Behavioral and Brain Sciences (under review)



Murder by Design:

The Evolution of Homicide

David M. Buss and Joshua D. Duntley



David M. Buss, Department of Psychology, University of Texas, Austin, Texas 78712. Email: URL:

Joshua D. Duntley, Department of Psychology, University of Texas, Austin, Texas 78712. Email:


Short Abstract

We propose Homicide Design Theory as a new theory to explain why people kill. Multiple homicide mechanisms have evolved as effective context-sensitive solutions to distinct adaptive problems. Killing historically conferred large fitness benefits: preventing premature death, removing rivals, gaining resources, depriving rivals of mates, aborting rival’s prenatal offspring, eliminating stepchildren, and winnowing future competitors of one’s children. Homicidal premeditations are part of evolved killer design, functioning to mobilize attention, rehearse scenarios, calculate consequences, and motivate behavior. Because getting killed inflicts temporally cascading costs on victims, selection has forged anti-homicide defenses, producing coevolutionary arms races between homicidal strategies and anti-homicide defenses.


Long Abstract

Why do humans kill other humans? Extant theories of homicide–those invoking social learning, media influence, cultures of honor, brain pathology, genetic defects, and evolutionary epiphenomena–fail to explain the varying motives for murder and the dramatically different circumstances in which murders occur. We propose a new theory of killing--Homicide Design Theory. Multiple homicide mechanisms have evolved as context-sensitive solutions to an array of diverse adaptive problems. Murder in circumscribed contexts conferred large fitness benefits. These include protecting self and kin from injury or death, depriving rivals of access to valuable mates, gaining access to contested resources, terminating prenatal offspring of rivals, eliminating genetically unrelated stepchildren, removing key antagonists, and winnowing future competitors of one’s children. Homicidal premeditations are hypothesized to be central components of evolved design for murder. Homicidal premeditations function to mobilize attention, construct and rehearse scenarios, calculate consequences, evaluate costs and benefits, and motivate decisions about whether or not to commit homicide. Getting killed, however, inflicts temporally cascading costs on victims. The fitness costs of premature death imposed selection pressure for the evolution of anti-homicide defense mechanisms that function to avoid death and deter killers. This led to a coevolutionary arms race between homicidal strategies and anti-homicide defense mechanisms, analogous to arms races between predators and prey. Homicide Design Theory generates several dozen novel testable predictions about the psychology of murder that are supported by empirical studies. The theory parsimoniously accounts for paleontological, archeological, ethnographic, psychological, and behavioral evidence that remains inexplicable on alternative theories.


Key words: homicide, murder, evolution, killing, victims, defenses, evolutionary psychology


"Now kill every male dependent, and kill every woman who has had intercourse with a man, but spare for yourselves every women among them who has not had intercourse." (Numbers 31)–Instructions by Moses after the conquest of the Midianites.

"To deprive others of their life is one of the most effective means of increasing one’s fitness" — Joseph Lopreato (1984)

1. Introduction

No offense against another human being inflicts greater costs than murder. The dead cease to contribute to their own affairs, and cannot actively influence the affairs of their families, friends, or enemies. Wherever written laws exist, killing is always singled out as a crime. No other infraction comes attached with greater punishment. Where written laws are absent, killing typically constitutes a major cause of death, sometimes accounting for the mortality of a third of all males (Keeley, 1996). Although cultures with written laws, hired police forces, and the prospect of imprisonment appear to have produced substantially lower homicide rates than cultures lacking these modern penalties, even in modern societies the odds of dying by the intentional hand of another run as high as one in twenty-six for certain sub-groups (Ghiglieri, 1999).

Homicide may be defined as the killing of one human being by another (Wilbanks, 1982, p. 153). In 1998, 18,936 murders occurred in the United States alone--a fairly typical year for killings (Federal Bureau of Investigation, 1999). These figures include only "successful" murders, and ignore the large, but currently unquantifiable, numbers of attempted murders that fail for one reason or another. These figures also don’t include homicides committed in self-defense. Nor do they include many of the assaults that do not lead to immediate death, but rather to the victim’s death days or weeks after a police report is filed. The official murder rates also fail to include the potentially large number of infanticides officially classified as Sudden Infant Death Syndrome (SIDS), a portion of which are turning out to be intentional homicides (Becroft, Lockett, Etzel, Montaña, Dearborn, Smith, Infeld, Dahms, & Carroll-Pankhurst, 1998). For all these reasons, current estimates of homicide rates are likely to substantially underestimate the numbers of deaths produced at the hands of other humans. Even given these underestimates, if we assume that the average American lives to be 72, the odds of being killed by the hands of another human being, are more than 1 in 200 (and are considerably higher for males).

Scientists have compiled a great body of statistics dealing with homicide, including the ages of killers and victims, and some of the circumstances in which killing occurs. The most compelling question, however, has thus far defied a cogent scientific answer: Why do humans kill members of their own species? This paper presents a new theory of homicide--a theory that departs from prior explanations of the taking of human life.

1.1. Homicide Prevalence and Importance

Some authors claim that the United States is an unusually violent culture, and compared to many other modern technologically advanced cultures, it is. American men aged 15 to 24, for example, kill at four times the rate of their Scottish counterparts, seven times the rate of their Canadian counterparts, and 40 times the rate of their Japanese counterparts (Fingerhut & Kleinman, 1990). During the 100 year period from 1900 through 1999, an estimated 1,056,296 people have been murdered in the United States alone (White, 1999). But within the larger context of cultures around the world, including traditional and tribal societies, the United States appears to be an unusually peaceable culture. In New Guinea, killings account for the deaths of 19.5% of adult men among the Huli; 25% among the Mae Enga; and 28.5% among the Dugum Dani (Chagnon, 1988). Among the Yanomamo of Venezuela, nearly 30% of all males die a violent death at the hands of another human being (Chagnon, 1988). Even among the so-called "peaceful" !Kung San of Botswana, the murder rate is higher than the most violent cities within the United States (Daly & Wilson, 1988). These so-called "gentle people" recorded 22 murders over a 25 year time span in a population of 1,500--a rate more than quadruple the rate in a typical year in the United States (Lee, 1984).

These reported within-culture rates of homicide typically do not include killings due to warfare and genocide. In 1971, for example, the Pakistani army killed an estimated 3 million people in the Bangladesh conflict (Wrangham & Peterson, 1996). More recently, genocides in Congo, Bosnia, Kosovo, and Indonesia have produced millions of deaths. Keeley (1996) reported estimates of the percentage of deaths due to warfare in more than a dozen prestate societies, ranging from the Tiwi tribe located on Melville and Bathurst islands to the Gebusi in New Guinea. The estimates ranged from a low of 7% for the Skateholm of Sweden in 4300 B.C. to 40% for the Qadan of Nubia in 10,000 B.C. Keeley (1996) calculates that a typical tribal society lost .5 percent of its membership per year to combat in warfare. Deaths due to warfare and genocide must be added to within-culture rates of homicide to gain accurate estimates of the likelihood of dying from homicide.

Killings galvanize human attention and interest like no other phenomenon. Television, movies, novels, plays, and paintings are filled with dramatic episodes of human killing. No culture with written laws fails to prohibit homicide. Some argue that laws are often made to prevent people from doing things that they might otherwise be inclined to do (Daly & Wilson, 1988). Human recorded history bears this out. It is replete with thousands of accounts of infanticide, fratricide, matricide, siblicide, duels, blood revenge, jealousy-induced mate killings, intrasexual rivalry killings, genocides, and organized warfare involving dozens, hundreds, thousands, and occasionally millions of deaths.

Given the prevalence of homicide and the dramatic nature of its consequences, it may seem astonishing to note that "we have only the most rudimentary scientific understanding of who is likely to kill whom and why" (Daly & Wilson, 1988, p. ix). A variety of theories, however, have been proposed to account for this most drastic form of violence, what has been called "the ultimate conflict resolution technique" (Daly & Wilson, 1988). These include social learning theories, theories that invoke "cultures of violence," theories that propose that homicide is a form of pathology, and theories that invoke principles of evolution and selection (Daly & Wilson, 1989).

We will argue that no existing theory provides an adequate explanation for the known empirical facts about homicide. A key part of the problem is that homicide has often been treated as a singular phenomenon, and presumed to be amenable to a single causal explanation. We argue that the umbrella category of "homicide" subsumes an astonishing variety of different types killing such as infanticide, step-child killing, intrasexual rivalry homicides, mate killing, and warfare killing. A comprehensive theory of homicide must account for these different types, including the different causal conditions that give rise to each.

We present a new theory of murder–Homicide Design Theory. We then show that this new theory can more parsimoniously account for the known facts about homicide and can generate a host of new, testable, empirical predictions not generated by prior theories. Finally, we briefly present results from several new empirical studies that support predictions from the theory--findings inexplicable on previous theories of homicide.

2. Prior Theories of Criminality Co-Opted to Explain Homicide

This section briefly describes prior attempts to explain homicide. Each theory is followed by a short critique that highlights inadequacies in accounting for known homicide data. Several introductory remarks must be made about these theories. First, many theories offered to explain homicide originally were not designed specifically for this task. Social learning theory, for example, was first proposed as a general theory of human behavior (Bandura, 1977), and only subsequently co-opted to explain particular patterns of homicide (Berkowitz, 1993). Second, most explanations of homicide deal with extremely delimited aspects of homicide, such as why men kill more than women or why some cultures have higher homicide rates than others. These theories therefore cannot be used to provide specific and differentiated explanations for qualitatively different types of homicide, such as infanticide, uxoricide (i.e., killing one’s wife), intrasexual rivalry homicides, and warfare homicides. Third, with a few notable exceptions, most extant theories have ignored evolution, natural selection, sexual selection, and adaptation entirely, focusing instead on local, parochial, and recent patterns. Finally, all fail to account for documented patterns of homicide.

2.1. Social Learning, Social Role, and Socialization Theories

Social learning theory (Bandura, 1977), social role theory (Eagly, 1995), and socialization theories (Berkowitz, 1993) share many assumptions, so can be treated together for the purpose of this brief review. The core premise of these theories is that people acquire social behavior by observing and imitating others–behavior for which they are rewarded or punished, shaping their subsequent behavioral repertoire. These social theories have been used to explain two aspects of homicide–(1) the fact that men kill more than women and (2) a more controversial and less established finding that seems to point to "contagion" or "imitation" or "modeling effects" of violence (Daly & Wilson, 1989).

Daly and Wilson (1989) provide a cogent critique of explanations of both findings on the grounds that the debate has occurred in a theoretical "vacuum." No one " . . . has proposed a specific psychological theory of the alleged ‘imitation’ that can explain what sort of publicized event will produce what sort of imitative violence, by whom, and after what delay" (Daly & Wilson, 1989, p. 103). Daly and Wilson conclude their critique of social learning theories by suggesting that "the evidence offered in its support has been shaky, consisting mainly of plausibility arguments, anecdotes (some of . . . dubious origin and likely to be apocryphal ‘urban myths’), and laboratory studies of questionable ecological validity" (Daly & Wilson, 1989, p. 103).

Regarding sex differences in homicide rates, most studies find that roughly 86% of all homicides are perpetrated by men, with more than 70% of the victims also being men (Berkowitz, 1993, p. 274). Social role theory explains this by arguing that social roles have been "traditionally assigned" to men and women differently: "Think of all the ways in which modern western society . . . teaches children that fighting is far more appropriate for men than women. Popular literature and the mass media consistently show men but not women fighting. Parents buy toy weapons for their sons and dolls for their daughters. Parents are more likely to approve of and reward aggressive behavior in boys than in girls. Again and again, directly and indirectly, youngsters learn that males are aggressive and females are not . . ." (Berkowitz, 1993, p. 395).

Although many of these observations are surely true--even a cursory look at modern movies verifies that more men than women are portrayed as aggressive killers--the explanation fails on two grounds. First, it assumes that the causal arrow runs from social environment to sex-linked aggressive propensities. It ignores a large body of evidence that suggests that boys preferentially seek out and respond to weapon-like toys and violent media images, that parents may be responding to already existing desires in their children, and that popular media merely exploit children’s existing preferences (Hoyenga & Hoyenga, 1993).

Second, these explanations fail when confronted with the cross-cultural evidence. Cultures that lack literature, mass media, television, and store-bought toys show precisely the same sex difference in rates of homicide: ". . . why invoke [features of] North American culture to explain a sex difference that is universal? Ironically, the sex difference in homicidal violence is smaller in modern North America than in any other human society yet described . . . [these theorists] have wasted their efforts positing culturally and historically particular explanations for a general phenomenon" (Daly & Wilson, 1989, p. 101-102).

In addition to these problems, social learning and social role theories run into a variety of fundamental difficulties. They offer no specific predictions about homicide in particular that could be tested and potentially falsified. Boys, girls, men, and women throughout their lives are exposed to an astonishing variety of "models" and "social roles," from serial killers to transvestites. From this collage of potential models and social exposures, what determines which models are imitated? Nothing in the theories of social learning, social roles, or socialization provides answers to this key question.

These social theories also ignore the fact that people kill despite being bombarded with powerful social sanctions designed to deter killing, religious exhortations and commandments not to kill, cultural messages indicating that "crime does not pay," and punishments in the form of life imprisonment and the death penalty. Finally, these theories fail to explain the origins of homicide to begin with. How did the modeling of homicidal behavior get started? Why would parents be motivated to incline their sons rather than their daughters toward killing? Why would popular media conspire to portray men more than women as killers, and why would subsequent media moguls be motivated to follow suit? Why does killing exist in the human behavioral repertoire at all?

These social theories, in short, are too general to be useful in explaining known facts about homicide (e.g., the cross-cultural evidence, the age distribution, the specific patterns of victims and perpetrators) and fail to generate new predictions about aspects of homicide yet undiscovered.

2.2. Cultural Theories of Homicide

Several cultural theories have been proposed to explain one dimension of homicide--why the rates of homicide vary across societies and across sub-groups within society. Most prominent are the "subcultures of violence" theory (Wolfgang & Ferracuti, 1967) and the "cultures of honor" theory (Nisbett, 1993).

The "subcultures of violence" theory proposes that within some cultures or sub-cultures, such as inner-city street gangs, individuals valorize violence or make it an obligatory means of defending and protecting one’s reputation. In other cultures or sub-cultures, such as middle-class academicians, individuals condemn violence and those who use it suffer damage to their reputation. According to the "subcultures of violence" theory, these differences explain why homicide rates vary from one subculture to another.

Daly and Wilson (1989) argue that the "subcultures of violence" theory is a pseudo-explanation, because it simply describes the behavioral differences it is invoked to explain. No attempt has been made to measure attitudes toward violence as independent predictors of actual homicide rates. And even if such tests were conducted, the direction of causality would be unclear. Discovering a link between attitudes toward violence and homicide rates would still fail to answer the question of why the subcultures differ to begin with. Furthermore, the "subcultures" theory fails to explain a host of known facts about homicide, such as (1) why men kill more than women in every culture for which we have solid data; (2) why killers are heavily concentrated in the age brackets of 15 — 24 in all cultures for which data are available, rather than more evenly distributed across the lifespan (Wilson & Daly, 1985); (3) why most individuals residing in subcultures of violence refrain from killing; (4) why some individuals not living in such subcultures nonetheless choose to kill; and (5) why infanticides and stepchild killings occur in predictable circumstances in all known cultures, and appear to be unlinked to valorization of violence.

More recently, Nisbett (1993) proposed a more penetrating theory to explain variation across cultures in rates of homicide. He proposed that the economic means of subsistence of a culture affect the degree to which a group originally develops a "culture of honor." In cultures of honor, insults are viewed as public challenges that must be met with direct confrontation. The theory proposes that differences in the degree to which honor becomes a central part of cultures ultimately rests with economics, and specifically with the manner of subsistence. In herding economies, one’s entire stock could be lost suddenly at the hands of thieves. Cultivating a reputation as willing to respond with violent force--for example, by actually responding with violence to a public insult--would presumably deter thieves who might otherwise encroach on one’s resources with impunity. In more settled, agricultural communities, in contrast, cultivating such a reputation is less important since one’s economic means of subsistence cannot be rapidly undermined.

Nisbett (1993) tested his theory using homicide statistics from different regions within the Unites States and experiments that measured differences in violence proneness in individuals from the North and South. Interestingly, he found evidence for domain-specificity. Southerners (historically more characterized by herding) do not endorse more positive attitudes toward the use of violence in general compared with Northerners (historically more characterized by sedentary agriculture). Southerners were more inclined, however, to endorse violence specifically for the purposes of protection and in response to insults.

The "cultures of honor" theory differs from the "subcultures of violence" theory in proposing a specific causal mechanism for the origins of differences between cultures and subcultures--economic means of subsistence. It has the further virtue of generating testable predictions about cultures that have not yet been examined. Nonetheless, the theory does not provide an explanation for the many patterns of homicide that do not directly involve reputational defense. Nisbett’s theory does not attempt to explain the existing patterns of infanticide, spousal homicide, or the killing of stepchildren, so these patterns of homicide are beyond its purview.

The key point, however, is that even if Nisbett’s theory is entirely correct in explaining rate differences in homicide across cultures, we still need a theory to explain the tremendous variety of specific and different forms of homicide, many of which appear unconnected with notions of honor.

2.3. Pathology Theories of Criminality

Pathology theories suggest that some individuals kill as a result of a brain defect, caused by factors such as head injury or alcohol damage. One version hypothesizes seizuring, which invokes damage to the amygdala, an area of the brain that is implicated in controlling social emotions such as jealousy and rage, and also has been shown to be the area of the brain most vulnerable to seizures (Ellis, in press). According to the seizuring idea, convulsive brain seizures may induce an individual to commit a suddenly violent and impulsive act, sometimes resulting in the death of another person. The evidence, however suggests that violence resulting from seizures is typically random rather than purposeful or patterned (Ellis, in press; Gazzaniga, Ivry, & Mangun, 1998).

A related theory proposes that criminality results from damage to the frontal lobe, and specifically the prefrontal cortex (Ellis, in press). Damage to the prefrontal region of the frontal lobes has been shown to be linked with flatness of emotion and affect, as well as indifference to the suffering of others (Asher, 1982; Gazzaniga, Ivry, & Mangun, 1998). According to the hypothesis implicating frontal lobe damage, criminal behavior, including homicide, may be caused by frontal lobe damage.

Each of these notions of pathology--seizure theory and frontal lobe damage theory--fail as general theories to explain known patterns of homicide. First, these forms of brain pathology are typically inferred from the very behavior that they are invoked to explain, although recent work using CAT scans and MRI technology may eventually circumvent this problem (Ellis, in press). Second, theories of pathology suggest that violent aggression, and hence homicide, should be directed randomly toward others, since the hypotheses contain no specifications about the nature of the victim or the contexts in which homicide will occur. For these reasons, brain pathology theories of criminality fail as penetrating explanations of the diverse and specific non-random patterns of homicide documented around the world.

Another theory of pathology invokes genetic abnormalities. Roughly one male out of every 700 to 1,000 is born with an extra Y-chromosome (XYY rather than the normal XY), and one male out of every 500 is born with an extra X-chromosome (XXY)(Hoffman, 1977, p. 447). Both genetic abnormalities result in males who are above average in height, develop more acne problems during adolescence, and score lower on standard tests of intelligence (Ellis, in press). Persons with these forms of genetic abnormality show an increased likelihood of criminal behavior, with prison populations showing five-times the proportion of these individuals compared with the general male population (Ellis, in press). Nonetheless, these genetic abnormalities are likely to explain only a tiny fraction of the homicides committed, since males with an extra chromosome only constitute 1 - 2% of the prison population (Ellis, in press).

2.4. Evolutionary Theories of General Criminality

Several evolutionary hypotheses have been proposed to explain criminality in general, although not homicide specifically. An application of r/K Theory suggests that criminals are r-strategists compared with non-criminals who are K-strategists; that is, criminals pursue short-term strategies involving high reproductive rates that presumably entail more risk-taking than the general population (Ellis, 1989; Rushton, 1995). The Conditional Adaptation Hypothesis suggests that unstable and resource-scarce environments during development shunt individuals into an opportunistic exploitative strategy involving criminal behavior (Harpending & Sobus, 1987; Harpending & Draper, 1988). Alternative Adaptation Theory suggests that genetic differences that influence the devotion of effort to mating instead of parenting leads to deception and criminality in the pursuit of opportunistic copulations (Rowe, 1995). And Expropriative Theory proposes that crime is an evolved tactic for obtaining resources (Cohen & Machalek, 1988; Vila & Cohen, 1993).

All share common features. All propose that genetic factors contribute to criminality in general. All propose that these genetic factors exist due to a history of natural selection. And all suggest that historically there have been adaptive benefits to pursuing strategies linked with criminal behavior that were sufficiently large to outweigh the costs, at least in certain contexts for certain individuals.

These hypotheses may eventually be able to predict some variance in who becomes a criminal (e.g., through molecular genetic techniques) and a few of the contexts that are likely to elicit criminal behavior (e.g., resource-scarce environments). Nonetheless, all share similar weaknesses as cogent and comprehensive explanations for the diversity of homicidal phenomena. First, they have not been developed to account for homicide specifically, and so make no specific predictions about when homicide, as opposed to some other form of criminal behavior, will occur. Second, these theories fail to predict most of the specific contexts in which homicide will occur. Third, they do not specify whether homicide per se (as opposed to criminality in general) has ever been adaptive, nor what the specific functional benefits of killing might have been. Fourth, they fail to explain why homicide is often perpetrated by individuals who do not seem to be pursuing a general strategy of criminality. There is no evidence, for example, that a normally law-abiding man who killed his wife in a jealous rage when confronted with her infidelity is pursuing the short-term opportunistic sexual strategy proposed by theories such as r/K theory and Alternative Adaptation Theory. Similarly, women who are otherwise "normal" and not pursuing a criminal strategy sometimes kill their infants when such infants are deformed, diseased, or born prematurely (Daly & Wilson, 1988). Tribal groups regularly engage in warfare, and this appears to be a "normal" activity in which most young males engage, rather than a specific subset of men pursuing a particular strategy (Keeley, 1996; Chagnon, 1988). For all these reasons, the existing evolutionary theories of general criminality are inadequate both for explaining known facts about homicide and for generating novel predictions about the contexts in which homicide will occur.

