R3.1. The concept of sexual selection

After Trivers' (1972) consideration of parental investment, it became logical to extend what we consider as sexual selection to processes following fertilization, although noting that this changes the original conception. I did so in the target article (sect. 4.6), and this was the theme of a recent article by Carranza (Reference Carranza2009), a source cited in two commentaries (Campbell and Pellegrini). Specifically, I considered the application of the game theory model of Clutton-Brock and Parker (Reference Clutton-Brock and Parker1995) to physical aggression between partners: again it should be stressed that models such as this one are quite different from classic sexual selection models (a point not appreciated by van den Berghe). Two commentaries (Cashdan and Kaighobadi & Shackelford) were concerned that I did not consider mate guarding derived from paternity uncertainty, the usual evolutionary explanation applied to partner violence. This is also a model that involves processes following fertilization. My reasons for omitting it in the target article are elaborated in section R3.2.

Of the two original processes involved in sexual selection, I focused on male competition because it was most relevant to the topic of my article, sex differences in aggression. While recognizing that female choice was the other process, I did not cover this for reasons of relevance. However, one issue concerning female choice is raised in two commentaries (Behme and Kempenaers & Forstmeier). It concerns the conflict between choosing aggressive or non-aggressive mates, and forms part of the extensively investigated topic of when women prefer men who are likely to show more parental effort and when they prefer men who typically show mating effort, and with it higher physical attraction and more pronounced aggression. These alternative male reproductive strategies were mentioned in my target article (sect. 2.7), but not in relation to women's preferences, which is another (extensive) topic. A full treatment of the evolution of sex differences in aggression would include this.

A number of commentaries seek to extend my central argument, based on inter-male competition, in interesting and relevant ways. These include intrasexual competition by females, which has been the subject of recent articles on sexual selection in animals (considered by Campbell and Pellegrini). One of these, Clutton-Brock (Reference Clutton-Brock2009), described instances where females compete for breeding territories and other resources necessary for reproduction, and for social rank: In some cases females can show a greater variance in reproductive rate than males. These and other examples of female competition are readily encompassed by traditional conceptions of sexual selection in terms of parental investment or reproductive rate (Trivers Reference Trivers and Campbell1972; Clutton-Brock & Vincent Reference Clutton-Brock and Vincent1991). They typically occur in polyandrous and monogamous species, or in eusocial breeders. Clutton-Brock (Reference Clutton-Brock2009) also pointed out a fundamental distinction between competition in the two sexes, that male competition is typically for access to the other sex and female competition is typically for resources necessary for reproduction. This is likely to provide an additional reason why female aggression is usually escalated to a lesser extent than is male aggression.

As Clutton-Brock (Reference Clutton-Brock, Kappeler and van Schaik2004, p. 26) noted: “one of the problems in writing about sexual selection today is that the term is used in so many different ways.” Campbell's consideration of her evolutionary explanation of less risky female aggression shows the difficulty of drawing the line in specific instances. I wonder whether it really is useful when considering sex differences in behavior to extend the concept of sexual selection to such an extent that, “We might say that, under this definition, almost all selection is sexual selection” (Carranza Reference Carranza2009, p. 750).

Since I was mainly concerned in the target article with inter-male competition, a central part of the original conception of sexual selection (Darwin Reference Darwin1871/1901), it was clear what I was dealing with, and what I was not. Campbell makes a compelling case for considering other features that might not have arisen from inter-male competition, and yet which are relevant to sex differences in aggression. Behme suggests that larger male size and musculature may be fitness-enhancing outside the realms of reproduction: Of course, this may or may not be the case, but unlike Campbell's example of greater female fear, no precise functions are suggested. Until they are, it is more parsimonious to regard the enhanced upper body musculature of the human male as part of a range of adaptive features that originated from inter-male competition.

Even those who propose a widening of the term sexual selection (e.g., Carranza Reference Carranza2009, cited by Campbell and Pellegrini) recognize the central importance of competition for mates in the process. Secondary differences, such as in feeding behavior or habitat use (Carranza Reference Carranza2009), may be relevant to sex differences associated with men's specialization for hunting and women's for gathering, although this arose much later in human evolution and, unlike male competition, is restricted to humans, rather than being widespread in many animals. It may well play a part in men's adaptations for group-cooperation and out-group aggression (Browne and Johnson & van Vugt).

