The variations nobody talks about 

The single most misleading thing about the Milgram study is the definite article. Between 1961 and 1962 he ran a programme of separate conditions, each with its own group of about forty participants, and the whole point was comparison. The percentages below are from his published series and should be read as approximate, because samples of forty produce wide margins.

Distance from the victim was the first lever. In the remote condition the learner was in another room and could not be heard except by pounding on the wall, and compliance to the final switch was high. With vocal protest audible through the wall it stayed around sixty-five percent. Put the learner in the same room, visible and audible, and full compliance fell to roughly forty percent. In the touch-proximity condition, where the teacher had to press the learner's hand onto a shock plate himself, it fell to about thirty percent. Physical involvement in the harm changes behaviour more reliably than any argument about it.

Distance from the authority was the second lever, and it was stronger. When the experimenter gave his instructions by telephone from another room, full compliance dropped to around twenty percent, and Milgram noted that some participants who claimed to be continuing were quietly administering the lowest shocks instead. Authority at a distance is easy to appear to obey. When two experimenters disagreed with each other in front of the participant, compliance collapsed almost entirely.

The third lever was legitimacy. The Yale laboratory was replaced by a modest office suite in nearby Bridgeport, presented as the work of a private research firm with no university attached, and compliance fell to roughly forty-seven percent. Lower, but far from nothing, which argues against the comfortable idea that participants were simply overawed by Yale. When an apparent ordinary man rather than the official experimenter gave the orders, compliance fell to about twenty percent.

Two further results are worth carrying. Forty women, run in the standard condition, produced a compliance rate close to the men's, though Milgram reported higher levels of expressed tension. And when participants were free to choose their own shock level, almost nobody went high, which is the cleanest possible evidence that the behaviour under instruction was not disguised sadism.

Taken together, the variations turn a shocking number into something more useful: a map of the conditions under which ordinary people stop. The most encouraging entry on that map involves nothing more complicated than one other person in the room saying no.