and then it was the real subject's turn.

Across the critical trials, subjects went along with the unanimous wrong majority about a third of the time. About three quarters of them did it at least once. A quarter never did.

That last figure is usually left out of the popular account, and leaving it out changes the meaning of the study. Asch was not demonstrating that people are sheep; he wrote that he found the independence of the minority as striking as the conformity of the majority. What he was demonstrating is that the pressure is real and physically uncomfortable, and that a substantial fraction of people will resolve it by publicly denying what their eyes are telling them.

The interviews afterwards separated two different things happening under one behavior. Some subjects knew the majority was wrong and went along anyway, because being the lone dissenter in a room of strangers was worse than being wrong — this is normative influence, and it operates on the fear of standing out. Others reported genuine doubt about their own perception, concluding that they must have misunderstood the task or been sitting at a bad angle — informational influence, which operates on the assumption that a lot of people are unlikely to all be wrong. Later neuroimaging work by Gregory Berns found activity in visual and spatial processing regions during conforming responses, which suggests that in some cases the group is not merely changing what people say but interfering with what they think they saw.

Then Asch found the switch. If a single confederate broke ranks and gave the correct answer, conformity dropped by roughly three quarters. It did not even have to be the correct answer: a confederate who gave a different wrong answer also freed the subject, because what he broke was not the majority's accuracy but its unanimity.

That is the most operationally useful finding in this chapter. Consensus does its work through the appearance of completeness. One visible crack is worth more than a good argument.

Everything built to simulate agreement is built to prevent that crack. The nineteenth-century claque, hired to applaud on cue. The laugh track. Purchased reviews and rating farms. Coordinated posting that makes a fringe position look like a groundswell. The sales technique of naming the neighbors who already bought. In each case the manufactured signal is unanimity, and the tell is the absence of visible disagreement in a situation where disagreement should exist. Real consensus among many people is untidy and includes complaints, edge cases, and people who liked it with reservations. Manufactured consensus is suspiciously clean.

If you are the one who can see the lines, the room is waiting for you to say so.