In his cross-species analysis of sexual selection, Clutton-Brock (Reference Clutton-Brock2009) stated that sex differences in behavior are probably more flexible than is commonly realized. This highlights the importance of ecological conditions for variations in the sex differences. Pellegrini refers to the degree to which sex differences in behavior are moderated by different ecologies. Related to this, in section 3.7, I described a cross-cultural test of the hypothesis that sex differences in size would be more pronounced in polygynous societies (Alexander et al. Reference Alexander, Hoogland, Howard, Noonan, Sherman, Chagnon and Irons1979). The original claim that this was the case was undermined by later re-analyses. A related issue concerns the relationship between mating systems and the ecological conditions under which they occur. I also mentioned the finding that polygny was associated with higher degrees of pathogen stress across 186 societies (Low Reference Low1990). Low (Reference Low1989) found that boys were taught to strive and compete more in polygynous societies. Thus, the degree of pathogen stress was related to the mating system, which in turn was related to the type of socialization. More recent studies have extended Low's analysis to characteristics of the social structure of modern states, including gender empowerment (Thornhill et al. Reference Thornhill, Fincher and Aran2009). This is a particularly interesting line of research suggesting that social roles can be explained in terms of ecological adaptations.

R3.2. Partner violence and paternity uncertainty

One particular evolutionary model involving post-copulatory conflicts of interest between the sexes has been used to explain male violence to women. This is viewed as a form of mate guarding, whose ultimate function is to ensure paternity. Since females do not have the adaptive problem of paternity uncertainty, they do not have this reason to mate-guard. This explanation is elaborated in two commentaries (Cashdan and Kaighobadi & Shackelford).

The logic of mate-guarding and its widespread existence in nonhuman animals (e.g., Parker Reference Parker1974b; Wilson & Daly Reference Wilson, Daly, Barkow, Cosmides and Tooby1992) is not disputed. It is likely to play a part in explaining men's sexual coercion to their partners (as Kaighobadi & Shackelford argue), which might also be linked to sperm competition (widely found in insects and birds). However, I know of only one study that provides direct evidence for sperm competition in humans (Baker & Bellis Reference Baker and Bellis1993), and as far as I know this has not been replicated.

I did not refer to the mate-guarding view as an explanation of sex differences in partner violence in my target article, although I was aware of work by Wilson and Daly (Reference Wilson, Daly, Barkow, Cosmides and Tooby1992; Reference Wilson and Daly1996) and others. I like the logic of mate-guarding, and the associated emotion (sexual jealousy) is clearly associated both with men's violence to other men and with men's violence to their own partners (Daly & Wilson Reference Daly and Wilson1988; Reference Daly and Wilson1990). It was in hindsight an omission not to have mentioned paternity uncertainty in relation to inter-male aggression motivated by sexual jealousy, since this would be in addition to the pre-copulatory inter-male competition typically covered by sexual selection. The reason I did not mention paternity uncertainty in relation to partner violence was not an oversight. It was because there are typically no sex differences in overall acts of physical aggression, at least in modern Western nations (Archer Reference Archer2000a; Reference Archer2002). There are sex differences in injuries, the consequences of these acts – a difference better explained by the size and strength differences of the two sexes than by mate-guarding.

Faced with the evidence of similar proportions of male and female physical aggression to partners, it would seem reasonable to conclude that whatever part mate-guarding plays in the male psyche, this alone cannot explain partner violence. It would then be of interest to examine not only behavior used by males to increase their chances of paternity, but also behavior used by females to maximize their partner's investment in the relationship. Buss's studies of mate-retention tactics (Buss Reference Buss1988a; Buss & Shackelford Reference Buss and Shackelford1997a) and Flinn's (Reference Flinn1988) study of mate-guarding in a Caribbean village did adopt such a “gender-inclusive” approach,” although others have only studied male-to-female violence (e.g., Buss et al. Reference Buss, Shackelford and McKibbin2008; Shackelford et al. Reference Shackelford, Goetz, Buss, Euler and Hoier2005a). A digression is necessary on the background to such selective studies that only consider partner violence as a one-way process.

Selective studies are based on the assumption that most partner violence is male-to-female. I presented a considerable body of evidence from Western nations that this is incorrect (both in the target article and in previous reviews). The sources cited by Cashdan were ideologically biased, as are many in this area. Her source of criticism of the Conflict Tactics Scale (CTS), the measure used by family researchers to study conflict between partners, is a selective narrative review that is more a polemic than an objective review (Dobash et al. Reference Dobash, Dobash, Wilson and Daly1992). A more representative, quantitative assessment of the evidence for the accuracy and reliability of the CTS (Archer Reference Archer1999) showed that it is much more reliable than Dobash et al. inferred from their review. Correlations between couples were r=.55 and .53, for men and women, respectively, for an aggregate of six studies. This is comparable with values obtained for spousal ratings of personality (McCrae & Costa Reference McCrae and Costa1990) and does not take into account attenuation by measurement error. In their study of 350 couples in New Zealand, Moffit et al. (Reference Moffit, Caspi, Krueger, Magdol, Margolin, Silva and Sydney1997) found that correlations of r=.58 and .54 were raised to .83 and .71 when latent correlations were derived from confirmatory factor analysis. Other evidence supports both the reliability and validity of the CTS and its modifications (Straus Reference Straus, Straus and Gelles1990; Reference Straus2004), as does the similarity in the findings from this measure and other methods, such as crime victimization surveys (e.g., Laroche Reference Laroche2005; Mirrlees-Black et al. Reference Mirrlees-Black, Budd, Partridge and Mayhew1998).

