Martin Daly
This chapter makes the case that the discovery and interpretation of basic facts about conflict and violence can be facilitated by having a sound theory of the nature of self-interest, and that the requisite theory is necessarily grounded in the theory of evolution by selection.
The basic facts about conflict and violence to which I refer are social, familial, demographic, epidemiological, and motivational. Who is more or less likely to come into conflict with whom, and over what substantive issues? What considerations affect the intensity of interpersonal disagreements and their potential for violence? Answers to these questions clearly differ according to specific social relationships—strangers, same-sex rivals, parent–offspring and other kin relationships, romantic partners, potential mates, in-laws, and so on—as well as in relation to the parties' ages, sexes, social positions, and reproductive histories. The ways in which these variables affect conflict are consequences, I suggest, of actors' perceptions of where their interests reside and of whether the actions and inclinations of others are then experienced as antagonistic to, or facilitating of, the satisfaction of those perceived interests.
It is essential to be clear about what one means by “interests.” The dictionary definition that is closest to my meaning here is an “advantage or benefit” to the party in question (see, e.g., http://www.oxforddictionaries.com/definition/english/interest). Defining interests in this way justifies the popular evolutionary theoretical equation of an individual's interests with the relative replicative success of that individual's alleles (loosely, “fitness”), but it also justifies a more psychological interpretation whereby one's interests correspond with one's preferred states, whether fitness-promoting or not. I will use the term “fitness interests” to refer to the former, evolutionary meaning, and “perceived interests” (with no implications about conscious awareness) to refer to the latter, psychological meaning. This distinction is essentially the same as that which Tinbergen (1963) made between “function” and “cause.”1
The cause/function distinction is easily blurred. “Strategy,” for example, is such a compelling metaphor for adaptively conditional responsiveness within an integrated hierarchy of functions that it is now used by evolutionists primarily in this functional sense (e.g., Shuster & Wade, 2003). This is unproblematic when the so-called strategists are brainless organisms such as plants (e.g., DeJong & Klinkhamer, 2005), but in its literal sense, strategy is a causal concept (and moreover one that implies conscious planning), so “reproductive strategy” is sometimes misunderstood as implying a literal pursuit of fitness rather than as the pursuit of goals that contributed to ancestral fitness. Trying to clarify one's meaning by adding that alleged efforts to “maximize fitness” are “unconscious” doesn't eliminate the potential for confusion. Consciousness or its absence is again a meaningful distinction only within the domain of proximate causation.
This sort of failure to carefully distinguish cause from function encourages two common fallacies. One is that Darwinism entitles us to expect that organisms will do whatever it takes to maximize their fitness, even in evolutionarily novel circumstances. Symons (1987, 1989, 1990) provided a series of critiques of such thinking and elicited great indignation from Darwinians whose oxen he gored, but his cogent analysis retains its bite to this day. The second fallacy that often follows from treating fitness as a goal is to think that because evolutionary explanations invoke fitness consequences, maladaptive behavior must require some alternative sort of explanation that is “not evolutionary.” The problem here is that maladaptive action and even extreme psychopathology are the outputs of brains/minds whose evolved structures and processes generate all behavior. An evolution-minded taxonomy of the sources of maladaptation is essential (e.g., Baron-Cohen, 1997; Williams & Nesse, 1991). Failures of adaptation can result not only from brain damage and insanity, but also from mismatches between adaptations and modern environments, from bad luck in domains where behavior is successful only on average, from hijackings of the actors' adaptations by other organisms, and from wasteful intragenomic conflicts (Nesse, Chapter 43, this volume; Tooby & Cosmides, Chapter 1, this Handbook, this volume).
The point of these cautions is not to deny that perceived interests and fitness interests often look very much alike. They do, of course, because the former have been shaped by selection to be means to the end of the latter. Unrelated same-sex individuals who are competitors for access to limited resources are threats to one another's fitness and are therefore apt to respond to one another as rivals. Being cuckolded harms a man's expected fitness, and his partner's sexual infidelities are therefore intensely disliked (Shackelford, Goetz, LaMunyon, Pham, & Pound, Chapter 15, this Handbook, Volume 1). However, it must never be forgotten that organisms are not seeking fitness in a literal sense, but simply responding to cues and pursuing goal states that were statistical predictors of expected fitness consequences in ancestral environments. Rats in a Skinner box will continue to work for hypothalamic stimulation that has been divorced from fitness-related outcomes, and vasectomized men will continue to lust after sexually appealing women.
