Granovetter was most taken by the situations in which people did things for social reasons that went against everything they believed as individuals. “Most did not think it ‘right’ to commit illegal acts or even particularly want to do so,” he wrote, about the findings of a study of delinquent boys. “But group interaction was such that none could admit this without loss of status; in our terms, their threshold for stealing cars is low because daring masculine acts bring status, and reluctance to join, once others have, carries the high cost of being labeled a sissy.” You can’t just look at an individual’s norms and motives. You need to look at the group.

His argument has a second implication. We misleadingly use the word “copycat” to describe contagious behavior—implying that new participants in an epidemic act in a manner identical to the source of their infection. But rioters are not homogeneous. If a riot evolves as it spreads, starting with the hotheaded rock thrower and ending with the upstanding citizen, then rioters are a profoundly heterogeneous group.

Finally, Granovetter’s model suggests that riots are sometimes more than spontaneous outbursts. If they evolve, it means they have depth and length and a history. Granovetter thought that the threshold hypothesis could be used to describe everything from elections to strikes, and even matters as prosaic as how people decide it’s time to leave a party. He was writing in 1978, long before teen-age boys made a habit of wandering through their high schools with assault rifles. But what if the way to explain the school-shooting epidemic is to go back and use the Granovetterian model—to think of it as a slow-motion, ever-evolving riot, in which each new participant’s action makes sense in reaction to and in combination with those who came before?