Cashdan also cites the influential but controversial work of Michael Johnson (see Dutton Reference Dutton2005; Dutton & Nicholls Reference Dutton and Nicholls2005). Again, I was aware of this, but chose not to cite it in the target article. Johnson (Reference Johnson1995) argued from a qualitative review of studies involving general population samples and samples of women victims from shelters, that there were two “types” of partner violence, one (characteristically male) involving high levels of sustained violence and control, found in shelter samples, and the other mutual lower-level violence in the absence of overall control, found in general samples. He termed these, respectively, “patriarchal” (later “intimate”) terrorism,” and “common (later “situational”) couple” violence. Johnson (Reference Johnson, Booth, Crouter and Clements2001) applied cluster analysis to two samples, a community sample and one selected for high levels of male-to-female violence, finding clusters corresponding to these samples. It is likely that these arose from the prior selection of the samples, which in turn corresponded to Johnson's earlier classification.

Johnson's other empirical study (Johnson & Leone Reference Johnson and Leone2005) used a sample drawn from a “violence against women” survey that cannot be regarded as an unbiased sample of violence by both sexes (Archer Reference Archer2000a; Reference Archer2002; Straus Reference Straus, Arriaga and Oskamp1999). In an analysis of a large-scale Canadian survey sampling physical aggression and controlling behavior by both sexes, LaRoche (Reference Laroche2005) found that 38% of those who fitted Johnson's “intimate terrorist” category (i.e., they were both violent and controlling) were women, again suggesting that the previous categorization was based on biased sampling. Using another sample not selected for high levels of victimization, and including reports by both men and women, Graham-Kevan and Archer (submitted) found that the “intimate terrorist” group contained similar proportions of men and women. Nonviolent victims – that is, those who do not use any physical aggression towards an “intimate terrorist” – were twice as likely to be men than women. These findings indicate that Johnson's typology should be viewed with some skepticism, rather than the uncritical acceptance it enjoys at present, and that it requires further investigation.

To return to paternity uncertainty, if this were a full explanation of partner violence, the following would be the case: (1) men but not women would show physical aggression to their partners; (2) the male-only nature of physical aggression would be consistent across cultures (since paternity uncertainty is universal); (3) males but not females would show controlling behavior to their partners; (4) male aggression but not female aggression (if and when it occurs) would be motivated by the desire to control the partner's behavior; (5) males would show more sexual jealousy than females; (6) physical aggression to a partner would be linked to sexual jealousy in males but not in females (again if and when it occurs).

The first of these expectations was mentioned above and was covered in the target article. Studies in Western nations show similar frequencies of acts of physical aggression among women and men, although women are more likely to be injured. The second was also covered in the target article, and as indicated, there is considerable and meaningful variability across nations, which is related to women's degree of societal power and to gender attitudes (Archer Reference Archer2006a). The third is refuted by a number of studies showing no sex differences in controlling behavior (Felson & Outlaw Reference Felson and Outlaw2007; Graham-Kevan & Archer, in press; Walby & Allen Reference Walby and Allen2004), or more control by women than men (Charles & Perreira Reference Charles and Perreira2007), all from large or representative samples. The fourth is refuted by evidence that control motives are linked to physical aggression in both sexes (Felson & Outlaw Reference Felson and Outlaw2007; Graham-Kevan & Archer, in press). Regarding sex differences in sexual jealousy, some studies have found none (Bookwala et al. Reference Bookwala, Frieze, Smith and Ryan1992; Felson Reference Felson2002; Mullen & Martin Reference Mullen and Martin1994; White Reference White1981), whereas others have found that women report more sexual jealousy than men (de Weerth & Kalma Reference de Weerth and Kalma1993; Felson & Outlaw Reference Felson and Outlaw2007; Nannini & Meyers Reference Nannini and Meyers2000; Webb Reference Webb2007). Physical aggression has been found to be related to sexual jealousy in both sexes in some studies (Felson Reference Felson1997; Reference Felson2002; Haden & Hojjat Reference Haden and Hojjat2006), or in men only (Archer & Webb Reference Archer and Webb2006; Webb Reference Webb2007), or in women more than in men (de Weerth & Kalma Reference de Weerth and Kalma1993).

Altogether, the above findings are not what one would expect if male mate-guarding, derived from paternity uncertainty, were responsible for partner violence. However much I agree that the paternity uncertainty hypothesis makes evolutionary sense, the evidence from modern Western nations indicates that women are also seeking to control men. Having said this, I accept the view that paternity uncertainty is likely to be more important in relation to sexual aggression to partners (Kaighobadi & Shackelford), and that men's mate-guarding is likely to show subtle changes in relation to the woman's fertility (as indicated by Sefcek & Sacco). I also acknowledge a different point made by Johnson & van Vugt, about male aggression towards women, that it is underestimated if men's aggression to out-group women, during warfare (and in other contexts), is ignored.

