NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Forum on Global Violence Prevention; Board on Global Health; Institute of Medicine; National Research Council. Contagion of Violence: Workshop Summary. Washington (DC): National Academies Press (US); 2013 Feb 6.

Cover of Contagion of Violence

Contagion of Violence: Workshop Summary.

Show details


Jeffrey Victoroff, M.D.

University of Southern California Keck School of Medicine

The Evolution and Mechanism of Aggression

Animals employ aggression in many ways to serve many intermediate goals, such as acquisition of nutrients, defense against predators, social control, and sexual success. However, the reason that aggressive behavior occurs throughout the kingdom Animalia is simply that natural selection favors the genetic replication of individuals who are more likely to survive and reproduce, and aggression facilitates inclusive fitness in multiple ways.

Some aggression is collective. For example, social insects such as ants and bees act aggressive collectively to advance and protect their fitness interests (Moffett, 2011). Social primates such as chimpanzees or spider monkeys coordinate their aggression in ways that enhance individual and (indirectly) group benefit (Mitani et al., 2010). Other social primates such as humans routinely engage in collective aggression. Examples include cooperative hunting, intergroup raiding and defense, police-like social control, and later in history, the emergence of specialization of subgroups who participate more than other community members in warfare. This type of collective aggression seems to have evolved, at least in part, because reciprocal altruism requires and rewards a costly signal demonstrating risk taking on behalf of the in-group (Barclay and Willer, 2007; Miller, 2007). Evidence shows that the very existence of human civilization derives from the innovation of such pro-social collective aggression (Bowles, 2009). Soldiers join armies and wage wars to demonstrate their commitment to reciprocal altruism. Genetic variants and cultural trends that tended to perpetuate or enhance such aggression have been favored, perhaps for millions of years of hominine evolution. It is possible, but remains to be shown that the success of Cro Magnons over Neanderthals was strongly related to this early modern human trait.

The human brain embodies aggressive potential. Multiple anatomical regions play complex interactive roles in mediating individual or collective aggression, including (1) brainstem (which help to monitor the environment, mediate arousal, and regulate the many neurotransmitters and neurohumoral agents that either increase or permit aggression); (2) the hypothalamus (which helps regulate multiple steroid and peptide hormones relevant to the mediation and moderation of aggression, such as serotonin, dopamine, cortisol, and the nonapeptides—arginine vasopressin and oxytocin); (3) the medial temporal lobar allocortex, especially the amygdala, which plays an important role in both detecting and learning threat; and (4) the cingulate gyrus (which mediates drive or motivation, as well as integrating cortical/liminal activity with subcortical/subliminal activity); and several regions of the neocortex (e.g., the dorsolateral prefrontal cortex, which mediates conscious or semiconscious decision making, and the ventromedial prefrontal cortex, which plays multiple roles, including assessment of social circumstances and inhibition of impulsivity) (Adams, 2006; Victoroff et al., 2011; Comai et al., 2012). Differences occur in the level of aggressiveness of individuals who are not accounted for by focal brain anatomy, but correlate instead with a variety of genetic polymorphisms that appear to underlie brain function, or with physiological variations such as hormone levels, central electrophysiology, or autonomic nervous system function.

Therefore, an innate capacity for and tendency toward aggressive behavior, embodied in the brain, represents an important and valuable outcome of the coevolution of genes and cultures.

For better or for worse—and perhaps contrary to the conventions of intuition—the same evolutionary, cultural, and neurobiological factors endow humans with a capacity for antisocial aggression, whether perpetrated by individuals or groups. More aggressive individuals, even psychopathic individuals, are merely expressing one perfectly adaptive life history option for maximizing fitness: the live-fast-die-young strategy (Wolf et al., 2007; Victoroff et al., 2011). Examples of this adaptive strategy include the commission of unsanctioned murder, rape, and participation in violent juvenile gang activity, and participation in violent adult organized criminal activity.