2.5. Killings as Evolutionary Epiphenomena and Maladaptation

Over the past few decades, several evolutionary scientists have offered explanations for homicide. Almost without exception, however, the evolutionary hypotheses offered posit that killing is unnatural and not part of human evolved psychology. The ethologist Eibl-Eibesfeldt (1989), for example, proposed that killing (e.g., in war) is a culturally imposed behavior that overrides an innate human inhibition to kill: "Our innate inhibitions against killing are also dampened in war. A cultural norm making killing a duty is superimposed upon the biological one forbidding murder, and this results in killing actually becoming a virtue. However, a massive indoctrination is required to make this superimposition . . ." (Eibl-Eibesfeldt, 1989, p. 711). Thus, according to this prominent ethologist, if natural selection has forged any psychology surrounding homicide, it is a natural proclivity against it. Killing, according to this view, is a cultural aberration and in no way part of human evolutionary psychology. This view leads to no detailed predictions about the various forms of homicide nor about the contexts in which they will occur.

By far the most comprehensive evolutionary explanation specifically advanced to account for patterns of homicide is that proposed in a series of publications by Daly, Wilson, and their colleagues (1988, 1995, 1998, 1999). Daly and Wilson’s contributions to the understanding of homicide and the contexts in which it occurs have been pioneering and seminal. But their explanations, we suggest, are inadequate. Daly and Wilson advance two distinct arguments about homicide--one may be called the "agnostic" argument and the other the "slip" or "epiphenomenon" theory. According to the agnostic argument, " . . . our evolutionary psychological approach in no way depends upon homicide per se being ‘an adaptation.’ It may or may not be the case that actual killing was a regular component of the selective events that shaped human passions" (Daly & Wilson, 1988, p. 12, emphasis added). They have maintained this position in more recent publications: "Using homicides as a sort of ‘assay’ of the evolved psychology of interpersonal conflict does not presuppose that killing per se is or ever was adaptive" (Wilson, Daly, & Daniele, 1995, emphasis added).

In the same publications, however, they advance the view that humans do not have an evolved psychology of homicide. Rather, they propose that homicides can be viewed as "dysfunctionally extreme byproducts," "epiphenomena", or "overreactive mistakes" of evolved mechanisms designed for other purposes--a kind of "spandrels" argument (see Gould, 1991; Buss, Haselton, Shackelford, Bleske, & Wakefield, 1998). According to this argument, "It is quite possible that actually killing one’s antagonist is more often than not an overreactive mistake--an act with negative consequences both for the killer’s net hedonic utility and for the actual expected fitness of which that utility is an evolved token" (Daly & Wilson, 1998, p. 438, italics original).

Although Daly and Wilson offer the "mistakes" explanation as a general explanation for homicide, they most frequently invoke it in the context of spousal homicides: "We propose that such homicides [killing estranged or unfaithful wives] are the dysfunctionally extreme byproducts of those same violent inclinations whose much more typical effects are successful deterrence and coercion" (Daly & Wilson, 1998, p. 449, italics added). It is clear from these quotes that Daly and Wilson are not proposing that that humans have evolved a distinct psychology of homicide, although they do not rule out such a possibility completely.

Subsequent evolutionary psychologists have adopted Daly and Wilson’s position (e.g., Crabb, 2000; Kenrick & Sheets, 1993). Crabb, for example, suggests that "psychological machinery for aggressive impulses would have served inclusive fitness well, with the caveat that extreme aggression leading to homicide may be disastrous for inclusive fitness because it may result in fatal retribution against the perpetrator" (Crabb, 2000, p. 226).

Although Daly, Wilson, and colleagues do not propose that humans have evolved psychological mechanisms specialized for killing, they do emphasize the importance and priority of an evolutionary psychological explanatory account: " . . . what is needed is a Darwinian psychology that uses evolutionary ideas as a metatheory for the postulation of cognitive/emotional/motivational mechanisms and strategies" (Daly & Wilson, 1989, p. 108-109). The goal of this article is to propose precisely such a theory. We briefly outline arguments for the a priori plausibility of evolved mechanisms specifically designed for conspecific killing, which simultaneously call into question the Daly and Wilson explanations for homicide.

3. The A Priori Plausibility of Evolved Mechanisms for Conspecific Killing

Killing obviously has many causes, and some theories, as noted above, seek only to provide partial explanations. Undoubtedly some homicides may be caused by "slips" due to violent brinkmanship or accidents in a dangerous game of coercion, as Daly and Wilson propose. A variety of sources of existing evidence, however, call into question the validity of the Daly-Wilson epiphenomenon theory of homicide as a general explanation for the majority of killings and simultaneously suggest that it is not unreasonable to propose that mechanisms dedicated to conspecific killing might have evolved.

The first source of evidence is comparative. A large number of species regularly commit conspecific killings in such predictable contexts that it is reasonable to advance the hypothesis that such species have adaptations designed to kill. Among insects, a wide variety of species engage in mate killings. Where mate killing and cannibalism is known to increase the number or viability of offspring (including mantids, black widow spiders, jumping spiders, and scorpions) males approach females cautiously and retreat quickly after copulation. Breene and Sweet (1985) showed that in the sexually cannibalistic black widow spider Latrodectus mactans, when males survive copulation, they often fertilize multiple females. Males of sexually cannibalistic species use diverse strategies to decrease their chances of being cannibalized: male scorpions sometimes sting the female after deposition of the spermatophore (Polis & Farley, 1979); male black widows (Gould 1984) and crab spiders (Bristowe, 1958) often restrain females in silk prior to copulation. Conspecific killing, as well as mechanisms to prevent getting killed, appear to be common among insects.

Among the 4,000 species of mammals, many also have well-documented patterns of conspecific killing. Male tigers, lions, wolves, hyenas, cougars, and cheetahs have been observed to kill the infants of rival males (Ghiglieri, 1999). These killings often have the effect of hastening the estrus of the mothers of those infants, at which point they often mate with the killers. Among primate species, of which humans are one, conspecific killings have now been well documented among langur monkeys (Hrdy, 1977), chacma baboons (Busse & Hamilton, 1981), red howler monkeys (Crockett & Sekulic, 1984), savanna baboons (Collins, Busse, & Goodall, 1984), mountain gorillas (Fosse, 1984), chimpanzees (Zuzuki, 1971; Bygott, 1972), blue monkeys (Butynski, 1982), and others (Hausfater & Hrdy, 1984). The killing of conspecific rival males has also been well-documented among the chimpanzees of Gombe (Wrangham & Peterson, 1996), perhaps the species genetically closest to human beings (98.4% of DNA in common), as well as in mountain gorillas, who share 97.5% of their DNA with humans (Fossey, 1984).

Thus, contrary to claims by ethologists such as Konrad Lorenz (1966), conspecific killing is widespread in the animal world, not limited to humans. The contexts in which many of these conspecific killings occur, such as males killing intrasexual rivals or the offspring of intrasexual rivals, provides evidence that some, perhaps many, primate species have evolved specific adaptations for killing. The comparative evidence, in short, supports the idea that mechanisms for conspecific killing may have evolved in a variety of mammalian and primate species. Based on this evidence, there is no reason to be skeptical about the possibility that mechanisms specifically designed for conspecific killing may have evolved in Homo sapiens, or may have become evolutionarily elaborated from earlier designs of ancestral species through natural or sexual selection.

A second source of evidence is the paleontological record, which reveals a highly patterned distribution of fractures and cranial traumas, and arrow tips and stone projectiles lodged in the rib cages of ancient humans (Grauer, 1998; Keeley, 1996; Walker, 1995). Fifty-nine human skeletons were found in a cemetery at Gebel Sahaba in Egyptian Nubia, for example, dating from 12,000 to 14,000 years ago (Late Paleolithic); more than 40% of the skeletons contained stone projectile points close to, or embedded in, them (Keeley, 1996). The excavator estimated that more than half of the 59 had died violently, many with multiple wounds (the highest number of wounds was 20). Another study of California prehistory documented that about 5% of the skeletons had arrowheads embedded in them, obviously only the most transparent evidence of violent intentional death (Keeley, 1996). The majority of such injuries appeared on male skeletons, and most appeared on the left sides of the crania and rib cages, suggesting a right-handed attacker (Grauer, 1998). Other evidence includes the scalping of skulls, which leaves characteristic cut-marks on the cranial bones founds in many regions in North America, as well as trophy skulls and other body parts such as teeth, hands, arms, or legs (Keeley, 1996).

There is also evidence of cannibalism, most recently coming from what is now southwest Colorado (Holden, 2000). Bone fragments show characteristic cuts. Analyses of ancient human feces (coprolite) revealed the presence of human myoglobin, which could only get there through the consumption of human muscle or heart tissue. More anecdotal evidence comes from reports from early missionaries, as illustrated by the following report. A Maori warrior was observed taunting the preserved head of an enemy chief: "You wanted to run away, did you? But my meri [war club] overtook you; and after you were cooked, you made food for my mouth. And where is your father? He is cooked:--and where is your brother? He is eaten:--and where is your wife? There she sits, a wife for me:--and where are your children? There they are, with loads on their backs, carrying food, as my slaves" (Vayda, 1960, p. 95). In Oceania as well, enemy captives were eaten. On Fiji, one chief kept a record of how many enemy bodies he had consumed by placing stones in a line behind his house, one stone for each victim. At the time of observation, the line contained 872 stones (Keeley, 1996).

Obviously, these data are fragmentary and not conclusive. But they add one important piece of circumstantial evidence suggesting that homicide has been a recurrent feature of human ancestral environments–a critical feature that is the sine qua non of our proposed theory of evolved homicide mechanisms.

A third source of evidence is archeological, stemming from the existence of weapons. While some tools have several functions (e.g., the bow and arrow can be used for hunting animals or killing humans), others seem specialized for homicide. Shock weapons such as maces and clubs serve no functions except to inflict injury or death on a human. Keeley (1996) argues that maces, lances, tomahawks, daggers, and swords are excellent for conspecific killing, but have no other apparent purposes. The graveyard of mass killings discovered at Talheim in Germany from around 5000 B.C. showed victims with holes in their skulls that correspond exactly to the shape of axes from the Early Neolithic.

A fourth source of evidence comes from fortifications. Fortifications are designed as defenses against attackers and invaders. They are seen in concentrations in northwestern Europe from the Early Neolithic, built as long as 7000 years ago. Palisades, moats, ditches, baffle gates, towers, drawbridge gates, ditches with sharpened sticks, stone walls, adobe walls, traps at gates, and pitfalls have been discovered throughout the world (Keeley, 1996). Fortifications and weapons both suggest a long human history of homicide.

A fifth source of evidence is pictorial, appearing on rocks and in cave paintings from Europe, the middle east, and Australia. Cave paintings discovered in Spain and France show men shooting other men with arrows (Dennen, 1995). In northern Australia, rock paintings dating back as many as 10,000 years depict males in combat, some hurling spears and boomerangs, some dodging these weapons, and some chasing others with raised weapons (Dennen, 1995). Some show figures pierced with spears, and some show men bending down to help those felled by spears.

A sixth source of evidence involves the cross-cultural and ethnographic record, which provides extensive documentation of tribal warfare (Keeley, 1996), genocide (Lee, 1984), blood revenge in foraging societies (Daly & Wilson, 1988), intrasexual homicide within groups (Chagnon, 1988; Daly & Wilson, 1988; Hart & Pilling, 1960; Keeley, 1996), infanticide (Daly & Wilson, 1988), and spousal killings (Daly & Wilson, 1988). Even in cultures held depicted as peaceful, homicide rates are high. Among the Netsilik Eskimo, for example, who appeared very peaceful to the anthropologists who studied them, there was a well-documented murder rate four times higher than in the United States (Keeley, 1996). Although there are obvious difficulties in inferring the prevalence of homicide in ancestral contexts from the current ethnographic and cross-cultural evidence, it is nonetheless clear that homicide is neither rare nor trivial from the standpoint of natural selection.

Finally, we note that homicides appear across cultures in quite predictable contexts and often involve premeditation, both of which call into question the epiphenomenon explanation of homicide. Spousal homicide occurs primarily in the context of a female sexual infidelity or defection from the relationship, male-male homicides occur predictably over issues of status, face-saving, and reputation, and infanticides occur when babies are deformed or diseased or when the mother lacks an investing father. All these predicable contexts for homicide, well-documented by Daly and Wilson (1988), call into question the "accidental" or "slip" or "epiphenomenon" explanation as a general theory of homicide.

Taken together, the comparative, paleontological, archeological, ethnographic, cross-cultural, and contemporary evidence all suggest that it is not implausible a priori to hypothesize that humans have evolved psychological mechanisms that are designed specifically for killing other humans in certain contexts. The remainder of this paper is devoted to presenting the basic outlines of a new evolutionary theory of homicide.

4. Précis of Homicide Design Theory

In sharp contrast to the theories reviewed above, we propose that there have been some highly specific and recurrent contexts in which the fitness benefits of killing outweighed the fitness costs. We propose that humans have evolved a number of distinct, context-sensitive psychological mechanisms devoted to homicide. We propose that these mechanisms are triggered by a delimited set of circumstances, and that they are designed to produce the death of conspecifics. A host of beneficial consequences, in the currency of fitness, historically flowed to killers as a result of murdering in some contexts (Buss & Duntley, 1998, 1999; Duntley & Buss, 1998, 1999). Most murders, according to Homicide Design Theory, are not slips, accidents, incidental byproducts, or epiphenomena. They are kills by design. Below we outline the major premises of Homicide Design Theory, enumerate specific propositions about the nature of the proposed cognitive/psychological mechanisms, and then present preliminary empirical evidence for the theory.

4.1. Premise 1: Over human evolutionary history, there has been recurrent selection in which the benefits of conspecific killing, on average, outweighed the costs in delimited contexts (i.e., rb > c, in the currency of fitness). [1] 

There are many potential benefits to killing another human that could have recurred over human evolutionary history.[2] The particular benefits reaped as a result of killing undoubtedly varied with the nature of the person killed, their relationship to the killer, the social context, and personal circumstances of the killer and victim. According to our theory, there are both lethal and non-lethal evolved strategies for solving many adaptive problems, which would suggest the possibility of evolved decision rules that guide which strategies to deploy in which circumstances. The purpose of this section is merely to present the argument that, over human evolutionary history, there would have been many potential benefits to killing another human being. We illustrate the 10 clearest and most likely fitness benefits that would have accrued to killers in ancestral conditions.

4.1.1. Benefit 1: Eliminating an intrasexual competitor. Because selection operates on genes for existing designs relative to genes for other designs around at a given time, eliminating an alternative design by killing a rival can provide direct reproductive advantages, even when the killer accrues no additional benefits from murder. A rival’s death is the killer’s gain, just by virtue of the fact that the rival is dead. A conspecific killer terminates the rival’s future reproduction, and begins a cascading set of fitness-reducing consequences for the rival’s kin, leading to a decrease in the frequency of the rival’s genes in subsequent generations. The killer simultaneously eliminates the direct costs that the rival otherwise would have inflicted if he or she were not killed, such as the costs of a the rival monopolizing reproductively relevant resources, potentially contributing to an increase in the frequency of the killer’s genes in subsequent generations.

4.1.2. Benefit 2: Gaining access to a rival’s material resources. By killing a rival, the killer can gain direct access to the victim’s material resources. These include the victim’s food, territory, tools, weapons, cookware, and shelter--resources that can aid in the killer’s survival and future reproduction.

4.1.3. Benefit 3: Gaining access to a rival’s fertile mates. Men who kill potentially can gain access to a rival’s reproductively valuable women, either for short-term sexual access or long-term mating. Increased sexual access to fertile women in ancestral environments would have constituted a large reproductive benefit that alone could have provided the selective impetus for the evolution of psychological mechanisms for homicide.

4.1.4. Benefit 4: Cultivating a reputation that deters exploitation and reduces the likelihood of costs inflicted by other potential rivals. Humans live in groups where status and reputation can boost access to survival and reproductively relevant resources, or consign a person to reproductive oblivion. By killing, and showing a willingness to kill, men can successfully deter other competitors who might encroach on one’s resources if such deterrence did not exist. Men unwilling to kill may cede their resources more willingly to men with credible reputations as killers.

4.1.5. Benefit 5: Protecting oneself from exploitation, injury, rape, or death at the hands of others. In some social contexts, humans face a choice of killing or being killed. Over human evolutionary history, those who failed to kill when their lives were threatened, or when some massive cost such as severe injury or rape was about to be inflicted, were likely to have been out-reproduced by those willing to kill to defend themselves. Indeed, if there were one context above all others that might prompt even the most pacific individual to commit homicide, it would be when there is a life-or-death zero-sum outcome between a would-be victim and their would-be killer.

4.1.6. Benefit 6: Protecting children and kin from exploitation, injury, rape, or death at the hands of others. Children and other kin are the vital "vehicles" necessary for transporting genes into the next generation. Protecting the survival of such vehicles would have been essential for successful reproduction. Fitness benefits would accrue to those who killed when their key vehicles were threatened with exploitation, debilitating injury, or death. Based on the logic of inclusive fitness theory, we predict a gradient of a willingness to kill to defend one’s vehicles, depending on genetic relatedness.

4.1.7. Benefit 7: Protecting one’s mate from exploitation, injury, rape, or death at hands of others. Valuable mates are always in short supply compared to the large numbers who desire access to them (Buss, 1994). A man who possesses a reproductively valuable mate incurs the envy of others and draws efforts at mate poaching (Schmitt & Buss, in press). Men in some contexts may have had to kill in order to protect their mates (e.g., Valero, 1970). The fitness benefits of killing in this context include retaining access to a valuable mate and deterring the mate poaching efforts of others in the future.

4.1.8. Benefit 8: Protecting existing resources such as territory, shelter, and food sources. The protection of valuable resources by killing extends to the resources needed for survival and reproduction, including shelter, food, tools, weapons, and choice territories. The recurrent fitness benefits of killing in some ancestral contexts in order to retain access to these reproductively valuable resources could have selected for a killer psychology.

4.1.9. Benefit 9: Eliminating resource-absorbing or costly conspecifics that do not contain copies of one’s genes. Step-children can absorb time, energy, attention, food, resources, and investments from a step-parent, but the survival and proliferation of step-children would not generally have increased the fitness of the stepparent. The diversion of a man’s or a woman’s resources to stepchildren--resources that alternatively could have been invested in one’s own genetic children--would have been costly in the currency of fitness. Killing stepchildren delimited contexts could have freed up these resources for more profitable avenues of investment. A male lion that kills a rival’s cubs prior to inseminating his new mate frees up her reproductive and parental resources for his offspring. Analogously, humans who killed stepchildren prevented the diversion of their new mate’s somatic and parental resources to a rival’s genetic vehicles instead of their own.

4.1.10. Benefit 10: Eliminating costly children who would otherwise interfere with investments of resources in vehicles better able to translate the investment into fitness. Evolution by selection creates decision rules that favor investing in vehicles that yield high fitness returns relative to other forms of investment. Children are sometimes born with defects, deformities, or diseases that make them poor vehicles for parental investment because of the poor return on investment (Daly & Wilson, 1988). Parents who killed such infants in certain circumstances would have avoided wasting parental effort on poor fitness vehicles, freeing up effort for children offering a better return on the limited resource of parental investment.

These are a sample of the most obvious and direct fitness benefits that could have flowed to ancestral human killers. Clearly they do not exhaust the list of specific potential fitness benefits of murder, which is longer and more varied. The key point is not that these benefits would always or necessarily have flowed to killers as a consequence of killing. The key point is that the potential benefits of conspecific killing are so numerous and substantial in magnitude that there is no a priori reason to be skeptical about the possibility that homicidal mechanisms could have evolved. Whether or not such benefits recurrently flowed to killers would have depended heavily on a variety of key contextual conditions, including the costs of pursuing a killer strategy--a topic to which we now turn.

4.2. Premise 2: Over human evolutionary history, selection has fashioned specific psychological mechanisms for killing in contexts in which the costs were low.

The costs of killing can be large and even catastrophic in the currency of fitness. By attempting to kill someone, the would-be killer risks getting injured or killed himself. Indeed, we propose that because killing has been a recurrent selective force, humans have evolved a number of anti-homicide mechanisms designed to prevent getting killed (see Premise 8 below). Successful killings may provoke costly retaliation by the victim’s friends, mates, or kin, inflicting damage or death on the would-be killer. People who perpetrate certain types of kills in some contexts may suffer ostracism, reputational damage, status loss, or decrements in mate value that interfere with the successful accomplishment of reproductively relevant tasks such as surviving, mating, and investing in children. These substantial costs historically would have deterred, and perhaps continue to deter, many homicides that otherwise might have occurred.

Premise 2 hypothesizes that selection has fashioned assessment mechanisms that gauge the likelihood of incurring such costs, as well as their magnitude, and subsequently compare these costs with representations of the cumulative benefits of killing, in comparison with alternative behavioral strategies that might be available. For example, the physical formidability of one’s rival compared with one’s own physical formidability undoubtedly affects risk assessment (Gilbert, 1989), and there is evidence that men frequently perform these comparative assessments (Fox, 1997). These assessments should affect decisions about whether to attempt a kill or instead pursue an alternative strategy. Evolved decision rules would favor attempts to kill rivals only when there is a reasonable probability of success, or when the risks were worth taking because of a bounty of benefits that the killer could accrue. Similarly, decision rules may favor killing in circumstances where the costs of failing to kill loom so large that killing may be attempted even under risky conditions where success is uncertain.

Decision rules about going to war, to take another example, will undoubtedly incorporate information about the relative size of the potentially warring tribes, with positive decisions more likely when one’s own group outnumbers the rival group (Tooby & Cosmides, 1993; Wrangham, 1999). Killing in contexts of anonymity (e.g., at night, or alone in the woods) might offer lower risk of retaliation than killing in front of a tribe of witnesses. The frequent use of "surprise raids" in tribal warfare involves selecting contexts in which the costs of killing are reduced (Wrangham, 1999).

We propose that uncertainty assessment is an important design feature of the cost-benefit analyses conducted by our killer psychology. Uncertainty resides in at least two sources of potential error. A first source of uncertainty lies with lack of complete information (e.g. lack of knowledge about the numbers, experience, and formidability of a rival group). A second source of uncertainty lies with the fact that some relevant variables may be difficult to predict with confidence, such as the psychological states of the intended victims (e.g., the degree to which the attempted kill is anticipated by the victim).

We hypothesize that evolved decision rules evaluate all of these assessments as input to decision rules that determine whether or not a homicidal strategy is implemented. These are not "blind instincts" or urges to kill regardless of context. They are sensitive psychological procedures that weigh the costs and benefits that historically have been linked with the strategic use of homicide as one solution among several potential solutions to particular classes of adaptive problems.