R3.3. Extending the range of human behavior covered by sexual selection

Several commentaries seek to extend the range of human behavior that could be explained by sexual selection, without necessarily extending the concept itself. These included artistic and other displays (Sefcek & Sacco), the behavior of homosexual males (Dickins & Sergeant), and the large sex difference in “conduct disorders” (Dickins & Sergeant) – that is, antisocial disorders in children and adolescents, which are two and a half times more frequent in males than in females. All of these seem appropriate extensions of the scope of sexual selection in its original meaning.

Schredl's interesting and unusual commentary extends the scope of sexual selection to the content of dreams. He reports that dreams reflect real-life sex differences in not only sexual activity but also physical aggression. These findings parallel research on homicidal fantasies: Men report more of these than women do, and their fantasies are more frequent and persistent (Crabb Reference Crabb2000; Kenrick & Sheets Reference Kenrick and Sheets1993). Schredl also notes that the sex differences in dream content have been stable since the 1940s and were unchanged during the time when societal gender roles changed. Consistent with this, Kenrick & Griskevicius also note that homicide rates do not differ with changes in gender attitudes from 1960s onwards. The sex difference in physical aggression in dreams becomes pronounced from around 17 years of age, consistent with analyses of violent crimes.

Several commentators suggest that my sexual selection analysis could have been further refined by paying attention to the particular forms of aggression that sexual selection would predict. Buss does so in relation to specific adaptive problems (associated with his modular framework: see sect. R4). Kempenaers & Forstmeier set out a number of specific predictions about male aggression derived from sexual selection. These provide useful hypotheses for future studies.

Corr & Perkins raised the issue of whether the sexes differ in different forms of aggression, distinguishing between competitive and provoked (and also predatory) aggression. There have been many suggested subtypes of aggression, and these are discussed elsewhere (Gendreau & Archer Reference Gendreau, Archer, Tremblay, Hartup and Archer2005). The more commonly used distinction between proactive and reactive aggression (Vitaro & Brendgen Reference Vitaro, Brendgen, Tremblay, Hartup and Archer2005) is likely to correspond to Corr & Perkins' competitive and provoked aggression. It would therefore be interesting to enquire whether men and women differ more in either proactive or reactive aggression. These forms are typically not distinguished in most of the available studies involving sex differences, although many of the items on the commonly used physical aggression scale of the Aggression Questionnaire (A.H. Buss & Perry Reference Buss and Perry1992) do seem to refer to reactive forms of aggression. Studies specifying reactive and proactive aggression in adolescents and adults (Archer & Thanzami, in press; Raine et al. Reference Raine, Dodge, Loeber, Gatzke-Kopp, Lynam, Reynolds, Stouthamer-Loeber and Liu2006) have involved only males. It is likely that men show more of both of these forms of aggression, and that it is its physical nature, rather than the presence or absence of an immediate provocation, that is important for the sex difference. Proactive and reactive aggression are often highly correlated, indicating that they co-occur in the same individuals (e.g., Brown et al. Reference Brown, Atkins, Osborne and Milnamow1996; Dodge & Coie Reference Dodge and Coie1987; Raine et al. Reference Raine, Dodge, Loeber, Gatzke-Kopp, Lynam, Reynolds, Stouthamer-Loeber and Liu2006), although they are related to different variables (Raine et al. Reference Raine, Dodge, Loeber, Gatzke-Kopp, Lynam, Reynolds, Stouthamer-Loeber and Liu2006; Scarpa & Haden Reference Scarpa and Haden2006).

R3.4. Beyond sexual selection: Inter-group competition

One suggested extension of sexual selection that requires longer consideration is inter-group conflict. The principles of sexual selection apply to a range of species, and are therefore not specifically designed for the human case where inter-group competition is widespread. I mentioned inter-group competition as a limitation to the existing psychological and criminological evidence on male aggression. Recent research (e.g., van Vugt et al. Reference van Vugt, De Cremer and Janssen2007; Wilson & Wrangham Reference Wilson and Wrangham2003; Wrangham Reference Wrangham1999; Wrangham & Wilson Reference Wrangham and Wilson2004) has considered coalitional aggression from an evolutionary viewpoint. Several commentators (e.g., Johnson & van Vugt) raise this issue as a limitation to a sexual selection view of sex differences in aggression that is based on aggression between individuals. How to integrate evolutionary approaches to inter-group and intra-group aggression is an important issue.