Individual and collective aggression varies. Some individuals and some groups indisputably exhibit relatively higher or lower frequencies and severities of individual (e.g., homicide) or collective (e.g., gang-related homicide) antisocial aggression. At least eight overarching factors contribute to the variation in observed aggressive behavior over time and space:


Individual genetic and epigenetic variation, for example, proaggressive polymorphisms such as the short allelic variant of the gene for monoaminoxidase, the A1 allele of the dopamine receptor D2 gene, the short variation of a repetitive sequence in the transcriptional control region of the serotonin (5-HT) transporter gene (SCL6A4, 5-HTT), or shorter repeat length of androgen receptor gene CAG.


Individual congenital/infancy factors, such as pre- or perinatal injury, fetal exposure to neurotoxins, birth complications, infantile illness, or larger body size at birth.


Individual variations in biological endophenotypes such as atypical neurotransmitter levels, transmission dynamics, cytokine levels, neuroendocrine traits and states associated with both steroid and peptide hormone levels, atypical central electrophysiology, atypical amygdalar responsiveness, atypical autonomic nervous system function, or a recently hypothesized reward deficiency syndrome linked to atypical monoamine function.


Individual postnatal environmental factors, such maternal deprivation, child abuse or harsh punishment, neurotoxicity (e.g., lead paint exposure), anabolic steroid exposure, prescribed agents, or substance abuse (note that a distinction is made between the comorbidity of substance abuse and aggression, described below, and the occurrence of aggression-promoting neurotoxicity due to substance abuse), stressful events or losses (with or without posttraumatic stress disorder or PTSD), exposure to media incitement or modeling, head trauma, or, rarely, infectious disease.


Occurrence of significant Axis I or Axis II mental disorders, including conduct disorder, schizophrenia, autism, antisocial personality disorders, borderline personality disorder, mood disorders, PTSD, dissociative disorders, substance abuse, posttraumatic encephalopathy, and combinations of such disorders, especially comorbid thought disorder and substance abuse.


Variation in personality traits, including callous-unemotional traits, impulsivity, antisocial traits, sensation seeking or risk taking, seeking or defense of dominance, fearlessness, or low frustration tolerance.


Variations in cognitive style or intellectual capacity, including the occurrence of learning disabilities or attention deficit disorders.


Group environmental factors, such as cultural tolerance or support of violent problem solving or revenge killings, socioeconomic factors reducing the availability of alternate life history strategies such as marriage and/or gainful employment, social learning of aggression, and social network factors exposing individuals to other innovators from whom they learn the efficacy of aggression. Such group factors inspire the coalescence of like-minded persons into groups sharing an identity. This dynamic transcends ideology and drives collective aggression of multiple types. For example, as Crenshaw opined (1986, p. 395):

Terrorist organizations become countercultures, with their own values and norms, into which new groups are indoctrinated…. They are in this respect similar to youth gangs or nonpolitical cults and sects.

Indeed, evidence suggests that while individuals with some genomes may be inherently at lower or higher risk of phenotypic aggression, and that certain environmental exposures increase the risk of adopting a violent lifestyle, it is most likely that aggressiveness is determined by bidirectional interactions between genes, epigenetic variations, endophenotypes, and environmental exposures (Veenema, 2009; Cohen, 2010; Nordstrom et al., 2011). For example, a child born with a neurobiologically based reward-deficiency syndrome may both be attracted to risk-taking behaviors and to substances of abuse. Seeking rewarding substances will not only plunge that child into a problematic social milieu, but may produce brain damage that will exacerbate his or her antisocial traits. A child who experiences early stress may develop the endophenotype of altered neuropeptide function, causing a lowered threshold for reactive aggression (Fries et al., 2005). Similarly, a child born with a suboptimal central processing system for emotional regulation may be more vulnerable to early victimization (Rudolph et al., 2011)—a known precursor of gang participation.

In essence, variations in cognitive style and emotional reactivity based on evolutionary diversity of adaptive life history strategies lead to variations in expression of both individual and, perhaps, collective aggression.

The Useful Metaphor of Contagion

Curiously, however, despite decades of research, variation in the rate of individual and collective aggression cannot be completely accounted for by factors that intuitively would explain the waxing and waning in violence. Neither economic markers, group humiliation nor dispossession, availability of weapons, exposure to psychotropic substances, nor any other plausible factors have been shown to account for all of the up and down swings in the rate of communal violence (Gilligan, 1997; Zimring and Hawkins, 1997; Fagan et al., 1998; Rutter et al., 1998).