4.3. Premise 3: The recurrent contexts in which the fitness benefits of killing outweighed the fitness costs, on average, were many in number and distinct in nature.

This premise is essential for Homicide Design Theory. Without it, a number of distinct homicide mechanisms could not have evolved. Furthermore, the contexts in which the benefits of killing outweighed the costs had to be sufficiently distinctive that they could not be handled by a single, more general homicide mechanism.

Consider the following contexts: Giving birth to an infant who is highly deformed, experiencing an impending attack from a neighboring tribe intent on killing, and walking in on a spouse who is having sex in flagrante delicto with someone else. The benefits and costs of killing in each of these contexts are highly distinctive. In the case of the deformed infant, the parent who commits infanticide can staunch the reproductive losses incurred by continuing to invest in a poorly equipped vehicle, while freeing up valuable parental resources for existing or future children who afford a better return on investment. Infanticide may also come with costs--the infant killer may suffer reputational damage, the loss of their mate, or punishments from their group. Killing in defense against a hostile tribe provides benefits in survival of self and family and preservation of land, food resources, and mates against rival encroachment. It also puts the would-be killer’s life in jeopardy and could produce a deluge of fitness costs if untimely death leaves the remaining family defenseless. Killing an unfaithful mate or a mate-poaching rival may be a last-ditch defense against paternity uncertainty and all of the "misdirected" paternal effort that might be expended on vehicles containing the genes of a rival male. But killing in response to sexual infidelity could also lead to reputational damage, difficulty securing mates in the future, social sanctions, and the loss of a caregiver for shared children. The benefits and costs of killing in each of these three contexts are qualitatively distinct. We propose that there will be a number of distinct contexts in which the benefits of killing, on average, outweighed the costs and provided a greater overall payoff than other strategies.

4.4. Premise 4: Evolved homicide mechanisms are many in number and specific in nature, each corresponding to the selective pressures created by each distinctive and historically recurrent benefit/cost context.

A single, simple psychological mechanism would be extraordinarily unlikely to govern information processing in all contexts of killing. We hypothesize that each context would require different assessment procedures for determining benefits, costs, and alternative courses of action. The personal risks to one’s life as a result of killing a deformed infant, for example, are typically trivial compared to the risks associated with killing a spouse’s lover, who may fight back and kill in self-defense against the aggrieved husband. The alternative strategies available in response to an attack from a neighboring village (e.g., attempt to kill, flee, submit) are substantively different from strategies available to a cuckolded husband who discovers a wife naked with another man (e.g., get a divorce, seek a retaliatory affair).

The decision rules that govern the alternative courses of action in these contexts would not be generalizable across contexts, any more than "mate choices" and "food choices" could both be governed by a single general set of preferences (Symons, 1987; Buss, 1999). According to our theory, evolved homicide mechanisms will be numerous and distinct to the degree that the selective contexts in which killing was historically beneficial were numerous and distinct.

4.5. Premise 5: Each evolved homicide mechanism has distinct functions.

Each homicide mechanism is proposed to have a distinct set of functions, corresponding to the unique benefits that killers would have accrued in each context. The distinct functions of a man’s step-child homicide mechanism, for example, would include eliminating his investment in a child containing an intrasexual rival’s genes and curtailing the investment of his current mate in a child produced by a rival male. One function of the warfare mechanisms hypothesized by Homicide Design Theory, in contrast, is to gain direct sexual access to new fertile mates. One function of the intrasexual rivalry homicide mechanisms is to eliminate a key source of strategic interference and to cultivate a reputation as someone to be deferred to and not interfered with. Each of the evolved homicide mechanisms, in short, is proposed to have unique functions.

4.6. Premise 6: Consciously articulated homicidal premeditations are central functional components of design for murder.

Although the human brain comprises only 2% of the average human’s body weight, it consumes roughly 20% of the body’s calories (Leonard & Robertson, 1992; 1994). Cognitive activity, simply put, is metabolically costly. Metabolic energy devoted to processing information about one adaptive problem precludes metabolic energy available for other adaptive problems. The costs of information processing are not merely metabolic. They include opportunity costs. Time allocated to processing information about one adaptive problem preempts opportunities for cognitive work on other adaptive problems. The nature, content, and duration of particular kinds of thought, according to this view, reveal important clues about the problems the human mind evolved to solve.

According to Homicide Design Theory, homicidal fantasies function in some ways that are analogous to sexual fantasies. As Symons notes in the latter context, "the mind is adapted to cope with the rare, the complex, and the future . . . the function of the mind is to cause behavior; even if only one impulse in a thousand is consummated, the function of lust nonetheless is to motivate sexual intercourse" (Symons, 1979, p. 206-207). Clearly, humans experience hundreds or thousands of sexual thoughts and fantasies for every one that is actually acted upon. But this ratio of thought to deed should not be interpreted as calling into question the important functions of those thoughts. Precisely the opposite. The rare and occasional reproductive payoff associated with one of those thoughts is likely to have been so large (e.g., motivating actual sexual access to a new fertile partner) that it outweighed all of the metabolic and cognitive opportunity costs of the thousands of un-acted-upon thoughts.

According to our theory, the first step in the emergence of a homicidal thought is the recognition of an adaptive problem that historically has been potentially solvable by killing, such as having one’s life threatened by a conspecific or being burdened by a deformed offspring. This does not imply that homicide historically has been the only solution to the adaptive problem under consideration, or even that it has been the most common solution to that adaptive problem. Indeed, one hypothesized function of scenario building is to evaluate the utility of alternative potential solutions, of which homicide is only one (see below).

Homicide Design Theory proposes that homicidal thoughts, fantasies, and premeditations are key components of the evolved design of homicide mechanisms. The theory proposes that homicidal thoughts or premeditations will occur universally or nearly universally in humans around the globe (i.e., will be species-typical) when they are evoked by predictable circumstances for which homicide historically was a potentially effective solution to an adaptive problem. At the most general level, we propose the following features of homicidal premeditations:

4.6.1. Context Specificity Assumption. Homicidal thoughts will occur in specific and predictable contexts. Selection is hypothesized to have fashioned cognitive mechanisms designed to recognize contextual cues to adaptive problems that were potentially solvable through conspecific killing.

4.6.2. Attention Mobilization Assumption. Homicidal thoughts function to direct attention to particular features of the adaptive problem, such as future costs if potential victim continues to live, the different reputational consequences if the potential victim is or is not killed, potential vulnerabilities of the victim, and the formidability of the victim and the victim’s kin group.

4.6.3. Scenario Building or Cognitive Simulation Assumption. Homicidal thoughts include scenario building, a cognitive simulation of carrying out the homicide. Following the creation of images and cognitive simulations, further scenario building involves calculating the costs and benefits, associated with actually carrying out the kill. Evaluating the costs include the risk of being caught, injured, ostracized, or killed, as well as evaluations of the impact of possible errors in assessment. Scenario building will often entail considering alternative strategies, including non-lethal ones, capable solving the adaptive problem. Finally, scenario building entails evaluating the multiple cascading consequences, including reputational ones, for self, kin, and group members of carrying out a particular strategy. In the vast majority of instances, these cost/benefit calculations will lead to the decision not to carry out a homicide, just as in the vast majority of sexual fantasies, the cost/benefit calculations lead to the decision not to carry out the sex act. Ironically, one of the proposed functions of homicidal scenario building is to inhibit acting on a homicidal impulse when the costs are too high.

4.6.4. Emotional or Affective Assumption: Affect, in our theory, serves several key functions. First, it can serve as a tangible trigger to a homicidal thought. Different evolved homicide mechanisms contain different affective triggers. The fear that serves as a trigger for self-defense homicide, for example, will differ from the jealous rage that triggers a sexual rival homicide. And both of these differ from the relatively affectless state that accompanies certain types of infanticide. Second, affect functions to tag particular persons and events for storage in memory, facilitating subsequent retrieval. Third, affect is hypothesized to guide decisions about whether to enact a particular homicidal scenario (e.g., a homicidal scenario which, when played out, leads to the emotional experience of fear will generally inhibit carrying out the kill; a major exception occurs with anti-homicide mechanisms, in which a fear of being killed might facilitate carrying out a preemptive homicide). These distinct affective states are hypothesized to function as tags used in the comparison of alternative strategies.

4.6.5. Simulation Rehearsal Assumption: Cognitive simulations of homicide will often be repetitive, with various components getting repeated, rehearsed, and refined over the temporal course of the premeditation (see Prentky et al., 1989).

4.6.6. Motivational Assumption: Affects linked to thoughts propel deeds. Analogous to Symons’s logic with sexual fantasies, even if one homicidal thought in a thousand is acted upon, one critical function of homicidal premeditation is to motivate carrying out an actual kill. This does not require, however, that an individual have the conscious intention or desire to kill a conspecific. Only the motivation to engage in behaviors whose function is to cause the death of a conspecific need be present for the operation of a psychology of homicide.

The features of homicidal thoughts just described, of course, are not independent of one another. We hypothesize that they are tightly integrated components of human information processing that makes up the psychology of homicide. It is also not our contention that the entirety of evolved killer psychology is manifest in conscious homicidal thoughts. Much of the information processing machinery involved in homicide undoubtedly operates without the killer’s awareness, just as most of human digestive machinery operates without the eater’s awareness. Furthermore, some homicides often occur on the spur of the moment, perhaps even in the absence of conscious premeditations. These presumptive facts do not negate the hypothesis that homicides are the designed output of psychological mechanisms. We hypothesize that the majority of homicides, despite the fact that some occur in the absence of assiduously elaborated homicidal premeditations, result from the implementation of an evolved killer psychology. Such implementation requires thought, but not necessarily conscious thought. Nonetheless, conscious homicidal thoughts provide one window into the psychology of homicide, and according to our theory, are functional components mechanisms that evolved to regulate decisions to kill or not to kill in response to particular adaptive problems.

A number of direct empirical predictions follow from these assumptions. First, the contexts in which homicidal thoughts occur should exhibit a predictable pattern across cultures worldwide, that is, exhibit universality. This does not imply that every person in every culture will have already experienced homicidal premeditations. Rather, the same contexts that trigger homicidal thoughts in Japanese or Jamaicans, for example, should trigger homicidal thoughts in Zambians or Zulus. Second, even though the vast majority of homicidal thoughts will not lead to an actual homicide, homicidal thoughts will typically precede carrying out an actual homicide. Third, the contexts in which homicidal thoughts occur should exhibit the same patterns as the contexts in which actually homicides occur. Fourth, the frequencies of particular kinds of homicidal thoughts should correlate positively with the frequencies of those same kinds of actual homicides.

In summary, homicidal premeditations are both windows into the psychology of killing and key functional components of the evolved homicide mechanisms. They serve a variety of functions, from scenario building to calculating cascades of consequences. One critical function of homicidal premeditations is to inhibit the translation of a homicidal impulse into actual behavior. Most homicidal thoughts do not lead to actual killing, because the cost-benefit appraisals suggest that the costs are too high, the benefits too low, or more effective solutions are available for solving the adaptive problem. Nonetheless, one of the critical functions of homicidal thoughts is motivational, preceding and prompting murder.

Finally, we propose that conspecific death is a key designed output of evolved homicidal mechanisms. The existence of recurrent cognitively-costly homicidal thoughts in specific circumstances are difficult to explain on theories that propose that strategies of violence have evolved solely for non-lethal deployment.

4.7. Premise 7: Men and Women Have Evolved Different Homicide Mechanisms

Selection has not forged identical homicide mechanisms in men and women. In its most general form, this assertion is unlikely to be controversial. The vast majority of actual homicides are committed by men, not women, and so it would not surprise evolutionarily informed scholars that men have evolved bodies and minds designed to kill more than women. But our theory goes well beyond the premise of a general sex difference in the likelihood of killing. We propose that men have evolved some homicidal mechanisms that are entirely lacking in women, such as those designed for warfare (see also, Tooby & Cosmides, 1988; Wrangham, 1999). We propose that women have evolved some homicidal mechanisms entirely lacking in men, such as certain types of infanticide. And we propose that even when men and women both possess similar homicide mechanisms at one level of abstraction (e.g., both sexes have evolved mechanisms to kill infants), those mechanisms will contain design features that differ substantially in women and men. In the case of infanticide, for example, we propose that only men have evolved a psychology of infanticide sensitive to the context of lack of certainty in their paternity, whereas only women have evolved a psychology of infanticide sensitive to the context of the lack of an investing father. Although both sexes have evolved infanticide mechanisms, the specific design features of these mechanisms are proposed to differ profoundly for the sexes.

Some of these hypothesized sex differences have a straightforward evolutionary logic that has essentially been articulated by others in the context of violence (e.g., Daly & Wilson, 1988; Ellison, 1985). Men historically had a higher ceiling on reproduction than women, since a man’s limit is imposed primarily by the number of women he can successfully fertilize (e.g., through polygyny or opportunistic short-term mating). At the same time, a man has a greater likelihood than a woman of dying with zero descendants. These recurrent historical differences in fitness variances have been hypothesized to lead to the evolution in men of a wide variety of risk-taking mechanisms. Risk taking, including the sort of violent risk taking involved in attempted homicide, would have paid greater reproductive dividends to men than to women, both in gaining great reproductive bounties and in avoiding reproductive oblivion.

This evolutionary reasoning can partially explain why men kill more than women, and perhaps even why men might have evolved more numerous or more frequently acted-upon homicide mechanisms. But it stops short of explaining the specific homicide mechanisms men possess, as well as their component design features. And it fails to explain why women have evolved homicide mechanisms or why there are contexts in which women are more likely to kill than men. Our theory, which includes many sex-linked premises, is meant to fill these theoretical gaps.

4.8. Premise 8: Victims incur extreme and heretofore unrecognized cascading fitness costs by being killed.

Although it would certainly not make headlines to announce that it’s costly to be killed, being a homicide victim has fitness consequences that no prior theories of homicide or human survival have enumerated or illuminated. To start with, being killed cuts off all avenues to the victim’s future reproduction. The victim loses access to his or her mate’s residual reproductive value, which may instead be coopted by a rival. Furthermore, additional mating opportunities through affairs, serial mateships, and multiple simultaneous spouses are lost, with all the attendant reproductive losses. The earlier the reproductive age of the victim, ceteris paribus, the larger the average reproductive costs will be to the victim if he is killed. A barely post-pubescent female victim, for example, would suffer a greater cost of being killed than her post-menopausal counterpart. Being killed, in short, curtails a victim’s future reproductive possibilities, which inflicts a massive fitness cost.

Second, being killed can damage the victim’s children in countless ways. Being dead, the victim is no longer present to invest in his or her children. The children become more vulnerable to exploitation or injury by others without the protection of the investing parent (Hill & Hurtado, 1996). The homicide victim loses the ability to affect the children’s future mating or any of the children’s future prospects, strategies through which reproductive success historically may have been increased. By being killed, the victim’s children become more vulnerable to losing the remaining spouse’s investment, which may get re-channeled to the children produced with a new mate. Finally, the victim’s children risk becoming step-children if the victim’s mate remarries. Since step-children suffer physical abuse and homicide at rates 40 to 100 times higher than children residing with two natural parents (Daly & Wilson, 1988), becoming a stepchild would have impaired the potential success of key fitness vehicles of the homicide victim.

A third set of detrimental fitness consequences of being murdered centers on damage to the victim’s extended kin group. By being dead prematurely, the victim loses the ability to protect or invest in kin. The entire collection of extended kin–brothers, sisters, aunts, uncles, nephews, nieces–become more vulnerable or exploitable as a result of the victim’s death. Without the victim’s presence and influence, specific kin members could lose mate value, impairing their future mating prospects. And the victim’s death could sever key ties between two different extended families, which originally forged an alliance through the marriage of the victim.

Last but certainly not least, all of the victim’s losses become potential gains for the rivals of the victims, and hence the rival’s genes. The victim’s mate, for example, becomes available to mate with one of the victim’s intrasexual rivals. The elimination of the victim from the status hierarchy opens a niche for a rival to ascend, a rise that previously may have been blocked by the presence of the victim. Children of a rival will thrive in competition with the victim’s children, who will be hampered by their parent’s death. In short, the costs of getting killed cascade to one’s children, grandchildren, great-grandchildren, and the entire collection of extended kin descendants or would-be descendants of the homicide victim.

Although human intuition tells us that it’s undesirable to be killed, no prior theories of human survival or homicide have articulated the large, specific, and multifaceted fitness costs. We propose that these multiple and far-reaching costs are critical to the evolution of homicide mechanisms. They describe a key portion of the selective benefits that would accrue to killers–fitness costs inflicted on intrasexual rivals can become fitness gains for the killers. Killers, in short, historically could have increased their fitness by inflicting these multiple and multifacted costs on conspecific competitors. At the same time, it is precisely these same costs that historically selected for anti-homicide mechanisms, leading to the next premise.

4.9. Premise 9: Because of the tremendous fitness costs associated with getting killed, a complex suite of anti-homicide mechanisms has evolved that function to prevent premature death.

If killing has been a recurrent feature of human evolutionary history and the costs of getting killed are as catastrophic as Premise 8 indicates, then it would be exceedingly unlikely that selection will have failed to forge mechanisms designed to prevent being killed. Just as humans appear to have evolved specific fears of snakes, spiders, and heights to prevent poisonous bites and fatal falls, we propose that humans have evolved specific anti-homicide mechanisms to prevent being killed by other human beings.

According to this premise, anti-homicide mechanisms would be sensitive to the precise contexts in which one’s life is at risk from a conspecific. One set of anti-homicide mechanisms may get activated in a fetus to combat evolved abortion mechanisms in mothers (Haig, 1993). Haig presents evidence that mothers have evolved abortion mechanisms designed to terminate investing in fetuses under certain conditions such as poor health of genetic abnormalities. Since the fetus has only one shot at life, it’s interests can depart from the mother’s. One adaptation that appears to have evolved for this function is the fetal production of human chorionic gonadotropin (hCG), a hormone the fetus secretes into the mother’s bloodstream. The hormone has the effect of preventing the mother from menstruating, and thus allows the fetus to remain implanted. Producing elevated levels of hCG, therefore, appears to be an adaptation in the fetus to subvert the mother’s attempts to spontaneously abort it. The female body appears to "interpret" high levels of hCG as a sign that a fetus is healthy and viable, and so does not spontaneously abort. Fetal anti-abortion devices, if future work confirms their existence, constitute evolved anti-homicide mechanisms designed to combat the very first lethal threat to existence.

A second hypothesized anti-homicide mechanism is "stranger anxiety," which appears nearly universally in infants around six to nine months, precisely the time when they become capable of crawling away from caregivers (Hrdy, 1999). Infants are not taught to fear strangers, and the seemingly irrational panic "derives from a built-in prejudice so deep it persists despite every reassurance the parent offers" (Hrdy, 1999, p. 416). It is reasonable to hypothesize that the intense fear of strangers exhibited so reliably by infants across cultures is an anti-homicide device, designed to evoke parental protection, that evolved in response to the likelihood that "infanticide may have been a chronic threat during hominid evolution" (Hrdy, 1999, p. 416).

We defer a full treatment of the evolution of anti-homicide mechanisms (see Duntley & Buss, in prep., a), but note here that adult anti-homicide mechanisms are hypothesized to have a number of design features. These include sensitivity to the contexts in which one’s life is in danger, such as trespassing on an enemy’s territory or being discovered having sex with an intrasexual rival’s wife. They include detecting a homicidal rage in another person or anticipating revenge or retribution from a person or group whom one has previously violated, and so invoke specific mind-reading abilities. They include emotions such as specialized terror of death and battlefield panic. And they include a variety of behavioral responses, such as emitting a piercing death scream to alert kin and allies, fleeing in terror, or furiously attacking a would-be killer.

Since an individual shares many genes with his kin, we predict that the psychology of anti-homicide will be activated by cues to the impending homicide of both an individual and his kin. We predict that the degree to which anti-homicide psychology will be activated will depend partly on the number of common genes, as well as coalitional allies, that an individual stands to lose as a result of the homicide. In effect, our psychology of anti-homicide evolved to protect our genes, wherever they may reside.

One central design feature hypothesized to characterize most or all anti-homicide mechanisms is a seemingly irrational overreaction to threats, which evolved according to the logic of Error Management Theory (Haselton & Buss, 2000). In making inferences in an uncertain world, there are two types of errors. One can falsely infer a problem or signal that does not exist, such as thinking that one’s life is in danger when it is not. Or one can fail to perceive a signal that is present, such as believing that one is safe when in fact one’s life is in danger. According to Error Management Theory, there have been many recurrent cost asymmetries associated with the two types of errors. In this example, it will generally be more costly to fail to infer real threats than to err on the side of over-inferring threats that are in fact non-existent. Recurrent cost asymmetries, according to this theory, result in the evolution of cognitive biases to err in the direction of less costly error (Haselton & Buss, 2000). Since getting killed is so catastrophically costly, we expect that human anti-homicide mechanisms contain cognitive biases that result in overestimating the likelihood of getting killed in certain circumstances.

This premise leads to a raft of testable empirical predictions about he design of anti-homicide mechanisms. First, humans are predicted to routinely exhibit biases in overestimating the likelihood that their lives are in danger across a variety of circumstances. Second, this bias should extend to close kin, and especially children, given their value as genetic vehicles. Fourth, the bias should extend to mates, and especially to the extent that they are reproductively valuable to the individual. Fourth, the degree of bias should decrease as a function of decreasing genetic relatedness of the person to the potential victim.

In summary, according to our theory, anti-homicide mechanisms include: (a) fear of strangers in infancy; (b) specialized mind-reading abilities to infer homicidal intent in others, (c) adaptively biased inference mechanisms that result in protective overestimates of the likelihood of getting killed, (d) selective preferences for habitats that afford protection and refuge (e.g., seeing without being seen), (e) anticipatory defensive maneuvers including development of arms, fortifications, and protective postures, (f) panic reaction to motivate fleeing when confronted by another human displaying homicidal intent, and (g) homicidal mechanisms designed to impel killing to prevent getting killed.

