behavior. These associations rely on castration and androgen replacement studies that have been carried out in very diverse ways (see Table 7). The results show that although androgens exert some influences on aggressiveness in some of these primates, the effects are variable and are often influenced by social factors (especially in females). For example, Mirsky (1955) studied the effects of implanting gonadectomized M. mulatta with pellets of testosterone or estrogen. None of the treatments produced major effects on position in the hierarchy or on dominant/subordinate behavior in unisexed groups of males or females.

Somewhat in contrast to the data for nonprimates and humans, Doering et al. (1980) reported that LHRF did not influence agonistic behavior in male chimpanzees (Pan troglodytes).

Brain (1984b) has critically reviewed the use of endocrine manipulations in controlling human aggression (accepting earlier statements that this epithet in humans is applied to a range of phenomena that are even more diverse than those seen in rodents). Castration has been applied to curb sexual aggression in Scandinavian and American populations. In spite of the considerable ethical problems associated with its use and the fact that it changes many aspects of physiology and behavior, such surgery has been claimed to produce impressively low rates of recidivism (one should, of course, examine the impacts of aging and perceptions of one's body here).

As noted earlier, therapies with hormones or antihormones are generally more ethically acceptable than castration because they seem (in theory at least) reversible. Estrogens have been used to control aggressive tendencies in intact men. The synthetic estrogen stilbestrol has been given orally to treat hyperirritable aggression and "excessive libido," but it has many unfortunate side effects including gynecomastia (development of breasts), fluid retention, and phlebothrombosis (production of blood clots), making its use problematic (Dunn, 1941). Chatz (1972) and Field and Williams (1970) advocated intramuscular or subcutaneous injections of long-acting estradiol BPC or estradiol valerate, which allowed the release of otherwise highly dangerous individuals. Both aggressive and sexual drive were essentially eliminated by such treatment.

Antiandrogens (e.g., cyproterone acetate) largely replaced castration (see later) and estrogen therapies in the treatment of European

(aggressive?) sexual offenders. Berner et al. (1983) recorded treating 21 inmates of a Vienna prison with a combination of cyproterone acetate (100 milligrams (mg) per day orally for 1–10 years) and supportive psychotherapy. Taking the drug had no effect on release from prison, and the rearrest rate for sexual offenses was 28 percent in individuals that were followed up. There was no comparison to inmates treated with psychotherapy only.

Progesterone derivatives (e.g., A-norprogesterone) and MPA have also been used (especially in the United States) in clinical therapy of human hostility. These compounds all reduce endogenous testosterone production or action and are said to produce variable ameliorative effects on behavior. They certainly have rather complex actions, are not without associated problems, and seem currently to be used with less enthusiasm for a variety of technical, legal, and ethical reasons.

In terms of actions, it is well established that cyproterone acetate blocks endogenous testosterone efficacy by competing with 5a-dihydrotestosterone for receptor sites (Mainwaring, 1975). In contrast, the antiestrogen tamoxifen binds to intracellular estrogen receptors, prevents estrogen uptake, reduces the estrogen surge characteristic of early pregnancy, and alters gonadotropin production (Watson et al., 1975); it also produces dose-related declines in cytosol high-affinity estrogen receptors in a variety of tissues including the hypothalamus (Bowman et al., 1982). Consequently, antihormones can have quite wide repercussions on the endocrine system. One has to add to this the rider that categories of antihormones are not homogeneous. For example, "antiandrogens" may be subdivided into "pure antiandrogens," "antiandrogens with antigonadotropic effects," and "progestins.'' Administration of a pure antiandrogen (e.g., flutamide) to an intact male increases LH production and consequently augments plasma testosterone. Cyproterone acetate is an antiandrogen with antigonadotropic effects that acts directly on the testis and results in a decline in plasma testosterone. Progestins alter liver steroid metabolism, augmenting the metabolic clearance rate of testosterone (Albin et al., 1973). Some progestins are without apparent actions on testosterone uptake and binding in target tissues, whereas others (e.g., MPA) have a minor inhibitory action in this respect (Suffrin and Coffey, 1973).