This observation has perhaps been the inspiration for a number of novel theories about the ultimate genesis of criminal violence, including the hypothesis that violence is contagious. That is, entirely apart from changes in age distribution, population frequency of proaggressive genetic polymorphisms, rates of child abuse, rates of poverty, or political oppression, scholars have noted what appears to be a cyclical pattern of the occurrence of community violence and have proposed that the spreads and contractions in the rate of such violence may represent a phenomenon that is strongly analogous to the spreads and contractions in the occurrence of infectious diseases.

An abundant literature has emerged over the past 60 years examining the plausibility that the concept of contagion usefully accounts for trends in a wide variety of social phenomenon. Contagious diffusion—or innovation followed by imitation—has been proposed as an important cause of biobehavioral trends as diverse as medical innovation, sexual behaviors, team behavior, mood, entry into first marriage, smoking, teenage suicide, and even everyday decision making.

Many forms of political aggression have also been attributed to contagion, including political unrest, political coups, civil wars, riots, ethnoreligious conflict, and terrorism (Bandura, 1973; Huff and Lutz, 1974; Li and Thompson, 1975; Bohstedt, 1994; Fox, 2004; Sedgwick, 2007; Nacos, 2009; Kathman, 2011). More specifically with regard to the waxing and waning of community violence, multiple authors have offered theoretical and empirical reasons to believe that such violence spreads in a contagious manner (Fagan and Davies, 2004; Patten and Arboleda-Florez, 2004; Fagan et al., 2007; Papachristos, 2009).

According to this hypothesis, individuals innovatively adapt their behavior to the goals and circumstances. Some will be innovators of violence. As other people observe the innovators, especially if the innovators are seen to achieve important life goals, imitation will occur (e.g., Fagan et al., 2007). Bandura (1973, p. 215) described the dynamics of this clearly:

Social contagion of new styles and tactics of aggression conforms to a pattern that characterizes the transitory changes of most other types of collective activities: New behavior is initiated by a salient example; it spreads rapidly in a contagious fashion; after it has been widely adopted, it is discarded.

(If those societies ever discard interpersonal and intergroup violence as a behavioral tactic.)

Authorities in quantitative sociology have proposed that, in such cases of contagious diffusion of behavior, independent of external factors one would expect phenomena such as the level of community violence to vary in a cyclical pattern that will roughly approximate a sigmoid curve. Yet the inevitability of historic or demographic factors playing a role has led to the prediction that that curve would be asymmetric. Indeed, Fagan and colleagues (2007) published evidence of just such a curve describing the otherwise inexplicable waxing and waning of handgun violence in New York City between 1968 and 2000 (see Figure II-7).

FIGURE II-7. Gun and nongun homicide rates per 100,000 persons, 1968–2000, New York City.


Gun and nongun homicide rates per 100,000 persons, 1968–2000, New York City. SOURCE: Fagan et al., 2007; used by permission.

Figure II-7 from Fagan et al. (2007) may be compared with our preliminary analysis of data from the Israeli/Palestinian conflict. With Janice Adelman, I recently analyzed both the occurrence of terrorist attacks and the level of Palestinian community support expressed for such attacks between January 1996 and January 2011. Our initial hypothesis was that, in accordance with Crenshaw's admonition (1986, 1995) that the behavior of terror groups is often (but not always) linked to social approval and communal support for that behavior, and her comment that this is particularly the case with regard to Hamas (Crenshaw, 2000), one would observe a time-lagged correlation between measures of communal support and measures of militant action, with attacks increasing as support for those attacks increased. We also speculated that major historical events, such as Ariel Sharon's stepping onto the Temple Mount and the construction of the security wall, would perturb the relationship between communal support and attacks. The data were more complex. While some upswings or downswings in attacks seemed explicable by reference to communal support or major historical/policy changes, an asymmetric quasi-sigmoid curve emerged that could not be accounted for entirely by these factors (see Figure II-8).

