In the early hours of March 13, 1964, a twenty-eight-year-old bar manager named Kitty Genovese was returning to her apartment in Kew Gardens, Queens, New York, when she was attacked in the street by a man named Winston Moseley. She screamed, and a neighbor's lights came on. Moseley retreated. Then, in two further attacks over the course of roughly thirty minutes, he returned and stabbed her again. She died before the ambulance arrived.

In the days that followed, the New York Times published a story that would reshape the course of social psychology. Reporter Martin Gansberg, under the headline "37 Who Saw Murder Didn't Call the Police," described a neighborhood that had watched a woman die without intervening. The story created a national sensation, and the narrative it offered — that nearly forty witnesses had observed every stage of the attack and had done nothing — became one of the most influential anecdotes in the history of behavioral science.

Like many influential anecdotes, it was not entirely accurate.

Decades later, researchers including Rachel Manning, Mark Levine, and Alan Collins would revisit the original case files, trial testimony, and contemporary police reports and publish a careful reassessment in American Psychologist (2007). Their findings complicated the original story considerably. The attack unfolded in segments, in different locations, at night, with limited sightlines. Many of those who heard sounds did not see the attack, did not understand what was happening, and some did call the police — apparently without being taken seriously. The building entrances, the layout of the street, the partial visibility from apartment windows: these factors meant the scenario was less a case of forty people watching passively and more a case of many people hearing fragments of a violent event they could not fully interpret.

And yet the Manning et al. reassessment does not exonerate the bystanders or dissolve the phenomenon the Genovese case helped launch. The broad structure of the event — multiple people aware of a crisis, none intervening decisively — remained, and it provided the impetus for two social psychologists at New York University to begin a series of experiments that would demonstrate, under rigorous laboratory conditions, that the failure to help is not a matter of indifference or cruelty but a predictable psychological outcome of social context itself.

The effect they identified, the bystander effect, has since become one of the most replicated and practically consequential findings in social psychology.

What the Bystander Effect Is

The bystander effect is the finding that the presence of other people reduces the likelihood that any individual will intervene in an emergency. The larger the group of potential helpers, the less likely any one of them is to act. This is counterintuitive: common sense suggests that more witnesses means more help, that crowds are protective, that numbers give people the confidence to intervene. The research consistently shows the opposite.

The phenomenon is not about the character of the observers. The experiments that established it have been run on university students, nurses, ordinary pedestrians, and trained professionals, across different countries and cultures, and the effect appears across all of them. The question is not who these people are but what the social situation does to the calculation — often unconscious and instantaneous — of whether to act.

Two mechanisms account for most of the effect, and they often operate simultaneously: diffusion of responsibility and pluralistic ignorance.

The Intellectual Lineage

John Darley was a social psychologist at New York University; Bibb Latane was at Columbia. Both had read the coverage of the Genovese case with the particular attention of people who recognized in it something that didn't fit the available psychological explanations. At the time, social psychology explained failures to help primarily in terms of personality — apathy, alienation, the dehumanizing effects of urban life. Darley and Latane suspected the explanation was situational rather than dispositional.

They drew on a tradition in social psychology that stretched back to Floyd Allport's 1920s work on social facilitation — the finding that the mere presence of others changes individual behavior — and to Kurt Lewin's field theory, which emphasized that behavior is always a function of the person in their environment, not the person alone. Solomon Asch's conformity experiments in the early 1950s had demonstrated that people would give obviously wrong answers to simple perceptual questions when surrounded by confederates who gave those wrong answers first. Stanley Milgram's obedience studies, conducted at Yale between 1961 and 1962, had shown that ordinary people would administer apparently dangerous electric shocks to strangers when instructed to do so by an authority figure.

Darley and Latane were working in this tradition: the tradition of demonstrating that situations, not just character, determine action. Their contribution was to identify the specific situational variable — the presence of other bystanders — and the specific mechanisms through which that variable operates.

Their foundational papers appeared in the Journal of Personality and Social Psychology in 1968 and 1970, and their book The Unresponsive Bystander: Why Doesn't He Help? (1970) synthesized the program of research into a coherent theoretical framework that has guided the field ever since.