The fitness interests of two individuals may overlap to varying degrees, or may be strictly antithetical, and the basic sources of commonality versus conflict of fitness interests are ancient and enduring. It is thus a reasonable working hypothesis that the evolved psychology of interpersonal conflict will be responsive to evolutionarily reliable cues of the extent to which fitness interests are shared, as well as the delimited contexts or circumstances in which they are shared or in conflict. The genetic relatedness between two parties, for example, is the most obvious and often the main determinant of the degree to which their fitness interests overlap, and we are therefore justified in postulating positive social responses to cues of genetic relatedness (e.g., DeBruine, 2002; Hames, Chapter 19, this Handbook, Volume 1; Krupp, DeBruine, & Jones, 2011). The other primary source of overlapping fitness interests is the fact that sexual reproduction gives individuals who are not themselves close kin a shared stake in the welfare of young who are close kin to both, and we are therefore justified in postulating effects of offspring on the quality of relations between mates, and between in-laws who are related through a mateship, as well as strong emotional responses to cues of fidelity and paternity (Shackelford, Goetz, LaMunyon, Pham, & Pound, this volume).
Intraspecific violence has attracted the attention of evolutionists for a variety of reasons. The fact that direct aggression is a risky competitive tactic has motivated extensive theoretical and empirical analysis of the factors that make animals more or less likely to resort to it (Hardy & Briffa, 2013; Maynard Smith & Price, 1973). Moreover, because intraspecific violence can be a significant source of mortality in some species, including human beings in small-scale, nonstate societies like those in which basic human attributes evolved (P. L. Walker, 2001; R. S. Walker & Bailey, 2013), its selective impacts warrant investigation. However, this review of the topic, and my own research on human violence in collaboration with the late Margo Wilson, are motivated not so much by the possible evolutionary significance of violence itself, as by the fact that it provides a useful assay of conflict, and a way to test theoretically derived hypotheses about the factors that affect conflict's intensity in different relationships.
It is, of course, possible and sometimes illuminating to investigate interpersonal conflict through its nonviolent manifestations. However, such research programs face potential threats to validity. Opportunities for direct observation of conflict behavior are limited, and researchers have relied heavily on questionnaires and interviews, but data derived from these self-report methods must be interpreted skeptically in any domain in which issues of social desirability, self-presentation, and self-justification are prominent. A cautionary example is provided by a large literature on violence between intimate partners and other family members that has relied on a self-report tool called the Conflict Tactics Scales (CTS; Straus, 1979). CTS respondents are asked to affirm or deny that they and their partners or other family members have performed each of a long list of “acts” in conflict situations within a specified period, usually the past year. It has long been apparent that the reliability of these measures is poor: When intimate partners are both tested, the correlations between their accounts of their respective actions are often negligible (R. P. Dobash, Dobash, Wilson, & Daly, 1992; Jouriles & O'Leary, 1985; Szinovacz, 1983). It is therefore unsurprising that CTS research has repeatedly generated “findings” that contradict those based on less equivocal manifestations of violence and that are almost certainly not valid; examples include an alleged absence of sex differences in intimate partner violence that is unique to CTS studies (reviewed and critiqued by R. P. Dobash et al., 1992), and supposedly identical levels of assaults on children by genetic and stepparents as reported by Gelles and Harrop (1991; see Daly & Wilson, 2008).
The poor validity of such measures is presumably due mainly to biased self-presentation. However, there is also some reason to doubt that people have the introspective ability that they would need to portray their social conflicts accurately, even if they were sincerely trying to be forthcoming (Nisbett & Wilson, 1977; Wegner, 2002). For these reasons, self-report data are of limited value in this area.
The experimental methods of behavioral economics afford opportunities for participants to select how cooperatively or competitively they will behave toward specific others in a variety of allocation decisions. To date, the literature on economic games has been overwhelmingly focused on stranger and/or anonymous interactions (see, e.g., Plott & Smith, 2008), but the methods are certainly amenable to the study of participants' “welfare trade-off ratios” (Cosmides & Tooby, 2013) with various sorts of relationship partners. The external validity of such methods remains questionable, however, not only because of the same self-presentation concerns that bedevil self-report studies, but also because of the likelihood that research participants may often be motivated to simply maximize joint profits in games played with close social partners, in the shadow of an expectation of post-experimental reallocation.