FIGURE II-8. Plot of support for attacks (Y1) and number of perpetrated attacks (Y2) over time.


Plot of support for attacks (Y1) and number of perpetrated attacks (Y2) over time. SOURCE: Victoroff and Adelman, 2012.

Comparing Figures II-7 and II-8, and acknowledging the major differences in the types of aggression and the methodology of data acquisition, one is at least tempted to consider the possibility that (a) contagion-like dissemination of aggressive behaviors may help to explain otherwise mysterious fluctuations, and (b) the early prediction of asymmetric quasi-sigmoid trajectories in the occurrence of such phenomena seems defensible.

Obviously, if there exists an asymmetric sigmoidal trend in the occurrence of communal violence of widely disparate types, that pattern of variation-with-time has a cause. For both theoretical reasons and because of our Palestinian data, I propose a hybrid model of the contagion hypothesis.

Violence waxes and wanes in part because of innovation and imitation. Yet it is vital to acknowledge the multiplicity of other factors that may contribute to or cause major changes in community rates of individual or collective violence. Innumerable such factors have been proposed or identified—from the population density of proaggressive genetic polymorphisms, to the occurrence of harsh discipline or parental abuse, to the rate of childhood heavy metal neurotoxicity, to the cohesiveness of the community, and to the structural stresses of deprivation and income inequality.

The neurobiological mechanisms by which these factors influence the central nervous systems of participants in violence are slowly being elucidated. At this point, it is premature to propose a weighting of factors or tight localization of systems, circuits, neurons, and neurohumors contributing to the causal pathway. Yet one conclusion has become inescapable: People do not simply choose to become violent by the rational exercise of free will.

That is, according to the discipline of behavioral economics, all decisions are rational. Humans are assumed to have evolved rational decision-making nervous systems, and indeed, the primate brain appears to contain systems that internally represent values, calculate risks and benefits, and make useful behavioral choices.

But classical microeconomic algorithms fail to provide accurate predictions of real human decision making for several reasons. First, contrary to a basic assumption of rational choice theory, empirical evidence shows that humans are not good intuitive calculators of risk and benefit. Second, there is considerable individual variation in human decision making, mediated not only by biases such as risk aversion, ambiguity aversion, and choice blindness, but by variations in numeracy—all neural operations potentially subject to genetic and epigenetic variation. Both intra- and interindividual variation occurs in emotional biasing of decisions. Emotional framing, for instance, makes brains process moral options with different tissue and outcomes. Innate and acquired variation may alter decision calculus. Fourth, recent research has identified systematic cultural-bound differences in punishment decisions. These recent findings may be especially relevant to the analysis of Islamist extremist behaviors; compared with most Westerners, those in some Middle Eastern societies appear more willing to engage in costly and irrational “antisocial punishment,” which may take the form of vengeance even when such actions are self-defeating.

All these real-world violations of economic prediction are sometimes excused as so-called bounded rationality—the claim that hominids simply lack the calculating capacity for reliable self-interested choice (Simon, 1955). It is possible that one major conceptual error underlying the misguided enthusiasm for rational choice models is the untenable assumption that human brains are serial processors (Simon, 1967). Although some authorities cling to the framework of quantitative predictive validity by calling their new calculus “neuroeconomics,” far too many deviations from rational economic predictions have been observed to sustain a hypothesis that economic balancing of expected risks and benefits plays a major role in the occurrence of most human actions. Newer models of decision making at least acknowledge an element of random or stochastic choice that only slowly drifts toward reward maximization (Soltani et al., 2006). Revising the rational choice/econometric theories of the past, a new generation of scholars is exploring the interaction between emotions and decision making.

The deep philosophical issue that lurks beneath these investigations is the popular assumption of free will. That issue cannot be addressed adequately in this brief essay. Suffice it to say that, from the biologist's perspective, the evidence for human free will is just as robust as the evidence for precognition. The illusion of free will is perhaps best regarded as a curious adaptation, perhaps confined to species with cortices, the value of which remains to be elucidated.