Benenson argues that the contrasting structures of male and female social groups lead to a more complex consideration of sexual selection than that proposed in the target article. In particular, she characterizes males as oriented primarily towards inter-group competition, with mechanisms for minimizing within-group aggression; female competition is characterized as being aimed at removing rivals. This is a very interesting and elegant extension of my target article, but it has the drawback of being inconsistent with evidence on adult friendship patterns.

Benenson's argument involves the contrasting nature of male and female groups, which has been noted for some time in the developmental literature (e.g., Archer Reference Archer and McGurk1992a; Maccoby Reference Maccoby1988; see also commentaries by Johnson & van Vugt and Pellegrini). Benenson extends this to claim that “human males invest more than females in, and exhibit more tolerance towards, same-sex peers.” She goes on to characterize males' same-sex relationships as being interconnected or group-based (in contrast to those of females, which are not). Benenson's citations in support her position rely heavily on child samples. Research on adult friendship patterns suggests a different female pattern to one characterized by the lack of a group basis, and fragile friendships. In many ways, both sexes have similarly patterned friendship networks, with certain reliable differences. Men's friendships tend to be centered on shared activities whereas women's involve shared feelings. In contrast to Benenson's statement that “Human females' same-sex dyadic relationships endure for shorter periods than those of males,” women's dyadic pairs are found to be more intimate, disclosing, and satisfying than those of men (e.g., Caldwell & Peplau Reference Caldwell and Peplau1982; Dindia & Allen Reference Dindia and Allen1992; Rubin Reference Rubin1985; Wright Reference Wright1988).

The social exclusion Benenson mentions refers to the outcome of female aggression in young children. It is, however, advantageous for both sexes to exclude competitors, and in the nonhuman world social exclusion results from direct aggression. In rodents, it is typically the fate of those who are displaced following territorial disputes (Archer Reference Archer and Crook1970). Disruptive young male rhesus monkeys may be driven from their social group before puberty by adult females (Suomi Reference Suomi, Tremblay, Hartup and Archer2005). Throughout human history, social exclusion from a group has been used as a form of punishment (e.g., Ruff Reference Ruff2001). Exclusion is therefore a consequence of aggression, and not connected to one particular form, indirect (or relational) aggression, as Benenson indicates. I considered this type of aggression throughout my target article. It works better in the denser and more elaborate social groups found among women and girls (Green et al. Reference Green, Richardson and Lago1996; Lagerspetz et al. Reference Lagerspetz, Björkqvist and Peltonen1988), than in the dyadic context that Benenson suggests is typical of female networks; here a different form of relational manipulation is found (Archer & Coyne Reference Archer and Coyne2005). In terms of immediate costs, such as physical retaliation, indirect aggression is less risky than direct aggression (Björkqvist Reference Björkqvist1994), but it can indeed be effective in removing a competitor. Recent evidence suggests that women's indirect aggression is targeted particularly at those viewed as posing a reproductive threat – that is, more overtly attractive women (Vaillancourt & Sharma Reference Vaillancourt and Sharma2008) – which is consistent with other analyses, including the perceptive evolutionary account of indirect aggression by Figueredo et al. To sum up, although Benenson's commentary opens up an interesting avenue for future discussion and research, it conflicts with the present evidence on men's and women's social networks.

Rather than seeking to characterize men's conflicts as inter-group and women's as within-group, it is more realistic to regard men as engaging in both inter- and intra-group conflicts, as Johnson & van Vugt do. These commentators suggest that I have underestimated the extent of sex differences in aggression by concentrating on within-group aggression. I did acknowledge inter-group aggression in passing, and I referred to some of the sources they cite, including warfare being the common cause of male deaths in pre-state societies. There still remains a considerable amount of human aggression that is within-group. It is this that is the major topic of psychological and criminological investigations, and can most readily be linked to sexual selection.

Nevertheless, an additional section on inter-group conflict would undoubtedly have strengthened and extended the scope of my article. The magnitude of the sex difference is clearly much larger for inter-group aggression. Male inter-group aggression is not readily explained in terms of classic sexual selection principles, but is likely to be based on sexually selected features (as Johnson & van Vugt suggest). Male sexually selected characteristics, such as upper body strength and willingness to escalate encounters to dangerous levels, form the basis of simple forms of inter-group violent contests. It is but a short step from the coalitional aggression found in nonhuman primates to the simpler forms of group violence characteristic of humans. Van Vugt et al. (Reference van Vugt, De Cremer and Janssen2007) identified a further psychological adaptation for such inter-group conflicts in men: Men are more likely than women to cooperate in the face of inter-group competition (see also target article). On the face of it, this would tend to decrease intra-group male competition. Yet, higher levels of male than female aggression are nevertheless reported in individual-level studies. It has to be noted, however, that studies on sex differences in inter-group aggression and conflict are only just beginning, and their data base is as yet very small compared to that covering within-group aggression, and it largely consists of laboratory studies. There are other studies (see Kenrick & Griskevicius) indicating that activating mating motives can lead men to conform less to group norms. The full picture regarding intra- and inter-group conflict awaits clarification by future research.

