1981). Although there is controversy over the extent to which the genetic mechanisms are the same for male and female aggression in Mus, the testing situation can change the rank order of selected lines (Hood and Cairns, 1988)—females from high-aggression lines exhibit their agonistic behavior mostly in sex-appropriate settings (e.g., postpartum tests). Similarly, studies of selected lines show that aggression may be modified by experience. Thus, although agonistic behavior shows some developmental continuity and cross-situational generality in inbred or selected strains, it is clearly not a single genetic phenomenon that can be studied in isolation from specific contextual cues, social environment, and development (e.g., Cairns et al., 1990; Jones and Brain, 1987).

The recent trend in behavior genetic research is less toward demonstrating the fact of the heritability of aggression and more on elucidating its genetic correlates and identifying genetic loci that underlie agonistic behavior. Here, the Y chromosome may contribute to individual differences in male aggression in the mouse, at least in some strains (Carlier et al., 1990; Maxson et al., 1989; Selmanoff et al., 1975). There also appears to be genetic sensitivity to the effects of early neonatal androgens on aggression in mice (e.g., Vale et al., 1972; Michard-Vanhee, 1988).

Research suggests that testosterone and its androgenic and estrogenic metabolites influence the probability of aggressive responding to environmental events and stimuli through organizational as well as activational mechanisms. Organizational effects are traditionally those exerted, generally permanently, by a hormone during some sensitive period of development. This type of mechanism appears to explain sex differences in anatomy and some aspects of sex differences in behavior. For example, testosterone present during a particular period of fetal development in mammals induces the development of the male reproductive tract and genitalia. If androgen levels are low, as is normally the case in females, this development does not occur and female genitalia develop instead. A similar control for male aggression has been demonstrated in a number of laboratory animal species. For example, female mice given a single injection of testosterone at

birth become much more sensitive to the aggression-enhancing effects of androgens as adults. Prenatal treatment of female rhesus monkeys with testosterone results in females that are male-like in their higher level of "rough and tumble" play as juveniles and more aggressive as adults. In humans, there is evidence for a similar, but reduced in magnitude, modulation of aggression by androgens. This research uses children that were accidentally exposed to inappropriate steroids during fetal development and assesses their behavior though observations, interviews, and psychologic assessments. In girls prenatally exposed to heightened levels of androgens, there is a trend for increased levels of aggression. In boys exposed to estrogens or antiandrogenic steroids during pregnancy, there is a trend for decreased aggressiveness (see Brain, Table 4 in this volume). However, generally these steroids also have had some effect on genital development and the behavioral differences may be due to altered body image or to the affected children being treated differently by parents or peers. Interestingly, prenatal testosterone also alters the development of parts of the preoptic area of the brain. Preoptic area structure and neurochemistry are sexually dimorphic in animals and in humans, and this brain area is also thought to have a role in aggressive behavior. However, a direct link between the sexual dimorphism of the preoptic area and human violent behavior remains elusive.

In animals, testosterone (or its metabolites) has effects on the probability of aggressive response to conspecifics or other environmental events. This is frequently referred to as an activational effect although, mechanistically, androgens are not stimulating aggressive behavior in vacuo; more accurately, they appear to be altering the response to aggression-provoking stimuli. In laboratory animals, particularly rodents, there is research that demonstrates the brain sites involved in this action and the importance of the biochemical mechanisms by which testosterone can alter neural activity. The strength of the modulation that testosterone exerts on aggressive behavior seems to decrease in more complex social animals. In nonhuman primates, the correlation between testosterone and aggressiveness or dominance frequently, but not in all studies, exists, but the activational effect of testosterone is more variable and harder to demonstrate. This trend is perhaps more exaggerated in humans. Positive correlations have been reported between androgen levels and aggressive or violent behavior in adolescent boys and in men, but these correlations are not high, they are sometimes difficult to replicate, and importantly, they do not demonstrate causation. In fact there is better evidence

for the reverse relationship (behavior altering hormonal levels). Stress (e.g., from being subject to aggression or being defeated) decreases androgen levels, and winning—even in innocuous laboratory competitions—can increase testosterone.

The results of manipulating androgens with antiandrogen therapy in violent offenders are also mixed and difficult to interpret because of confounding influences on the data collection. Some critical reviews have concluded that antiandrogens show promise as an adjunct therapy for violent sex offenders. These may be relying more on the clearer relationship between testosterone and sexual motivation than between testosterone and violence. Another interesting approach will be to study the effects of anabolic steroids, but these studies have just begun and face very difficult methodologic problems. In general, most investigators conclude that there can be an influence of androgens on violence but that it is only one component accounting for a small amount of the variance.

Gonadal steroids have also been postulated to be involved in the increased irritability and hostility seen in some women with premenstrual syndrome (PMS). However, the endocrine evidence in support of this view is weak, and most recent papers find that individual differences in estrogens, progestins, and other hormones across the menstrual cycle do not explain the variability in intensity of PMS symptoms.

Adrenal steroids (glucocorticoids, such as cortisol and corticosterone) and the pituitary hormone ACTH (adrenocorticotropic hormone) have also been found to be related to aggressive behavior in animals. However, the strongest relationship is a negative one. Chronically increased levels of corticosteroids decrease aggressiveness, and ACTH increases submissiveness and avoidance of attack. These two effects are difficult to separate endocrinologically, but they appear to be mediated by different mechanisms. Correlations between dominance and corticosteroid levels in primates may more directly reflect variations in the ability to adapt to stress.

In summary, there is no simple relationship between steroids and aggression, much less violence. The strongest conclusion is that in humans, androgens can influence and be influenced by aggressive behavior. However, they are only one of many influences and not the determining factor. The opposite relationship (i.e., the environment and behavior influencing hormone secretion) is the stronger of the two linkages.