Seven people call the short line the long one, and you give in

They show you two cards. On the first there is a line. On the second there are three numbered lines, and one of them is exactly the same length as the line on the first card; the other two differ by an inch or more, plain to the eye. They ask you which. It is so easy it is almost embarrassing, and there are seven other people in the room who will answer before you do.

Everyone gets the first round right. The second too. On the third, the first person says a number that is wrong. The second says the same wrong number. The third, the fourth, the fifth, the sixth and the seventh all say that same number. And then it is your turn.

The set-up is Solomon Asch's, a psychologist at Swarthmore College, and he published it in 1951. The seven in front of you were actors and you are the only one who knew nothing; they had seated you next to last on purpose, so that you would hear almost everyone before opening your mouth. Out of eighteen rounds, in twelve of them the majority answered wrongly on purpose.

About three out of four participants joined the error at least once. Counting all the rigged rounds, roughly a third of the answers followed the majority. To know how much that weighs you have to look at the control group, which did the same task alone and in writing: there the errors did not reach one per cent.

Asch came away with more than the percentage. When he explained the set-up at the end, the participants who had given in did not all tell the same story. Some said they had seen the line perfectly well and had said something else so as not to be left on their own. Others, fewer of them, held that it really had looked to them like what the group was saying. Most were in the first case: they knew the answer and said another one. They knew the answer, they said another one, and in telling it afterwards they did not use the word error: they used the word courtesy.

Asch then repeated the set-up changing one single thing: he put among the actors one who gave the correct answer. He added nobody: he instructed one of the seven, in the same room of eight. With that person inside, following the majority collapsed to around a quarter of what it had been.

And afterwards he ran the test that really shows how this works. Instead of an ally who got it right, he used one who got it wrong too, but on his own account: the majority said "two" and he said "three". Someone who was not on your side, who was not agreeing with you and who was, on top of that, more lost than anyone. There too it collapsed: from a third of the answers to fewer than one in ten.

What protected the one who had doubts was that the room had stopped speaking with a single voice. The moment the block breaks somewhere, anywhere, seven against one stops weighing, because they are no longer seven against one. There is a third variation that confirms it from the unpleasant side: when the ally was there and then went over to the majority halfway through the session, the participant started giving in again as at the beginning. The protection was not a conviction that had settled in anyone. It was the company, and it left with the company.

And there is a figure of Asch's that makes this far cheaper than it looks: the size of the majority barely matters. With one single person saying the opposite the effect hardly exists; with two it appears; with three it is already almost complete, and adding twelve more hardly moves it. It does not take a crowd to drag you along. It takes three, and for none of them to break ranks.

Up to here it is an experiment. What follows is what you do with it, and it is not about the times you have doubts, which are few and which you remember. It is about the many times you are in the room not doubting anything.

Think of the last meeting where somebody said something that struck you as odd and nobody commented. If you said nothing because, after all, you were not sure either, you were playing actor number four. You did not need to have the right answer: Asch's dissenter did not have it either and he worked just the same.

But the useful part comes a little later, and it is the one almost never mentioned in meetings.

Breaking the consensus first is expensive. You are left exposed, you have to hold the room's gaze and sit through the three seconds of silence that follow. Being second costs almost nothing, and it is the one that decides. If nobody backs the person who spoke, they back down, apologise for interrupting, and the meeting carries on exactly as before. If one single person says "hang on, let him finish", unanimity is already broken and the conversation moves elsewhere.

Notice that the second person does not even have to agree. They can think the first one is completely wrong. They can say so. The only thing needed is for the room to hear two voices instead of one, because from then on the ones who were keeping quiet have somewhere to stand.

The part of second cannot be prepared. You take it or you do not, and it is decided in the three seconds that follow somebody speaking out of turn. It is worth knowing in advance that the part exists, because when it comes there is no time to reason it out. In some meeting this week, somebody is going to speak and nobody is going to back them. Three words will do, and nobody else is going to say them.