Still, rather than considering the overwhelming evidence of rationality violations as debunking a myth of rational man, I propose that such discoveries open the door to refinements that will lead to a better, wiser, biologically informed psychoneuroeconomics. Human actions on this earth are determined by actions in brains. Actions in brains are determined by physical laws, the investigation of which remains in its early infancy. That incipient science is beginning to attend far more to the importance of emotions. Emotions are central to determining who among us ultimately acts violently and who deals with life's challenges without resort to violence.

Emotions are subjective feelings in response to either internal or external stimuli. They derive both from innate, inborn, and probably genitivally and epigenetically determined personality traits and from many fetal, childhood, and later developmental and environmental influences. When one of three boys in a family joins a gang and the others do not, it is insufficient to point to political structural factors, economic factors, or even parenting as the trigger. The gang joiner is different. He or she probably exhibits different types and degrees of emotional and physiological responsiveness to loss, to perceived threat, to perceived injustice, to out-group exposure, and to the rewards of peer acceptance. I predict that empirical research will identify both emotional and biological traits that distinguish gang participants from the nonparticipant siblings.

Discoveries in this domain potentially have implications for the prevention of violence. If, for example, early childhood depression or traumatic brain injury were shown to account for a significant amount of the variance in gang participation, the community could redouble its efforts to provide comprehensive maternal health care, and to protect young brains from injury (e.g., by evidence-based revisions in return to play guidelines in youth sports).

Cause(s) of Violence

Based on the foregoing discussion, it becomes clear why any unidisciplinary approach to the terrible dilemma of nonsanctioned human violence will founder. The metaphor of contagion (which shares a great deal of its empirical authority with findings from studies of network theory), may help to account for some aspects of trends in violence over time. Yet, rather than searching for “the” cause of a violent event, it may be useful to consider the interaction of multiple causes.

The epidemiology of primary injury prevention also offers a potentially useful way to conceptualize this kind of causality: the Haddon Matrix. Haddon matrixes were originally devised by physician William Haddon as a new way to analyze the causes of injuries and the multiple potential avenues for prevention. This framework alerts policy makers that factors influencing injuries—including violent injuries—are subject to the influence of three overlapping tiers of potential intervention: individual behavioral, environmental, and public policy. Since Haddon's introduction of this conceptual framework, it has been applied to diverse forms of injury, including

  • tiger escapes,
  • crocodile attacks,
  • burns,
  • amusement park accidents,
  • chemical warfare terrorism,
  • the 2005 London bombings,
  • workplace violence, and
  • youth gun violence.

For purposes of illustration, I have prepared two preliminary Haddon matrixes addressing the largely overlapping phenomena of violent youth gangs and violent political extremists (Tables II-3 and II-4). There appear to be important shared characteristics of these superficially different social problems. Both involve in-groups of persons who share an identity that is in conflict with one or more out-groups. Another similarity is that individuals who elect to participate are, in essence, electing a live-fast-die-young life history strategy (Victoroff et al., 2011). Another may be that both gang members and extremists are preoccupied with collective blame of others. It seems plausible that both urban youth and extremists may be propelled, to some degree, by audiovisual media depictions or incitement of violence (Tsfati, 2002; Atran and Stern, 2005; Gunter, 2008; Wright, 2008; Anderson et al., 2010). Moreover, it was recently reported that, just as elevated testosterone (T) levels are associated with antisocial behavior among adolescent boys, evidence suggests that elevated T may be associated with support for extremist violence among adolescent boys (Victoroff et al., 2011).

TABLE II-3. Haddon Matrix for Violent Urban Youth Gang Membership: Known or Suspected Risk Factors for the Occurrence and Dissemination of Gang-Related Violence.


Haddon Matrix for Violent Urban Youth Gang Membership: Known or Suspected Risk Factors for the Occurrence and Dissemination of Gang-Related Violence.

TABLE II-4. Haddon Matrix for Violent Extremist Groups: Known or Suspected Risk Factors for the Occurrence and Dissemination of Political Violence.


Haddon Matrix for Violent Extremist Groups: Known or Suspected Risk Factors for the Occurrence and Dissemination of Political Violence.

