The first effects of testosterone on animal aggression are exerted in a critical perinatal period. At this early age, testosterone acts as an organizing principle that makes the brain sensitive to its own effects that are exerted much later, beginning with the pubertal period (Fig. 2.2). This “organizing” role of testosterone was first recognized in conjunction with sexual behavior. It was stated already in the 1960s that “Sex hormones of the embryo affect ultimate sexual behavior in two ways: indirectly by their control of morphogenesis of the genitalia and, in some species at least, directly by their neural-organizer effect” (Money 1965). It was soon discovered that testosterone has similar effects on aggressive behavior: neonatally administered testosterone increased, while neonatal castration decreased adult aggression in mice (Bronson and Desjardins 1968, 1969; Edwards 1969; Peters et al. 1972). This increase in aggression was found to be conditional; instead of leading to an overall increase in aggressiveness, neonatal testosterone in fact sensitizes the neural system to its later effects: the aggression-heightening effects of testosterone in adulthood are facilitated by perinatal exposures to testosterone (Hutchison 1974; Klein and Simon 1991). The early effects of testosterone are so strong that it can induce male-like aggressive behavior in female rodents, provided that these are treated with testosterone in adulthood (Gandelman et al. 1979). Moreover, intrauterine position, e.g., the sex of neighboring embryos, also has a role in establishing adult aggressiveness by virtue of the diffusion of testosterone through the amniotic fluid (Quadagno et al. 1987; Ryan and Vandenbergh 2002). The early organizing effects of testosterone with regard to its later behavioral effects are bound to a specific time window of animal development; this time window is species dependent and short; it encompasses a few hours or days shortly before, during, or after birth (Dixson 1993; Klein and Simon 1991; Motelica-Heino et al. 1993).

These findings demonstrate that there is critical time period in the early development of animals, when testosterone “transforms” the brain such that it becomes sensitive to its own aggression-promoting effects in adulthood. Similar changes were seen in humans, where the prenatal period is also considered crucial for the sexual differentiation of brain and behavior (Warne et al. 1977; Rubin et al. 1981). This assumption is supported by three lines of evidence: (1) congenital adrenal hyperplasia, a condition that is associated with high testosterone levels in early periods of life, results in male-like aggressive behavior in females (Berenbaum et al. 2000; Hines et al. 2002); (2) prenatal exposure to androgens via treatments received by mothers increased aggressiveness in both male and female offsprings (Reinisch 1981); and (3) the length ratio of the second and fourth fingers (a somatic marker of prenatal testosterone exposure called 2D:4D digit ratio) associates negatively with aggressiveness, suggesting that larger testosterone exposure during embryonic life (when digit length is determined) is associated with higher aggression during adulthood (Butovskaya et al. 2013; Hampson et al. 2008; van der Meij et al. 2012). Moreover, girls of opposite-sex twins show a more masculine pattern of aggression than girls of same-sex twins, which may be considered a human analog of the intrauterine position effect seen in animals (Cohen-Bendahan et al. 2005).

Taken together, the findings reviewed above suggest that the developmental effects of testosterone are rather similar in animals and humans. Contradictory findings also exist. Voracek and Stieger (2009), for instance, found no association between 2D:4D digit ratio and aggression, and the first author of the above paper questioned the validity of the findings of van der Meij et al. (2012) based on data-analytic and statistical power issues (Voracek 2013). Such contradictions may be solved by taking aggression types and genetic constitution into consideration. Millet and Dewitte (2009) for instance found that 2D:4D digit ratio predicted aggression in the dictator game (a variant of the ultimatum game presented in Chap. 1) only when an aggression cue was present; in neutral situations, the low (more masculine-like) digit ratio predicted prosocial behavior. In line with this, Benderlioglu and Nelson (2004) found that a low 2D:4D digit ratio predicted aggressiveness in females under high provocation only. These authors deduced aggressiveness from the follow-up letters of subjects solicited to give donations for a fictitious charity organization. Under these conditions, digit ratios predict aggressiveness in women but not in men. In a study investigating trait aggression, the expected correlation with digit ratios was found in males but not in females (Bailey and Hurd 2005). Congenital adrenal hyperplasia by contrast increases aggression in females but not males (Pasterski et al. 2007). In another study, trait aggressiveness was predicted by digit ratios in conjunction with dopamine D4 receptor polymorphism (Butovskaya et al. 2012). Such findings may be considered contradictory, but in our view the differences reflect genetic background-, gender-, and aggression type-related differences in the involvement of testosterone.