The discovery of anti-homicide mechanisms affords a powerful and fresh source of potential empirical evidence to test the central tenets of Homicide Design Theory. The key to whether an evolved psychology of anti-homicide provides compelling evidence for the existence of an evolved killer psychology rests largely on the specificity of psychological design features for anti-homicide. The greater the number and degree of elaboration apparent in the design of psychological mechanisms for anti-homicide, the greater the likelihood that they evolved in response to the presence of an equally specific and elaborated evolved homicidal psychology. Discoveries of specificity in one will lead to investigation for corresponding specificity in the other. This leads to the final premise, involving a co-evolutionary arms race between killers and victims.

4.10. Premise 10: The evolution of anti-homicide mechanisms makes killing less beneficial and more costly, resulting in selection for increasingly refined homicide mechanisms and increasingly elaborate defense mechanisms, producing a within-species co-evolutionary arms race between homicidal strategies and defensive anti-homicide strategies.

The principle of co-evolution typically has been used to describe reciprocal evolutionary changes in interacting species, such as predators and prey or parasites and hosts. It has been used less often to explain reciprocal evolutionary changes within a single species, although there are important exceptions to this generalization (e.g., Buss, 2000; Haig, 1993; Holland & Rice, 1998). Premise 10 invokes the principle of co-evolution to describe the reciprocal evolutionary changes set into motion when killing became an important part of human selective environments and anti-homicide mechanisms appeared on the scene.

Anti-homicide mechanisms lower the on-average benefits that otherwise would accrue to would-be killers in two ways. First, the evolution of anti-homicide mechanisms reduces the average success of attempted homicides. Potential infanticide victims who cry to alert parents, male rivals who anticipate an attack, and women who secure the protection of close kin in order to defend against an enraged husband intent on uxoricide all are acting to reduce the average success of a killer strategy. Second, anti-homicide mechanisms can inflict heavy costs on would-be killers. Alerted parents, defensive rivals, and protective kin can injure or ostracize a prospective killer. In the extreme case, when people kill in self-defense in order to prevent getting killed (or to prevent kin from being killed), the cost inflicted on the would-be killer are grave indeed. For both reasons–reduced success and the risk of costs--the emergence of anti-homicide mechanisms has the effect of lowering the on-average benefits that otherwise would have been obtained by pursuing a homicidal strategy.

The evolution of anti-homicide mechanisms, in turn, would prompt reciprocal evolutionary changes in killer psychology. First, when it becomes more costly to attempt a kill, the detection and evaluation of costs become increasingly important inputs to decision rules about whether or not to attempt a kill. Ceteris paribus, this would make killing a less frequently chosen solution to a variety of adaptive problems, thus reducing overall homicide rates or holding them temporarily in check.

Second, there would be further selection for refined homicidal strategies designed to circumvent a victim’s anti-homicide defenses. These refinements may include the evolution of deceptive tactics, such as feigning friendliness or concealing homicidal intent so as not to arouse a would-be victim’s anti-homicide mechanisms, and then surprising the victim with a sudden attack. Another refinement would be an assessment of the comparative formidability of self and intended victim to identify circumstances in which the victim’s anti-homicide defenses will be ineffective. The co-evolution of these and other design features in killers would have had the effect of lowering the costs to killers inflicted by the anti-homicide mechanisms of victims, increasing the average success of homicidal attempts, and thus increasing the average success of homicide as a strategy. If there were no further co-evolution, the homicide rate would then increase over time as killers evolved to circumvent the anti-homicide defense mechanisms of victims.

Given how costly it is to be killed, however, selection would immediately favor refinements in anti-homicide psychology. Consequently, defense mechanisms would become increasingly sophisticated and elaborate. People would evolved heightened sensitivity to a larger array of subtle signals of homicidal intent. Thresholds for detecting homicidal intent would decrease. People would become increasingly suspicious to counteract being deceived. Behaviorally, people would avoid circumstances in which their lives were at risk and more prone to resort to lethal violence as a defense. Adaptive anti-homicide biases, evolving according to the logic of Error Management Theory, would be heavily favored.

In summary, a core premise of Homicide Design Theory is that homicide and anti-homicide mechanisms co-evolved, each imposing selective pressure on the other, resulting in a coevolutionary arms race throughout human evolutionary history. Like predators and prey, conspecific killer mechanisms and victim defense mechanisms evolve to become increasingly elaborate, responding to the constantly evolving contrapuntal adaptive problems imposed by each.

5. Hypothesized Evolved Homicide Mechanisms

In this section, we outline our theory of specific homicide mechanisms, noting selection pressures hypothesized to have given rise to each mechanism, the proposed design features of each mechanism, the sex-differentiated components of each mechanism, and specific empirical predictions that follow from each proposed mechanism. We begin with infanticide because it is the most well documented form of conspecific killing.

5.1. The Evolution of Infanticide

Early in her research on langur monkeys, Hrdy (1977) noted that the infant monkeys she was observing started to vanish. Langurs live in matrilineal groups consisting of several adult females and their infants, with only a single adult male permitted within each. They subsist primarily on leaves, and are not carnivorous. Hrdy witnessed at separate times four adult males enter a matrilineal group and defeat or scare away the resident male. In each case, the victorious male proceeded to kill the youngest infants in the group, averaging three infanticides per male. The new male then proceeded in the ensuing months to mate with the mothers of the infants they had killed. A new male who failed to kill genetically unrelated infants would typically have had to wait for several years before the female became fertile because of lactation-produced amenorrhea. Since males reign on average for two years, a male who failed to commit infanticide would fail to reproduce. The infanticide came at a tremendous cost to the mothers, who often attempted to defend their young. But after the killings, they showed no hesitation in mating with the killer males.

Many animal biologists and primate anthropologists grant that such brutal infant murders make eminent reproductive sense (Ghiglieri, 1999). Hrdy, for example, titled her 1977 article on lemur infanticide "Infanticide as a Primate Reproductive Strategy." More recently, Ghiglieri concluded that "Overall new males who did not kill infants but instead waited sired fewer infants than males who did kill them. This is all sexual selection needs to install infanticide as a male reproductive strategy" (Ghiglieri, 1999, p. 131). Similar observations have been found for gorillas. One out of 7 gorillas born ends up a victim of infanticide. And 40% of infant morality in the Virunga Volcano gorilla groups has been attributed to infanticide by Diane Fossey (Goldsmith & Zimmerman, in press, chap. 13, p. 12). Similar observations have been made among lions, tigers, cheetahs, and cougars (Ghiglieri, 1999).

When it comes to humans, we observe a predictable set of circumstances in which people commit infanticide, but surprisingly, most scientists stop short of proposing that humans have a specific evolved infanticidal psychology. Daly and Wilson note that "There is no reason to suppose that an evolved parental psychology should be such as to value every offspring equally and indiscriminately" (Daly & Wilson, 1988, p. 42). They continue: "parental psychology has been shaped by selection to make adaptive decisions about the magnitude of parental commitments" (p. 44). Then they say that "We thus expect child-specific parental love to be variable in its strength" (p. 73). Even Hrdy, who originally discovered infanticide among Langurs, notes strikingly analogous patterns in humans, but fails to propose the evolution of specific mechanisms for infanticide. Instead, she argues that "what evolved was a high threshold for responding in a solicitous way toward an offspring not likely to be genetically related–the equivalent of emotional earplugs" (Hrdy, 1999, p. 237). The metaphor of "emotional earplugs," however, does not offer a satisfactory explanation of the highly predictable contexts in which human infanticides occur. Nor can "a high threshold for responding in a solicitous way" explain the premeditated and deliberate manner in which infanticides are carried out, the specific contexts in which they occur, or the brutal battering that some of the infants suffer at the hands of their killers.

Daly and Wilson present an impressive array of evidence suggesting that parents kill their infants under very predictable circumstances. These include: (1) When there is paternal uncertainty, such that the infant might not be the putative father’s child; (2) when the infant is deformed, diseased, severely underweight, or of otherwise questionable quality such that it might not survive or thrive; and (3) when external circumstances, such as food scarcity or lack of an investing father render this not a propitious time for a woman to reproduce.

Among the Ayoreo women residing in Bolivia, for example, women kill their infants when they lack a mate who could provide for her and her offspring, when the infant is born deformed, when twins are born (typically one is killed), and when an infant is born too soon after a previous birth, which jeopardizes the survival of the older child (Bugos & McCarthy, 1984). Similar patterns have been documented among the Ache of Paraguay (Hill & Kaplan, 1988), among the Yanomamo (Chagnon, 1983), and in fact all over the world (Scrimshaw, 1984).

The key point is that over the long course of human evolutionary history, there would have been tremendous fitness costs associated with continuing to invest in an infant that was deformed, diseased, unlikely to survive, or unlikely to reproduce.

After much discussion of the evidence, several theorists have come to the conclusion that humans have evolved mechanisms for murdering infants, just as lemurs, cougars, and lions have evolved mechanisms for murdering conspecific infants. Ghiglieri notes that "Infanticide seems written into our genes, but it is generally manifested in young, unwed, desperate mothers who murder neonates" (Ghiglieri, 1999, p. 135). He notes that "most infanticide has been decided by people trying to produce more children in the long run by sacrificing an infant now" (Ghiglieri, 1999, p. 136). He argues that "Men kill stepchildren for the same reasons other primate males kill infants: to increase their reproductive success by sweeping the reproductive arena clean of the offspring of male competitors" (Ghiglieri, 1999, p. 137). Even Daly and Wilson, authors of the "byproduct hypothesis," come quite close to postulating an evolved infanticide mechanism: "an evolved assessment mechanism should be such as to terminate any hopeless reproductive episode as early as possible, rather than to squander parental effort in an enterprise that will eventually be abandoned" (Daly & Wilson, 1988, p. 75).

We build on the conceptual claims by Ghiglieri, Daly, and Wilson by specifying additional design features in the evolved psychology of infanticide. Specifically, it is important to cleanly separate hypotheses about male and female infanticide mechanisms, since the proposed evolved design of infanticide psychology differs sharply for the sexes.

5.1.1. Female infanticide mechanisms. Women differ from men in the magnitude of their obligatory parental investment (Trivers, 1972). Women carry a nine month gestational obligation, which involves heavy metabolic costs, decreased mobility, increased physical vulnerability, and substantial mating opportunity costs. Compared to men, women's fertile years are more sharply compressed into a smaller fraction of their overall lifetime. Given the small number of genetic "vehicles" a woman can produce, it would be unlikely that selection would have failed to forge mechanisms to terminate maternal investment in some infants in order to preserve her finite time and resources for others.

We propose that infanticide solved five key adaptive problems for ancestral women: (1) avoiding wasting reproductive resources on phenotypically or genotypically sub-par offspring (e.g., those who were deformed, diseased); (2) avoiding investing in offspring when external conditions substantially imperil the infant’s survival (e.g., food scarcity; lack of an investing father, harsh season); (3) preventing competition with older or more viable offspring (e.g., twin births or when birth spacing is too close); (4) increasing the mate value of the mother, for whom children either interfered with an existing or emerging mateship or lowered her odds of successfully attracting a desired new mate; and (5) maintaining an existing long-term relationship (e.g. a child that obviously does not look like the woman’s long-term mate, a cue to infidelity).

The fitness benefits of committing infanticide in all five of these conditions are predicted to decrease as a function of the age of the mother. Young mothers, because of their greater residual reproductive value, can better afford to wait for more auspicious resource and mating circumstances to replace the dead offspring. Older mothers, having lower residual reproductive value, cannot wait, and so are predicted to be more reluctant to kill an infant under each of the five conditions.

We predict that mothers faced with these circumstances will entertain homicidal premeditations, conducting cognitive simulations of behaviors leading to killing, with the attendant cost-benefit and affective evaluations described earlier. Our theory predicts that younger women confronted with these adaptive problems will be more likely than older women to entertain such infanticidal premeditations. Other theories of infanticide, such as that infanticide is merely an incidental byproduct of more general mechanisms of parental solicitude, do not make these predictions and would have difficulty explaining their existence and their context-specific nature.

5.1.2. Male infanticide mechanisms. The adaptive problems solved by male-perpetrated infanticide, according to our theory, are in some ways dramatically different from those of females, and hence the contexts in which men and women kill infants should differ correspondingly. A man’s mate value, for example, does not decline as much as a woman’s as a result of having existing children by a former mate (Buss, 1994). Furthermore, since a smaller portion of a man’s than a woman’s total effort is typically allocated toward parenting, the costs of investing in sub-par offspring are commensurately lower for men than for women.

Conversely, there are several adaptive problems solved by male-perpetrated infanticide that we hypothesize were recurrent for men, while being weaker or entirely absent in women. These include: (1) avoiding investing in an infant known to be sired by a rival male (e.g., mate was already pregnant prior to the initiation of the current mateship); (2) terminating investment in an infant for whom there is paternity uncertainty (e.g., appraisal of possible infidelity by the woman); (3) preserving the mate’s resources for the man’s current or future offspring; and (4) hastening the commencement of the mate’s ovulation and hence conceptability.

In sum, we predict that the contexts triggering both infanticidal premeditations and actual infanticides will be sharply sex-differentiated. Non-evolutionary theories of infanticide would have great difficulty explaining the existence of such context-specific premeditations and their hypothesized sex-linked occurrence. Similarly, evolutionary theories that hypothesize that infanticide is a byproduct of more general mechanisms such as "differential parental solicitude" or "a high threshold for responding in a solicitous way" contain no premises that can explain the context-specific and sex-specific design features of infanticide.

We have begun our discussion of evolved homicide psychology with infanticide for several important reasons. First, infanticide is well documented in the mammalian world, and those who have studied species ranging from lemurs to lions appear to have no doubt that these species have evolved specific infanticide mechanisms (e.g., Hrdy, 1999). The adaptation logic for such non-human infanticides is apparently so clear and compelling, whether it’s a male killing an infant he did not sire or a female killing an intrasexual rival’s infant, that the hypothesis of evolved infanticidal mechanisms is apparently uncontroversial.

Second, many of the same patterns are observed in human infanticide, and some theorists, even those who are skeptical of the existence of evolved homicide mechanisms in other contexts, have alluded to the possibility that humans have evolved mechanisms for murdering infants. In fact, of all the scientists with whom we’ve discussed our theory, the infanticide component appears to be the evolved homicide mechanism that most seem ready to accept as plausible (e.g., Margo Wilson, personal communication).

We mention these points for a specific reason: Infanticides represent a relatively trivial proportion of all homicides--merely 2-4% of all human killings. If it’s reasonable to hypothesize that selection has forged a specific psychology of infant murder given the relative rarity of the event, then it is reasonable to hypothesize that selection has forged specific mechanisms for murder in other contexts that appear to be far more frequent. Spousal homicides in the United States, for example, are roughly five times more common than infanticides.

Selection operates on the recurrent frequency of selective events in conjunction with the fitness consequences of those events. Homicides have particularly dramatic fitness consequences, as we argued in Premise 1 (potential fitness benefits to killers) and Premise 8 (potential fitness costs to victims). Therefore, if selection can forge specific evolved mechanisms of infanticide given the relative rarity of infant killing, then it is reasonable to hypothesize that selection could have operated even more powerfully to forge mechanisms for murderous contexts that are many times more common. One of the most common is intrasexual rivalry homicide, a topic to which we now turn.

5.2. Killer Males: Intrasexual Rivalry Homicide

By far the most frequent type of homicide worldwide is male-on-male intrasexual killing. In Daly and Wilson’s study of Canadian killers, 69% of the killings fell into the male-male quadrant (Daly & Wilson, 1988). Similar statistics have been found among the !Kung Bushmen (Lee, 1979; Shostak, 1981), the Yanomamo of Venezuela (Chagnon, 1983), the Gebusi of New Guinea (Knauft, 1985), and the Ache of Paraguay (Hill & Hurtado, 1996).

The !Kung Bushmen, often described as "the harmless people," provide a good illustration of the prevalence of male-on-male homicides. A key ethnography noted of the !Kung that "It is not in their nature to fight . . . Bushmen cannot afford to fight with one another and almost never do because their only real [lethal] weapon is their arrow poison, for which there is no antidote. But even were they to disregard this danger, Bushmen would try not to fight because they have no mechanism in their culture for dealing with disagreements other than to remove the causes of their disagreements . . . The !Kung call themselves zhu twa si, the harmless people" (Thomas, 1959, p. 21-24).

Richard Lee recorded 22 homicides among the !Kung over a 35 year period (Lee, 1984), which corresponds to a homicide rate of 29.3 killings per 100,000 per year (Ghiglieri, 1999). This homicide rate exceeds that for modern day Los Angeles or New York City, which have homicide rates of 24.5 and 16.1, respectively (Ghiglieri, 1999, p. 116). All the killers were men, as were 86% (19) of their victims. Nearly all the !Kung killers used poisoned arrows to carry out the deed. The homicides appear to have been intentional, not slips or accidents. When the ethnographer interviewed the !Kung killers about why they chose to use poisoned arrows rather than a less lethal weapon to harm their rival, a common response was "We shoot poisoned arrows because our hearts are hot and we really want to kill somebody with them" (Lee, 1984, p. 95).

Daly and Wilson (1988) compiled same-sex homicide statistics from 35 different studies representing a broad span of cultures from Miami to the BaLuyia of Kenya. Although homicide rates vary widely from culture to culture, the proportion of same-sex homicides that involve male killers and male victims are remarkably consistent, ranging from a low of 85% in Denmark from 1933 through 1961 to a high of 100% in Iceland (1946-1970), among the Tzeltal Mayans in Mexico (1961-1965), and among the Munda of India. The average across samples is approximately 95%; 19 out of twenty same-sex killings are committed by men. Even this figure underestimates the magnitude of the adult-on-adult homicide sex difference, since infanticides are included. When infanticides are excluded, the ratio of male-to-female same-sex killers jumps to almost 40 to 1 (Ghiglieri, 1999, p. 146).

In every culture for which there are data, the rate at which men kill other men vastly exceeds the rate at which women kill other women: "there is no evidence that women in any society have ever approached the level of violent conflict prevailing among men in the same society" (Daly & Wilson, 1988, p. 149; emphasis original). Any comprehensive theory of homicide must account for both facts–why men kill so much more often than women worldwide and why other men comprise the majority of their victims.

Daly and Wilson (1988) explain this large sex difference by invoking sexual selection for risky male tactics, originally traceable to the greater reproductive variance between males and females. For example, "If status has persistently contributed to reproductive success, and a capacity for controlled violence has regularly contributed to status, then the selective advantage of violent skills cannot be gainsaid" (Daly & Wilson, 1988, p. 133). It is important to stress that Daly and Wilson are not proposing an evolved psychology of homicide. Instead, they argue that homicides "have to be understood as the rare, fatal outcomes of a ubiquitous competitive struggle among men for status and respect" (Daly & Wilson, 1988, p. 146). In a recent publication, Daly and Wilson are even more explicit in arguing that intrasexual homicides are epiphenomenal byproducts and over-reactive mistakes (Daly & Wilson, 1998).

Beyond general statements about "the selective advantage of violent skills," the evolution of "risky competitive tactics," and the expressed view that killing antagonists may be, and may always have been, an over-reactive "mistake," Daly and Wilson say little about the precise nature of the psychological mechanisms involved in same-sex killings. Nor do they enumerate the specific reproductive benefits that would have accrued to intrasexual killers. They do document, however, a range of contexts in which men kill other men, interpreted within a general theory of conflicts in reproductive interests. They argue that male-male killing is reducible to a fairly small number of underlying motives such as redressing a public insult or other blows to reputation, acquiring material resources (e.g., robbery murders), and sexual rivalry (e.g., killing a rival who has had sex with one’s wife).

According to Homicide Design Theory, the underlying psychological mechanisms involved in male-male killings are specific evolved homicide mechanisms that are not well described in general phrases such as "risky competitive tactics" or "over-reactive mistakes." Our theory proposes that most intrasexual homicides are not mistakes, but rather are kills by design. Mechanisms for male intrasexual murder, according to our theory, have evolved to solve at least five distinct adaptive problems–protection, reputation management, resource acquisition, mate encroachment, and mate acquisition.

5.2.1. Self and kin protection: Killing in self-defense. Although we lack a videotape of the past four million years, much available evidence (reviewed earlier) suggests that conspecific killing has been a recurrent feature of human ancestral environments. This inference is endorsed even by scientists who do not propose that humans have evolved specific homicidal mechanisms. Daly and Wilson, for example, note after reviewing the then-available paleontological evidence that "What is certain is that men have fought and killed one another for tens of thousands of years" (Daly & Wilson, 1988, p. 144). Since that time, substantially more paleontological evidence has been gathered, all pointing to the same conclusion (Grauer, 1995; Grauer & Stuart-MacAdams, 1998).

If conspecific killing has been a recurrent feature of human ancestral environments, it reasonable to infer that the risk of being killed formed a significant adaptive problem. Given the devastating fitness costs of being killed, selection would favor the evolution of defenses designed to lower the likelihood. Clearly, there are may potential solutions. Potential victims could offer to cede all of their mates, resources, territory, and goods to the would-be killer, a strategy that might lower the odds of being killed. They could run or hide to escape from would-be killers. They could migrate to a distant territory. Or they could attempt to gather kin and non-kin allies in sufficient numbers to deter a would-be killer. According to Homicide Design Theory, these and other anti-homicide strategies have evolved in humans (Duntley & Buss, in prep., a).

Another potential anti-homicide strategy is murder. Although there are costs associated with this strategy, including risking retaliation from the kin of the dead victim, the benefits in certain circumstances could easily have outweighed the costs. By killing his attacker, the self-defense killer prevents getting killed, thereby preventing all the costs to self, children, and collateral kin described earlier. The self-defense killer also preserves, rather than cedes, resources, territory, and mates, all of which could have been lost to an intrasexual rival. He or she can eliminates future sources of "strategic interference," including costs that the would-be killer and his children would inflict in the near and distant future. And by killing an intrasexual competitor, one inflicts a severe cost on that rival, including damage to the rival’s children and extended kin group. For all these reasons, we hypothesize that humans have evolved homicidal mechanisms to solve the adaptive problem of lethal threats from conspecific rivals.

Several specific, although perhaps not surprising, predictions follow from this theory: (1)

Humans worldwide will entertain thoughts of killing another human being in order to prevent being killed themselves, when faced with a life-threatening rival; (2) Humans worldwide will entertain thoughts of killing another human being in order to prevent their children from being killed; (3) There will be a gradient of willingness to kill to prevent the death of a kin member, falling off as the degree of genetic relatedness decreases (e.g., given the risks of attempting to prevent a kin member from being killed, the benefits of killing must conform to Hamilton’s c < br rule, whereby the fitness costs must be less than the fitness benefits, weighted by degree of genetic relatedness); (4) Among cultures, subgroups, or specific individuals within a group that recurrently face the adaptive problem of getting killed, intrasexual homicidal premeditations will be more common than among those who have not faced this adaptive problem; (5) among those who have never before entertained the thought of killing, homicidal premeditations can be easily evoked by confronting an individual with the credible threat of being killed (or close kin being killed).