Browne's commentary also concerns coalitional aggression, although not specifically about the nature of men's and women's groups. He argues that the sex differences outlined in the target article would prevent women from being as effective or as suited to combat as men. To be convincing, Browne's argument needs to include the sort of “male warrior” adaptations discussed by Johnson & van Vugt. One finding that may be relevant in this context is the early sex difference in preferences for games and entertainments involving killing (Benenson et al. Reference Benenson, Carder and Geib-Cole2008).

There is a counterargument to Browne's position. Both physical and psychological sex differences – even those with large effect sizes such as upper body strength – overlap, so that there will always be some women who possess these characteristics more than some men do. As stated, Browne's argument seems to neglect such individual variability, and the importance of motivation, general physical training, and specific combat training. Consequently, it may well be that most men are more fitted for combat than most women, but this still leaves the few exceptions. It is likely that throughout history gender attitudes have converted these overlapping differences into categorical ones by generalizing from the typical sex difference to one that is prescribed by social rules. The argument that women can form effective combat units is usually illustrated by the example of the fighting units of the West African Dahomey kingdom in the eighteenth and nineteenth centuries (as noted by Eagly & Wood). It is also likely that women have fought effectively alongside men, particularly in twentieth-century liberation armies. The position that women lack psychological adaptations for inter-group conflict could simply mean that it is more difficult, but not impossible, to integrate women as effectively as men into combat roles. How the infrequent cases discussed by Eagly & Wood, and more extensively by Goldstein (Reference Goldstein2001), have overcome such adaptations provides an interesting challenge for future evolutionary researchers in this field.

R4.1. Tinbergen's four questions

The first question, regarding “survival value,” corresponds to the functional evolutionary explanations found in evolutionary psychology. As set out in Buss's commentary and elsewhere (e.g., Buss Reference Buss2004; Buss & Duntley Reference Buss, Duntley, Schaller, Kenrick and Simpson2006; Buss & Shackelford Reference Buss and Shackelford1997b), Buss's approach is to list a number of survival and reproductive problems, and to argue that there are specific mental modules designed to solve these problems. Tinbergen (Reference Tinbergen1963) described three other types of explanation in addition to survival value (or function). One was evolution, or phylogenetic history, which is typically applied to the evolution of display characteristics in nonhumans (see Archer Reference Archer1992b, pp. 149–75), and in humans, to nonverbal signals (e.g., Andrew Reference Andrew1963; van Hooff Reference van Hooff and Hinde1972). The evolutionary background of the mechanisms underlying aggression, and how they might differ in the two sexes, is particularly important for distinguishing an ethologically based evolutionary approach from the modular approach, and also for addressing issues raised in other commentaries.

From an ethological viewpoint, the modular approach seems to have omitted consideration of the process of natural selection. New adaptive mechanisms are built into already-evolved structures, rather than being designed as specific mental modules for new adaptive problems. The neuroendocrine mechanisms underlying aggression and aggressive displays, and the effector organs used in fighting, illustrate this process. Humans still possess the same basic emotions derived from a phylogenetically ancient fight-flight system: more recently-evolved neural circuitry has been incorporated into older systems rather than forming new and separate structures.

Development and causation are the other types of explanation described by Tinbergen (Reference Tinbergen1963), and considerations of the process of evolutionary change imply that the links between these and function are not necessarily straightforward. Any particular adaptive function may be fulfilled by one of several possible mechanisms, and may develop in different ways. Many examples can be found in studies of adaptive behavior in animals, such as kin recognition and mate attraction. The only requirement of the developmental and causal mechanisms underlying adaptive behaviour is that they produce a particular end-product regularly in the environment in which that behaviour evolved.

Thus, causal mechanisms are not necessarily mapped on to specific adaptive problems (as in the modular view), although this does occur in some cases. Perhaps the mechanism closest to this is the ethological concept of the fixed-action pattern, whereby a specific stimulus evokes a specific response (e.g., Lorenz Reference Lorenz and Martin1971). Mechanisms underlying aggression may range from this sort, as in the case of pain-induced aggression (Archer Reference Archer1988; Reference Archer1989–90), through emotional reactions, which can be seen as filling “the gap between fixed action patterns and impeccable rationality” (Johnson-Laird & Oatley Reference Johnson-Laird and Oatley1992, p. 206), to more specific decision-processes, responsive to local cost-benefit contingencies. The mechanisms underlying human aggression are a complex mixture of these processes, as evidenced by theories that emphasize emotions (Archer Reference Archer, Bateson and Klopfer1976; Reference Archer1988; Baumeister et al. Reference Baumeister, Smart and Boden1996; Berkowitz Reference Berkowitz1962Reference Berkowitz2008; Dollard et al. Reference Dollard, Doob, Miller, Mowrer and Sears1939), and those stressing rational goal-directed behavior (Tedeschi & Felson Reference Tedeschi and Felson1994). Aggressive behavior has a very old phylogenetic history, so that more recently evolved mechanisms are built upon, and arise out of, older systems, rather than being created anew for each particular adaptive problem.