Early treatment with testosterone induces male-like aggression patterns in female rodents, which may be considered abnormal by virtue of the second criterion of abnormal aggression (see Chap. 1, Sect. 1.​2.​2). Particularly, the criterion “disregard of species-specific rules” applies here and renders male-like behavior abnormal in females. As a similar phenomenon is noticed in congenital adrenal hyperplasia, the first criterion of abnormal aggression may also be met (mimicking an etiological factor of human abnormal aggression). Nevertheless, aggression is not a major symptom of this condition, which cannot be considered an aggression-related disorder. Moreover, aggressiveness is increased by this condition in females but not in males, a difference likely related to the reversal of sex roles. In contrast to the human case, early testosterone administration increases adult aggressiveness in rodents belonging to both sexes. Therefore, the resemblance between the human condition and the model is only vague. Beyond this equivocal similarity, there is no information on the nature of aggression elicited by early testosterone manipulations. This does not mean that such treatments result in normal forms of aggression in rodents; it only means that the issue was not investigated so far.

The relationship between early androgen hyperfunction and abnormal aggression was almost as poorly studied in humans as in rodents. There are only two studies suggesting a link. Liu et al. (2012) showed that early androgen exposure (as assessed by digit ratios) increased externalizing behavior in boys but not in girls. A similar association was recently found between digit ratios and intimate partner violence (Romero-Martínez et al. 2013). The only “flaw in the scheme” is that neither externalizing behavior nor intimate partner violence is an aggression-related psychopathology. Yet, a 24-year-long study of externalizing symptoms demonstrated that these often develop into, while intimate partner violence is frequently associated with antisocial behaviors; moreover, psychopathy (Fowler and Westen 2011; Reef et al. 2011). This suggests that early androgen hyperfunction may lead to abnormal human aggression, but the issue remains poorly studied.

Perinatal testosterone hyperfunction (1) increases energy metabolism (Quadagno et al. 1977; Hill et al. 2012); (2) increases adult testosterone levels, the aromatization of testosterone to estrogens, and estrogen receptor expression (Ryan and Vandenbergh 2002; Schlinger and Callard 1989; Vaillancourt et al. 2012); (3) stimulates cell growth and differentiation in brain areas related to male-typical behaviors and aggression, e.g., the hypothalamic sexually dimorphic nucleus and its human equivalent, the 3rd interstitial nucleus of the anterior hypothalamus, the ventromedial hypothalamus, and amygdala (Hines 2010; Ryan and Vandenbergh 2002); and (4) upregulates the noradrenergic and dopaminergic innervation of the prefrontal cortex and downregulates the serotonergic innervation of the same area (Butovskaya et al. 2012; Dominguez et al. 2003; Stewart and Rajabi 1994). In addition it masculinizes the serotonergic innervation of the medial preoptic area in female rats (Simerly et al. 1985). Early testosterone treatments also affect the neuronal connectivity (Hines 2010). Importantly, findings in rodents and humans were highly compatible. Some of these testosterone-associated changes may be epiphenomena, but others may be causally linked to changes in aggressiveness.

Albeit castration was used for millennia to pacify domesticated male animals that are otherwise difficult to handle, the first scientific study attesting the role of the testes in aggression was performed by Arnold Berthold in 1847: he showed that castration decreased, while testis transplantation restored aggressiveness in domesticated roosters (cited based on Soma 2006). The particular role of testosterone was demonstrated considerably later by Seward (1945) and Beeman (1947) who showed that puberty-associated increases in testosterone are associated with the emergence of aggressiveness, while aggression suppressed by castration is restored by testosterone, respectively. These basic findings were amply replicated by subsequent research, which revealed further aspects of testosterone action. Seasonal variations in testosterone production covariate with aggressiveness (Butterfield and Crook 1968); current levels of circulating testosterone are in tight correlation with the propensity to behave aggressively (Rose et al. 1971); testosterone injected after the winning or the losing of aggressive encounters increases aggressiveness and decreases submissiveness, respectively, in subsequent encounters; and finally, social rank correlates positively with plasma testosterone in social groups (Lucion et al. 1996; Solomon et al. 2009; Trainor et al. 2004). This set of congruent findings resulted to a strengthened belief that there is a clear-cut causal relationship between testosterone production and aggression in animals—as opposed to humans, where this interaction is much less certain.