All these predictions are testable. No other theory of homicide, to our knowledge, generates these specific predictions, nor would any previous theory of homicide be able to explain them if they are confirmed.

Although both sexes historically have faced the adaptive problem of being killed by a conspecific, death at the hands of rivals has been a more frequent adaptive problem for males than for females. Furthermore, the contexts in which one’s life and one’s children’s lives were at risk from other humans are and were sharply sex differentiated. The person most likely to kill a man, for example, is an intrasexual rival. In contrast, the person most likely to kill a woman is a current or former romantic partner. Men across cultures face the problem of getting killed in a variety of contexts by rivals, including being the victim of a jealous husband or boyfriend who becomes enraged, having honor and reputation challenged, and being attacked in a surprise raid by a neighboring tribe. Women are rarely at risk of being killed by intrasexual rivals, even from women whose husbands they are caught sleeping with.

For these reasons, two additional predictions, both sex-linked, can be derived from our theory: (6) men will be more likely than women to have homicidal premeditations about killing in self-defense; (7) men will have these homicidal self-defense premeditations in a wider variety of contexts than will women (e.g., when being caught in flagrante delicto with another man’s wife).

5.2.2. Killing to protect resources. The direct threat of being killed is merely one end of a continuum of reproduction-impairing costs that can be inflicted by rivals. Rival encroachment on one’s resources, in some circumstances, would have been disastrous to survival and reproductive success. Food resources stolen in ancestral times would sometimes have pushed a person to the brink of starvation. Stolen stone tools, spears, bows, or arrows could leave a person stripped of the means for food acquisition. Weapons stolen could leave a person bereft of the means to defend self and family. Expropriating a shelter could render the victim vulnerable to the hostile forces of nature (e.g., predators, cold, heat) as well as more exposed to hostile conspecifics. Expropriated land could deprive the victim of the water to survive or the game to thrive. Having one’s resources encroached upon with impunity could have led to a ruinous reputation as someone who could be exploited without fear of retaliation–a selective force so powerful in itself that we consider it separately below.

Clearly, there are many possible solutions to the adaptive problems posed by rival resource infringement other than killing the rival. Non-lethal violence, assembling a kin coalition in defense, retaliatory thefts, and migration away from hostile humans, for example, undoubtedly were and are common solutions. Nonetheless, we hypothesize that males have evolved decision rules that include killing as one solution to the problem of rival resource encroachment. Although attempting to kill a resource poacher obviously carries risks, it also can bring benefits such as (1) preserving reproductively vital food, tools, weapons, shelter, and habitat for oneself and one’s kin; (2) depriving an intrasexual competitor of access to those resources; (3) eliminating future strategic interference that the rival would otherwise inflict if he were alive; and (4) cultivating a reputation that deters other conspecifics from attempting to encroach on one’s resources.

Specific empirical predictions follow from the hypothesized resource protection killing mechanism: (1) Humans will have homicidal fantasies about someone who has robbed or plundered a significant personal resource; (2) Males will have homicidal thoughts in these contexts more than females; (3) The probability of carrying out the kill will depend on a variety of cost-benefit calculations such as the risk of being killed, anticipated reputational consequences, costs to kin, alternative sources available for replacing lost resources, and the viability of non-lethal alternative solutions such as migration. The higher the perceived costs of killing, the lower the odds that a man will attempt to enact the homicidal premeditation.

5.3.3. Killing for reputation management. The importance of reputation, honor, face-saving, and humiliation in male-male killings has been noted by a number of authors (e.g., Chagnon, 1988; Coon, 1971; Daly & Wilson, 1988; Ghiglieri, 1999; Wolfgang, 1958). Roughly 37% of one sample of 560 cases of murder were classified into the label of "Altercation of relatively trivial origin; insult, curse, jostling, etc." (Wolfgang, 1958; cited in Daly & Wilson, 1988, p. 125).

Daly and Wilson provide the most detailed evolutionary explanation of the importance of reputation: "the emphasis on ‘triviality’ obscures a still more important point. A seemingly minor affront is not merely a ‘stimulus’ to action, isolated in time and space. It must be understood within a larger context of reputations, face, relative social status, and enduring relationships. Men are known by their fellows as ‘the sort who can be pushed around’ or ‘the sort who won’t take any shit,’ as people whose word means action and people who are full of hot air, as guys whose girlfriends you can chat up with impunity or guys you don’t want to mess with . . . In most social milieus, a man’s reputation depends in part upon the maintenance of a credible threat of violence . . . one’s interests are likely to be violated by competitors unless those competitors are deterred. Effective deterrence is a matter of convincing our rivals that any attempt to advance their interests at our expense will lead to such severe penalties that the competitive gambit will end up a net loss which should never have been undertaken" (Daly & Wilson, 1988, p. 128). Ghiglieri concurs with Daly and Wilson, arguing that "men who win in this murderous game of face-saving do not simply win a single contest. By killing once, men gain or enhance a reputation of ferocity, which later helps them wrest resources from other men without conflict" (Ghiglieri, 1999, p. 143).

Homicide Design Theory is in alignment with these ideas, and proposes that human males have evolved features of rival homicide mechanisms designed specifically for status and reputation management. There are at least three distinct adaptive problems that are solved by killing in this context: (1) establishing a reputation that serves a deterrent function, preventing rival males from resource encroachment, rape of kin, or other forms of abuse; (2) establishing a reputation for ferocity that facilitates the submission of other men, and hence the ceding of their resources; (3) causing competitors to migrate, or at least give wide berth, thereby reducing the number of active competitors in the area.

Several specific predictions follow: (1) being humiliated publicly will catalyze homicidal fantasies; (2) the more public the humiliation and the more severe the reputational consequences, the more likely the homicidal fantasies; (3) homicidal fantasies as a consequence of public humiliation by a rival will more frequently occur in men than in women; (4) creating a reputation as a killer will have the effect of inducing others to cede resources and territory. As with other hypothesized evolved homicide mechanisms, translating a homicidal fantasy into a homicide attempt will often hinge on specific cognitive simulations, the magnitude of rage and humiliation suffered, the viability of alternative solutions to reputational damage, and specific cost-benefit calculations associated with the competing scenarios.

Most people will not act on their homicidal fantasies, especially in modern societies that contain police forces, a legal system that has largely usurped the role of retribution from individuals, and jail sentences that might act as deterrents. Even in traditional cultures lacking police forces, an individual may choose an alternative strategy for dealing with public humiliation, or might lack the physical prowess or coalitional backing to combat a rival who has delivered the insult. Chagnon, for example, notes the following case, about a man named Rerebawa, who has had an affair with his older brother’s wife. As Rerebawa was giving Chagnon a tour around the village, he passed the older brother, "grabbed him by the wrist, and dragged him to the ground: ‘This is the brother whose wife I screwed when he wasn’t around!’ A deadly insult, one that would usually provoke a bloody club fight among more valiant Yanomamo. The man did nothing. He slunk sheepishly back into his hammock, shamed, but relieved to have Rerebawa release his grip."

The relevant background is that several months earlier, when the older brother discovered the affair, he attacked Rerebawa with a club. But Rerebawa, being substantially stronger, beat his brother with the blunt side of an ax. The brother was so intimidated by the thrashing and threat of further violence that he kept away from the village for several days. The key point is that humiliation, insults, and other forms of reputational assault do not invariably result in homicide. But they frequently trigger homicidal thoughts, with all the attendant cost-benefit calculus, some of which lead to attempts at murder.

This homicide mechanism is proposed to be highly sex-linked, present in men but largely absent in women. The rationale for this sex linkage involves the more severe cascading costs to men than to women of reputational assaults. A woman’s mate value might be lowered as a result of public insult, but a humiliated man might lose entirely his ability to attract or hold onto a mate, given the importance women place on status in selecting a mate (Buss, 1994).

5.2.4. Killing to acquire resources. Those who hold valuable reproductive resources rarely relinquish them willingly. Precisely the opposite. The more valuable the resource, the more ferociously people guard it. This principle is most easily seen in the context of human mate guarding. The intensity of effort men devote to guarding their mates and repelling potential rivals has been shown empirically to be a function of the mate’s youth and physical attractiveness, both cues to reproductive value (Buss, 1988; Buss & Shackelford, 1997). The intensity of guarding as a function of resource value relevant to survival and reproduction can be expected to apply to a wide variety of items, including prime habitats, water sources, places to see without being seen, edible animals, precious plants, tools, weapons, and social status. The benefits to those who guard resources intensely come at a cost to those who lack these resources. Humans have evolved mechanisms for gaining access to reproductively relevant resources held by others.

There exist several potential solutions to this suite of adaptive problems. One could migrate to different places that are less saturated with human competitors. One could attempt to covertly steal the resources. One could issue threats in the hope that those holding resources will cede them without further violence. But an additional strategy of resource acquisition is to kill those holding the critical resource. According to Homicide Design Theory, intrasexual homicide evolved as one context-contingent strategy for gaining access to reproductively relevant resources possessed by others.

The conditions under which this strategy would be deployed hinge on a variety of contextual variables, including the intensity with which the rival, kin, and coalition members are guarding the resources; the formidability of the prospective killer and his kin and coalition; the viability of alternative methods of resource acquisition, including migration, stealing, and threats; and the prospective killer’s ability to successfully guard or hold onto the resources so acquired after the killing has occurred (e.g., vulnerability to incursion by others).

Specific predictions follow: (1) thoughts of killing will occur in contexts where an intrasexual rival is monopolizing a critical resource and hence preventing someone else from gaining access to it; (2) thoughts of killing intrasexual rivals will increase as a function of the value of the resource monopolized; (3) thoughts of killing will be more common in contexts where alternative routes to resource acquisition are not available (e.g., if migration, stealing, or bluffs are ineffective); (4) when resources are scarce, imposing the threat of pushing an individual below the threshold of survival, thoughts of killing intrasexual rivals will increase.

5.2.5. Killing to solve the problem of mate encroachment. From our review of infanticide, it is apparent that for many species there have been substantial fitness benefits of killing the infants of intrasexual rivals. Killing a rival’s infant creates a temporally extended cascade fitness benefits for murderers, including eliminating future competition for one’s own children. Homicide Design Theory proposes that infanticide is held in check primarily by the coevolution of anti-infanticide mechanisms, both in infants and in their parents (Duntley & Buss, in prep., a). But if the selective benefits of infanticide are apparent, wouldn’t it be more beneficial to kill intrasexual rivals directly prior to their production of any or additional offspring?

Since sexual access to fertile females defines the limiting resource for males in their production of offspring, rivalry over sexual access to mates becomes a key adaptive problem. This can be partitioned into at least two distinct adaptive problems–mate encroachment and mate acquisition. The problem of mate encroachment occurs primarily in long-term mating contexts. When a person has expended effort to acquire a long-term mate, and a rival attempts to lure that mate away (either temporarily or permanently), the adaptive problem becomes especially severe. For a man, a sexual encroachment on his mate jeopardizes his certainty in paternity, while an encroachment causing a permanent mate defection causes the total loss of access to the woman’s reproductive value. For a woman, a sexual encroachment by a rival on her mate per se causes relatively little damage, whereas an encroachment causing a permanent mate defection can cause the total loss of access to her husband’s resources (Buss et al., 1992). For a variety of reasons, these losses are generally more reproductively costly for males than for females, and hence our theory predicts sex differences in evolution of homicide as a solution to the problem of mate encroachment.

Humans appear to have evolved a host of adaptive solutions to the problem of mate encroachment, ranging from vigilance to violence (see Buss, 2000). One evolved solution, according to Homicide Design Theory, is killing rivals who threaten to encroach, or who succeed in encroaching, upon one’s mate. The use of murder as a solution to the mate encroachment problem is hypothesized to depend on a variety of circumstances. These include one’s personal formidability, formidability of one’s kin and coalition, the formidability of the rival’s kin and coalition, the reproductive value of one’s mate, opportunities for mate replacement, extent of reputational damage caused by the encroachment, and others.

Successfully killing an intrasexual rival would have produced many benefits tributary to fitness. These include, in principle, maintaining access to one’s mate, preserving social status and reputation, eliminating the rival’s future encroachment on one’s mates, curtailing the rival’s reproduction, and eliminating a source of competition for one’s children in the next generation.

Intrasexual rivalry homicide strategies are hypothesized to have the following design features: (1) it will be triggered by attempts by another male to lure, attract, or "hit on" one’s long-term more often than one’s short-term mate; (2) the more severe the encroachment, the higher the likelihood of homicidal premeditations; (3) sexual encroachment will be more likely to trigger men’s than women’s intrasexual rivalry homicidal thoughts, since paternity uncertainty, but not maternity certainty, is threatened by sexual infidelity; (4) encroachment on a mate of higher reproductive value will evoke more homicidal thoughts than encroachment on a mate of lower reproductive value (e.g., one who is post-menopausal); (5) jealousy will be a primary emotion triggering, and accompanying intrasexual homicidal thoughts due to mate encroachment. These are all testable predictions based on our theory of an evolved rival homicide psychology–predictions not generated by any previous theory of homicide.

5.2.6. Killing to solve the problem of mate acquisition. Males compete to attract and retain reproductively valuable females (Darwin, 1871; Symons, 1979; Trivers, 1972). At the most abstract level, there are two general methods of successful competition–(1) besting a rival through mate attraction tactics designed to embody or fulfill what the potential mate desires, and (2) inflicting costs on a rival that eliminates him from successful contention (Buss & Dedden, 1990). The tactics for inflicting costs on rival males are highly variable, and in humans can range from verbal derogation tactics designed to damage reputation to inflicting physical damage through violence. We propose that one strategy of mate acquisition within men’s evolved arsenal is the outright killing of the intrasexual rival. This hypothesis was anticipated by Charles Darwin in his description of the process of sexual selection. He noted that "the sexual struggle . . . is between individuals of the same sex, generally the males, in order to drive away or kill their rivals . . . " (Darwin, 1871, emphasis added).

Killing as a mate acquisition strategy is more likely to occur in the context of between-group warfare than within-group intrasexual rivalry (see below). Nonetheless, the problem of mate acquisition can also be solved by killing intrasexual rivals within one’s group. This might happen in several contexts: (1) when a married woman conspires with her lover to kill her husband; (2) when a man’s status within the group is sufficiently enhanced by killing a rival, resulting in increased success at attracting potential mates (Chagnon, 1988); and (3) when lethal violence by a man against a rival serves as a deterrent to his mate or mates against defecting.

Killing an intrasexual rival within one’s group, of course, can carry considerable costs, such as the risk of injury death as a result of the operation of the intended victim’s anti-homicide mechanisms. Killing a within-group rival also could lower the formidability of one’s coalition for the purposes of defense against, or attack on, outside coalitions. And like many other forms of killing, retribution from kin, again stemming from coevolved anti-homicide mechanisms, can lower the average benefits associated with killing a rival in the service of mate acquisition. For these reasons, we suggest that intrasexual rivalry mechanisms have evolved to assess these costs, and decision rules have evolved to attempt a kill only when these costs are unlikely to be incurred.

5.2.7. Killing to reduce a rival’s strategic interference. At one level of abstraction, the functions of rival homicide can be reduced to acquiring resources or defending against their incursion. There are many more forms of incursion, or "strategic interference," than those outlined above. First, a man might interfere with another man by blocking his ascension within a status hierarchy. Stunted status ascension could have the effect of precluding access to mates and other key resources that would otherwise have been obtained. Second, a rival could inflict strategic interference by causing him to be ostracized from a desired coalition. Third, a rival could usurp dyadic alliances. Male intrasexual rivalry mechanisms are hypothesized to be sensitive to these key sources of strategic interference, triggering homicidal premeditations, leading to actual killings in some circumstances.

5.3. Spousal Homicide

Within the United States, roughly 1,500 women are murdered each year by their mates or former mates, a category that includes husbands and boyfriends and ex-husbands and ex-boyfriends (Crowell & Burgess, 1996). This represents approximately 30% of all women who are killed, a stable percentage from year to year, as contrasted with only 3% of male-victim homicides being committed by wives or girlfriends (Ghiglieri, 1999). Canadians show a similar rate of male-perpetrated mate murder, despite the relative absence of handguns, but Canadian women murder their mates only half as often as American women do (Ghiglieri, 1999). Why do men sometimes kill their mates? Why are nearly a third of all women victims of homicide killed by the ones who profess to love them?

The most frequently endorsed explanation is that proposed by Daly and Wilson (1988), who argue that mate killings are "slips," or maladaptive byproducts of evolved male mechanisms of coercive control: "Men . . . strive to control women . . . women struggle to resist coercion and to maintain their choices. There is brinksmanship and risk of disaster in any such contest, and homicides by spouses of either sex may be considered slips in this dangerous game" (Daly & Wilson, 1988, p. 205, italics added). Recently, they elaborate: "We propose that such homicides [killing estranged or unfaithful wives] are the dysfunctionally extreme byproducts of those same violent inclinations whose much more typical effects are successful deterrence and coercion" (Daly & Wilson, 1998, p. 449, italics added).

Homicide Design Theory proposes instead that men have evolved a specific mate-killing mechanism designed to (1) inflict heavy costs on intrasexual rivals by depriving them of sexual access to a reproductively valuable mate whom the killer has lost irrevocably; and (2) staunch the fitness costs associated with cuckoldry. The costs to a man of his mate’s infidelity and defection are multiple and severe. They include losing access to the woman’s reproductive value, risking the devotion of personal resources to a rival’s offspring, having the mate devote all of her parental efforts to a rival’s offspring, and incurring reputational damage. The reputational damage alone could cripple a man’s ability to attract future mates or hold on to other current mates, impair his future ascension in status hierarchies, and increase his risk of being exploited injured, or killed.

The fitness costs a man incurs as a result of his mate’s infidelity or desertion do not come merely from the direct costs associated with the loss, such as the magnitude of future effort that will have to be expended to attract and retain a replacement mate. They also come from the fact that one man’s loss of a desirable mate constitutes an intrasexual rival’s gain. Since selection acts through the relative replicative success of competing designs, this "double-whammy," we hypothesize, constituted a powerful selection pressure that forged a male psychology of context-contingent mate killing. Below we consider this selective logic in greater detail, and offer specific predictions that contrast our theory of mate killing with the "slip" or "epiphenomenal byproduct" theory.

5.3.1. Paternity uncertainty. A woman’s sexual infidelity creates uncertainty in her husband’s paternity of children. The risk of genetic cuckoldry is not merely hypothetical. Estimates from modern populations based on blood and DNA fingerprinting studies conducted over the past 30 years estimate the rates to be between 9% and 13% (Baker & Bellis, 1995). Mistaken paternity historically would have inflicted multiple fitness costs on a male. First, all of the effort the man devoted to selecting, courting, and attracting his spouse goes down the fitness drain. Second, all the effort he has expended to date in maintaining and retaining her, including the vigilance costs, direct guarding, provisioning, and fending off rivals similarly become wasted. Third, the cuckolded man suffers the opportunity costs of foregone mating opportunities, both short-term and long-term, as a result of investing in an unfaithful wife. Fourth, he risks his parental effort becoming channeled to a rival’s offspring in the mistaken belief that they are his own. Fifth, his mate’s parental effort becomes invested in a rival’s offspring rather than his own. Sixth, the new genetically unrelated child becomes a half-sibling rather than full sibling of whatever existing children he has or future children he will have, which results in his children experiencing greater competition and conflict than would otherwise be the case.

Seventh, the cuckolded man can suffer large reputational damage, hindering his current position and future ascension in the status hierarchy and impairing his ability to attract future mates. Men will be more likely to scoff at him. Women will assume that he lacks the ability to prevent other men from encroaching, and so lower their perceptions of his value as mate. An inability to prevent rival encroachment, for example, may signal an inability to protect the woman physically and lower his social status in the eyes of other men as well as women--key components of men’s mate value. Adding all these factors together results in formidable fitness costs associated with being cuckolded.

The hypothesis that mate killing has evolved as one among several solutions to the cuckoldry problem confronts two obvious challenges. First, killing a mate does not solve the problem of sunk costs. The initial courting costs a man incurs, for example, are not recouped by killing an unfaithful mate. Nonetheless, killing her can curtail some costs, such reputational damage. More importantly, it can deprive intrasexual rivals of access to a her reproductive value and simultaneously terminate a rival’s prenatal child. A second challenge is more formidable: If primary fitness benefits to killing an unfaithful mate include depriving rivals of access to a reproductively valuable resource and terminating prenatal offspring, why don’t men just go out and kill the mates or children of their rivals?

Our hypothesis is that the cost-benefit calculus of killing a rival’s mate and killing one’s own mate are likely to have been dramatically different. First, any woman who has not rejected a man as her mate is a potential future mate. Just because a rival currently has sexual access to her does not necessarily mean that the rival will maintain access to her. Injury, change in status, or death of the rival all could result in the woman becoming available to re-mate. Second, men fiercely guard their mates from rivals and inflict grave costs on those who attempt to destroy their reproductively valuable resources. This leads to a specific prediction: A man will be more likely to kill a mate who has cuckolded him or defected from the relationship when he believes that he has lost her irrevocably, but before the woman has firmly established a new long-term mateship with another man who would be willing to protect her. Killing under these circumstances would have been considerably less costly than attempting to kill a woman who is already in, or who has already established, a long-term mateship with another man.

Much evidence already exists that selection has fashioned male defenses designed to guard against paternity uncertainty. These include psychological mechanisms such as male sexual jealousy (Buss, 2000; Buss et al., 1992; Daly, Wilson, & Weghorst, 1982; Symons, 1979), behavioral tactics such as mate guarding and retention (Buss, 1988; Buss & Shackelford, 1997), and physiological mechanisms such as increased sperm volume, increased sperm insemination in contexts in which cuckoldry might have occurred, and possibly different sperm morphs designed to combat rival male sperm (Baker & Bellis, 1995). According to Homicide Design Theory, men have also evolved mate killing as one extreme strategy for solving the cuckoldry problem, and specifically for inflicting costs on rival males by effectively killing their prenatal children.