While it is tempting to propose a distinction—that youth gangs are not ideologically driven while extremist groups are—this overstates the difference. Although urban youth gangs may not base their violent plans on a coherent, articulated religious or political ideology, they clearly base their behaviors, in part, on a shared weltanschauung in which the local world is viewed as a hostile, hopeless, and insecure place in which conventional values are irrelevant, injustice is rampant, social Darwinism determines success, and in-group age-related gang affiliation offers identity, fictive kinship, and physical protection. Thus, gangsters evading the police in a crack house in Detroit or extremists evading drones in the souks of North Waziristan perhaps share the worldview of a beleaguered oppositional counterculture.

This is not by any means to say that urban youth gangs and violent extremist groups are identically structured or motivated. In conversations with members of Los Angeles gangs and with members of Hamas, both similarities and differences emerge. One important difference between these types of groups is that gang violence is often reactive—an immediate response to a confrontation with a rival, often unplanned, while terrorist aggression is more likely to be proactive and premeditated. Even so, just as extremists usually plan their attacks, gangs sometime premeditate attacks or lie in wait. Another difference is that urban street gangs usually comprise age-stratified networks, while extremist networks are less likely to be age stratified. Another difference is that, while members of urban gangs would die to defend their fictive brothers, they are not devoted to advancing an ideological goal that would benefit a significant part of society as much as they are devoted to advancing their own personal interests, including entrepreneurial ambitions. That is, while young people ultimately join any type of violent group out of personal interest, that interest is conscious and explicit in the case of urban gangsters, but not among violent extremists. Another possible difference relates to perceived progress toward life goals, hope, and well-being; while frustration is expressed by both gang members and extremists, acknowledging a lack of quantitative empirical evidence, it is my impression that there is more fatalism and anomie among typical urban gang members and more hope (realistic or not) of a changed world among extremists. This is perhaps consistent with the empirical observation of relatively high rates of depression and anxiety among members of youth gangs—a phenomenon that perhaps relates, in part, to another difference: youth gang members are at a high rate of victimization in their communities (Taylor et al., 2008).

In these preliminary Haddon matrixes, the content of each cell is based on empirical research, but does not provide weightings in terms of scientific defensibility for each proposed causal factor. Instead, these are known or suspected causal factors for which the literature supports further investigation (for the sake of concision in this brief essay, I will not offer citations for each factor in each cell; such citations are available on request).

It is immediately apparent that known or suspected causal factors might sometimes be assigned to more than one cell. For example, heritable environmentally induced epigenetic changes favoring impulsivity, aggression, hyperresponsiveness or threat, irritability, group allegiance, or urge for antisocial punishment could be classified as Vectors or Agents that disseminate violent gang behavior, but, once present in somatic cells, might also be classified as Individual factors. One must rush to acknowledge the relative paucity of empirical findings supporting the influence of the proposed factors in Tables II-3 and II-4. The biggest challenge is to identify the vector. That is, if violent behaviors, like injuries, occur in contagious clusters, it would be valuable to identify the mechanism of transmission.


In conclusion, natural selection is responsible for the human population as it is. Human brains mediate both individual and collective violent behaviors because those behaviors proved active in the ancestral environment. Self-recruitment to aggressive groups—whether Seal Team 6, al Qaeda, or the Crips—occurs largely for the same reasons: late adolescents and young adults, men more than women, have brains that arrive at the largely unconscious conclusion that being seen to participate in violence against an out-group makes one worthy of the rewards of reciprocal altruism. Violence and altruism are not polar opposites, but two sides of the same coin of evolved, adaptive human nature.

It is self-evident that innate and acquired biological factors interact with environmental factors to determine who will become a violent criminal, a gangster, or a terrorist. That indisputable observation does not relieve us of the responsibility to determine how this occurs, and what elements of the causal algorithm are susceptible to what cost-effective interventions. Political agendas probably represent a barrier to the mitigation of violence. Yet rigorous empirical research holds the promise of informing better violence prevention policies. This will be the work of generations.



The author wishes to gratefully acknowledge Janice Adelman, Ph.D., for her valuable contribution to the analysis of the data on Palestinian terror attacks and their support.

Copyright 2013 by the National Academy of Sciences. All rights reserved.
Bookshelf ID: NBK207258


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (2.2M)

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...