In fact, however, the testosterone-aggression link may be perceived as being just as volatile in animals as in humans. Albeit castration reduced aggressiveness in most studies published so far, there still exist carefully performed studies where this effect was not seen (California mice: Trainor and Marler 2001; lizards: Moore and Marler 1987; marmosets: Dixson 1993; prairie voles: Demas et al. 1999; rock hyrax: Manharth and Harris-Gerber 2002; Siamese fighting fish: Weiss and Coughlin 1979; song sparrows: Soma et al. 2000; Siberian hamster: Jasnow et al. 2000; wood rats: Caldwell et al. 1984); moreover, castration increased aggression in some species under certain circumstances (in laboratory mice, castration increases aggression towards lactating females: Haug et al. 1984; in mandarin voles, castration leads to an overall increase in aggressiveness: He et al. 2012). Thus, surprising effects of castration were seen in all major vertebrate classes and in a variety of taxa within mammals, including rodents. Interestingly, testosterone production was not totally disrupted from aggression control in these species; e.g., winning-related increases in testosterone facilitated aggression in subsequent encounters in California mice, despite the fact that castration did not reduce aggression in this species (Fuxjager et al. 2011).

Similarly, strange effects were noticed in studies investigating the correlation of aggressiveness with plasma testosterone levels, or the effects of testosterone administration. For example, basal plasma levels of testosterone did not show correlation with aggressiveness in Siberian dwarf hamsters (Castro and Matt 1997), wild-type rats maintained in the laboratory (Everts et al. 1997), the red jungle fowl studied in the laboratory (Johnsen and Zuk 1995), and in laboratory rats in the visible burrow system (Blanchard et al. 1995). Again, the relationship between testosterone and aggression was not totally disrupted in these models. In the study by Johnsen and Zuk (1995), basal levels did not, while acute increases in testosterone did correlate with aggressiveness; in the study by Blanchard et al. (1995), submission was not correlated with low testosterone when group members living in the visible burrow system were preselected for high aggressiveness, but the expected relationship was found when group members were not preselected for aggressiveness. As it regards testosterone administrations, these did not affect aggression in some studies (ground squirrels, postmating period: Millesi et al. 2002; male lambs, familiar environment: Ruiz-de-la-torre and Manteca 1999; European stonechats, nonbreeding season: Canoine and Gwinner 2002); moreover, testosterone decreased aggression in Siberian hamsters (Jasnow et al. 2000). As shown above for other models, aggression was not totally disrupted from regulatory control by testosterone, as treatments ineffective in the nonbreeding season or in familiar social environments became effective in the breeding season and in unfamiliar social environments (Canoine and Gwinner 2002; Millesi et al. 2002; Ruiz-de-la-torre and Manteca 1999).

The studies briefly reviewed above (and a series of others not cited here) show that there is no clear cause–effect relationship between testosterone production and aggressiveness. Even more strikingly, strange effects were noted in species where testosterone clearly controls aggression. For example, castration dramatically decreased, but did not abolish biting attacks in castrated mice; in addition, the surgery did not affect the frequency and duration of offensive threats (Clipperton-Allen et al. 2011). In male laboratory rats, Albert et al. (1986) reported that the presence of females increased aggression in both intact and castrated males. Thus, the absence of testosterone did not prevent males from being aggressive, not even in species where the relationship between aggressiveness and testosterone was believed to be clear-cut.