What the Research Shows

The Smoke-Filled Room

In their first major experiment, published in 1968, Darley and Latane had participants complete a questionnaire in a room that, after several minutes, began to fill with smoke through a vent. Participants were either alone or in groups of three. When alone, 75 percent of participants left the room and reported the smoke within two minutes. When in a group of three, only 38 percent reported the smoke, and of those groups in which no one reported, participants typically remained in the room for the full duration of the study while smoke continued to pour in, frequently glancing at each other and then back at their questionnaires.

The participants in the group condition were not less concerned about fire. They were caught in a social dynamic in which each person, uncertain about whether the situation was an emergency, looked to others for cues — and found others who were also looking around and appearing calm. Everyone took the calmness of others as evidence that there was no real danger. This is pluralistic ignorance: a situation in which each individual privately doubts or fears something but publicly signals the opposite, causing each member of the group to incorrectly infer that the others are relaxed because they have good reason to be.

The Epileptic Seizure Study

In 1968, Darley and Latane conducted what became the most cited experiment in the bystander literature. Participants sat in individual cubicles and communicated with others through an intercom system, believing they were discussing the stresses of college life. The "others" were pre-recorded confederates. At some point, one of the voices began to describe symptoms of an epileptic seizure — choking, confusion, increasingly distressed appeals for help.

Participants who believed they were the only one who could hear the seizure intervened 85 percent of the time, and quickly. Participants who believed four other people could also hear it intervened only 31 percent of the time, and many did not intervene at all.

The elegant design of this experiment was that there was no actual group — each participant was physically alone. The mere belief that others existed who could also hear the emergency was sufficient to depress helping behavior by more than half. The social context, not the physical presence of other people, drove the effect.

The 1981 Meta-Analysis

In 1981, Latane and Nida published a meta-analysis in Psychological Bulletin reviewing over 50 studies involving more than 6,000 participants. The analysis confirmed the effect with remarkable consistency: people in groups were significantly less likely to help than people alone, and the effect size was substantial across diverse experimental paradigms, emergency types, and populations. The probability of helping was inversely related to group size in a pattern that was close to the theoretical predictions of their social impact model. The bystander effect was not an artifact of any single laboratory methodology.

Fischer and Danger

A significant qualification emerged from a large-scale meta-analysis by Fischer et al. published in Psychological Science in 2011. Analyzing 105 studies, Fischer and colleagues found that the bystander effect was reliably weaker — and sometimes reversed — in situations involving clear physical danger to the victim. When the emergency was unambiguously violent rather than ambiguous (a medical event or an unclear situation), groups were not consistently worse at helping than individuals. The proposed mechanism is that clear danger resolves both ambiguity (there is no question what is happening) and the fear that intervening will be socially embarrassing (the social cost of doing nothing is high). Dangerous situations appear to override the inhibitory dynamics that typically operate in ambiguous emergencies.

Group Identity and the Reversal

Mark Levine and colleagues at Lancaster University extended the bystander literature in a direction that had been largely neglected: the role of group identity. In a 2005 study published in the British Journal of Social Psychology, Levine et al. manipulated whether bystanders shared a group identity with the victim. When bystanders were made aware that they and the victim shared the same team allegiance (in this case, support for the same football club), the bystander effect reversed: helping increased with group size rather than decreasing. A follow-up study by Levine, Cassidy, and Brazier (2011), published in Social Psychology, extended this finding, showing that shared social identity — whether based on team allegiance, nationality, or other group membership — can transform the dynamic entirely. The bystander effect is not a fixed feature of group size but of the relationship between the individuals in the group.

The Two Mechanisms

Diffusion of Responsibility

The primary mechanism identified by Darley and Latane is diffusion of responsibility: when multiple people witness an emergency, the sense of personal obligation to act is distributed across the group. No single person feels fully responsible because everyone present shares the responsibility. If there are ten witnesses and any one of them could call for help, the individual burden is experienced as one-tenth rather than total. This diffusion is automatic and largely unconscious.

