The Bystander-Effect: A Meta-Analytic Review on Bystander Intervention in Dangerous and Non-Dangerous Emergencies Peter Fischer University of Regensburg Joachim I. Krueger Brown University Tobias Greitemeyer University of Innsbruck Claudia Vogrincic University of Graz Andreas Kastenmu ��ller Liverpool John Moores University Dieter Frey University of Munich Moritz Heene, Magdalena Wicher, and Martina Kainbacher University of Graz Research on bystander intervention has produced a great number of studies showing that the presence of other people in a critical situation reduces the likelihood that an individual will help. As the last systematic review of bystander research was published in 1981 and was not a quantitative meta-analysis in the modern sense, the present meta-analysis updates the knowledge about the bystander effect and its potential moderators. The present work (a) integrates the bystander literature from the 1960s to 2010, (b) provides statistical tests of potential moderators, and (c) presents new theoretical and empirical perspectives on the novel finding of non-negative bystander effects in certain dangerous emergencies as well as situations where bystanders are a source of physical support for the potentially intervening individual. In a fixed effects model, data from over 7,700 participants and 105 independent effect sizes revealed an overall effect size of g ���0.35. The bystander effect was attenuated when situations were perceived as dangerous (compared with non-dangerous), perpe- trators were present (compared with non-present), and the costs of intervention were physical (compared with non-physical). This pattern of findings is consistent with the arousal-cost-reward model, which proposes that dangerous emergencies are recognized faster and more clearly as real emergencies, thereby inducing higher levels of arousal and hence more helping. We also identified situations where bystanders provide welcome physical support for the potentially intervening individual and thus reduce the bystander effect, such as when the bystanders were exclusively male, when they were naive rather than passive confederates or only virtually present persons, and when the bystanders were not strangers. Keywords: bystander effect, bystander intervention, dangerous emergencies, helping, meta-analysis On the 12th September, 2009, Dominik Brunner was murdered at a German train station after he helped little children against two per- petrators. He has not chosen to look the other way, but sacrificed himself when others were in need. ���Dominik Brunner Foundation It is a denial of justice not to stretch out a helping hand to the fallen that is the common right of humanity. ���Seneca (5 BC���65 AD) The bystander effect refers to the phenomenon that an individ- ual���s likelihood of helping decreases when passive bystanders are present in a critical situation (Darley & Latane ��, 1968 Latane �� & Darley, 1968, 1970 Latane �� & Nida, 1981). Many sad real-life examples illustrate this effect: In 1964, Kitty Genovese was raped and murdered in Queens, New York, while several of her neigh- bors looked on. No one intervened until it was too late.1 More recently, in 2009, Dominik Brunner was murdered at a German train station by two 18-year-olds after he tried to help children who were attacked by these young criminals. Several passersby wit- nessed the murder, but nobody physically intervened. In support of this anecdotal evidence, an influential research program conducted 1 The precise number of bystanders, what they saw, and how they interpreted the situation are still under dispute (Manning, Levine, & Col- lins, 2007). This article was published Online First May 2, 2011. Peter Fischer, Department of Psychology, University of Regensburg, Regens- burg, Germany Joachim I. Krueger, Department of Cognitive, Linguistic, and Psychological Sciences, Brown University Tobias Greitemeyer, Department of Psychology, University of Innsbruck, Innsbruck, Austria Claudia Vogrincic, Moritz Heene, Magdalena Wicher, and Martina Kainbacher, Department of Psy- chology, University of Graz, Graz, Austria Andreas Kastenmu ��ller, Department of Psychology, Liverpool John Moores University, Liverpool, England Dieter Frey, Department of Psychology, University of Munich, Munich, Germany. The meta-analysis was supported by DFG (German Science Foundation) Project FI 938/2-1. Correspondence concerning this article should be addressed to Peter Fischer, Department of Experimental Psychology, Social and Organiza- tional Psychology, University of Regensburg, 93040 Regensburg, Ger- many. E-mail: peter.fischer@psychologie.uni-regensburg.de Psychological Bulletin �� 2011 American Psychological Association 2011, Vol. 137, No. 4, 517���537 0033-2909/11/$12.00 DOI: 10.1037/a0023304 517