Killing a mate who has sexually defected carries several fitness benefits. It would prevent his existing or future children from having to compete with genetically unrelated children or partially related children. In some social circumstances, the killer reduces the reputational damage as a result of the social derision incurred apparently universally by cuckolds, and hence maintains his ability to attract a replacement mate. But most important, killing an unfaithful mate inflicts costs on the man’s intrasexual rivals. It deprives the rival’s children of what would have been her maternal effort, thus seriously hampering the rival’s children’s chances of surviving and thriving. It deprives the rival of current sexual access as well access to her residual reproductive value. It inflicts costs on the rival by causing him to have wasted all the time and effort he expended to lure her into an affair. And perhaps most of all, killing the unfaithful mate, in some cases, results in literally killing the rival’s prenatal children–essentially a form of rival infanticide.

Killing an unfaithful mate, of course, can be an extremely costly solution in many circumstances. It could deprive his existing children of maternal investment. It could deprive the killer of future sexual access he might have achieved with his partner. It could evoke retribution from her kin. According to our theory of mate homicide, human evolved psychology contains design features that assess and weigh these costs. And in most circumstances, the costs of killing a mate would have far outweighed the benefits, especially when killing is compared to other possible solutions to the adaptive problem of cuckoldry. In fact, most men worldwide do not kill mates who are discovered to have been sexually unfaithful. Our theory proposes that mate killing is a low base-rate solution, one strategic outcome of decision rules that get activated only in highly particular circumstances--where the costs of killing are evaluated to be comparatively low and the benefits high, relative to alternative courses of action.

Our theory offers several specific predictions that contrast sharply with those of the "slip" or "epiphenomenal byproduct" theory of mate killing:

(1) most killings of unfaithful wives will be intentional and designed, not accidents or "slips";

(2) most men who kill unfaithful wives will not be evaluated to be "psychotic" or legally insane;

(3) men whose mates are discovered to be sexually unfaithful will have thoughts, fantasies, or premeditations of killing them, not merely beating or coercing them (the "byproduct" theory cannot explain why the terminus of a mate killing fantasy is the spouse’s death, not her control);

(4) sexual infidelity will be a sex-linked trigger of mate homicide thoughts, occurring more frequently in men than in women;

(5) men who have no existing children with an unfaithful spouse will be more likely to experience homicidal premeditations than men who do have existing children with her (due to the increased fitness costs of depriving his existing children of her maternal investments);

(6) as the likelihood increases that a man’s existing children with an unfaithful spouse can survive without her, the likelihood that men will experience homicidal premeditations will increase;

(7) men whose wives have existing children sired by another man will be more likely to kill an unfaithful wife, since killing her thus inflicts greater costs on intrasexual rivals;

(8) men residing in cultures or subcultures where cuckolds suffer greater reputational damage will be more likely to think about, and carry out, killing a sexually errant mate;

(9) size and proximity of the wife’s kin group will affect decision rules to kill an unfaithful mate, with increases in the likelihood of kin retribution deterring would-be mate killers due to the heavy costs of killing under these circumstances (this prediction also holds for many of the homicide hypotheses, since the number of kin linked to the victim will often enter into the cost-benefit calculus).

All these are testable predictions. The empirical outcome of would adjudicate the two central competing theories of mate killing. If the majority of these empirical predictions are confirmed, the balance of evidence would shift to Homicide Design Theory and away from the "killing as incidental byproduct" theory.

5.3.2. Mate defection. Many costs associated with being cuckolded apply to being abandoned. For several reasons, the outright defection of a mate from a marriage could inflict a greater fitness cost on the husband than a sexual infidelity. If a woman remains with a man she has cuckolded, her husband may continue to gain access to her reproductive value, whereas with a total defection, he loses that access entirely. Whereas a wife’s infidelity gives a husband’s intrasexual rival partial access to her current reproductive value, abandonment may grant the rival total access to her residual reproductive value. For these reasons, we expect that thoughts of mate homicide would be even more strongly evoked in men who are abandoned than those who are cuckolded.

Nonetheless, not all selective forces suggest that mate homicide will be more common after abandonment than infidelity. Men whose partners abandon them do not incur at least one cost that may be incurred as a result of cuckoldry–the mistaken diversion of the man’s own parental investment toward a rival’s offspring. A cuckolded husband, in contrast, could devote years or decades of parental effort toward genetically unrelated children, a possibly catastrophic cost the abandoned man does not incur. Furthermore, being abandoned, as opposed to being cuckolded, frees up time and resources to pursue alternative mates, albeit with possible reputational damage as a result of public knowledge of the abandonment.

For some costs, of course, there may be no a priori rationale for determining whether being cuckolded or being abandoned is worse. In both cases, men suffer reputational damage. Being abandoned can create harmful social inferences--that the deserted partner is lower in mate value than his partner, that he has some fundamental flaw that became revealed over the course of the mateship, or that he has declined in mate value sufficiently for her to trade up incrementally on the mating market. Being cuckolded can create these damaging social inferences as well, with perhaps the added mockery that goes along with the judgment that the man is a fool for allowing a rival into his wife’s naked embrace.

We hypothesize that there are two primary adaptive problems that are solved by killing a wife who has permanently defected. First and most important, killing a wife who has permanently defected can inflict a heavy costs on an intrasexual rival. It deprives a rival entirely of access to a mate whom the man failed to retain. He also effectively aborts any of the rival’s prenatal children his former mate may be carrying. The fitness benefits of killing a mate who has defected, of course, would have been substantially greater in the small group living contexts in which humans evolved than in modern urban contexts, since the number of "effective" rivals of the killer would have been small.

Several other ancillary costs are inflicted on rivals by murdering a wife who has defected. A defecting wife can reveal private and potentially damaging information to others, including hidden flaws, hidden weaknesses or strengths, and specific vulnerabilities. She can reveal specialized knowledge the abandoned husband has accrued, including secret plans, tactics, and strategies harbored by the husband, as well as scandalous or shameful information about the man or his kin. All this private information could be revealed to a rival, giving him a potentially large strategic advantage in intrasexual competition. Killing the wife essentially deprives intrasexual rivals of access to this damaging information, above and beyond depriving the rival of access to the woman’s reproductive value and parental effort.

The second key adaptive problem solved is curtailing the reputational damage that occurs as a consequence of being abandoned. One of the key concerns of O.J. Simpson when his wife, Nicole Brown Simpson left him, was the effect on his reputation. According to some evidence, he killed her when it became clear that she was not going to come back to him combined with evidence suggesting that she was having sex with another man. One story does not a case make, of course, but Simpson is not unique in his concern with the reputational damage linked with being abandoned. By killing a defecting woman, in some contexts, a man can restore his reputation as someone who cannot be trifled with, and for whom any kind of social defection comes at a steep price, deterring friends, coalition members, other wives (if he’s polygynous), and future wives from defecting.

Killing a wife who has defected should be more likely to occur in certain contexts and to be carried out by specific men. When the defecting woman has not yet formed a new mateship, for example, the costs to the killer are lower since she may lack the physical protection that might be offered by a new mate (her kin, of course, can also deter a would-be killer). Moreover, the higher the mate value of the defecting woman, the greater the costs that will be inflicted on rivals as a consequence of killing her. Hence, women higher in mate value (e.g., younger, more physically attractive) should be more vulnerable to getting killed than women lower in mate value (e.g., older, less physically attractive). One countervailing force, however, is the evolution of anti-homicide mechanisms. The kin of a young woman, for example, should show special alacrity in deterring potentially murderous mates, due to her reproductive value to the each member of the extended kin network.

Empirically, reproductive-aged women are indeed more likely to be killed than post-reproductive aged women (e.g., Daly & Wilson, 1988; Shackelford, in press a, b). Proponents of the "byproduct" theory offers this explanation: "we propose that such homicides are the dysfunctionally extreme products of violent inclinations whose lesser manifestations are effective in coercion . . . Uxoricide risk is indeed maximal for the youngest . . . This finding may strike the reader as evidence against the proposition that men ‘value’ young wives maximally, but the paradox disappears when one views uxoricides as the dysfunctional extremes of ‘normal,’ nonlethal coercive violence" (Wilson, Daly, & Scheib, 1997, p. 448-450). "This pattern [men killing young wives] may seem to belie the proposition that male minds place high ‘value’ on young wives, but again, as with the estranged husband who pursues and kills a woman he can’t abide losing, violent inclinations seem best understood as coercive tools for controlling wives about whom they feel proprietary–and the lethality is a rare and dysfunctional outcome of the most extreme feelings" (Wilson & Daly, 1998).

We suggest a simpler explanation--men have evolved adaptations that are designed to incite murder of mates in particular circumstances, especially those mates who are high in reproductive value. The greater the reproductive value of the defecting wife, the greater the reproductive benefits flow to an intrasexual rival, and hence the greater fitness benefits that accrue to the killer. Killing a young wife who has irrevocably defected, simply put, confers more fitness benefits on the killer than does killing an older wife, since the cost inflicted on the relevant intrasexual rivals are commensurately greater.

Human beings are notoriously difficult to kill. According to Homicide Design Theory, humans have evolved a large collection of anti-homicide mechanisms designed to avoid getting killed (Duntley & Buss, in prep., a). Although the "accidental death" argument is sometimes used by killers as a defense ("I just wanted to coerce her, I didn’t mean to kill her"), and some mate killings may be byproducts of mechanisms designed for coercive control, this argument strains plausibility when faced with the evidence of premeditated, well-planned, intricately designed uxoricides. Even in so-called impulse killings that are not premeditated, one must explain why the man chose to use "deadly force" rather than non-lethal coercive force. The evolved mate killing hypothesis provides a more parsimonious explanation for the available data, including the greater rates of killing young wives who have defected, without resorting to special "dysfunctional" explanation for a very predictable and patterned form of homicide. Our theory also offers a suite of novel testable predictions that, if confirmed, could not be explained by the byproduct theory of mate killing.

None of these arguments, of course, deny the plausibility of the hypothesis that men have evolved mechanisms, including the use of non-lethal violence, that function to deter mates from infidelity or defection. Indeed, there is abundant evidence for precisely such mechanisms (e.g., Buss, 2000; Daly, Wilson, & Weghorst, 1982). Rather, our argument is that the mechanisms that result in mate homicide differ from those involved in mate coercion, control, and deterrence.

5.3.3. Which rivals suffer from a man killing a mate who has defected? All of the multiple costs inflicted on rivals by killing a mate who has deserted are not incurred by all intrasexual competitors equally. They are inflicted most heavily on the rivals with whom the killer is in most direct competition. To understand the logic of this hypothesis, the logic of assortative mating for mate value must be introduced.

As a general rule, men and women assortatively mate on the basis of overall mate value (Buss, 1994; Frank, 1988; Symons, 1979; and others). The "10’s" tend to pair off with other "10s," the "8’s" with other "8’s," and "4’s" with other "4’s." There are discrepancies in mate value, of course, although these are linked with a higher likelihood of infidelity and defection (see Buss, 2000, for a review of this evidence). Nonetheless, strong positive assortment is the rule. As a consequence, in intrasexual mate competition, men are primarily competing with men who are closest to them in mate value. In the modern world, for example, the burger flipper and gas station attendant (low status, low resources, and hence low in mate-value) are not in direct competition with successful movie stars or rock stars for mating with attractive models. They compete, instead, with those in their local mating pool who are most similar to them in mate value.

Selection, according to this logic, has designed psychological mechanisms to inflict costs on those rivals with whom one is in most direct competition. The evolved mechanisms are designed for "local skirmishing," not for competing "in general" with all conspecifics (see Frank, 1985, for similar arguments in the context of status and hierarchy negotiation). These phenomena are witnessed in many domains. Finding out that a member of one’s department who is similar in rank and reputation has received a higher pay raise, for example, may cause more psychological anguish than finding out that someone in another department or another university has received a higher pay raise.

As a consequence, calculating the fitness benefits that would have accrued to a killer from murdering a mate who has defected does not involve dividing the benefits over the entire population of other males. The benefits that accrue stem from inflicting costs on the intrasexual rivals with whom the killer is more directly and immediately in competition–those in the local pool who are most similar in mate value. These more concentrated benefits essentially increase the selection pressure for the evolution of mate killing mechanisms.

5.4. When Women Kill Their Mates

According to Homicide Design Theory proposes, women have evolved mate killing mechanisms that differ profoundly in design from men’s mate killing mechanisms. To start with, the fitness benefits to women of killing a partner who is sexually unfaithful would historically have been relatively trivial, even without considering costs. A sexually unfaithful husband does not jeopardize a woman’s certainty that she’s the mother of her children. Nor does a sexually unfaithful husband cause the potential misdirection of a woman’s own parental efforts toward children who are genetically not her own. Sexual infidelity per se, therefore, should be less likely to trigger homicidal thoughts or deeds in women.

Nonetheless, there are some delimited circumstances in which it might historically have been advantageous for a woman to kill a partner: (1) when the mate is physically abusing her to an extent that it threatens her survival, and as a consequence, her children’s well-being and survival (i.e., anti-homicide mechanism); or (2) when the mate is inflicting severe costs on her children, such as through physical abuse or sexual abuse.

In these delimited contexts, killing a mate could solve key survival and reproductive problems by curtailing the massive costs being inflicted on the woman or her children by the mate. Killing a mate in some circumstances also could inflict a cost on female intrasexual rivals by depriving them of access to the husband and his investments, although we expect that selection has been weak for killing under these circumstances. Reproductive variance is lower among women than among men, so the benefits of depriving a rival female of a particular male are commensurately lower. Because sperm are cheap and eggs are expensive, men are rarely the limiting reproductive resource for women that women are for men.

In sum, selection pressure is hypothesized to have been considerably weaker on women than on men for the evolution of a psychology of mate killing. Men are proposed to have evolved specific mate killing mechanisms that are absent in women. Women, in turn, are hypothesized to have evolved mechanisms that lead to killing in the very delimited contexts in which their lives and the lives and well-being of their children are in serious danger from a mate.

5.5. Evidence Bearing on Hypothesized Mate Killing Mechanisms

A variety of sources of evidence can be used to test predictions from the theory of evolved mate killing mechanisms and to differentiate it from the predictions generated by the Daly-Wilson byproduct theory of mate killing. Before examining this evidence, we note that that some (we think a relatively small) proportion of mate homicides may result from accidents, whereas others (we think the majority) are caused by evolved mechanisms designed for murder. If the competing theories are to have any conceptual and explanatory power, however, they must generate specific predictions that do not flow from the other. Below we outline several sources of existing evidence that distinguish our theory from the byproduct theory.

One of the key differences pertains to the motivational processes underlying violence and homicide proposed by the two theories. The byproduct theory clearly states that "lethal and nonlethal wife assault are motivationally similar" (Wilson & Daly, 1998, p. 300); both are presumed to be motivated by the goal of coercion. In sharp contrast, according to Homicide Design Theory, mate killing and the nonlethal assault of a mate are motivationally distinct. We concur with Wilson and Daly that the primary motivation for nonlethal spousal violence is coercive control. But we differ in arguing that the central motive for lethal mate killing is not control or coercion, but rather the literal death of the partner. Below we summarize evidence that bears on these fundamentally different theoretical positions.

If men kill their mates accidentally as a byproduct of using violence as a means of coercive control, as predicted by the byproduct theory, then the majority of mate killings should occur in mateships that show a substantial amount of male violence. That is, among those relationships in which a wife or partner gets killed, there should be prior evidence that the males have been using violence as a means of coercive control that somehow has escalated. A large-scale study appears to refute this prediction from the byproduct theory. A study of 15,537 cases of domestic battery, which included both spousal beatings and spousal homicide, found that only one in 33 spousal homicides showed a prior history of domestic violence (Staff, 1990). The study also found that, of the 110 cases in which one partner made threats to kill the other partner, not a single one later led to death. As Ghiglieri concludes, "men who decide to kill their wives simply go ahead and kill them without giving as many advanced signals as one might expect" (Ghiglieri, 1999, p. 153). If the byproduct theory were correct, then mate homicide should typically occur in the context of a history of marital violence. It does not.

A second source of evidence comes from a study of Canadian wife killings conducted by Daly et al. (1997). They found that when the wife had one or more co-residing children who were sired by a previous partner, and the present partner had not sired any children by the current wife, the rate of wife killing was more than ten times that which occurred when co-residing children were entirely sired by the present partner. When the household contains at least one child of each (one from current and one from previous partner), the rates of uxoricide are only slightly higher than when all co-residing children are from the present partner. Daly and Wilson do not provide an explanation for this important and dramatic set of findings. Instead they conclude that "marital violence may be elevated in stepfamilies," a domain-general claim that, while true, fails to explain the dramatic context effects within different types of stepfamilies(Wilson & Daly, 1998, p. 226).

More specifically, the Daly-Wilson explanations for wife beatings and wife killings are identical–male sexual proprietariness that uses violence for coercive control, which sometimes reaches "dysfunctional" extremes of uxoricide: " . . . most men who coerce, pursue, and threaten women do not go so far as to kill them, and those who do may be considered dysfunctional over-reactors in a game of brinkmanship" (Wilson & Daly, 1998, p. 302. The empirical findings, however, are not compatible with a single explanation for both phenomena. The rates of women seeking shelter from assaulting husbands is virtually identical if they have children solely from a previous partner and if they have at least one child each from the current and previous partner, and both are equally elevated compared with women who have children only with the present partner (see Wilson & Daly, 1998, Figure 8.5, p. 227). In sharp contrast, the presence of a single child by the present partner, even when the wife has one or more from a previous partner, dramatically reduces the uxoricide risk compared to when the only children present are from a previous partner. This pattern of findings supports the notion that non-lethal violence against a mate and uxoricide are fundamentally different phenomena. The data contradict the idea that a single, general theory of coercive control can simultaneously explain spouse beatings and spouse killings. By killing a wife who has produced children only by a previous partner, a man inflicts costs primarily on his intrasexual rivals. In contrast, when she has at least one child sired by him, a would-be wife killer would be inflicting a heavy cost on his own fitness as a result of killing her, since he would be depriving his own child of the woman’s maternal efforts.

This result was predicted by our theory of evolved mate killing mechanisms prior to our discovery of this finding by Daly and Wilson (Buss & Duntley, 1998). Killing a wife who has existing children sired by an intrasexual rival, yet none sired by the current mate, inflicts a great cost on previous and future intrasexual rivals. It deprives the prior rival of the maternal effort that she would otherwise direct toward his children. And it deprives the future rival of a valuable reproductive resource. In summary, existing evidence suggests that a single theory of coercive control cannot explain both non-lethal spousal violence and spousal homicide.

5.6. Warfare and Genocide

Warfare may be defined as "a conflict between social groups that is resolved by individuals on one or both sides killing those on the opposite side" (Ghiglieri, 1999, p. 160-161). Homicide Design Theory includes detailed hypotheses about the evolution of killing in the context of warfare. A number of other evolutionary theories of warfare have been proposed, notably by Alexander (1979), Chagnon (1988), Hamilton (1975), Ghiglieri (1999), Tooby & Cosmides (1988, 1994), and Wrangham (Wrangham, 1999a, b; Wrangham & Peterson, 1996). Our theory coordinates with many of the components of these theories, but a detailed analysis of these theories and our own theory of warfare is beyond the scope of this paper (see Duntley and Buss, in prep., b).

Here we merely mention some fundamental premises of our theory of war, since many homicides throughout human history have been committed in the context of war. Keeley (1996) estimated that 0.5% death rate per year occurred from raiding alone, based on 21 societies for which data on death were available. Combined with the historical record, paleontological evidence, archeological evidence, and ethnographic evidence, it is reasonable to hypothesize that warfare has been a powerful and pervasive selective force over human evolutionary history. Indeed, Wrangham (1999b) hypothesizes that lethal raiding evolved in both humans and chimpanzees, sharing a common evolutionary origin around five to six million years ago. Our theory, like those of Alexander, Wrangham, Ghiglieri, and others, also assumes that group against group killing has been a recurrent feature of human evolutionary history that has selected for features of psychology that are specific to warfare. The theory rests on five key premises.

5.6.1. Premise 1: Men have evolved warfare mechanisms that are lacking in women. Men, but not women, have benefited in some contexts from forming coalitions with the express purpose of killing members of other groups. Recurrent benefits of coalitional killing have resulted in the evolution of a warfare-specific psychology in men that is lacking in women (Tooby & Cosmides, 1988, 1994). As Tooby and Cosmides note, "females have more to lose, and less to gain" from engaging in coalitional warfare (Tooby & Cosmides, 1988, p. 8).

5.6.2. Premise 2: Primary fitness benefits of warfare are multiple, including the acquisition of new mates, resources, and territory, as well as eliminating key competitors. As several prior authors have suggested (e.g., Dennen, 1995; Tooby & Cosmides, 1988, 1994; Wrangham, 1999a, b), the primary fitness benefits that recurrently accrued from warfare included the (1) acquisition of new women as mates, (2) immediate short-term sexual access to fertile women, possibly through rape, (3) acquisition of resource-rich territory, (4) cultivating a reputation that deters exploitation or encroachment, and (5) eliminating sources of intrasexual competition for these and other scarce resources. Wrangham (1999a) notes additional benefits of lethal raiding, including (6) inspiring fear in neighboring groups, perhaps one mechanism by which deterrence occurs, (7) protection from territorial incursion, and (8) expanding one’s borders and hence "safe zones." To this list one may add (9) securing honor, status, or glory within one’s own group, which could lead to a cascade of reproductive benefits and (10) inflicting costs on rival groups in revenge for having sustained costs. Wrangham (1999a) makes the important point that warriors need not be required to identify the specific benefits that they might accrue through warfare for the evolution and maintenance of warfare mechanisms (the same is obviously true of all of the evolved homicide design features proposed by Homicide Design Theory). It is only necessary that the one or several of these benefits flow to warriors, and that the benefits of this strategy, on average, outweigh the costs over the relevant instances.