The effect is well-illustrated by the intercom seizure experiment: the participants who believed others were listening were not callous people who had decided not to help. Many showed clear signs of distress — they were pale, trembling, clearly agitated when the experimenter entered at the end of the study. They had not decided that the seizure didn't matter. They had experienced a progressive reduction in their felt personal obligation as the group grew larger, until the obligation was insufficient to overcome the uncertainty about what to do.

Pluralistic Ignorance

The secondary mechanism operates through social information. In an ambiguous situation — and most genuine emergencies are initially ambiguous — people look to others to determine whether the situation requires action. If a person collapses on a street, the first question that runs through each witness's mind is whether the person is truly in trouble or merely drunk, resting, or playing. Observers instinctively check each other's faces and behavior for interpretive cues.

The problem is that everyone is doing this simultaneously, and everyone is trying to appear calm and composed while doing it — partly from the social norm against overreacting, partly from the embarrassment of misreading a situation. The result is a group of people each privately uncertain and publicly projecting calm, which each person reads as evidence that the others are calm because they know something — that the situation is not serious. The group collectively arrives at an interpretation none of its members individually holds, which is the definition of pluralistic ignorance.

Together, these mechanisms create a trap: ambiguity prompts people to look at others, others project calm, calm is misread as certainty, responsibility is felt to be shared, and no one acts while everyone watches.

Four Case Studies

Case Study 1: Kitty Genovese (1964)

The Genovese case, even revised by Manning, Levine, and Collins (2007), retains its significance as a real event in which the diffuse social structure of an urban neighborhood — people behind closed doors, in separate apartments, hearing fragments of a violent event at night — produced a situation in which no one intervened decisively. The Manning reassessment is important for correcting the mythology of the original press coverage, but it does not argue that help was ably provided. It argues that the situation was more complex than the headline suggested, which is a different claim. The bystander effect does not require forty passive witnesses watching from windows. It operates in precisely the messy, ambiguous, partially-informed conditions that the Genovese case actually presented: multiple people, imperfect information, diffuse obligation.

The case mattered because it launched a research program. And the research program demonstrated that the failure it illustrated was not a moral failure of New York City residents in 1964 but a structural feature of human social psychology that operates across cultures, conditions, and centuries.

Case Study 2: Wang Yue (2011)

On October 13, 2011, a two-year-old girl named Wang Yue toddled into a street in Foshan, in Guangdong Province, China, and was struck by a van. The van driver paused and then drove forward, running over her a second time. Over the next several minutes, at least eighteen pedestrians and cyclists passed the small, bleeding child without stopping or calling for help. A second vehicle also struck her. She was eventually pulled from the road by a woman who collected recyclable rubbish, Chen Xianmei. Wang Yue died in hospital eight days later.

The incident was captured on surveillance footage and, when it circulated online, provoked a national reckoning in China about the moral condition of contemporary Chinese society. Commentators debated whether fear of legal liability — a genuine concern given several high-profile cases in which Good Samaritans had been sued by the people they rescued — had distorted the normal human impulse to help. This legal context is significant: it represents a socially specific inhibitor that compounds the bystander effect's mechanisms. Diffusion of responsibility was operating (eighteen other people were present or passing), pluralistic ignorance was operating (no one was intervening, so the situation might not require intervention), and on top of these standard mechanisms, a country-specific legal deterrent was also present. The Wang Yue case is perhaps the most documented modern example of how multiple inhibitory factors can interact to produce collective inaction of a kind that is, in retrospect, almost incomprehensible.

Case Study 3: London Underground Studies

Mark Levine and colleagues used an unusual natural laboratory to test bystander dynamics: closed-circuit television footage from public areas in Lancaster and London. Their 2011 analysis of footage from public conflicts — situations in which people were being threatened or attacked — found that intervention by third parties was actually quite common in public spaces, occurring in the majority of incidents captured on camera. The more bystanders present, the more likely at least someone intervened. This finding, published in Psychology, Crime and Law in 2011, appears to contradict the standard bystander effect and was initially received with some surprise.