Thus, despite the limitations of nonexperimental research, the investigation of interpersonal solidarity versus conflict must still rely on spontaneous real-world manifestations thereof, such as legal proceedings, bequests, divorces, and violence. Unfortunately, none of these manifestations can provide a completely unbiased window on the real distribution of conflict. The subsets of people who launch civil lawsuits, who register marriages and divorces, and who die intestate are surely not random samples of human decision makers, and persons charged with assault are not a random subset of actual assailants. It is for this reason that Margo Wilson and I were first drawn to the study of lethal interpersonal violence. It is a truism among criminologists that homicide is the crime that is least vulnerable to biased detection and recording. Although there are surely exceptions necessitating continued caution, by and large the bodies are found and the causes of death are investigated. Consider, for example, the question of whether child maltreatment is disproportionately perpetrated by stepparents. In a U.S. study, Wilson, Daly, and Weghorst (1980) were the first researchers to estimate rates of reported physical abuse of children in stepparent-plus-genetic-parent versus two-genetic-parent, and found a large excess in the former. It is easy to imagine, however, that stepparents' abusive acts might be especially likely to be detected or recorded, and for that reason, Wilson and colleagues (1980) noted and stressed that the differences between household types were much larger in the relatively rare lethal cases than in nonlethal abuse, a result that is hard to reconcile with an explanation in terms of biased detection. Greater stepparental overrepresentation in lethal beatings than in nonlethal maltreatment has proven to be a robust result, and provides some of the best evidence that “Cinderella effects”—elevated risks at the hands of stepparents—are genuine (Daly & Wilson, 2008).
According to the leading theory of social evolution, Hamilton's (1964) inclusive fitness theory, the fundamental basis of shared fitness interests is genealogical kinship (Hames, Chapter 19, this Handbook, Volume 1). An obvious hypothesis, then, is that perceived interests will follow suit and social motives will be effectively “nepotistic.” Creatures cherish close kin because they are vehicles of inclusive fitness and selection has favored valuing their welfare if they can be identified with any degree of reliability. The flip side is that we should be more reluctant to harm kin than nonkin.
Bohannan's (1960) study of homicides among the Tiv of central Nigeria during the colonial period provides a nice illustration of such nepotistic restraint. The society was strictly patrilocal, such that men's daily interactions were predominantly with close patrilineal kin. A highly successful Tiv man might have as many as 20 wives, not all of whom he could easily guard, and competition among agnates (patrilineal kin) was intense. But Bohannan was able to show that close kinship typically prevented that competition from becoming lethal:
In a community in which 83% of the adult males are agnatic kinsmen of one another, the chances that a woman's lover will be a kinsman of her husband are obviously extremely high.… In the eight cases in which men killed their wives' lovers, only [two cases] show any kinship relation between the husband and the lover. Any fieldworker in Tivland realizes that adulteries between women and their husbands' kinsmen occur frequently. Tiv do not suggest that such adultery does not occur. They insist, however—and the cases prove them right—that a wife's adulteries must not be allowed to disturb relationships among kinsmen. (Bohannon, 1960, p. 42)
The tendency for patrilineal kinsmen to form political and military alliances that compete against other patrilineages is of course well known to anthropologists, who have dubbed such coalitions “fraternal interest groups” (e.g., Göhlen, 1990; Otterbein, 1968). Anti-Darwinians such as Sahlins (1976) have asserted that these practices have “nothing to do with biology,” and while such arguments prove only their authors' incomprehension and hostility, there are legitimate questions about the degree to which fraternal interest groups are effectively nepotistic. How well do clan memberships and patronyms serve as proxies for actual relatedness? And does terminological kinship “crowd out” other potential cues of relatedness in the social and psychological control of nepotism?