R4.2. Development

From the previous section, it should be clear that I do not agree that “evolutionary theory predicts the early emergence of sex differences in direct aggression” (Cashdan). Rather, in the target article I stated: “the course of development cannot be precisely specified from an evolutionary explanation” (sect. 2.1.2). Nevertheless, certain possibilities can be ruled out: for example, evolved sex differences in aggression could not be totally reliant on culture.

Eagly & Wood comment that I attributed to them the view that “sex difference should start small and increase with age through childhood, coincident with the cumulative influence of socialization.” This was a statement taken from Table 1, summarizing predictions derived from sexual selection and social role theories. In the text, it was elaborated, where I stated that this was a reasonable inference from social learning accounts of sex differences in aggression, which emphasize the cumulative influence of socialization through childhood. Eagly & Wood note that social role accounts have “remained silent” on the issue of age trends in aggression. Yet social role (and “biosocial”) accounts incorporate social learning as one of the mechanisms underlying the transmission of social roles, thus making it reasonable to infer that they are in agreement with the processes described in social learning accounts, in particular the gradual learning over time of aggressive behavior from role models.

Tremblay & Côté make the same point about social learning in their commentary, noting that “Counterintuitively, the frequency of physical aggression declines during periods (middle childhood and adolescence) when exposure to physical aggression in the neighborhood and the media increases dramatically.” As they put it, no social learning or social role theorist has suggested that infants learn to aggress from their parents and subsequently unlearn this when they become exposed to influences from outside the family and from the media. From this perspective, it is difficult to understand why Behme thinks that “it is difficult to see how a theory of sexual selection can explain that the amount of aggressive behaviour decreases steadily throughout childhood.” She goes on to argue that “we should expect the highest values of aggression […] at the time when mating occurs.” There is no requirement for the various traits associated with inter-male competition to arise at the same time during development, only that they are all in place by the time of maximum inter-male competition. Behme's comment again assumes that functional origins and development are necessarily closely linked. It also conflates the assessment of aggression based on frequency of acts (which is greatest at 2–3 years of age) with their damaging nature (which peaks in young adulthood). It is the effectiveness of physical aggression that is likely to result in success in reproductive-related competition, not how often feeble motor acts occur. As noted by Tremblay et al. (Reference Tremblay, Japel, Pérusse, McDuff, Boivin, Zoccolillo and Montplaisir1999), and in subsequent work (e.g., Côté et al. Reference Côté, Vaillancourt, LeBlanc, Nagin and Tremblay2006), the decline in physical aggression with age during childhood comes about through the learning of alternative responses to physical aggression, a process that differs among individuals and between the sexes, and is influenced by a range of maternal characteristics (Tremblay & Côté).

Pellegrini poses the question of why there is no early sexual dimorphism in size and strength when there is early dimorphism in behavior. His interesting suggestion is that sex differences in physical activity and locomotor play are developmental precursors for the size and strength differences that result from the actions of testosterone at puberty.

Several authors highlight the interactive nature of the development of sex differences in aggression, and note the differential environmental influences on males and females. Tremblay & Côté emphasize the importance of maternal influences, particularly a number of high risk factors such as early childrearing and smoking during pregnancy – these predict high levels of physical aggression by males. They comment that even though such mothers do not themselves show the high levels of physical aggression more characteristic of males, “they have all the risk factors needed to place their male children on a trajectory of chronic physical aggression.” Their sons will fit the pattern of showing a reproductive strategy characterized by a bias towards mating rather than parental effort (sect. 2.7 in the target article). In a similar vein, the work of Boden and his colleagues showed that an early adverse family environment was more predictive of partner violence for men than for women. Men showed less continuity in their aggression than women did (which, Boden notes, would not be expected from a social role view).

R4.3. Causation

Tinbergen's remaining question, in addition to function, evolution, and development, concerns causation, the immediate internal and external influences on behavior. Some commentators (Cashdan, Finkel & Slotter, and Pound et al.) questioned the inference that there should not be a sex difference in anger, the emotion that underlies aggression, although others (e.g., Campbell) accepted it. The reason given in the target article is that both sexes have reasons for becoming angry but the costs and benefits of the behavior resulting from the anger are different. This would fit the finding (discussed below) that the sex difference in physical aggression is mediated by risky impulsivity. It would also be consistent with Daly and Wilson's (1988, pp. 163–86) emphasis on risky strategies underlying the evolved greater male propensity to violent conflicts. Daly and Wilson stated at the beginning of this chapter: “Men become embroiled in dangerous competitive interactions far more often than do women” (p. 163). I suggested that it is the interactions resulting from anger that are more dangerous in men than in women.

