Nine Times Out of Ten, Somebody Steps In

The 38 witnesses to Kitty Genovese never existed. Then 219 CCTV recordings from three cities showed how often bystanders actually intervene.

The bystander effect is probably the most successful idea psychology ever exported to the general public. People who never took a psychology course can still state it: the more people watching, the less likely anyone is to help. It has a founding story and a name attached, plus sixty years of textbooks.

The founding story is the part that fell apart first.

In March 1964 the New York Times reported that thirty-eight respectable, law-abiding citizens in Queens watched a killer stalk and stab Kitty Genovese over half an hour, and that not one of them called the police during the attack. That article launched a research programme. It was also, it turns out, wrong. In 2007 Rachel Manning, Mark Levine and Alan Collins went back to the archive material and published what they found in American Psychologist. No evidence for thirty-eight witnesses. None that the people who were there watched the murder, or that they stayed inactive. Two of the paper's authors are social psychologists working in the very field the story built.

Their word for what the case became is "parable" — a story that got told because it taught something people already wanted to believe about cities and strangers.

Then somebody watched the tapes

Correcting a founding myth does not tell you what actually happens on a street. For that you need real incidents, and until recently nobody had a way to watch enough of them. Lab studies could tell you what an undergraduate does when smoke comes through a vent in a psychology building. They could not tell you what happens outside a nightclub at one in the morning.

Then closed-circuit cameras got everywhere. In 2020 a team led by Richard Philpot published a study in American Psychologist that used them: 219 recordings of real aggressive incidents — arguments escalating, shoving, fights — from public spaces in Amsterdam, Lancaster and Cape Town. Every clip was coded for whether anyone intervened.

In about 91 percent of the incidents, at least one bystander did something. Not called for help afterwards. Did something during: gestured, physically separated the parties, blocked or held back the aggressor, consoled or pulled away the victim. The average number of people who intervened per incident was 3.76.

The three cities differ enormously in how safe the public believes them to be. Cape Town and Lancaster are not comparable places by any ordinary measure. The intervention rates came out similar anyway.

The direction that surprises people

In this data, the more bystanders were present, the more often somebody acted. That runs the opposite way to what most people remember from the textbook.

It is not, strictly speaking, a refutation of the classic laboratory result. The lab studies measured the probability that a given individual helps, and that probability really does drop as the group grows — responsibility gets diffused, and everybody waits to see what everybody else does. Philpot's team measured a different quantity: the probability that at least one person in the crowd acts. Those two can move in opposite directions, and apparently they do: your own odds of being the one who steps forward fall in a crowd, while the odds that somebody helps you go up.

The popular version collapsed both into one sentence, and the sentence was wrong in the way that mattered most to anyone worried about being the person on the ground.

What the study does not show

CCTV footage is not a random sample of human trouble. Cameras are aimed at busy public places, usually well-lit ones, often town centres at night. An incident that happens where a camera is pointing is by definition an incident with an audience. Nothing here tells you about a stairwell, a private flat, or a quiet road at four in the morning — and those are the situations where a victim most needs the answer.

The 219 incidents were also, by selection, visible conflicts. Aggression that stays quiet, coercion, the slow kind of harm — none of that gets coded by someone watching a screen for shoving.

And this is film, not an experiment. More bystanders went together with more intervention; nobody added people to a street corner to see what would happen. The association is real and it points the encouraging way, but it is an association.

What survives scrutiny is narrower than the headline. When a conflict turns visibly aggressive in a busy public space, in three cities on two continents, someone almost always steps in, and a bigger crowd went with more help rather than less. That is a real finding about a real and common situation; the stronger claim — that nobody is ever left alone — is not supported by it.

Why the belief matters more than the base rate

We have written before about how people say yes to requests roughly twice as often as the asker expects. The bystander story is the same error with the stakes raised. People underestimate the willingness of strangers, and the underestimation is not free — it changes what they do.

Someone who believes nobody will help is slower to shout, slower to ask, slower to make what is happening to them other people's business. That hesitation is real behaviour caused by a false belief, and the false belief was taught in good faith by teachers reading from textbooks that were reading from a 1964 newspaper article whose claims nobody checked against the archive until 2007.

Ninety-one percent is not everyone. Someone reading this will be the exception, and no statistic helps them. But the number is a great deal closer to the truth than thirty-eight people watching from their windows and doing nothing — for which, when somebody finally went and looked, there turned out to be no evidence at all.

Sources. Manning, R., Levine, M., & Collins, A. (2007). The Kitty Genovese murder and the social psychology of helping: The parable of the 38 witnesses. American Psychologist, 62(6), 555–562. Philpot, R., Liebst, L. S., Levine, M., Bernasco, W., & Lindegaard, M. R. (2020). Would I be helped? Cross-national CCTV footage shows that intervention is the norm in public conflicts. American Psychologist, 75(1), 66–75.