Napoleon Chagnon's studies of Yanomamö horticulturalists in Venezuela speak to these issues with unusual clarity. The Yanomamö use an “Iroquois” kinship system, in which all men who can trace their patrilineage to the same male ancestor at the same generational depth are called “brother.” But this does not mean they treat their patrilateral parallel cousins and their “real” brothers identically. To the contrary, genetic relatedness as estimated from recited genealogies is a better predictor of who lines up with whom in a violent conflict situation than is mere kin terminology (Alvard, 2009; Chagnon, 1981; Chagnon & Bugos, 1979). Relatedly, kin stick together in space. When Yanomamö villages grow unmanageably large, disputes arise, mainly over adulteries, and the villages then “fission,” with each man having to choose in which of the two new smaller villages he will reside. Because of generations of cross-cousin marriage, virtually everyone is related to everyone else, often by multiple loops, such that the “correct” characterization of particular people's relationships (and hence their entitlements to marry) may even be contested, and yet it is a striking fact that the average genetic relatedness of pairs residing in the same village is elevated by a fissioning event (Chagnon, 1981). In sum, the Yanomamö behave as if they know their degrees of relatedness to one another more accurately than one could infer from knowing only the kin terms that dyads use to address one another, and they use that awareness to form alliances both for protection and for aggressive exploitation of others.
It isn't only in tribal societies with semichronic warfare between villages that you might be wise to stick close to your close genetic relatives. In tough times, close kin can be one's salvation. Several analyses of the sources of differential mortality in disasters have reached the conclusion that when mortality rates skyrocket, the density of kin in your immediate vicinity is a major determinant of your chances of surviving, sometimes mattering even more than your age, sex, or wealth (Grayson, 1993, 1996; McCullough & Barton, 1991). Part of the reason for this—but only part—is that those people who have no relatives to protect and avenge them are fair game for homicide and even cannibalism.
In modern mass society, the importance of kinship has surely diminished, so one might reasonably wonder whether nepotistic biases are still detectable in action. Decades ago, as newcomers to the study of family violence, Margo Wilson and I were startled to encounter allegations that close kin relations are routinely violent! Not only were psychoanalysts seemingly convinced that parricidal and filicidal impulses are universal components of human nature, but even number-crunching social scientists were on board. According to the two best-known investigators of family violence in the United States, for example:
The family is the most frequent locus of all types of violence ranging from slaps, to beatings, to torture, to murder. Students of homicide are well aware that more murder victims are members of the same family than any other category of murderer–victim relationship…In fact, violence is so common in the family that we have said it is at least as typical of family relations as is love. (Gelles & Straus, 1979, p. 188)
This would be astonishing if it were true. But it is not true. Intrafamilial homicides are in fact quite rare in the United States, as they are elsewhere (Daly & Wilson, 1982, 1988b).
A small part of the obfuscation here derives from an excessively broad definition of “family.” Since Wolfgang (1958), mainstream criminologists who partition the victim–killer relationship have typically recognized just three categories: strangers, acquaintances, and family members. This taxonomy, extraordinarily naïve from an evolutionary perspective, has persisted through a half century of studies. Some researchers now at least distinguish “intimate” or “romantic” partners (who almost always comprise a large majority of “family” victims) from “other relatives,” but this latter basket category still typically includes the killer's children, parents, other genetic kin, step-relations, and in-laws (e.g., Kubrin, 2003). This is of course unsatisfactory because it conflates relationships that are qualitatively distinct, not only in degrees of their genetic relatedness, but even in whether they are consanguineal or marital—distinctions critically important to an evolution-minded theorist. The conflicts between intimate partners are utterly different in their substance and triggers than those between other family members, and the same goes for conflicts involving genetic versus marital relatives, as well as conflicts among the various more specific kinds of genetic relatives (Daly & Wilson, 1988a, 1988b).
But even if “family” is defined in the encompassing way that most criminologists thoughtlessly favor, Gelles and Straus's assertions about its dangers are still clearly false: Family homicides are not more numerous than those in any and all other categories of victim–killer relationship. Instead, in the United States and also in the world as a whole, the two broad categories of stranger homicides and lethal contests between acquainted nonrelatives both substantially outnumber killings of family members. According to the FBI's data for the years 2000 through 2010, for example, marital partners and genetic relatives together comprised fewer than 20% of the victims of solved U.S. homicides, whereas about a quarter were strangers; more than half were persons who were known to, but not related to, their killers (Puzzanchera, Chamberlin, & Kang, 2013). Moreover, even these numbers almost certainly exaggerate the prevalence of family murders. Over 40% of cases remained unsolved and therefore lack any victim–killer relationship code, and because unsolved homicides are disproportionately cases that occur in public places rather than homes and that exhibit evidence of other criminal victimization (especially robbery and gang violence), it is very likely that they are proportionately more often committed by strangers and acquaintances than solved homicides.