Levine and colleagues reconciled the finding with the existing literature in several ways. Unambiguous public violence, consistent with the Fischer et al. (2011) meta-analysis finding, tends to override the ambiguity mechanism. Interventions in public are often incremental — a word, a gesture, positioning oneself between the parties — rather than dramatic confrontations, which means the social cost of intervening is lower than laboratory analogies suggest. And crucially, when group size increased in their natural dataset, the overall probability of intervention increased even if the probability for any single individual may not have. The London studies do not refute the bystander effect; they specify the conditions under which the inhibitory mechanisms are weaker than the standard experimental paradigm suggests.

Case Study 4: The Online Bystander Effect

The bystander effect does not require physical co-presence. As social interaction has migrated to digital platforms, researchers have examined whether the same mechanisms operate in online environments — and found that they do, with some modifications. Markey (2000) was among the first to demonstrate, in a study of online chat rooms, that requests for help received fewer responses when the group was larger; the probability that any individual would respond fell as the number of people in the chat increased. More recently, studies of cyberbullying have found that large numbers of witnesses to online harassment rarely intervene. Bastiaensens et al. (2015), in a study published in Computers in Human Behavior, found that bystanders to cyberbullying reported lower perceived responsibility to intervene when they believed many others had also seen the incident.

The mechanisms translate directly. The internet creates diffusion of responsibility at scale: when a harassing post is seen by ten thousand people, the burden on any one person to act is experienced as infinitesimal. Pluralistic ignorance also operates: if no one visibly responds or challenges the harassment, the silence reads as normalized — as evidence that others either condone it or do not find it serious. Online environments add a complication: anonymity can either increase or decrease helping behavior depending on the nature of the situation and the salience of group norms. Where offline bystanders are inhibited partly by the embarrassment of intervening publicly, online bystanders face less immediate social exposure — yet helping behavior remains depressed, suggesting that diffusion of responsibility may be the more powerful of the two mechanisms.

The bystander effect sits within a broader landscape of social psychological phenomena, several of which either contribute to it, overlap with it, or offer contrasting predictions.

Concept Definition Relationship to Bystander Effect
Diffusion of Responsibility The dilution of individual felt obligation in group settings The primary mechanism through which the bystander effect operates
Pluralistic Ignorance Each member of a group privately holds a view different from the one they publicly express, leading the group to a shared mistaken inference The secondary mechanism; explains why ambiguity is resolved in the wrong direction by groups
Social Loafing The tendency for individuals to exert less effort on collective tasks than on individual tasks Shares the diffusion mechanism; extends it to effort rather than emergency intervention
Conformity The adjustment of one's behavior or beliefs to match those of a group Overlaps with pluralistic ignorance; the pressure not to be the only one who reacts is a conformity dynamic
Deindividuation The loss of individual identity and self-awareness that occurs in large groups Related but distinct; deindividuation predicts reduced inhibitions (sometimes increased antisocial behavior), while the bystander effect predicts reduced prosocial behavior
Moral Disengagement Bandura's concept of psychological mechanisms that allow people to behave unethically without self-condemnation Partially overlaps; in bystander situations, people may rationalize non-intervention, but the primary bystander effect mechanisms operate before moral evaluation
Social Identity Theory Tajfel and Turner's framework in which group membership shapes self-concept and behavior toward in-group vs. out-group members Directly moderates the bystander effect; shared identity with the victim reverses the diffusion of responsibility effect (Levine et al., 2011)

When the Bystander Effect Does Not Apply

The bystander effect is one of the most robust findings in social psychology, but its robustness is conditional. Research over the past two decades has identified several circumstances in which the standard predictions do not hold or are substantially weakened.

Unambiguous danger. The Fischer et al. (2011) meta-analysis found that the effect is reliably weaker when the emergency involves clear physical harm. Dangerous situations override ambiguity — one of the two mechanisms driving the effect — and may create sufficient social pressure that the norm against non-intervention becomes stronger than the norm against overreacting.

Small, cohesive groups. The bystander effect is driven partly by the diffusion of personal responsibility. In small, cohesive groups where accountability is high and where members feel personally known to each other, responsibility is harder to diffuse. Experimental work by Rutkowski, Gruder, and Romer (1983) found that bystanders who felt cohesion with their co-witnesses were more rather than less likely to help, reversing the standard prediction.