5.6.3. Premise 3: Psychological design features for warfare. The psychological design features of male warfare psychology include:

(1) strategies for forming cooperative warfare coalitions (Tooby & Cosmides, 1988; Alexander, 1979, 1987);

(2) methods for increasing coalitional number, since numerical superiority is an excellent predictor of success (cf. Wrangham, 1999a, b);

(3) preferential selection of coalition partners depending on specific characteristics (e.g., physical formidability, bravery)(DeKay, Buss, & Stone, 1998);

(4) procedures to create and enforce coalitional solidarity (e.g., status rewards for bravery and sacrifice for the group; reputational damage from cowardice);

(5) procedures to minimize within-coalition conflict (e.g., eliminating sources of sexual conflict and jealousy);

(6) procedures for enforcing individual compliance with coalitional contract, such as inducing obedience and conformity (Dennen, 1995) and punishing cheaters (Tooby & Cosmides, 1988);

(7) self-assessments, other assessments, and comparative assessments of group size and formidability, with sensitivity to numerical superiority (Wrangham, 1999a, b);

(8) motivational mechanisms to induce attack (e.g., working group into an emotional frenzy); and

(9) ethnocentrism, xenophobia, and pseudospeciation that lead to derogation, demonization, diabolization of opposing group (e.g., that they are sub-human animals) (Dennen, 1995).

5.6.4. Premise 4: Evolution of anti-homicide mechanisms to combat coalitional warfare from other groups. Being a member of the group that becomes the victim of a coalitional attack would have inflicted severe fitness costs. This created a selective force that led to the evolution of anti-homicide warfare mechanisms. Indeed, Alexander (1979, 1987) has proposed that the primary function of group formation in humans, in terms of their fitness significance for the individuals within them, was protection from other aggressive human groups–a prime candidate for an anti-homicide mechanism. He notes that "in no other species do social groups have as their main jeopardy other social groups of the same species" (Alexander, 1987).

In addition to coalition formation as an important anti-homicide device, other hypothesized anti-homicide mechanisms to combat other groups include:

(1) heightened psychological sensitivity and vigilance to the contexts in which one’s group is vulnerable to attack from other groups (e.g., being numerically weaker);

(2) adaptive cognitive "errors" that produce overestimates of the likelihood of enemy attack (see earlier discussion of Error Management Theory);

(3) specialized fear or paranoia about being attacked by a hostile group;

(4) preemptive protective strategies, such as migration, ceding resources;

(5) promoting strategic alliances through trade or intergroup marriage (Wrangham, 1999b);

(6) coordinating surprise attacks to minimize costs of being victimized by opposing coalition;

(7) reputation management as non-exploitable (e.g., willing to suffer large costs in order to inflict damage on an attacking coalition);

(8) fleeing in response to immediate onslaught; and

(9) a psychology of vengeance that functions simultaneously to inflict costs on others and to deter rival groups from future attacks.

Warfare can simultaneously be a manifestation of homicide and anti-homicide mechanisms. When the chiefs of headhunting tribes of Borneo were interviewed about why they continued on "this senseless practice of going on the warpath, the best reason he can give is that unless he does so his neighbors will not respect him and his people, and will fall upon them and exterminate them" (McDougall, 1915). Coevolved anti-warfare mechanisms, according to Homicide Design Theory, led to the further coevolution of warfare mechanisms designed to circumvent the anti-warfare psychology, as follows.

5.6.5. Premise 5: The co-evolution of warfare psychology in response to anti-warfare mechanisms. As anti-warfare psychology evolved, it created further selection for the evolution of increasingly context-specific warfare psychology. These design features include: Surprise raids to circumvent anti-warfare mechanisms; exaggerated displays of threat and ferocity to extort resources without having to incur the risks of going to war; and heightened vigilance to contexts in which potential victim might engage in a preemptive strike.

The theory that humans have evolved elaborate suites of warfare and anti-warfare mechanisms obviously does not imply that warfare is inevitable or occurs rigidly or invariantly across contexts and cultures. The engagement of warfare mechanisms, like those of other hypothesized evolved homicide mechanisms, are predicted to be highly context-dependent. Identifying the specific design features and contexts in which they are activated remains the most viable and hopeful means of reducing their engagement. Interested readers are referred to Duntley and Buss (in prep., b) for a detailed exposition of this branch of Homicide Design Theory, along with specific empirical predictions and preliminary empirical tests.

5.7. Empirical Evidence Relevant to Homicide Design Theory

In this paper, we have enumerated more than four dozen specific testable empirical predictions that follow from Homicide Design Theory and cited dozens of empirical findings that are consistent with the theory and inconsistent with competing theories. Our research program over the past three years has been devoted to testing the novel predictions about which there exist no relevant data. We have conducted more than a dozen empirical studies of homicidal premeditations (N = 2,900), anti-homicide premeditations (N = 500), hypothetical scenarios in which people estimate the likelihood that they would kill in certain circumstances (N = 1,000), and free listings of the reasons that would impel people to resort to murder (N = 300). We have also conducted a study of 800 people actually charged with murder and analyzed FBI homicide statistics involving thousands of convicted killers.

A detailed report of the dozens of new findings is beyond the scope of this paper, which has focused on explicating the premises and logic of the Homicide Design Theory. The research is being prepared for publication in the primary empirical journals. Nonetheless, we note that the results thus far provide strong support for a number of the specific empirical predictions. We find that more than 82% of women and 91% of men (in a sample close to 3,000) report having had vivid homicidal premeditations. Similar percentages occur in our cross-cultural sample from Singapore. Men tend to have more frequent, more detailed, and more recurring homicidal thoughts than women, sex differences also found by Kenrick and Sheets (1993) and by Crabb (2000). Homicidal thoughts are often recurring over periods of days, weeks, months, and sometimes years. They involve detailed scenarios about means, motives, and opportunities.

Fantasies about killing intrasexual rivals far outnumber fantasies about any other category of victim, corresponding to the empirical fact that most homicides involve intrasexual rivalry (Daly & Wilson, 1988). Men are far more likely to have these fantasies than women, also corresponding to the actual rates of homicide. Indeed, the percentages of homicidal fantasies in the two-by-two matrix of sex of killer and sex of victim show a remarkable correspondence with the known percentages in that matrix. The order in both cases is male killer/male victim (65%, 59%), male killer/female victim (22%, 18%), female killer/male victim (10%, 17%), and female killer/female victim (3%, 6%), where the first number in parentheses refers to actual homicides and the second to our data on homicidal fantasies.

The catalysts of the homicidal fantasies are also highly sex-linked. Men are more likely than women to cognitively rehearse killing a rival who has sex with their mate, a rival who has humiliated them in public, a rival who has physically attacked them, or a rival who has physically dominated them. Women are more likely to fantasize about killing a rival who is more physically attractive, particularly if their mates have indicated an attraction to those rivals.

The second most common victims of homicidal fantasies are killing romantic partner, again corresponding to the second most common category of actual homicides. The causes of homicidal fantasies are sharply sex-differentiated in ways that correspond precisely to the predictions. Among those who report thoughts of killing a romantic partner, more than twice as many men as women report a partner’s sexual infidelity as being a catalyst and more than three times as many men as women report getting irrevocably dumped as a catalyst. These findings accord with findings from studies of uxoricide. In one study of all uxoricides committed in a Canadian city over 22 years, for example, female-initiated separation emerged as a motive in the testimonial evidence in 63% of the cases (Daly, Wiseman, & Wilson, 1997). Women, in contrast, are far more likely to report a partner’s emotional infidelity and physical abuse at the hands of a partner as a cause of their homicidal thoughts.

Among those who reported thoughts of killing a family member who was not a romantic partner, reports of killing a non-genetic relative were nearly four times as frequent as thoughts of killing a genetic relative. The most common victims in these premeditations were stepparents. The most common causes cited involved the stepparent inflicting costs on the person (or the person’s genetically related siblings) and absorbing all of the resources of their genetic parent. One woman, for example, reported thoughts of killing her stepmother because "She is an evil repugnant witch who only wants my father’s money and she despises me because I seem to be the only one who realizes this."

We have documented a cognitive bias suggesting that people overestimate the likelihood of getting killed in a variety of circumstances (e.g., being discovered in bed with a rival’s mate), as predicted by the error management component of the theory. For example, men give far higher estimates of the likelihood that a rival who finds them in bed with the rival’s partner will try to kill them than to the likelihood that they will try to kill a rival they find in bed with their partner. These and related findings were predicted in advance by applying Error Management Theory logic to the evolution of anti-homicide psychology. To our knowledge, no competing theories of homicide predict cognitive biases of this sort. Our studies of 800 killers, homicide scenarios, free listings of reasons for killing, and FBI statistics on homicide also reveals support for several specific predictions.

The ultimate merits of Homicide Design Theory in contrast to competing theories of homicide will ultimately rest with the weight of evidence collected by the community of scientists who study homicide. Although the authors of any theory are obviously not unbiased, we suggest that unbiased reviewers of the evidence already existing in the published literature will come to the conclusion that prior theories of homicide are inadequate to explain what is already known.

6. Objections to Homicide Design Theory

In the context of presenting our theory to colleagues both formally at professional conferences and informally, we have encountered a number of objections. Below we address the more substantive of these objections.

6.1. Objection 1: The costs of killing would have been too high for a psychology of homicide to evolve. The costs of attempting to kill another human being are obviously substantial, including the risk of being injured or killed while attempting murder and the risk of retribution from the victim’s kin, friends, or coalitional allies. The fact that it is so costly to kill, ironically, is actually evidence for, rather than against, the idea that homicide has been a recurrent selective force over human evolutionary history. The tremendous costs associated with pursuing a killer strategy did not, and could not, have arisen in a vacuum. The heavy fitness costs to victims of homicide provided powerful selective forces leading to the evolution of anti-homicide mechanisms–both mechanisms designed to promote peace (e.g., de Waal, 1989) and mechanisms that make homicide an extremely hazardous strategy to carry out. According to our theory, these costs have held homicide in check, resulting in a lower frequency of homicide than would have occurred without these co-evolved anti-homicide mechanisms. They led additionally to the coevolution of additional complex design features for a homicidal strategy, such as (a) preferences for exploitable victims, and (b) preferences for contexts where costs will be least likely to be incurred (e.g., when victim is less formidable and victim’s kin less numerous).

Selection operates according to net benefits relative to costs in the currency of fitness averaged over the relevant instances. All evolved strategies carry costs. According to Homicide Design Theory, the benefits of killing in certain contexts provided access to fitness benefits, centrally the elimination of intrasexual competitors and the resources that would flow to those competitors (including potential mates), to a far greater extent than has been previously recognized. The benefits of killing a rival’s prenatal or not-yet-conceived children, for example, produce a cascade of benefits for the killer’s children–a benefit not recognized in prior analyses of homicide. As Tooby and Cosmides (1994) point out in the more delimited context of warfare, even when there is a substantial risk of death, a strategy of killing can evolve if the benefits get distributed to those who succeed in the strategy (and obviously if the fitness benefits outweigh the fitness costs of the strategy, on average, recurrently over the relevant events).

We are not attempting to minimize the costs of killing; precisely the opposite. The costs of pursuing a killer strategy are so high precisely because killing has been so beneficial to killers over human evolutionary history. If it had not been so beneficial, we would not see today the formidable array of evolved anti-homicide mechanisms that keep killing in check. Demonstrating that anti-homicide mechanisms inflict heavy costs on potential killers provide evidence for, not against, the central assertion that homicide has been a recurrent feature of human ancestral environments.

6.2. Objection 2: Homicide occurs too infrequently to be acted on by selection.

This objection can be countered with two points. First, even if homicide has been a low base-rate event in the lives of our ancestors (e.g., a 1% probability of getting killed), the fitness consequences nonetheless would have been tremendous. The cascading fitness costs incurred by the victim and the victim’s kin and the cascading fitness benefits accrued by the killer and the killer’s kin, for reasons outlined earlier, would have provided more than enough selective impetus for mechanisms designed for killing conspecifics to have evolved. Indeed, a design feature that provides a mere 1% fitness advantage over competing designs can spread from 1% to 99% of the population in a mere 285 generations. It is not an extravagant claim to suggest that design features that terminate rival designs could confer at least a 1% fitness advantage over those competing designs.

Second, homicides may not have been rare, and the fitness advantage may have been considerably higher than 1%. To use one Amazonian group as an example, among the Yanomamo, as many as 30% of all males are killed by someone else (Chagnon, 1988; see also Valero, 1970). The Yanomamo are not alone in these astonishing rates, nor are they the highest. Keeley’s (1996) analysis suggests that the percentages of death due to warfare alone in various prestate societies reached as high as 1.5% per year, which would add up to a cumulative probability of dying at the hands of another human of higher than 30%. These rates of homicide, given the associated fitness consequences to individuals and their rivals suggests that selection easily could have operated to fashion psychological adaptations devoted to killing. Furthermore, high rates of conspecific killings have also been documented in chimpanzees, our closest primate relative. Wrangham (1999) notes that adult chimpanzee killings have been reported from four independent study sites. Goodall (1986) estimates 30-40% adult male mortality from intraspecific coalitionary aggression. Although this is likely to be an overestimate, Wrangham comes to the conclusion that "the killings, in relation to total observed adult deaths . . . appear to be demographically significant" (p. 10).

A final rate issue pertains to the type of homicide. In our professional presentations of Homicide Design Theory, many have found our analysis of the evolution of a psychology of infanticide highly plausible, but express greater skepticism about the evolution of some of the other proposed homicide mechanisms (e.g., Margo Wilson, personal communication, June 1998). However, infanticide represents a tiny fraction of all homicides, less than 2 percent. Mate homicides are than five times more frequent, and male intrasexual rivalry homicides are more than thirty times more frequent. If it is plausible that selection can fashion infanticide mechanisms in humans, given their low base-rate, there is no reason to believe that selection could not have fashioned other homicide mechanisms, given that these other homicides occur at far greater rates.

6.3. Objection 3: For each adaptive problem to which homicide is a hypothesized solution, there are many obvious, more effective, less costly non-lethal alternative strategies.

For most of the adaptive problems discussed in this paper, we are not arguing that homicide is the only evolved solution; precisely the opposite. Our theory is that homicide is one designated option within several evolved psychological decision trees. Consider the adaptive problem confronted by a young unmarried woman who gives birth, but who has failed to induce the father of her child to commit or invest. Clearly, there would have been several potential solutions to this adaptive problem–attempt to secure another mate who might be willing to invest, solicit help from kin, or attempt to raise the child alone. A fourth solution would be to kill the infant and to wait until circumstances are more favorable to bear and rear a child. The hypothesis that infanticide is one evolved solution to the adaptive problem this young woman is facing does not in any way contradict the hypothesis that there are alternative evolved solutions.

Whether the evolved decision rules produce in infanticide or some alternative course of action in any particular instance, of course, depends heavily on context. Specifically, what is the woman’s appraisal of her ability to attract a mate willing to invest in a child who is not his own? How extensive, resource-rich, and willing-to-invest are the members of the woman’s extended kin network? These and other contexts are hypothesized to be central components of the evolved psychological apparatus and therefore will affect the likelihood of choosing infanticide rather than some other alternative as the behavioral solution to the relevant adaptive problem.

The same logic applies to each of evolved homicide mechanisms proposed Homicide Design Theory. Even with the self-defense homicide hypothesis, there are alternative solutions to the adaptive problem of being confronted by a person who shows intent to kill you. A potential victim can run, hide, migrate, attempt to placate the killer, enlist allies, or attempt to murder the attacker. Which solution a person adopts depends heavily on contextual factors that influence cost-benefit appraisals (not necessarily conscious) of the possible solutions–factors such as comparative physical formidability, relative weaponry, availability of escape routes, and so on. The key point is that existence of non-homicidal solutions to particular adaptive problems is not a cogent argument against homicide being one of the evolved solutions.

6.4. Objection 4: Homicide merely represents the end of a continuum of violence, "the tip of the iceberg of coercive control rather than a motivationally distinct phenomenon" (Daly & Wilson, 1999, p. 67).

According to our theory, and in sharp contrast to the Daly and Wilson position, most homicides are manifestations of evolved psychological mechanisms that are distinct from mechanisms designed for coercive control and other nonlethal methods of solving adaptive problems. Like the rain, the burden of proof for these competing theories falls on both alike. We suggest that there are already sufficient grounds for doubting the theory that homicide is merely an extreme manifestation of the same mechanisms responsible for coercive control.

Before enumerating these grounds, it is important to keep distinct several issues. We fully acknowledge that some of the same circumstances that provoke non-lethal violence also can provoke homicide. Finding a rival in bed with one’s partner, for example, may for one man evoke a non-lethal beating and for another man evoke a homicidal rage. Being a stepchild, to take another example, is a circumstance that evokes both an elevated risk of being physical abused and an elevated risk of being killed (Daly & Wilson, 1988). But similarity or of some of the circumstances that provoke homicide and non-lethal violence does not warrant the conclusion that the underlying psychological mechanisms of the two phenomena are identical, motivationally or cognitively.

Let’s first consider motivation, and use mate killing as the example to illustrate the logic of our argument. According to our theory of uxoricide, one key motive for murder is the destruction of a mate who has defected, resulting in depriving intrasexual rivals of access to what would be a extremely valuable reproductive commodity. A common phrase spontaneously reported by both killers (Daly & Wilson, 1988) and in our studies of spontaneous homicidal premeditations is "If I can’t have her, no one can." In contrast, the motive for non-lethal beatings is to try to coerce the woman back into the relationship. Thus, the two phenomena are motivationally distinct, even though one central circumstance is the same, namely that the mate has departed.

Obviously, the decision by a deserted man to kill the woman, beat her, try to damage her reputation, or leave her alone entirely will depend on a host of other circumstances–the perceived likelihood that she can be coerced back into the relationship, whether she has produced children with him, extensity and strength of her kin group (Figueredo, 1995), and so on.

Competitive tests of our theory against alternative theories that argue that homicide and non-lethal violence are generated by the same psychological mechanisms (continuity theories) must rest with identifying whether or not there are psychological mechanisms unique to homicide. According to Homicide Design Theory, the existence and pervasiveness of homicidal premeditations provides one direct source of evidence bearing on the alternative theories. The continuity theories have difficulty accounting for these findings: (1) men have homicidal premeditations of killing their mates when their mates have left them irrevocably; (2) in our studies of homicidal scenario building, men think of killing their partner, not merely injuring her; (3) homicidal thoughts are recurrent and pervasive, and can consume large quantities of cognitive capacity. All of these findings are well explained by Homicide Design Theory, but are entirely puzzling on the coercive control theory of mate killing.

In sum, although some of the same circumstances that are causally connected with non-lethal violence are also causally connected with homicide, there is already compelling evidence to suggest that the two are motivationally and psychologically distinct.

6.5. Objection 5: Distinct Objections to the Theory of Evolved Mate Killing.

During professional presentations, critics have raised two objections to the mate killing component of our theory. First, why it ever would have been advantageous to kill a mate, who after all has positive reproductive value? Second, wouldn’t the costs associated with the risk of retribution from the dead wife’s kin have imposed have outweighed whatever benefits might otherwise have accrued to the killer?

Our theory, combined with ancillary information, provides a powerful refutation of these objections. The multiple fitness costs one inflicts on an intrasexual rival, including the effective infanticide of the rival’s prenatal children, alone could confer in principle the necessary fitness benefits for mate killing mechanisms to have evolved. And since the best prediction of future sexual behavior is past sexual behavior (Buss, 1994), a wife found to be unfaithful is both likely to have been so in the past and likely to be so again in the future. Thus, killing her in some circumstances solves a raft of adaptive problems, confers fitness benefits on the killer and his future children, and inflicts large costs on the rival and his would-be or future children (including reducing the number of fitness vehicles his rival’s children have to benefit from and invest in).

Regarding the deterrent effect of retaliatory costs inflicted by the mate’s kin (Figueredo, 1995), we hypothesize that this was indeed a selective force that co-evolved in response to the adaptive problem of having a woman kin member killed by her mate! Without this evolved anti-homicide mechanism, mate killing would have been far more beneficial than it was. But unless one supposes that people lack evolved kin-protection mechanisms and just "figure it out," then the very existence of a context-contingent retaliatory mechanism in people to inflict severe costs on men who kill their mates constitutes evidence for the recurrence of mate killing over human evolutionary history.

It is likely that the majority of ancestral social groups were composed of a core of men who were related, and females who were imported from other groups. Most females would have lived among a group of people to whom, on average, she was less genetically related than they were to one another. In such an environment, males who killed their partners would have been much less likely to suffer the costs of retaliation from the woman's kin.

Furthermore, there is no reason to believe that the "kin retaliation" argument against the theory of evolved mate killing would not apply with equal force to other forms of killing. Why doesn’t the prospect of kin retaliation deter men, for example, from killing their intrasexual rivals? Again, it’s important to stress that we think it likely that kin retaliation has acted indeed as one deterrent to mate homicide. But we propose that it has evolved as one among many anti-homicide mechanisms, and has co-evolved along with mate killing psychology. The existence of these anti-homicide mechanisms impose co-evolutionary selection pressure on the evolution of killer sensitivity to circumstances in which the benefits of mate killing will be unusually high and the costs unusually low (e.g., if perpetrator detection can be evaded). We suggest that precisely such context-sensitivity is part of the co-evolved mate killing mechanism.

7. Conclusions

Evolution by selection operates through the relative replicative success of competing designs. At the highest level of abstraction, there are two general strategies that can be favored by selection (Buss & Dedden, 1990). One is to acquire reproductively relevant resources more adeptly than competitors. The second is to inflict reproductive costs on competitors. Homicide is a remarkably effective strategy for inflicting costs on competitors. Its consequences cascade down generations, affecting entire kin lineages. In circumstances where a killer acquires the resources plundered from a conspecific competitor, murder provides the double-benefit of eradicating a rival entirely while gaining access to his prized reproductive resources. Because killing conspecifics is so extraordinarily beneficial, it is simultaneously extraordinarily costly to victims. Therefore, when conspecific killing recurs, selection will rapidly favor the evolution of anti-homicide mechanisms, including killing to prevent getting killed, resulting in a coevolutionary arms race between homicide and anti-homicide strategies.

People have killed other people in all known cultures and have done so throughout human recorded history and prehistory. Humans kill in a wide variety of distinct circumstances, from a young mother who kills a deformed infant to a tribe that conducts a retaliatory war raid to avenge a prior attack. The known empirical facts provide a complex mosaic that must be explained by any adequate theory of homicide.

Prior theories of homicide fail at this task. Part of the failure may be attributed to the fact that homicide is not a singular phenomenon, but rather a diverse set of different phenomena. Beating a stepchild to death, for example, is so different from declaring war that we should not expect the explanation for one to successfully account for the other. Nonetheless, theories of homicide have tended to focus on single-cause explanations, which then collapse when the broader empirical picture is considered. Theories that invoke culture-specific causes such as media violence, for example, cannot explain why identical patterns of homicide are documented in cultures lacking these causes. Theories that invoke sex-neutral causes, as most do, cannot explain the substantial sex differences in both the rates and the circumstances in which men and women kill.

