The rebel who breaks the room 

In one variation the participant was told he would work alongside two other teachers, both confederates. One read the word pairs, one recorded responses, and the real participant delivered the shocks. At 150 volts the first confederate announced that he would not continue and left the panel, sitting down across the room despite the experimenter's insistence. At 210 volts the second did the same. The participant was then instructed to carry on alone.

In that condition roughly one in ten participants went to the final switch. Set against the sixty-five percent of the standard configuration, it is the largest swing Milgram produced by adding anything to the room, and it was produced not by an argument but by an example. Nobody lectured the participant on ethics. Two people simply demonstrated that refusal was available, survivable and did not end the world.

Milgram's explanation drew on the conformity research of his former mentor Solomon Asch. An authority's definition of a situation is powerful partly because it is unopposed. A lone participant has no way to know whether stopping is reasonable or hysterical, and in the absence of any comparison he defers to the person who seems to know what is normal here. A peer who quits supplies the missing comparison. It also redistributes the social cost, because the awkwardness of being the difficult one is now shared.

There is a mirror-image condition that keeps the finding honest. When confederates administered the shocks and the participant only performed a subsidiary task, compliance was very high. The same mechanism that makes dissent contagious makes participation contagious, and there is nothing morally uplifting in the process itself. Groups do not push towards decency. They push towards whatever the visible members are already doing.

That is why this chapter, rather than the famous percentage, is the one worth carrying into an ordinary week. The transferable claim is not that people are sheep, and not that you would have been the hero. It is narrower and better evidenced: the first person to say plainly that something should stop changes the odds for everyone else in the room, and the second person to say it changes them again. This is also the finding most easily overstated, so state it carefully. One rebellious peer is not a guarantee against anything. It is a large improvement on silence.

None of which addresses the question the participants themselves raised afterwards, which is what the experiment did to the people who took part in it.