Asking the same question in a more general way: does a considerable body of contradictory findings question the role of testosterone in aggression? From a simplistic view it certainly does: the assumption that testosterone controls aggression is negated by these findings. However, testosterone does not control aggression: it affects the properties of neuronal circuits, which are under multiple influences. In other words, the neural and behavioral responses to social challenges are not defined by testosterone, but by the overall outcome of the multitude of concurrent influences. Interactions of this kind were systematically mapped in a few species, and several theories were advanced to explain the discrepancies reviewed above (Fig. 2.3).

The first theory of this kind, the “challenge hypothesis” put forward by Wingfield et al. (1990), made a distinction between breeding season-related changes in testosterone production and those elicited by male–male interactions. The theory was based on data obtained in monogamous birds. It was hypothesized that maintaining high testosterone levels for the whole breeding season is detrimental in terms of both increased metabolic costs resulting from enhanced behavioral activity and reduced parental care, a “by-product” of enhanced testosterone production. Metabolic and parental trade-offs are solved by limiting testosterone production to the early phases of sexual competition and periods of social challenge in the parental period. Thus, the disruption of the testosterone-aggression link is ostensible; in fact, testosterone does control aggression, but its production is controlled by “needs.” This influential hypothesis was supported by a number of field and laboratory observations in birds and mammals (Cavigelli and Pereira 2000; Ferree et al. 2004; Muller and Wrangham 2004; Ros et al. 2002) and was exploited to explain the testosterone-aggression link in humans (Archer 2006).

The second theory addressing the disruption of the testosterone-aggression link was also based on data obtained in birds. This theory stipulates that nonbreeding season-related aggression (which is rather high in many species) is not controlled by testosterone but by adrenal steroids, e.g., dehydroepiandrosterone (Soma 2006; Soma et al. 2008). The mediator of this mechanism is melatonin, which is secreted in high amounts during the winter. Under the influence of this hormone, dehydroepiandrosterone is converted to estrogen (one of the active metabolites of testosterone) in the brain. Albeit based on bird data in the first place, this theory also assumes validity for other vertebrates (Soma et al. 2008). Albeit humans are no seasonal breeders, dehydroepiandrosterone blood levels correlate with certain forms of aggressiveness, suggesting that the mechanisms operating in animals are present in humans (Buydens-Branchey and Branchey 2004; Dmitrieva et al. 2001; van Goozen et al. 1998a, b, 2000a, b). Noteworthy, however, dehydroepiandrosterone does not affect aggression in a species that is critical for this hypothesis—the Siberian hamster, a seasonal breeder which shows high levels of aggression in the nonbreeding season (Jasnow et al. 2000; Scotti et al. 2008).

Studies in Syrian and Siberian hamsters gave rise to the third hypothesis. It was suggested that in these species, aggression is controlled by melatonin per se (Demas et al. 2004; Jasnow et al. 2002; Wang et al. 2012a, b). It remains to be elucidated how melatonin affects aggression. One series of evidence suggested that melatonin downregulates neuronal nitric oxide synthase activity independent of gonadal steroid hormones; according to this hypothesis, melatonin increases aggression in the nonbreeding season by reducing brain nitric oxide levels (Bedrosian et al. 2012; Wen et al. 2004). It is possible that the effects of melatonin are supported by testosterone synthesized from dehydroepiandrosterone under the effects of melatonin (Demas et al. 2004; Scotti et al. 2008, 2009). Taken together, aggressive behavior is independent of testosterone in certain species of hamsters, where aggression seems to be controlled by melatonin, likely by its effects on both nitric oxide synthesis and the conversion of adrenal steroids into testosterone. The consequences of this line of research for understanding aggression—beyond certain hamster species—are dual. Firstly, it directed attention towards the role of nitric oxide in aggression (Nelson et al. 2006); secondly, and more importantly for the present work, it revealed that the control of aggression may be overtaken by melatonin under certain conditions. Although this phenomenon received little attention in human research, some reports suggest that the aggression-promoting effects of melatonin are present in humans. For example, demented patients treated with melatonin showed increased aggression (Haffmans et al. 2001). Melatonin secretion may also be involved in self-injurious behavior (Herman 1990).