So homicide is less of a family affair than some would portray it. But how can we take the matter further? In order to assess whether and to what extent kinship might be associated with a reduction in lethal conflict, we need some sort of reasonable null model that would generate an “expected” incidence of related victims under the assumption that kinship is not relevant. Daly and Wilson (1982, 1988b) tackled this issue of base rate expectations in two ways. First, they analyzed a set of Detroit homicides in which victims and their killers resided in the same household, and generated the required expected values for various relationship categories from survey-based household composition data for the city. The result was that a homicide was more than 10 times as likely between genetically unrelated coresiding persons, regardless of whether they were intimate partners or other nonrelatives such as roommates, than between coresiding genetic relatives.
Daly and Wilson's other approach to this issue exploited the fact that some homicides have multiple perpetrators. The logic was this: Although we cannot know the base rate of the average individual's interactions with kin versus nonkin, we can postulate that if kinship were without effect on conflict versus solidarity, such that both arise in frequencies proportionate to the availability and intensity of interactions with others, then the distribution of the social relationships between those who collaborate in homicide should approximate the distribution of victim-killer relationships. With that as our null hypothesis, we assembled every data set containing the requisite information on both co-killer relationships and those between victims and killers that we could find. The resultant samples represented the urban United States, several horticultural or agricultural societies, and an historical registry from late medieval England. In every case, we found that the average relatedness of collaborative killers greatly exceeded that between killers and their victims. In the U.S. urban samples and in 13th-century England, for example, the average relatedness between pairs of co-killers ranged from .08 to .09, and that between killers and victims between .01 and .03.
In more traditional societies, fraternal closeness entails a bitter irony: Brothers may be natural allies, but the very fact of close kinship also forces them into intense rivalries. Not only are close kinsmen likely to be the sole claimants to a title or to the family farm, but where extensive genealogical links dictate who can and cannot marry whom, as is often the case in nonstate societies, brothers must find their brides within the same limited pool of legitimate marriage partners. It is little wonder, then, that stories of fratricide abound in such societies, often in close association with origin myths, as exemplified by the story of Cain and Abel. Can fraternal solidarity survive these social structural pressures? In the analyses described above, the sample included four strongly patrilocal indigenous peoples in India: the Bison-Horn Maria, Bhil, Munda, and Oraon. In these societies, as among the Tiv studied by Bohannan, most of a man's routine social interactions are necessarily with his agnates. Primogeniture in land inheritance created fierce rivalries between brothers, and fratricides in these societies constituted about 10% of all homicides. (In contrast, fratricides comprise about 2% of homicides in modern countries with low homicide rates such as Japan, and fewer than 1% in the United States; Daly, Wilson, Salmon, Hiraiwa-Hasegawa, & Hasegawa, 2001.) But even so, men in these patrilineal Indian societies were substantially less likely to kill close agnates than to join forces with them in lethal quarrels with other unrelated or distantly related men; whereas the average victim–killer relatedness for the four societies as a whole was .07, the average relatedness between collaborating killers was .24.
Citing anthropological and historical sources, Daly and Wilson (1988b) concluded their discussion of fratricide by suggesting that it became a problem only after the invention of agriculture led to the inheritance of land. (We were then unable to find even one account of a fratricide in hunter-gatherer ethnographies, but in a personal communication in about 1990, Thomas Headland then drew a single such case to our attention.)
Where the temptation to fratricide really gets out of hand is in the circumstance where brothers are rivals for a position of enormous value, and yet agnatic kinsmen are not themselves a crucial source of a man's power.…In a feudal society…vassalage at least partially replaces kinship as a basis of loyalty and power, and rivalrous power blocs may line up behind related pretenders to the same throne. Here, surely, is a situation designed to overwhelm brotherly affection, and indeed the history of royal families in feudal empires is a seemingly endless tale of fraternal bloodletting. (Daly & Wilson, 1988b, p. 31)
It seems that we underestimated nepotistic restraint, however, for there is now evidence that it tempered even feudal disputes over royal succession. S. B. Johnson and Johnson (1991) analyzed the historical struggles among rival claimants to the Earldom of the Orkney Islands, and found an apparent mitigating effect of close kinship: When two men had simultaneous claims to rule, brothers were almost always able to partition the perquisites amicably, whereas a homicide was typically required to resolve analogous competitions between more distant agnatic kin. Dunbar, Clark, and Hurst (1995) extended these analyses by showing that the infrequent killings of agnatic kinsmen occurred only in the context of clear and substantial incentives, quite unlike killings of nonkin, which were often the dénouements of relatively minor disputes with little at stake. Analyses of lethal conflict over the English crown tell a similar story (S. B. Johnson & Johnson, 1991; McCullough, Heath, & Fields, 2006): Although both near and distant kin were certainly slain in battles over succession, fratricides, parricides, and filicides were nevertheless inhibited relative to the enthusiasm with which aspirant kings went about “culling the cousins.”