Although there are strong indications that neural androgen receptors are implicated in some forms of aggressive behavior, not all forms of "violence" depend on such actions, and the details have been less systematically investigated than in the case of sexual behavior (e.g., Massa et al., 1983). Certainly, we do not

know which neural androgen receptor populations are implicated in which aspects of behavior, and we know relatively little about the enzyme changes and transformations involved in androgen-mediated violent behavior. Sheard (1987) maintains that treatment with medroxyprogesterone is "the most common pharmacological approach [to the treatment of aggression] in the USA." The material has also been much used in Canada. This compound has been used to treat "aggression" in temporal lobe epilepsy. For example, O'Connor and Baker (1983) used MPA as an adjuvant in the treatment of three males (22–40 years of age) diagnosed as having chronic schizophrenia. In a double-blind study involving staff evaluations of behavior with the Brief Psychiatric Rating Scale, two of the patients who were assaultive showed significant dose-related (25–75 mg per week) drug improvements.

Many of the early attempts to correlate levels of testosterone with aggression in hostile and nonhostile prisoners (reviewed in Brain, 1984b), for example, have proved difficult to replicate. This seems related to the facts that the behavioral measures (e.g., rating by courts or the individuals per se) were often vague and divorced in time from the endocrine measurements (generally a single plasma determination of testosterone, a hormone that is secreted in a highly fluctuating manner, using samples that were taken in a "stressful" fashion, potentially reducing testosterone secretion). Rather obvious complications such as the incidence of homosexual activity in the populations and alcohol consumption were not controlled.

What then of the current position? Archer (1990) performed a limited meta-analysis on the five available studies that attempted to associate aggression as measured on the Buss-Durkee Hostility Inventory (Buss and Durkee, 1957) with plasma testosterone. The analysis suggested a very low but positive relationship between testosterone levels and overall Buss-Durkee Inventory score for the 230 males tested over the five studies. Social environment was more highly correlated with testosterone level than this score, and there was a closer association between aggression and the hormone when external assessments of the subject's behavior (rather than self-assessments) were made.

One should note that the Buss-Durkee scale is intended to measure aggressive feelings rather than aggressive actions. Indeed, Buss-Durkee factor II (the item correlated with testosterone in the above studies) is a composite of several measures, and there is little evidence of its relevance to violent or dominant behavior. Unfortunately, violent individuals rarely fill out questionnaires at

the time of their violent actions! However, Langevin (personal communication, 1990) suggests that the perceptions of aggressive offenders are often very different from their actual behavior. This is an obvious area for future study.

Langevin et al. (1985) performed a comprehensive pilot study on predictive factors of sexual aggression in which hormones were examined as factors (total testosterone, LH, FSH, estradiol, dehydroepiandrosterone sulfate (DHAS), androsterone, cortisol, and prolactin). Adrenal production of sex hormones (notably DHAS) seemed important in sexually aggressive males, and it was felt that sex hormones other than testosterone may prove of relevance to sexual aggression (see also the data on lower vertebrates). It may be possible to distinguish sadists (abnormal LH and FSH) from rapists (elevated DHAS, cortisol, and prolactin). Bain et al. (1987) failed to find significant hormonal differences among murderers, assaulters, and controls but did suggest that further study of the complex interactions of these factors is necessary. There were indications that changes in LH and LHRF might be implicated in some forms of violent behavior. Bain et al. (1988) studied baseline values of eight hormones in sexually aggressive males and found no significant group differences. In an ACTH stimulation test, however, sexual aggressives had lower baseline values of DHAS than controls. These results appear more clear-cut than most, probably because they focus on sexual aggression, distinguish subcategories of this behavior, and seem prepared to measure a range of hormonal factors. One would still like to establish whether these approaches extrapolate to other populations and situations.

The so-called challenge tests of hormonal function were clinically developed initially to assess the integrity of the endocrine system (primarily with a view to detecting pathologies). Consequently, the function of the HPA axis could be challenged by a stimulatory dose of ACTH or by suppression with a synthetic glucocorticoid such as dexamethasone. It is thought that measurement of hormones after such treatments gives one an indication of the reactiveness of the particular endocrine system and may pick up differences between individuals that are not apparent in the "basal" (unchallenged) condition. One should comment that infrahuman animal studies suggest that maximal information is extracted when one has both basal and challenge test data. It is possible that varied endocrine reactivities are unrelated to ongoing associations between hormones and behavior.