Taking into account the above, the response to the question asked in the title of this section is dual. Yes, testosterone affects aggression by interacting with brain mechanisms that control aggression. No, testosterone is not the only mechanism that controls aggression; moreover, there are contexts that render aggression independent of testosterone production. Under the influence of these contexts, alternative mechanisms are activated that enable aggressiveness to be expressed at high levels despite low levels of plasma testosterone.

Testosterone seems to play a role in mice selected for aggressiveness. Short attack latency mice show higher levels of testosterone than long attack latency mice (the timid selection line) (van Oortmerssen et al. 1992; Compaan et al. 1993a, b, 1994a, b). Similar findings were obtained in Turku aggressive mice and cows selected for fighting ability (Plusquellec and Bouissou 2001; Sandnabba et al. 1994). The impact of testosterone seems weaker than that of genetic selection, because castration did not cancel genetically determined behavioral differences; however, aggression was decreased more strongly in short attack latency mice, suggesting that testosterone has a role in their abnormal aggression (van Oortmerssen et al. 1987). Subsequent research showed that the perinatal organizational effects of testosterone and adult brain aromatase activity are also altered in selection lines, further supporting the notion that abnormal aggression in these lines has a testosterone component (Compaan et al. 1992, 1993a, b, 1994a, b; Sandnabba et al. 1994; Sluyter et al. 1996).

The interaction between testosterone and abnormal aggression was also studied in the alcohol model of abnormal aggression. Two findings deserve attention: (1) the early organizational effects of testosterone and the effects of alcohol on aggression seem to interact because neonatal testosterone administration increased, while neonatal castration decreased the effects of alcohol on aggression in adulthood (Lisciotto et al. 1990; Winslow et al. 1988); (2) in adult animals (not treated in the neonatal period), there was a significant positive interaction between plasma testosterone and the effects of alcohol on aggression albeit this interaction was somewhat less robust (DeBold and Miczek 1985; Winslow and Miczek 1988).

In the case of other models, the role of testosterone in abnormal manifestations of aggression was not directly investigated (e.g., by castration or testosterone administration). Noteworthy, however, testosterone levels were measured in some of the models. Testosterone production was not altered in the maternal separation, postweaning social isolation, rat and hamster early subjugation, low-anxiety, and anabolic-androgenic models (Akbari et al. 2008; Amstislavskaya et al. 2013; Cunningham and McGinnis 2008; Ferris 2003; Ferris et al. 2005; Ortiz et al. 1984; Salas-Ramirez et al. 2008, 2010; Shimozuru et al. 2008; Veenema et al. 2006, 2007a, b; Veenema and Neumann 2009). In the high-anxiety model, basal levels remained unchanged, but the testosterone response to aggression was abolished (Veenema et al. 2007b). In the pubertal stress model, changes were seen not in testosterone levels per se but in the testosterone/corticosterone ratio increases (Márquez et al. 2013). In other models, e.g., the glucocorticoid dysfunction and repeated victory models, plasma testosterone levels likely increase but contradictory findings also exist (Balasubramanian et al. 1983; de Boer et al. 2003; Feek et al. 1989; Gao et al. 1996; Kalra and Kalra 1977; Poggioli et al. 1984).

In summary, the roles of testosterone were adequately studied in two models of abnormal aggression: in mice selected for aggressiveness and in the alcohol model. Testosterone appears to promote abnormal forms of aggression in both models. Indirect evidence suggests that the same is for the high-anxiety selection line, pubertal stress, glucocorticoid dysfunction, and repeated victory models, while a role for testosterone appears unlikely in other models. We mention, however, that this evidence is indirect.

The role of testosterone in human aggression is questioned and contested based on a series of arguments as follows: (1): females readily show aggressiveness despite their low testosterone levels; (2) aggression is not increased at puberty when testosterone levels increase; (3) high- and low-aggression individuals do not consistently differ in serum testosterone; (4) aggression does not increase in hypogonadal males when exogenous testosterone is administered to support sexual activity; (5) castration or antiandrogen administration to males is not associated with a consistent decrease in aggression; and (6) the replication of positive findings is often difficult. This set of contra arguments was formulated based on Albert et al. (1993); the reviews by Archer (1991) and Book et al. (2001) were also used. While studies that attest the role of testosterone in human aggression cannot be ignored, and these show that testosterone does control aggression, all the contra arguments listed above are valid.