The reasons for solidarity and conflict between opposite-sex mates are peculiar to that relationship, but as with genetic kinship, those reasons ultimately derive from overlapping fitness interests: Sexual reproduction creates a situation in which unrelated mates combine their prospects for direct fitness, and the young that they produce constitute a powerful source of shared goals. Couples who are faithful monogamists, and who engage in little or no nepotistic investment in their distinct collateral kindreds, attain a commonality of fitness interests that surpasses that of the closest nonclonal genetic relatives (Alexander, 1987). This gibes with, and perhaps explains, the commonplace observation that the tastes, ambitions, and worldviews of longstanding couples often converge.
Despite this powerful source of shared fitness interests, the solidarity of mates is easily and often undermined. Your genetic kin are yours for life, and no betrayal can erase the fact that their reproduction enhances your genetic posterity; that is presumably why the psychology of forgiveness and reconciliation seems to cut kin more slack than friends. Not so for mateship: If a couple is not sexually monogamous, then that which enhances one party's fitness may be systematically damaging to the other party's, and this of course explains the peculiar emotional potency of issues related to infidelity and cuckoldry (Gangestad, Thornhill, & Garver-Apgar, Chapter 14, this Handbook, Volume 1; Shackelford, Goetz, LaMunyon, Pham, & Pound, Chapter 15, this Handbook, Volume 1). Unfortunately, fidelity alone cannot eliminate partner conflict. Even in a population in which couples invest only in joint progeny whose well-being will contribute equally to both parents' direct fitness, selection will still favor those parents who shirk and let their partners pay the lion's share of child rearing costs, unless two conditions are met: There is absolutely no chance that either partner will ever remate (e.g., in the event of one's death) and there is absolutely no chance that either partner can ever promote his or her inclusive fitness by diverting resources to nondescendant kin. These conditions are not likely to be met in any pair-forming biparental species, and they are certainly not met in human beings.
An evolution-minded analysis of the relationship between mates points to at least the following six, more or less distinct, sources of conflict, many of which clearly match lay notions of the most important threats to a happy marriage:
Except for the sexually asymmetrical risk of cuckoldry, all of these apply in principle to both women and men, although not necessarily with equal force.
A large majority of the couple conflicts that culminate in a homicide, across the gamut of human societies, are precipitated by male sexual jealousy, if that term is defined to encompass men's resentment of both partner infidelity and partner desertion (Daly & Wilson, 1988b; Daly, Wilson, & Weghorst, 1982). To the best of my knowledge, this conclusion has not been contravened. Some authors have suggested that it was overstated, but the data that supposedly challenge it are invariably informationally impoverished; for example, it may be noted that the police coded only a minority of some set of spousal homicides as “jealousy” cases, while a larger number were coded instead as “arguments” instead, but of course the latter label is mute about the substance of conflicts. In later writings, we have preferred to characterize the issue as one of “proprietariness” rather than “jealousy” to focus attention on men's tendency to construe wives as property, and hence to react similarly to their (suspected) infidelities and their efforts to terminate marriages, both of which are resented as violations of husbands' property rights (Wilson & Daly, 1992, 1996, 1998). This conceptualization captured the imagination of many feminist researchers who had been wary, at best, of evolutionary psychology, and helped create a space for interdisciplinary dialogue (see, e.g., Campbell, 2012; R. E. Dobash & Dobash, 2012; H. Johnson, 2012).