McEwen and Pfaff (1985) have emphasized that the effects of

hormones on hypothalamic neurons can involve neurotransmitter effects and neuromodulator actions including LHRF and prolactin. They speculate that such interactions can be involved in processes such as the regulation of aggression. So far as humans are concerned, Tiwary (1974) speculated that LHRF was involved in behavioral change in a young child after testosterone treatment. Such studies are, however, contentious because there is considerable debate concerning the ethics of giving a synthetic analogue of LHRF (gaserelin, Zoladex, ICI) to a pedophile in England (Brahams, 1988). The treatment was said to suppress sexual urges in a way not evident with cyproterone acetate or MPA.

Some more recent studies in our species have made use of competitive sporting situations. Although there are problems with these data, they are suggestive. Mazur and Lamb (1980) studied testosterone responses 1 to 2 hours after performance in a tennis doubles match for a cash prize, after obtaining a similar prize by lottery, and after success in medical degree examinations. Testosterone levels were elevated in the successful tennis players (compared to the losers) and the recipients of an M.D. degree but not the lottery winners. They suggest that when a male achieves enhanced status via his own efforts, both mood and testosterone levels are elevated. Elias (1981) measured cortisol, testosterone, and testosterone binding globulin in 15 males at three times in relation to wrestling bouts. Concentrations of both hormones increased during the bouts, but the binding globulin decreased in concentration. Winners of these competitive matches showed greater increases of both hormones than losers of the bouts. Salvador et al. (1987) carried out a pilot study on young male judo competitors in which plasma testosterone and cortisol were also measured. Winning or losing per se did not change the levels of these hormones. These authors did find, however, that previous personal success altered the response. Members of a regional team showed increases in testosterone levels postfighting that were not seen in individuals who had not been selected to represent their locality. The authors also emphasized that physical exercise alone could increase plasma testosterone levels.

Gladue et al. (1989) studied changes in testosterone and cortisol (assessed in saliva by immunoassays) in 40 young male U.S. subjects (18–34 years of age) in response to a nonathletic laboratory reaction time task. Subjects were randomly assigned to "winning" or "losing" categories by varying the computer feedback they received. Within the winning and losing categories, contests could be "close" or "decisive." Postcompetition mood was also assessed.

Winners had higher overall testosterone levels than losers, there being no significant difference between close and decisive wins. Winning and losing had no measurable effect on saliva cortisol measures. Mood was depressed in decisive losers compared to all other categories. The data suggest that the perception of winning or losing differentially influences testosterone output as a consequence of changing mood and apparent status.

Julian and McKenry (1989) suggested (on the basis of sociobiologic theory) that lower levels of aggression are most adaptive for men, particularly at midlife. They consequently studied 37 middle-aged (39–50 years of age) professional males from the American Midwest and used radioimmunoassay-determined serum measures of testosterone as the dependent variable. It was found (by using stepwise multiple regression analysis) that low testosterone titers related to enhanced marital and parental relationships and androgynous behavior. High levels of testosterone were significantly related to emotional expressiveness.

One should perhaps add that there have been repeated (rather anecdotal) suggestions that the anabolic steroids used illegally by some athletes may have a much more profound effect on aggression and hostility than they do on muscle development. There have even been claims from Sweden that the self-administration of these substances played a role in the homicidal behavior of weight lifters or body builders (especially when combined with alcohol ingestion). It seems appropriate to suggest that this claim should be properly evaluated in controlled studies. Any clear association would be a powerful incentive to regulate these substances more widely than at present.

Jeffcoate et al. (1986) studied relationships between dominant behavior (assessed by attendant females) of four males sharing a boating holiday of 14 day duration. This very preliminary study suggested a positive relationship between testosterone level on the boat and the rating of dominant behavior. Hellhammer et al. (1985) measured salivary testosterone levels in young males before, during, and after films containing erotic, sexual, stressful, aggressive, or neutral material. Short-term increases were found 15 minutes after exposure to erotic or sexual films. A decrease was found after exposure to stressful material, but the aggressive film produced no change.

Archer (1990) has emphasized the essentially correlational nature of the existing evidence linking androgenic hormones and measures of aggression in humans. He suggests that future research might involve more extensive longitudinal studies (as in