Although less aggressive than males overall, females readily engage in aggressive conflicts in a variety of situations (aggressive response to hypothetical conflict situations: Reinisch and Sanders 1986; aggressive responses in the Taylor Aggression Paradigm: Böhnke et al. 2010a, b; bullying in prisons: Archer et al. 2007; criminal violence and homicide: Heide and Solomon 2009; dating violence: Cercone et al. 2005; intimate partner violence: Williams et al. 2008; substance dependence-related aggression: Bacskai et al. 2011). Gender differences were significant in most studies, but were not large. For example, the physical aggression score on the Buss-Perry Aggression Scale was 16.92 ± 0.69 in males and 14.63 ± 0.63 in females; gender differences were even smaller in substance-dependent subjects (males, 23.43 ± 0.49; females, 22.77 ± 0.89). In the study by Archer et al. (2007), males scored significantly higher on the physical aggression score of the Response to Victimization Scale (10.51), but the score of females was also considerable (7.58). Moreover, the score for displaced physical aggression was somewhat (nonsignificantly) higher in females (7.53) than in males (7.09). In the study by Böhnke et al. (2010a, b), unprovoked males and females showed comparable levels of aggression in the Taylor Aggression Paradigm; when provoked, males responded more aggressively, but females also increased their aggressive responses. Moreover, dating-related physical assault was significantly more frequent in female than in male undergraduate students (Cercone et al. 2005). Thus, females are almost as aggressive as males; this assumption is supported by meta-analyses showing that gender differences in aggression are consistent but are moderate in magnitude; the small effect size holds true for both verbal and physical aggressions (Hyde 2014).

The pubertal raise in testosterone production is accompanied by a rise in aggression in animals but not in humans. Moreover, physical aggression decreases in puberty as compared to childhood. While children primarily use physical aggression to resolve disputes, they gradually shift their behavior towards verbal (intimidation) and relational aggression to avoid the risk of retribution (Bjoerkqvist et al. 1992; Tremblay 1999). As a consequence, the positive correlation between testosterone and aggression seen in 12–13-year-old boys is lost by the age of 15–16 (Turner 1994), and plasma testosterone is linked to social dominance but not to physical aggression in early adolescence (Schaal et al. 1996).

Regarding the correlation between testosterone levels and aggression in males, we will avoid a “battle of references” (the counting of positive and negative findings) by taking examples from studies where the correlation was found significant. A common characteristic of such studies is that while the correlation is significant overall, individual correlation points are rather dispersed; consequently, some subjects are aggressive on the background of low testosterone, while other subjects show low aggression scores despite elevated plasma levels of testosterone. Such dispersed correlations or highly overlapping plasma testosterone ranges were found in a number of paradigms (aggressive delinquency: Banks and Dabbs 1996; Buss-Durkee Hostility Inventory: Ehrenkranz et al. 1974; Point Subtraction Aggression Paradigm: Carré et al. 2010; reactive aggression: Benderlioglu et al. 2004; response to anger faces: Wirth and Schultheiss 2007; venturesomeness in personality disorders: Coccaro et al. 2007a, b).¹ The conclusion deriving from this analysis is supported by the meta-analysis of Book et al. (2001): testosterone is a significant but weak contributor of aggressive behavior. Finally, a number of studies attest that increased testosterone production in women (due, e.g., to polycystic ovarian syndrome), and hypogonadism in males, and testosterone-based corrective treatments have no significant effects on aggressiveness (Finkelstein et al. 1997; O’Connor et al. 2002; Shufelt and Braunstein 2009; Stanworth and Jones 2008; Weiner et al. 2004).

These are serious arguments in favor of the idea that the role of testosterone in human aggression is minimal. Nevertheless, a substantial body of evidence demonstrates that under particular conditions, the effects of testosterone are rather strong (see Sect. 2.2.5). This conflict prompted a series of speculations on how this endocrine factor may affect aggression strongly such that its overall effects remain small.