Shared group identity with the victim. As noted above, the Levine et al. work on social identity constitutes perhaps the most theoretically important qualification of the bystander effect. When bystanders identify with the victim — when the victim is perceived as "one of us" rather than a stranger — the diffusion of responsibility is counteracted by the pull of in-group solidarity. The implication is significant: the bystander effect is not about human nature in the abstract but about the social structure of the group that witnesses an emergency.

Expertise. When bystanders have specific expertise relevant to the emergency — medical training, for instance, or fire safety training — they are substantially more likely to intervene, regardless of group size. This partially explains why first-responder training programs that include bystander intervention components (CPR training, mental health first aid) appear to increase intervention rates in the populations trained.

Personal relationship with the victim. The bystander effect is reliably attenuated when the victim is known to the observers. A person who collapses on a city street among strangers is in a different situation from a person who collapses at a family dinner. The diffusion of responsibility requires a degree of anonymity between observers and victim to operate fully.

High perceived self-efficacy. Bystanders who feel confident in their ability to help — who believe they can actually do something useful — are more likely to act. This is consistent with Bandura's social cognitive theory and with research on bystander training: one mechanism through which training programs increase intervention is not merely by reducing diffusion of responsibility but by increasing bystanders' belief that their intervention will be effective.

Applications

Bystander Intervention Training

The single most important applied implication of the bystander effect research is that it is possible to train people to overcome the effect's inhibitory dynamics. Programs such as Green Dot (developed by Dorothy Edwards at the University of Kentucky) and Bringing in the Bystander (developed at the University of New Hampshire) target sexual violence prevention by teaching potential bystanders to recognize warning signs and to act despite the inhibitory social pressures the research has identified.

These programs work primarily by making the diffusion-of-responsibility and pluralistic-ignorance mechanisms explicit: by teaching people that their instinct to wait and see what others do is a known psychological trap, not a reliable guide to action. Direct instruction about the bystander effect, and practice in the specific behavioral responses that constitute intervention, consistently shifts behavior in the direction of increased helping in follow-up assessments.

Emergency Response Design

The bystander effect has practical implications for the design of emergency response systems. A 911 call that begins "someone should probably call..." is less likely to result in a fast response than one made by a person who feels personally responsible. Emergency response trainers have long understood the importance of direct assignment of responsibility — telling specific individuals to call emergency services, perform CPR, direct traffic — rather than issuing general appeals to a group. The Darley-Latane research provides the theoretical grounding for why direct assignment works: it eliminates the diffusion mechanism by concentrating responsibility on a named individual.

Online Platform Design

Understanding the online bystander effect has implications for how social media platforms and online communities are designed and moderated. Platforms that make visible the number of people who have seen a post but not responded create conditions for maximum diffusion of responsibility. Interventions that make individual witnesses feel personally addressed — through direct notifications, individual prompts, or mechanisms that make their specific inaction visible — may counteract the diffusion dynamic.

References

  1. Darley, J. M., & Latane, B. (1968). Bystander intervention in emergencies: Diffusion of responsibility. Journal of Personality and Social Psychology, 8(4), 377–383. https://doi.org/10.1037/h0025589

  2. Latane, B., & Darley, J. M. (1968). Group inhibition of bystander intervention in emergencies. Journal of Personality and Social Psychology, 10(3), 215–221. https://doi.org/10.1037/h0026570

  3. Latane, B., & Nida, S. (1981). Ten years of research on group size and helping. Psychological Bulletin, 89(2), 308–324. https://doi.org/10.1037/0033-2909.89.2.308

  4. 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. https://doi.org/10.1037/0003-066X.62.6.555

  5. Fischer, P., Krueger, J. I., Greitemeyer, T., Vogrincic, C., Kastenmüller, A., Frey, D., Heene, M., Wicher, M., & Kainbacher, M. (2011). The bystander-effect: A meta-analytic review on bystander intervention in dangerous and non-dangerous emergencies. Psychological Bulletin, 137(4), 517–537. https://doi.org/10.1037/a0023304