To be useful, an evolution-minded analysis of the relationship between mates should do more than just provide compelling terminology; it should help scientists generate fruitful new hypotheses. The “stepchild problem” provides an illustrative case in point. Whereas a couple's children create a commonality of fitness interests and therefore facilitate consensus on difficult issues such as the ideal uses of the couple's resources, children of former unions have precisely the opposite impact (Daly & Wilson, 1996). This may seem an obvious hypothesis, but although the presence of stepchildren had long been known to be associated with elevated rates of divorce (Becker, Landes, & Michael, 1977; White & Booth, 1985), no one had assessed whether their presence might also be associated with elevated rates of marital violence before Daly, Singh, and Wilson (1993) showed that women with children sired by previous partners were disproportionately heavy users of a women's shelter. In a Canadian study (Daly, Wiseman, & Wilson, 1997), we subsequently showed that such women also incurred a much greater risk of uxoricide than mothers whose children were all sired by the current partner, a result that has been replicated in a U.S. sample (Brewer & Paulsen, 1999). Meanwhile, Jacquelyn Campbell and collaborators were developing tools to assess the risk that intimate partner violence will escalate to lethality, a very difficult task because the risk factors for lethal and nonlethal partner violence are largely the same (Wilson, Johnson, & Daly, 1995). The stepfamily findings persuaded these researchers to include this measure in their assessment battery, and it proved to be one of their most useful. Given that a woman has already been a victim of recurrent physical assaults by her male partner, the evidence to date pinpoints three statistical predictors that he will eventually kill her: a history of his threatening suicide, a gun in the home, and the presence of a child sired by a predecessor (Campbell, 2012; Campbell et al., 2003).
Another important general point that the study of intimate partner homicide can be used to illustrate is that there is seldom a single privileged “evolutionary hypothesis” that can be contrasted with those generated from other perspectives. Instead, evolutionary psychologists and biologists can and often do generate competing hypotheses that are equally Darwinian. Thirty years ago, relationship-specific demographic patterns of homicide risk were virtually unstudied, and no one had investigated how age might be related to differential risk. Reasoning that the value that men place on their female partners should be a function of reproductive value (and influenced, perhaps unduly, by three notorious cases in which high-status Canadian men had hired “hit men” to kill their middle-aged wives), I hypothesized that women's risk of being slain would increase prior to or at menopause, both because of planned disposals and, more importantly, because angry men would be less inhibited by the danger of doing an older wife serious harm. But Margo Wilson predicted that the risk would be maximal for young wives of the highest reproductive value, reasoning that violence against wives is functionally controlling, and that men are most inclined to exert coercive control when their partners are especially attractive to rival males and thus especially likely to confront temptations to defect. Margo's hypothesis, which assumes that uxoricides are mainly functionless by-products of anger and are rarely strategic disposals, was the evident winner: We found a substantial negative relationship between age and uxoricide risk in Canada, with no hint of an elevation at menopause (Daly & Wilson, 1988b; Wilson, Johnson, & Daly, 1995), and this pattern has proven to be replicable in the United States, the United Kingdom, and Australia (Mercy & Saltzman, 1989; Shackelford & Mouzos, 2005; Wilson & Daly, 2001).
A further major risk factor for intimate partner homicide revealed by Daly and Wilson's (1988b) early analyses was common-law or de facto status: Both partners were victimized at rates very much higher in coresiding couples than in registered marriage couples, and this, too, has been replicated in other English-speaking countries (Mouzos & Shackelford, 2004; Shackelford & Mouzos, 2005; Wilson & Daly, 2001). Suggested reasons for this include socioeconomic confounds, conflicts arising from lower commitment and higher rates of infidelity in common-law unions, and the much higher incidence of stepchildren in the homes of co-residing couples who have not registered their marriages; a distinct age pattern such that risk in co-residing couples is maximal in middle age, which has also been replicated internationally, lends some support to the stepfamily hypothesis. Surprisingly, however, the large and seemingly robust difference in homicide rates between registered and common-law marriages has been shrinking since 1990 in the United States, and had disappeared completely by 2005 (James & Daly, 2012), and the same appears to be true in Canada (James, 2011). James and Daly (2012) could find no evidence that the two types of unions have been converging in other attributes, and the reasons for this striking change remain to be discovered, as indeed do the reasons for large changes over time, including recent declines, in intimate partner homicide rates as a whole. An evolution-minded understanding of couple conflict will surely remain a crucial element of any satisfactory future explication of these trends.