Nevertheless, it is possible to see how the motivational mechanism behind angry aggression could have been changed so that men and women differed in the extent to which they become angry. In their “culture of honor” studies, Nisbett and Cohen (Reference Nisbett and Cohen1996) found that it was the angry response to certain provoking situations that differentiated people from an honor culture, and those who were not. If men and women differed in the same way, this would produce a sex difference in anger to honor-related situations, even though general measures of anger might not reveal such a difference. In this respect, the final paragraph of the commentary by Pound et al. is the most pertinent. If it is the state of becoming angry that entails the elevated risk, we should predict a sex difference in the ease of becoming angry in men and women, but only in certain circumstances (e.g., competitive exchanges with members of the same sex). On the other hand, when men perceive physical aggression to have high costs, it is likely that they still experience anger: it is the behavior that is different from when the costs are lower. Applying this to women, the calibration of the level for high costs could simply be set lower. Thus, women's behavior (but not their initial emotional reaction) would be more inhibited by the costs of physical aggression, as Campbell (Reference Campbell1999) suggested. There are pros and cons to both sides of this argument, and while I found the commentary by Pound et al. almost persuasive, I had lingering doubts. Their suggestion provides several interesting avenues for future research.

As indicated in section R4.1, causation is a separate issue to function, and in my view it needs to be addressed in a different way than listing adaptive problems and assuming modules in the mind corresponding to each one (Buss). In line with this, I considered possible mediators of the sex difference in aggression, but noted that there were currently no empirical tests of these. Campbell's commentary describes such a test. In devising a measure of risk, she was careful to include only risks that would have potentially dangerous consequences, as these had been identified as the relevant forms in evolutionary analyses (e.g., Campbell Reference Campbell1999; Daly & Wilson Reference Daly and Wilson1988). Controlling for this measure of risk-taking did eliminate the usual sex differences in physical and verbal aggression.

Finkel & Slotter suggested that my argument concerning the mediation of sex differences in aggression could be extended with reference to “I³ Theory.” This is an analysis of impelling and inhibiting motivational forces. It involves a binary classification of variables that should be included in any motivational model of aggressive behavior. My article concerned those variables that mediated sex differences in aggression, and therefore would not take into account all the inhibiting and impelling forces controlling aggression. I have two observations on I³ Theory. The first is that it seems to be a recasting of a cost-benefit analysis, which was discussed in the target article, and has been applied to the study of several forms of aggression (Archer et al., submitted; Archer & Southall Reference Archer and Southall2009; Rutter & Hine Reference Rutter and Hine2005). The second is that it appears not to be a theory in the sense that specific hypotheses can be generated, but a list of variables that make aggression more or less likely. This may be useful for organizing our thinking of the various influences on forms of aggression, such as partner aggression (Finkel Reference Finkel2007), but it would seem to add little or nothing to how the motivational system underlying aggression is organized. For example, Finkel & Slotter's penultimate paragraph merely recasts what I (and others) have written about the variability in partner violence – that different balances in the societal power of men and women alter the immediate cost-benefit calculus for resorting to physical violence against a partner.

In the target article, I stated that the rising levels of testosterone in male adolescents were not associated with an increase in male aggression, as occurs in other species. Elsewhere (Archer Reference Archer2006b), I argued that the behavior-hormone relationships in humans fit a pattern identified previously by Wingfield et al. (Reference Wingfield, Hegner, Dufty and Ball1990) in birds, known as “the challenge hypothesis.” According to this, testosterone rises at puberty, but to moderate levels that support growth and reproductive functioning and behavior, but it does not directly facilitate aggression. Both sexual arousal and challenges involving young males do lead to short-term rises in testosterone levels, and these facilitate competitive behavior, including aggression. Terburg, Peper, Morgan, & van Honk (Terburg et al.) argue that behavioral studies indicate that testosterone is a mediator of aggression; however, the studies they cite are correlational, and there is abundant evidence that aggressive behavior can lead to increased testosterone levels (Archer Reference Archer2006b; Kemper Reference Kemper1990). Terburg et al.'s commentary goes on to make the important point that interactions between testosterone and cortisol – found in several studies – have been neglected in such analyses: The influence of testosterone appears to be stronger in individuals with low cortisol levels, although in some cases the two hormones increase together (Cohen et al. Reference Cohen, Nisbett, Bowdle and Schwarz1996).

Terburg et al. also mention findings of a positive association between aggression and digit ratio (2D:4D), regarded as a marker for prenatal testosterone levels. From this, they suggest that prenatal testosterone may affect the sensitivity to later circulating testosterone (as it does in the classic studies of the organizing effects of perinatal testosterone). Thus, Terburg et al. draw our attention to two specific and testable variables (cortisol and prenatal testosterone) that might moderate the association between testosterone and aggression from puberty onwards.