In 1964, a newspaper said that 38 people saw a woman attacked in New York, but nobody helped. Later, researchers found that this story was not correct. Far fewer people really saw the attack, and some people did help.

Even so, experiments show that when many people are around, each person is less likely to help. But in real street videos, someone stepped in to help in about nine out of ten fights.

In 1964, The New York Times reported that 38 neighbors watched a woman being attacked but did nothing. The story inspired psychologists John Darley and Bibb Latané to study what is now called the bystander effect. In their 1968 experiment, students heard someone apparently having a seizure over an intercom. When they believed they were the only listener, 85 percent went for help before the seizure ended.

When they thought four others could also hear, only 31 percent did. Responsibility seemed to be shared, so each person felt less need to act. However, a 2007 study of court records showed that the "38 witnesses" story was exaggerated. In addition, a 2020 analysis of security camera footage found that at least one person intervened in about 90 percent of public conflicts.

The 1964 murder of Kitty Genovese in New York long stood for big-city coldness. According to The New York Times, 38 neighbors watched and none called the police. It prompted John Darley and Bibb Latané to study the bystander effect. In their 1968 experiment, 85 percent of the students who thought they alone could hear another participant having an apparent seizure sought help before it ended, against 31 percent of those who thought four others were listening. They blamed diffusion of responsibility.

A 2011 meta-analysis of more than 7,700 participants confirmed that others' presence reduces individual helping, though the effect was weaker in clearly dangerous emergencies. The Genovese story has crumbled. In 2007, Manning and colleagues used court records to show that far fewer than 38 people had clearly seen the attack, and some had tried to help.

A 2020 study of 219 public conflicts caught on security cameras in three countries found that at least one bystander intervened in about nine in ten cases, and more bystanders made intervention more likely. Both findings can be true. Each person may act less readily in a crowd, while the crowd as a whole is more likely to contain a helper.