Other categories of interpersonal relationships have specific sources of conflict that are manifested in distinct patterns of homicide risk. Trivers's (1974) famous parent–offspring conflict theory laid bare the reasons why parents neither love their young more than they love themselves, nor cherish all of their children equally. The insights that the theory yields about variability in the intensity of parent–offspring conflict as a function of the parties' ages have been supported in a number of studies of age-related trajectories of both maternally and paternally perpetrated filicides, as well as parricides (Daly & Wilson, 1988a; Wilson & Daly, 1994). Trivers's theory also succeeds when its predictions are pitted against Sigmund Freud's notorious theory that parent–offspring conflict is essentially a matter of same-sex rivalry, a theory that is a priori implausible, since human parents and their same-sex offspring rarely compete for mates at all and sons do not lust after their own mothers, as Freud supposed (Daly & Wilson, 1990a).
The relationship that engenders the largest number of homicides by far is that between unrelated male rivals (Daly & Wilson, 1988b, 1990b), and an evolutionary perspective is essential for understanding the cases themselves and their highly variable incidence. That is a topic that we and others have dealt with extensively elsewhere (Courtwright, 1996; Daly & Wilson, 1988b, 1990b, 1997, 2001, 2010; Eisner, 2003; Pinker, 2011; Wilson & Daly, 1985), and that I will not discuss further here.
When using homicides as an “assay” for testing adaptationist hypotheses about the variables that aggravate and alleviate interpersonal conflicts, my collaborators and I have adopted a stance of agnosticism about whether the lethality itself is an adaptive function of the fatal attacks or is a nonadaptive minority outcome that would have reduced the killers' fitness, on average, even in ancestral environments (e.g., Daly & Wilson, 1988b). There is no question that violent capability is an adaptation, nor that the intensity of violent action is modulated by our evolved psychology (e.g., Sell, 2011), and it may well be the case that the human mind contains adaptations for killing. But an evolutionarily informed theory of relationship-specific conflicts is a valuable source of hypotheses about homicide risk regardless of whether most killings are by-products of adaptations for domination and coercion or are instead reflections of adaptations for lethality.
This agnostic stance has been criticized by Buss (1999, 2000) and by Duntley and Buss (2008, 2011). These authors argue that because many killings are intentional rather than being accidental “slips,” because most people say that they would kill in certain dire circumstances such as to protect their children, and because large numbers of people admit to homicidal fantasies, lethality is therefore unlikely to be a nonadaptive by-product of adaptations designed to achieve other results. They propose instead that people have been equipped by natural selection with a suite of relationship-specific psychological adaptations both for killing and for avoiding being killed. Unfortunately, identifying aspects of the mind that might have been designed for the specific function of killing is not that easy, and in my view, fantasy, intent, and professed willingness to kill are all beside the point. Even larger numbers of male undergraduates report fantasies of video game playing than of killing (Kai, unpublished, cited by Wilson, Daly, & Pound, 2009) although there are certainly no adaptations for video game playing, and high proportions of people profess willingness and formulate intentions to do myriad things that were never targets of selection, ranging from watching their favorite TV shows to having their pets spayed. Finally, with respect to certain specific relationships for which Duntley and Buss have proposed dedicated homicide “modules,” such as wife-killing and stepchild-killing, I can find no ethnographic evidence indicating that these killings might have been either fitness-promoting, on average, in small-scale, face-to-face societies like those in which humans evolved, or frequent enough to be plausible candidates for dedicated evolved psychological machinery. Thus, although I grant that lethal violence has surely been an agent of selection in human evolution, and that killers may even have enjoyed fitness advantages that have had selective consequences (Chagnon, 1988), we have no sound basis for concluding that most, or indeed any, homicides reflect “homicide adaptations” (see also Durrant, 2009).
And in a certain sense, it doesn't matter: This controversy can be set aside in the present context because the predictions that one would make about the patterning of homicide risk are largely unaffected by its resolution. Regardless of whether killing is typically a by-product of adaptations or a more direct reflection of what the relevant adaptations are designed to achieve, we should expect similar effects of cues of infidelity, reproductive value, kinship, and other conflict-related variables. New evidence and ideas may eventually clarify and resolve these points of contention, but of this we can already be sure: Homicides will continue to provide a rich source of data for testing evolution-minded hypotheses about interpersonal conflict.