Prior efforts to explain killing by invoking evolutionary factors have been hastily dismissed by many, sometimes for good reasons, but often based on profound misunderstandings. In an otherwise brilliant book on war before civilization, Keeley (1996) dismisses "a biological explanation for warfare" because of "the incredible plasticity of human conduct" (p. 158). As evidence, he cites the fact that some peoples have changed from peaceable to warlike, or vice-versa, within a few generations. Examples of human mutability are abundant. The mistaken assumption behind Keeley’s analysis is that if warfare is "biological," then somehow it must be invariant, universally expressed, difficult to suppress, and impossible to modify. Because human behavior is so obviously variable and changeable with context, Keeley and others have concluded that killing can’t be "biological," so it must be "shaped by learning and decision making" (Keeley, 1996, p. 158).

As Wrangham points out, "This error seems remarkable, because behavioral ecologists [and evolutionary psychologists] have long stressed that that psychological adaptations are expected to respond in a contingent way to appropriate contexts" (Wrangham, 1999b, p. 21). Nonetheless, this "remarkable error" appears to be widespread among social scientists. In a commentary on an article by Steven Pinker suggesting that infanticide could have evolved, a well known sociologist wrote: "Your article on infanticide illustrates how silly evolutionary explanations of human behavior have become. When mothers protect their newborns . . . it’s because that behavior is evolutionarily adaptive. And now, when a few mothers kill their newborns, that’s evolutionarily adaptive too. Any behavior is ‘explained’ by evolutionary selection: homicide, love, suicide, hate, nurturance, aggression–you name it. Thus, nothing is explained" (C.S. Fischer, Nov. 2, 1997, New York Times Magazine). Obviously, nothing is "explained" merely by pointing to a behavior and slapping the label "adaptive" on it, but that fails "remarkably" to capture the logic of the articulation of precise testable evolutionary psychological hypotheses such as those proposed in this paper. The notion that humans have evolved a complex collection of psychological mechanisms that are selectively engaged according to the exigencies of circumstances, and that these different mechanisms could give rise to behaviors as diverse as helping and homicide, is logically coherent and internally consistent.

Homicide Design Theory discards the false dichotomy of "biological" versus "learned." It proposes that evolution by selection has produced a mosaic of psychological mechanisms that are designed to be highly sensitive to social circumstances. The proposal that humans have evolved infanticide mechanisms, for example, does not imply that humans blindly kill babies. They are hypothesized to do so only under delimited conditions (e.g., when infant is deformed) that differ for men and women (e.g., paternity uncertain is a context unique to men) and according to circumstance (e.g., available resources). The proposal that humans have evolved warfare mechanisms does not imply that humans blindly go into combat or invariantly express an "instinct" for battle. Only some humans do it (e.g., men, not women), and only under delimited social conditions (e.g., in self-defense) or ecological conditions (e.g., nearing starvation). The variability and specificity of contexts in which killing does and does not occur suggests the need for a theory that can explain that variability and specificity. Clearly, a "biological" theory that proposed an invariant killer instinct cannot succeed. Neither can vague hand-waving about "learning" or "decision making."

The paleontological, archeological, and ethnographic evidence is strong that homicide has been a recurrent feature of human evolutionary history. Given the dramatic fitness costs inflicted on victims of homicide, it would be surprising if selection had failed to fashion mechanisms designed to prevent getting killed. Indeed, the logic of the evolution of anti-homicide mechanisms may have been presaged by Charles Darwin: "as . . . there is a frequently recurring struggle for existence, it follows that any being, if it vary however slightly in any manner profitable to itself . . . will have a better chance of surviving, and thus be naturally selected" (Darwin, 1859). Homicide Design Theory suggests that the survival strategies that have evolved in humans to combat the "hostile force of nature" include defenses against realistic threats of being murdered by other humans.

According to our theory, these include mechanisms such as fear of strangers in infancy; specialized mind-reading abilities to infer homicidal intent in others; adaptively biased inference mechanisms that result in protective overestimates of the likelihood of getting killed; selective preferences for habitats that afford protection and refuge (e.g., seeing without being seen); anticipatory defensive maneuvers including development of arms, fortifications, and protective postures; panic reaction to motivate fleeing when confronted by another human displaying homicidal intent, and homicidal mechanisms designed to impel killing to prevent getting killed.

No prior theories predict the existence of these well-designed anti-homicide mechanisms, in part because no prior theory of homicide has identified such a large range of evolved homicide mechanisms that have imposed specific and recurrent problems of survival.

Once anti-homicide mechanisms started to evolve, they would immediately have imposed selective pressures on the further evolution of killer psychology. The evolution of anti-homicide mechanisms increases the costs of pursuing a killer strategy, holding the homicide rate in check. At the same time, anti-homicide mechanisms impose selection pressure that favors the further refinement of killer psychology. These refinements include sensitivity to the specific contexts in which the immediate and future costs of killing might be low and as well as meta-mind reading abilities such as concealing homicidal intentions from victims to surprise them or to avoid activating their anti-homicide mechanisms. From Biblical account of Trojan Horses to modern accounts of sneak tribal raids, deception has been used to prevent the activation of intended victim’s anti-homicide mechanisms.

Because killing has been a recurrent feature of human living for so long, we expect the evolved psychology of killing and of avoiding getting killed have coevolved, each become highly elaborated, each containing many specific design features not explicable on any other theory. One of the most important collections of psychological design features center on homicidal premeditations.

Cognitive space is a precious and limited human resource. Information processing capacities, including attention, memory, retrieval, judgment, and decision-making, are finite and metabolically expensive. Cognitive effort allocated toward one adaptive problem creates opportunity costs that preclude effort allocated to other adaptive problems. Our research and that of others on homicidal fantasies has found that the vast majority of men and women have experienced a number of thoughts about killing. While some are fleeting, many are detailed, elaborate, and recurring (e.g., Crabb, 2000; Kenrick & Sheets, 1993; Duntley & Buss, in prep.). These premeditations are highly patterned, sharply sex-differentiated, and pervasively occurring in precisely the contexts in which humans have faced grave adaptive problems.

The existence of highly patterned, sharply sex-differentiated, pervasively occurring, cognitively taxing, apparently-designed homicidal thoughts cannot be explained by prior theories of killing. The available evidence for highly context-specific, sex-linked, premeditated forms of actual murder that recur across cultures and over time cannot be explained by prior theories of homicide. The apparent existence of well articulated anti-homicide mechanisms, corresponding in specificity to the contexts in which human life is in danger, cannot be explained by prior theories of homicide. When combined with the paleontological, archeological, ethnographic, psychological, and behavioral evidence, it is not unreasonable to suggest that selection over human evolutionary history has produced specific mechanisms whose function is to produce the death of other humans.


Special thanks go to Victoria Beckner, Randy Diehl, Sam Gosling, Martie Haselton, Bill Swann, and Andy Thompson for providing insightful suggestions on earlier drafts.



[1] This equation is usually called "Hamilton’s rule," which outlines the conditions under which a design feature can evolve. The benefits (b) in reproduction conferred by design feature , multiplied by the coefficient of relatedness (r), must exceed the reproductive costs (c) in order for that design feature to evolve. Back

[2] The use of the term "benefit" in this context refers solely to benefits in fitness from the standpoint of natural selection, and does not correspond to any personal, intuitive, or societal construals of benefit. Back



Alexander, R. D. (1979). Darwinism and human affairs. Seattle: University of Washington Press.

Alexander, R.D. (1987). The biology of moral systems. New York: Aldine.

Bandura, A. (1973). Aggression: A social learning analysis. Englewood Cliffs, NJ: Prentice-Hall.

Baron, J.N., & Reiss, P.C. (1985a). Same time, next year: Aggregate analyses of the mass media and violent behavior. American Sociological Review, 50, 347-363.

Becroft, D.M., Lockett, B.K. , Etzel, R.A., Montaña, E., Dearborn, D.G., Smith, P.G., Infeld, M.D., Dahms, B.B. & Carroll-Pankhurst, C. (1998). SIDS or Murder. Pediatrics, 101, 953.

Black, D. (1993). The social structure of right and wrong. San Diego: Academic Press.

Brinnigan, A. (1998). Criminology and the holocaust: Xenophobia, evolution, and genocide. Crime and Delinquency, 44, 257-276.

Bristowe, W.S. (1958). The world of spiders. London: Collins.

Bugos, P.E., & McCarthy, L.M. (1984). Ayoreo infanticide: A case study. In Gl Hausfter & S.B. Hrdy (Eds.), Infanticide: Comparative and evolutionary perspectives (pp. 503-520). NY: Aldine de Gruyter.

Buss, D. M. (1995). Psychological sex differences: Origins through sexual selection. American Psychologist, 50, 164-168.

Buss, D. M. (1996). The evolution of human social strategies. In E. T. Higgins & A. Kruglanski (Eds.), Social Psychology: Handbook of Basic Principles. New York: Guilford Press.

Buss, D.M. (2000). The dangerous passion: Why jealousy is as necessary as love and sex. New York: Free Press.

Buss, D.M., & Dedden, L.A. (1990). Derogation of competitors. Journal of Social and Personal Relationships, 7, 395-422.

Buss, D.M., & Duntley, J. (1998). Evolved homicide modules. Paper presented to the Annual Meeting of the Human Behavior and Evolution Society, Davis, California, July 10.

Buss, D.M., Haselton, M.G., Shackelford, T.K., Bleske, A.L., & Wakefield, J.C. (1998). Adaptations, exaptations, and spandrels. American Psychologist, 53, 533-548.

Buss, D.M., Larsen, R., Westen, D., & Semmelroth, J. (1992). Sex differences in jealousy: Evolution, physiology, and psychology. Psychological Science, 3, 251-255.

Busse, C., & Hamilton, W.J., III. (1981). Infant carrying by adult male chacma baboons. Science, 212, 1281-1283.

Butynski, T.M. (1982). Harem male replacement and infanticide in the blue monkey (Cercopithecus mitis Stuhlmann) in the Kibale Forest, Uganda. American Journal of Primatology, 3, 1-22.

Bygott, J.D. (1972). Cannibalism among wild chimpanzees. Nature, 238, 410-411.

Chagnon, N. (1983). Yanomamo: The fierce people (3rd ed.). New York: Holt, Rinehart, & Winston.

Chagnon, N. (1988). Life histories, blood revenge, and warfare in a tribal population. Science, 239, 985-992.

Collins, D.A., Busse, C.D., & Goodall, J., (1984). Infanticide in two populations of savanna baboons (pp. 193-216). In G. Hausfater & S.B. Hrdy (Eds.), Infanticide: Comparative and evolutionary perspectives. New York: Aldine.

Coon, C.S. (1971). The hunting peoples. New York: Atlantic-Little, Brown.

Covell, K. (1999). Cultural socialization and conceptions of war and peace: A cross-national comparison. In Raviv, A., & Oppenheimer, L. How children understand war and peace: A call for international peace education. Pp. 111-126. San Francisco, CA: Jossey-Bass Inc, Publishers.

Crabb, P.B. (2000). The material culture of homicidal fantasies. Aggressive Behavior, 26, 225-234.

Crockett, C.M., & Sekulic, R. (1984). Infanticide in red howler monkeys (pp. 173-192). In G. Hausfater & S.B. Hrdy (Eds.), Infanticide: Comparative and evolutionary perspectives. New York: Aldine.

Crowell, N.A., & Burgess, A.W. (Eds.)(1996). Understanding violence against women. Washington, C.C.: National Academy Press.

Daly, M., & Wilson, M. (1988). Homicide. Hawthorne, NY: Aldine.

Daly, M., & Wilson, M. (1989). Homicide and cultural evolution. Ethology and Sociobiology, 10, 99-110.

Daly, M., & Wilson, M. (1999). An evolutionary psychological perspective on homicide. In M.D. Smith & M.A. Zahn (Eds.), Homicide: A sourcebook of social research (pp. 58-71).. Thousand Oaks: Sage Publications.

Daly, M., Wiseman, K.A., & Wilson, M. (1997). Women with children sired by previous partners incur excess risk of uxoricide. Homicide Studies, 1, 61-71.

DeKay, W.T., Buss, D.M., & Stone, V. (1998). Coalitions, mates, and friends: Toward an evolutionary psychology of relationship preferences. Unpublished manuscript, Department of Psychology, University of Texas, Austin, Texas.

Dennen, J.M.G.V.D. (1995). The origin of war. Groningen, The Netherlands: Orion Press.

Dyer, G. (1985). War. New York: Crown.

Duntley, J.D., & Buss, D.M. (in prep., a) The evolutionary psychology of anti-homicide: Why the mind is designed to prevent getting killed. Department of Psychology, University of Texas.

Duntley, J.D., & Buss, D.M. (in prep., b). The evolutionary psychology of warfare. Department of Psychology, University of Texas.

Eibl-Eibesfeldt, I. (1989). Human ethology. New York: Aldine de Gruyter.

Ehrenreich, B. (1997). Blood Rites: Origins and History of the Passions of War. New York: Metropolitan Books.

Ellison, P.T. (1985). Lineal inheritance and lineal extinction. Behavioral and Brain Sciences, 8, 672.

Figueredo, A. J. (1995). Preliminary report: Family deterrence of domestic violence in Spain. Department of Psychology, University of Arizona.

Fingerhut, L.A., & Kleinman, J.C. (1990). International and interstate comparisons of homicide among young males. Journal of the American Medical Association, 263, 3292-3295.

Forsberg, R. (1998). Toward a theory of peace: The role of moral beliefs. Dissertation Abstracts International, 58, 11-A.

Fossey, D. (1984). Gorillas in the mist. Boston: Houghton Mifflin.

Fox, A. (1997). The assessment of fighting ability in humans. Paper presented to the Ninth Annual Meeting of the Human Behavior and Evolution Society, University of Arizona, Tucson, Arizona (June 4-8).

Frank, R. (1985). Choosing the right pond. New York: Oxford University Press.

Frank, R. (1988). Passions within reason. New York: Norton.

Gazzaniga, M., Ivry, R., & Mangun, G. (1998).Cognitive Neuroscience : The Biology of the Mind.

Ghiglieri, M.P. (1999). The dark side of man: Tracing the origins of violence. Reading, MA: Perseus Books.

Gould, S.J. (1984). Only his wings remained. Natural History, 93,10-18.

Gould, S.J. (1991). Exaptation: A crucial tool for evolutionary analysis. Journal of Social Issues, 47, 43-65.

Grauer, A.L. (1995). Bodies of evidence : reconstructing history through skeletal analysis. New York : Wiley-Liss.

Grauer, A.L., Stuart-Macadam, P. (Eds.)(1998). Sex and gender in paleopathological perspective. New York, NY: Cambridge University Press.

Haig, D. (1993). Genetic conflicts in human pregnancy. Quarterly Review of Biology, 68, 495-532.

Hamilton, W.D. (1975). Innate social aptitudes of man: An approach from evolutionary genetics. In Robin Fox (ed.) ASA Studies 4: Biosocial Anthropology. Pp. 133-53. London: Malaby Press.

Hart, C.W., & Pilling, A.R. (1960). The Tiwi of North Australia. New York: Hart, Rinehart, & Winston.

Haselton, M.G., & Buss, D.M. (2000). Error management theory: A new perspective on biases in cross-sex mind reading. Journal of Personality and Social Psychology, 78, 81-91.

Hausfater, G., & Hrdy, S.B. (Eds.)((1984). Infanticide: Comparative and evolutionary perspectives. New York: Aldine.

Hill, K., & Kaplan, H. (1988). Tradeoffs in male and female reproductive strategies among the Ache, part 2. In L. Betzig, M. Borgerhoff Mulder, & P. Turke (Eds.), Human reproductive behavior: A Darwinian perspective (pp. 291-305). New York: Cambridge University Press.

Holden, C. (2000). Cannibalism: Molecule shows Anasazi ate their enemies. Science, 289, 1663.

Holland, B. & Rice, W.R. (1998). Chase-away sexual selection: antagonistic seduction versus resistance. Evolution, 52, 1-7.

Hoyenga, K.B., & Hoyenga, K.T. (1993). Gender-related differences: Origins and outcomes. Boston: Allyn & Bacon.

Hrdy, S.B. (1977). Infanticide as a primate reproductive strategy. American Scientist, 65, 40-49.

Hrdy, S.B. (1999). Mother nature: A history of mothers, infants, and natural selection. New York: Pantheon Books.

Johnson, J. (1969). Organic psychosyndromes due to boxing. British Journal of Psychiatry, 115, 45-53.

Keeley, L.H. (1995). War before civilization. New York: Oxford University Press.

Kenrick, D.T., & Sheets, V. (1993). Homicidal fantasies. Ethology and Sociobiology, 14, 231-246.

Knauft, B.M. (1985). Good company and violence: Sorcery and social control in a lowland New Guinea Society. Berkeley: University of California Press.

Lee, R.B. (1984). The Dobe !Kung. New York: Holt, Rinehart and Winston.

Leonard W.R., Robertson M.L. (1992). Nutritional requirements and human evolution: a bioenergetics model. American Journal of Human Biology, 4, 179-195.

Leonard W.R., Robertson M.L (1994). Evolutionary perspectives on human nutrition: the influence of brain and body size on diet and metabolism. American Journal ofHuman Biology, 6, 77-88.

Mann, C.R. (1996). When women kill. Albany, NY: State University of New York Press.

McDougall, W. (1915). An introduction to social psychology. Boston, MA: Luce.

Niehoff, D. (1999). The Biology of Violence. New York: Free Press.

Nisbett, R.E. (1993). Violence and U.S. regional culture. American Psychologist, 48, 441-449.

Otterbein, K. (1979). The evolution of war. New Haven, CT: HRAF Press.

Otterbein, K. & Otterbein, C. (1965). An eye for an eye, a tooth for a tooth: A cross-cultural study of feuding. American Anthropologist, 67, 1470-1482.

Polis, G.A. & Farley, R.D. (1979). Behavior and ecology of mating in the cannibalistic scorpion Paruroctonus mesaensis Stahnke (Scorpionida: Vaejovidae). Journal of Arachnology, 7,33-46.

Prentkey, R.A., Burgess, A.W., Rolous, F., Lee, A., Hartman, C., Ressler, R., & Douglas, J. (1989). The presumptive role of fantasy in serial sexual homicide. American Journal of Psychiatry, 146, 887-891.

Schmitt, D.P., & Buss, D.M. (in press). Human mate poaching: Tactics and temptations for infiltrating existing relationships. Journal of Personality and Social Psychology.

Scrimshaw, S.C.M. (1984). Infanticide in human populations. In Hausfater, G., & Hrdy, S.B. (Eds.)((1984). Infanticide: Comparative and evolutionary perspectives (pp. 439-462). New York: Aldine.

Seabury, P. & Codevilla, A. (1989). War Ends and Means. New York: Basic Books.

Shackelford, T. K. (in press, a). Are young women the special targets of rape-murder? Aggressive Behavior.

Shackelford, T. K. (in press, b). Cohabitation, marriage, and murder: Woman-killing by male romantic partners. Aggressive Behavior.

Shostak, M. (1981). Nisa: The life and words of a !Kung woman. Cambridge, MA: Harvard University Press.

Staff. (1990). Haphazard homicides at home. Science News, 137, 110.

Suzuki, A. (1971). Carnivority and cannibalism observed in forest-living chimpanzees. Journal of the Anthropological Society of Nippon, 74, 30-48.

Symons, D. (1979). The evolution of human sexuality. New York: Oxford University Press.

Theilgaard, A. (1984). A psychological study of the personalities of XYY- and XXY men. Acta Psychiatrica Scandinavica (suppl.) 69, 133.

Thomas, E.M. (1959). The harmless people. New York: Knopf.

Tooby, J. & Cosmides, L. (1988). The evolution of war and its cognitive foundations. Institute for Evolutionary Studies, Technical Report #88-1.

Tooby, J., & Cosmides, L. (1994). Cognitive adaptations for threat, cooperation, and war (Abstract). Ethology and Sociobiology, 10, 29-49.

U.S. Department of Justice. (1993). Crime in the United States, 1992. Washington, D.C.: U.S. Government Printing Office.

Valero, H. (1970). Yanomama, the narrative of a white girl kidnapped by Amazonian Indians, as told to Ettore Biocca [Translated from the Italian by Dennis Rhodes] New York: Dutton.

de Waal, F.B.M. (1989). Peacemaking among primates. Cambridge, MA: Harvard University Press.

Walker, P. (1995). Documenting patterns of violence in earlier societies: The problems and promise of using bioarchaeological data for testing evolutionary theories. Paper presented at the Annual Conference of the Human Evolution and Behavior Society, Santa Barbara, CA: July 2.

White, M. (1999). Historical Atlas of the Twentieth Century.

Wilbanks, W. (1982). Murdered women and women who murder: A critique of the literature. IN N.H. Rafter & E.A. Stanko (Eds.), Judge, lawyer, victim, thief, gender roles, and criminal justice. Boston: Northeastern University Press.

Williams, G.C. (1992). Natural Selection. New York: Oxford University Press.

Wilson, M., & Daly, M. (1985). Competitiveness, risk-taking, and violence: the young male syndrome. Ethology and Sociobiology, 6, 59-73.

Wilson, M., & Daly, M. (1998). Lethal and nonlethal violence against wives and the evolutionary psychology of male sexual proprietariness. In R.E. Dobash & R.P. Dobash (Eds.), Violence against women: International and cross-disciplinary perspectives (pp. 199-230). Thousand Oaks, CA: Sage.

Wilson, M., Daly, M., & Scheib, J. (1997). Femicide: An evolutionary psychological perspective. In P. Gowaty (Ed.), Feminism and evolutionary biology (pp. 431-465). New York: Chapman & Hall.

Wolfgang, M.E. (1958). Patterns in criminal homicide. Philadelphia, PA: University of Pennsylvania Press.

Wolfgang, M.E., & Ferracuti, F. (1967). The subculture of violence. London: Tavistock.

Wrangham, R. (1999a). Is Military incompetence adaptive? - Apes and the origins of human violence, Evolution and Human Behavior, 20, 3-17.

Wrangham, R. (1999b). Evolution of coalitionary killing. Yearbook of Physical Anthropology, 42, 1-30.

Wrangham, R. & Peterson, D. (1996). Demonic Males. Boston: Houghton Mifflin.

Wright, R. (1995). The biology of violence. The New Yorker, March 13, p. 68-77.