by Bibb Latane �� and John Darley provided strong empirical evidence for the existence of the bystander effect in a variety of experimental settings (see Latane �� & Nida, 1981, for a review). In both a theoretical and a practical sense, the bystander effect has played an increasingly important role in our understanding of helping behavior. References to the effect can be found in nearly every introductory (social) psy- chology textbook. Various television shows continuously report and try to replicate the effect, and knowledge of the effect is now firmly anchored in public awareness. Although the evidence for the inhibitory bystander effect is striking, there are also counter-examples. Sometimes, the presence of bystanders can facilitate acts of moral courage. In Munich in 2001, for example, a young man from Turkey helped a young Greek who was chased and beaten by a group of skinheads. The young Turk risked his life while many other bystanders were watching. Similar results were found in laboratory experiments, where the bystander effect vanished when the emergency was a particularly dangerous one (e.g., Fischer, Greitemeyer, Pollozek, & Frey, 2006). Given the variation in the size and the direction of the bystander effect, we asked on a theoretical level and an empirical level whether there are specific situations that might reduce or even revert the traditional inhibitory effect of bystanders on helping. To answer this question, we conducted a meta-analysis across both the classic and recent studies. There are three pertinent reasons why research on the bystander effect should be submitted to a meta-analytic integration. First, the last systematic review (Latane �� & Nida, 1981) is now dated, and it did not meet current standards for meta-analysis. Second, potential moderator variables have not been considered beyond qualitative description, and the role of some of these moderators (e.g., the ambiguity of the emergency) is unsettled. Third, and perhaps most importantly, reversals of the traditional, inhibitory bystander effect have recently been reported, especially for dangerous emergencies. For these reasons, we conducted a full-scale meta-analysis. Doing so, we aimed to address the following theoretically moti- vated questions: (a) Is the bystander effect reduced or reversed in situations of dangerous emergencies? (b) Are there specific situa- tions where bystanders can increase helping because they are seen as welcome physical support in dangerous emergencies? (c) Are there other theoretically and practically important moderators of the bystander effect? (d) Does the recent wave of bystander studies (i.e., post-1981) offer new insights into the magnitude of the effect (e.g., has the effect increased or declined over time)? Review of Bystander Research Definitions and Psychological Accounts Early research consistently showed that the presence of passive bystanders reduces the likelihood that individuals will intervene and help a victim in a critical situation (Darley & Latane ��, 1968 Latane �� & Darley, 1968, 1970 Latane �� & Nida, 1981). To account for the effect, Latane �� and Darley (1970) proposed a five-step psychological process model. They postulated that for intervention to occur, the bystander needs to (1) notice a critical situation, (2) construe the situation as an emergency, (3) develop a feeling of personal responsibility, (4) believe that he or she has the skills necessary to succeed, and (5) reach a conscious decision to help. Latane �� and Darley (1970) identified three different psycholog- ical processes that might interfere with the completion of this sequence. The first process is diffusion of responsibility, which refers to the tendency to subjectively divide the personal respon- sibility to help by the number (N) of bystanders. The more by- standers there are, the less personal responsibility any individual bystander will feel. Likewise, the individual bystander will only feel responsible for a fraction of the cost to the victim associated with non-intervention. The second process is evaluation apprehen- sion, which refers to the fear of being judged by others when acting publicly. In other words, individuals fear to make mistakes or act inadequately when they feel observed, which makes them more reluctant to intervene in critical situations. The third process is pluralistic ignorance, which results from the tendency to rely on the overt reactions of others when defining an ambiguous situation. A maximum bystander effect occurs when no one intervenes because everyone believes that no one else perceives an emer- gency (cf. Latane �� & Nida, 1981). The bystander literature has remained somewhat ambiguous with regard to the relevant psychological processes. Latane �� and Nida (1981), for example, distinguished the processes of diffusion of responsibility, social influence, and audience inhibition, which are rather close but not fully identical to the processes assumed by Latane �� and Darley (1970). This discrepancy might be one reason for some of the process ambiguity in bystander research. Finally, it should be noted that there are also explanations of the bystander effect that are derived from evolutionary psychology or game theory. Among these are reciprocal altruism (Axelrod & Hamilton, 1981 Trivers, 1971), competitive altruism (Hardy & Van Vugt, 2006), inclusive fitness (Hamilton, 1964a, 1964b), and the volun- teer���s dilemma (Krueger & Massey, 2009). The Classic Bystander Research Paradigm A typical bystander study proceeds as follows: Participants work either alone or in the presence of one or more other partic- ipant(s) (passive bystanders) on an allegedly important task (e.g., filling out questionnaires, waiting for the experimenter). They suddenly witness a staged emergency (e.g., the experimenter be- comes injured, a perpetrator offends someone, a thief steals some- thing). Their responses to these emergencies are recorded, typi- cally in terms of their probability of intervening and the time it takes them to do so. Results in the multiple-bystander condition are then compared with results in the single-bystander condition. By applying this classic paradigm, bystander effects have been found in many domains. For example, bystanders decrease helping in serious emergencies, such as an injury (Latane �� & Darley, 1968), an asthma attack (Harris & Robinson, 1973), or physical illness (Darley & Latane ��, 1968). However, the bystander effect also occurs in less critical situations, such as a stranded motorist (Hur- ley & Allen, 1974) or other technical problems (Misavage & Richardson, 1974). The effect occurs even in cases of mundane mishaps, as when pencils spill to the ground (Latane �� & Dabbs, 1975) or when a door needs to be answered (Levy et al., 1972). The Bystander Literature Before 1981 After the tragic death of Kitty Genovese, Latane �� and Darley began to investigate the social psychological conditions that keep 518 FISCHER ET AL.

people from helping. Their influential research program yielded a variety of important empirical and theoretical insights. Above all, their work showed that the bystander effect is a robust phenome- non that occurs in many experimental and field situations. Summarizing the existing body of evidence at the time, Latane �� and Nida (1981) concluded that Darley and Latane �����s original conjectures and findings had been corroborated. After summariz- ing Latane �� and Nida���s main findings, we turn to studies conducted after 1981, quantitatively address possible moderating effects, and consider bystander effects in dangerous emergencies. Besides fo- cusing on the role of the number of bystanders, Latane �� and Nida looked at seven different characteristics of bystander/emergency situations: attributes of the incident (e.g., incident occurred in rural vs. urban areas), whether the study was conducted in the laboratory or in the field, to what extent the incident was ambiguous, by- stander/participant attributes (e.g., bystander competency, sex of participant), victim attributes (e.g., sex of victim), attributes of other bystanders (e.g., friends vs. strangers), and to what extent bystanders could communicate with each other. Overall, the au- thors identified four different contexts of the bystander effect: (1) all bystanders are in danger (e.g., a room becomes suddenly filled with smoke Latane �� & Darley, 1968 or a fire bell started to ring Ross & Braband, 1973), (2) a victim is in danger (e.g., a person has an asthma attack Harris & Robinson, 1973 a person simulates a seizure Darley & Latane ��, 1968 or a person falls from a bookshelf Latane �� & Rodin, 1969), (3) villain acts (e.g., a perpetrator steels money Latane �� & Elman, 1970 a case of beer is stolen Latane �� & Darley, 1970 books are stolen Howard & Crano, 1974), and (4) non-emergency incidents (e.g., answering the door Freeman, 1974 help with a broken car tire Hurley & Allen, 1974 or leaving a tip Freeman, Walker, Bordon, & Latane ��, 1975). Latane �� and Nida���s distinction between emergency/villain acts (with implied high danger in case of intervention) and non-emergency situations (with implied low danger in case of intervention) is of special interest for the present meta-analysis, because we postulate on the basis of more recent findings (e.g., Fischer et al., 2006) that the strength of the bystander effect systematically varies with the bystander���s expected danger when deciding whether to help. How- ever, no firm answer can yet be given to this question because Latane �� and Nida