  6. Levine, M., Cassidy, C., & Brazier, G. (2002). Self-categorization and bystander non-intervention: Two experimental studies. Journal of Applied Social Psychology, 32(7), 1452–1463. https://doi.org/10.1111/j.1559-1816.2002.tb01446.x

  7. Levine, M., Prosser, A., Evans, D., & Reicher, S. (2005). Identity and emergency intervention: How social group membership and inclusiveness of group boundaries shapes helping behavior. Personality and Social Psychology Bulletin, 31(4), 443–453. https://doi.org/10.1177/0146167204271651

  8. Levine, M., & Crowther, S. (2008). The responsive bystander: How social group membership and group size can encourage as well as inhibit bystander intervention. Journal of Personality and Social Psychology, 95(6), 1429–1439. https://doi.org/10.1037/a0012634

  9. Bastiaensens, S., Vandebosch, H., Poels, K., Van Cleemput, K., DeSmet, A., & De Bourdeaudhuij, I. (2015). 'We all feel the same pressure': A qualitative exploration of bystander motivations to intervene in cyberbullying. Computers in Human Behavior, 48, 486–494. https://doi.org/10.1016/j.chb.2015.02.013

  10. Markey, P. M. (2000). Bystander intervention in computer-mediated communication. Computers in Human Behavior, 16(2), 183–188. https://doi.org/10.1016/S0747-5632(99)00056-4

  11. Rutkowski, G. K., Gruder, C. L., & Romer, D. (1983). Group cohesiveness, social norms, and bystander intervention. Journal of Personality and Social Psychology, 44(3), 545–552. https://doi.org/10.1037/0022-3514.44.3.545

  12. Latane, B., & Darley, J. M. (1970). The Unresponsive Bystander: Why Doesn't He Help? Appleton-Century-Crofts.

Tags

bystander effect social psychology diffusion of responsibility cognitive bias behavioral science

Frequently Asked Questions

What is the bystander effect?

The bystander effect is the finding that the presence of other people reduces any individual's likelihood of intervening in an emergency. First demonstrated by John Darley and Bibb Latane in 1968 following the Kitty Genovese murder, it holds that as the number of bystanders increases, each person's felt responsibility for acting decreases — a phenomenon called diffusion of responsibility. The effect is robust across experimental conditions and real-world contexts.

What did Darley and Latane's original experiments find?

In their landmark 1968 studies, Darley and Latane staged emergencies for participants who believed themselves to be either alone or in a group. In the smoke-filled room experiment, 75% of lone participants reported the smoke within two minutes; with two passive confederates present, only 10% did. In the intercom seizure study, 85% of participants who believed they were the only listener sought help; with five supposed bystanders, only 31% did — and response time increased dramatically with group size.

What is diffusion of responsibility?

Diffusion of responsibility is the primary mechanism behind the bystander effect. When multiple people witness an emergency, the moral obligation to act is perceived as shared among all present — each individual feels less personally responsible for the outcome. In a group of ten, each person may implicitly calculate that their individual responsibility is one-tenth of what it would be if they were alone. This mathematical division of felt obligation reduces each person's motivation to act.

Does the bystander effect apply online?

Yes. Research on cyberbullying and online harassment has found a digital bystander effect: as the number of people who have viewed an abusive post increases, the probability of any individual intervening or reporting it decreases. The mechanisms — diffusion of responsibility and pluralistic ignorance — operate in online contexts much as they do in physical ones, despite the absence of face-to-face social pressure.

When does the bystander effect not apply?

Fischer et al.'s 2011 meta-analysis found the bystander effect is significantly weaker in genuinely dangerous situations — when the victim is visibly in peril, bystanders are more likely to help regardless of group size, possibly because the clarity of need overrides pluralistic ignorance. Levine et al. found the effect reverses when bystanders share a group identity with the victim: strangers benefit less from large groups, but fellow group members benefit more. Personal relationship, expertise, and unambiguous danger all attenuate the effect.

Share this article