A simplistic approach would suggest that the conflict is explained by errors in investigation. However, the large number of studies and the high scientific level at which many of these were performed excludes this possibility. The volatility of findings is not due to the volatility of methodologies, even if there was a permanent need to improve these. It occurs that the reason of contradictions lies in the complexity of the phenomenon. In fact, the question is not whether testosterone controls aggression or not; the real question asks about the conditions and circumstances that make testosterone operative or ineffective.

Taken together, these theories suggest that the testosterone-aggression relationship cannot be expected to be linear. Studies where no association between testosterone and aggression was found may fall short at considering conditions that influence testosterone efficacy. In other words, negative findings may have turned out positive if conditions other than testosterone were also taken into account.

It was recognized very early that aggression is stressful and leads to glucocorticoid production in a variety of species (fish: Earley et al. 2006; lizards: Woodley et al. 2000; mice: Bronson and Eleftheriou 1965; quails: Ramenofsky 1985; rats: Schuurman 1980). In the rat resident-intruder test, the increase in glucocorticoids was similar in the two contestants at the beginning of the encounter, but values rapidly returned to basal levels in the winner while remaining high in the loser once dominance relationships were established.

Disparate early findings suggested that corticosterone secreted in response to the encounter might have an impact on ongoing behavior, as ACTH and corticosterone treatments increased aggressiveness while corticosterone synthesis inhibition decreased it (Brain et al. 1971; Bronson and Eleftheriou 1965; Munro and Pitcher 1985; Mainardi et al. 1987). Similar findings were obtained later in fish, where aggressiveness tightly correlated with corticosterone levels that preceded the aggressive encounter and acute treatment with glucocorticoid receptor antagonists decreased aggressiveness (Chang et al. 2012; Schjolden et al. 2009). In two parallel studies, Hayden-Hixon and Ferris (1991a, b) identified certain subregions of the hypothalamus as the main sites of the effect of corticosterone on attack behavior. In view of the fact that the hypothalamus is tightly involved in the control of biting attacks in all the species studied so far (Haller 2013), these findings suggest that corticosterone directly influences brain mechanisms that control attacks.

The existence of a “positive feedback loop” between aggression and HPA-axis activation was challenged in the early 1980s by the discovery of the genomic mechanisms of glucocorticoid action. The general assumption by that time was that glucocorticoids exert their effects only by the genomic mechanisms. As these mechanisms need time to take effect (at least 30 min but usually much more) and, in addition, the increase in glucocorticoids per se needs time (about 5 min), it was assumed that glucocorticoids cannot influence the aggressive acts that triggered their release, because by the time when plasma glucocorticoids increased and the genomic mechanisms were activated, most of the aggressive acts of dyadic encounters were already carried out. As a result, there remained nothing else to be influenced by glucocorticoids than submission, i.e., the consequence of losing the encounter. This belief was reflected in a series of studies, even in those where glucocorticoid synthesis blockade clearly decreased aggressiveness, and glucocorticoids increased aggressiveness (Munro and Pitcher 1985). Later on, a series of studies demonstrated that the rapid effects of glucocorticoids are real and several non-genomic mechanisms were also identified (de Kloet et al 2008a, b; Groeneweg et al. 2011; Haller et al 2008).

In a series of studies on the rapid interactions between glucocorticoids and aggression, we showed first that psychosocial encounters per se are sufficient to increase plasma glucocorticoids, i.e., the mere sensory contact with opponents is able to activate the HPA-axis (Haller et al. 1995). Similar findings were obtained in free-ranging animals as well (Landys et al. 2010; Silverin 1993). As aggressive encounters are preceded by a series of nonsocial and social behaviors that can be perceived as preparations for fights but are not aggressive themselves, these findings demonstrated that glucocorticoids are increased before the actual fights are started. In the following, we showed that natural variations in glucocorticoid synthesis, e.g., those related to ultradian and diurnal variations, are paralleled by changes in aggressiveness when rats are challenged in the resident-intruder test (Haller et al. 1998a, b, c, 2000a, b). Particularly, rats were significantly more aggressive in the increasing phase of their corticosterone fluctuation than their counterparts in the decreasing phase of their corticosterone fluctuations. Moreover, glucocorticoid enhancement-induced aggressiveness was normalized when the natural increase in glucocorticoid secretion was counteracted by glucocorticoid synthesis blockade or by glucocorticoid receptor antagonists. As the time frame of these hormonal and behavioral changes was less than 20 min, these findings indirectly supported the notion that glucocorticoids rapidly promote aggressiveness via non-genomic mechanisms. In the next step, we studied the interaction between glucocorticoid secretion and biting attacks that were elicited by the electrical stimulation of the hypothalamic attack area (Kruk et al. 2004). We found that the stimulation of the attack area dramatically increased corticosterone secretion while corticosterone decreased the electric threshold of attacks, i.e., facilitated the behavioral response. This suggested that there was a fast positive feedback loop between the adrenocortical stress response and brain mechanisms involved in aggressive behavior, the time frame of this interaction being approximately 10 min. Finally, we demonstrated that within 2 min, corticosterone administration restored aggressiveness that was downregulated by corticosterone synthesis inhibition (Mikics et al. 2004). This fast effect of corticosterone was resistant to protein synthesis inhibition, clearly demonstrating both the rapid aggression-promoting effect of glucocorticoids and the non-genomic nature of the mediating mechanism.