did not statistically test whether there are differ- ences in effect sizes between emergency (high-danger) versus non-emergency (low-danger) situations. In addition, we test whether the expected effect of emergency danger on the bystander effect depends on various moderators investigated by Latane �� and Nida. Most importantly, Latane �� and Nida (1981) concluded that help- ing is reduced when the number of bystanders increases or when the situation is ambiguous (e.g., Clark & Word, 1974 Solomon, Solomon, & Stone, 1978). The effect occurs both in the laboratory and in the field (e.g., Shaffer, Rogel, & Hendrick, 1975). Latane �� and Nida found that the bystander effect occurs for both sexes of participant and victim (e.g., Latane �� & Dabbs, 1975) as well as across nearly all age groups (except very young children Staub, 1970). Moreover, Latane �� and Nida found that the bystander effect tends to be stronger in cities than in rural areas (e.g., Merrens, 1973). The competence of the bystanders yielded mixed results. Sometimes highly competent bystanders reduced the bystander effect (e.g., Horowitz, 1971), and sometimes they increased it (e.g., Darley & Latane ��, 1968). Mixed conclusions were also drawn for age of bystanders (Ross, 1971) and similarity between them (Smith, Smythe, & Lien, 1972). In contrast, more helping was found for bystanders who were friends instead of strangers (Latane�� & Rodin, 1969). Finally, Latane �� and Nida stressed the importance of communication possibilities among bystanders in predicting the bystander effect rather counter-intuitively, they found that in- creased communication possibilities increased the bystander effect. In sum, from the perspective of the focal individual who is supposed to help, Latane �� and Nida (1981) found a substantial bystander effect in groups. Whereas 75% of participants helped when they faced a critical incident alone, only 53% did so when other bystanders were present. Also, from the victim���s perspective, the likelihood of receiving help was lower when his or her need was witnessed by a group (70% helping) versus by a single person (82% helping) however, this difference was attenuated to non- significance when bystanders could not communicate with one another (for a critical discussion, see Krueger & Massey, 2009). The authors concluded that the effect is robust and that there are few limiting boundary conditions, such as very young bystander age (Staub, 1970), low situational ambiguity (Clark & Word, 1972), low competence to intervene (Bickman, 1971), or reduced communication among bystanders (Latane �� & Darley, 1976). Although these conclusions were justified in light of the empir- ical evidence available at the time, Latane �� and Nida (1981) did not systematically employ quantitative meta-analytical methods to sta- tistically test potential moderators. The present meta-analysis pro- vides quantitative moderation analyses, integrates new studies on the bystander effect, and provides a new theoretical framework for the bystander effect in dangerous emergencies. Latane �� and Nida���s theoretical distinction between emergency and non-emergency sit- uations is of special interest in this meta-analysis, as we expect the bystander effect to be stronger in the latter than the former. The Bystander Literature After 1981 We found 15 new articles on the bystander effect with over 40 effect sizes. The present meta-analysis adds these effect sizes to the effect sizes of the classic research and tests for theoretically important moderation effects. The new studies address a variety of critical issues, such as bystander intervention in dangerous emer- gencies (Fischer et al., 2006 Harari, Harari, & White, 1985), high versus low ambiguous emergencies (Kalafat, Elias, & Gara, 1993), or bystander intervention for a Black victim versus a White victim (Gaertner, Dovidio, & Johnson, 1982). Some recent studies inves- tigated the bystander effect in new media contexts (e.g., online response via e-mail, sharing of virtual knowledge Barron & Ye- chiam, 2002 Blair, Thompson, & Wuensch, 2005 Lewis, Thomp- son, Wuensch, Grossnickle, & Cope, 2004 Markey, 2000 Voel- pel, Eckhoff, & Forster, 2008), social control behavior (Chekroun & Brauer, 2002), donation behavior (Wiesenthal, Austrom, & Silverman, 1983), or effects of group cohesiveness on bystander intervention (Rutkowski, Gruder, & Romer, 1983). Finally, with a focus on potential interventions against the bystander effect, re- search after 1981 has investigated the effects of trained versus untrained bystander groups (Shotland & Heinold, 1985), different levels of bystander competencies (Cramer, McMaster, Bartell, & Dragna, 1988 Pantin & Carver, 1982), and the effects of remind- 519 THE BYSTANDER EFFECT: A META-ANALYTIC REVIEW