Rates of killing rivals of the same species vary widely, and are particularly high in humans compared to most animals. The Demonic Males Hypothesis states that predispositions for violence by human males have been favored by natural selection thanks to an evolutionary ecology in which the killing of helpless rivals was advantageous. The hypothesis thus suggests that human males have a ‘demonic’ tendency compared to males of most primate species. According to this idea, men categorize others into in-group and out-group members, and when they encounter out-group members who cannot effectively defend themselves, they tend to conduct violent attacks intended to kill the victim. This behavior is putatively psychologically rewarding even without any resources being acquired as a result, because natural selection has led men to enjoy the prospects and practice of such violence. The reasons parallel those concerned with the desire for sexual interactions. In neither case can the actors predict the outcome in terms of increased biological fitness, but in both cases the behaviors have on average been beneficial in the past.

The Demonic Males Hypothesis emerges from considering why humans share patterns of violence with chimpanzees. Shared patterns include participation predominantly by males, an intense personal and group concern with status, variable subgroup composition, defense of group integrity, inter-group fights that include surprise attacks, a tendency to avoid mass confrontation, and voluntary participation in killing. In both species, supposed benefits to violence include a reduction in the fighting power of a neighboring group, which tends to lead to increased access to resources for the group that kills more. The ultimate causes for selection favoring violence include group territoriality and variable subgroup size. Among humans these elements can be seen in a variety of systems, including social networks among hunter-gatherers, farming tribes, and inner-city gangs. I suggest that a key feature of these systems predisposing towards collective violence of the chimpanzee style is that killers are immune from punishment, because the member groups live in a state of anarchy with respect to each other.

Application of these ideas to religious violence demands recognition of features of modern human society that create substantially greater complications than occur in chimpanzees or nomadic hunter-gatherers. Aside from their emotionally supportive, intellectual and spiritual aspects, I see religions as systems of indirect reciprocity that create cooperation. The cooperative behavior fostered by religion is used both within the group (altruistically) and against rival groups. In that light the following are especially relevant to consideration of religious violence.

First, human societies are characterized by multiple and cross-cutting kinds of group, such as religions, kinship, ethnicity, professional organizations and nations. This means that it can become hard for individuals to recognize moral boundaries between in-group and out-group, and it raises the challenge for scholars of identifying the circumstances that lead individuals to perceive killing as morally justified. Religions tend to harden out-group distinctions, because their social benefits depend partly on member individuals being totally committed to the religious code.

Second, individuals are typically embedded in hierarchies in which leaders dictate the actions of their subordinates. This means that subordinate individuals are routinely forced to participate in violence against their will or better judgment. It also means that patterns of violence can be distorted away from those predicted by adaptationist theory, thanks to the pyschological idiosyncracies of particular leaders who order violence (such as James Jones). The combination of these effects mean that specific cases of killing might not conform to the straightforward predictions of evolutionary theory, and tends to increase the likelihood of violence.

Third, instead of making judgments about the vulnerability of potential victims by assessing the relative size of rival subgroups, individuals in complex societies have to assess their chances (of killing rivals cheaply) using multiple sources of information. These sources include many whose validity cannot be personally verified, allowing leaders to bias perceptions selfishly in favor of violence. Leaders may use statements designed to manipulate their perception of the vulnerability of potential victims, such as the reported exhortations to Hutus in 1994 that if they all work together, they can eliminate the Tutsi. Effective use of language or other symbols can therefore increase the potential for violence, especially by large and hierarchically ordered coalitions such as religious groups. It can also decrease it.

Fourth, the accumulation of power by a leader shifts the calculus of aggression from one concerned with whether it makes sense to kill, to one concerned with whether it makes sense to fight. This means that violence follows different patterns from the simple expectations based on imbalance-of-power theory.

Fifth, human groups create histories that mean that the probability of aggression (or cooperation) is influenced by perceptions or memories of prior interactions. Long histories and biased accounts can lead to unreasonably great desire for revenge. The stories created by religions are particularly long-lasting and emotionally powerful, and thus have the capacity to potentiate greater violence. While these features complicate the analysis of violence, they do not appear to undermine the utility of the Demonic Males hypothesis in providing a framework from which human patterns are elaborated, and in explaining more than alternative hypotheses such as the Blank Slate.