Taken together, the findings reviewed above show that (1) the HPA-axis is activated by encounters with unfamiliar individuals before the actual fights start and (2) glucocorticoids secreted in response to the encounter rapidly promote aggressiveness. Actually, aggressiveness is expressed concomitantly and under the supporting influence of glucocorticoids. In a subsequent study we showed that corticosterone does not affect aggressiveness in established colonies of rats, i.e., in colonies where the general level of aggressiveness is low and animals are not under the pressure of an acute social challenge (Mikics et al. 2007).

Thus, glucocorticoids rapidly and non-genomically promote aggressiveness specifically under conditions of a social challenge.

While acute surges in glucocorticoids promote aggressiveness in socially challenged animals, the effects of chronic increases in glucocorticoids are opposite.

As shown above, the aggressive encounter is stressful for both contestants; however, plasma glucocorticoids rapidly decrease after winning, but remain elevated in losers (fish: Höglund et al. 2002; lizards: Summers et al. 2003; mice: Bronson and Eleftheriou 1965; monkeys: Yodyingyuad et al. 1985; rat colonies: Blanchard et al. 1995; rat dyadic encounters: Schuurman 1980; snakes: Schuett and Grober 2000). In an experiment involving opponents with contrasting levels of aggressiveness, we showed that the effects of losing do not depend on the prefight status of individuals; plasma corticosterone levels were rapidly reduced in intruders that won the encounter, while the opposite happened in residents that were defeated in their home cage (Haller et al. 1996).

A series of studies show that elevated levels of glucocorticoids decrease aggression and promote submissiveness. This was demonstrated in a variety of species treated with glucocorticoids for prolonged periods (from 3 days to 3 weeks) (decreased aggressiveness in fish, lizards, birds, and rats: Meddle et al. 2002; Hayden-Hixson and Ferris 1991a, b; Overli et al. 2002; Tokarz 1987; increased submissiveness, fish, mice, and rats: DiBattista et al. 2005; Leshner et al. 1980; Hayden-Hixson and Ferris 1991a, b; reduction of home-range lizards: DeNardo and Sinervo 1994). Similar findings were obtained in experiments where subjects were exposed to stressors. Prolonged and strong single stressors (e.g., restraint for 2–6 h) or chronic social and nonsocial stressors decreased aggressiveness (Blanchard et al. 1995; Corum and Thurmond 1997; Wood et al. 2003; Yohe et al. 2012). The effects of post-encounter increases in glucocorticoids appear to be very strong and not necessarily related to the subjective experience of losing. For example, rats showed increased submissiveness if injected with corticosterone after an aggressive encounter on the previous day (Timmer and Sandi 2010). Thus, the primary mechanisms of losing are hormonal and not cognitive in rats.

Thus, durable increases in circulating glucocorticoids—which may or may not result from defeat experience—decrease aggressiveness and promote submissive behaviors.

In the studies reviewed here, plasma cortisol was measured before and/or after the subjects performed a test where they executed aggressive acts. In this respect, these studies belong to that category of human studies which were called above “manifest aggression studies.”