Causes and Consequences of Violence Against Women
A vital part of understanding a social problem, and a precursor to preventing it, is an understanding of what causes it. Research on the causes of violence against women has consisted of two lines of inquiry: examination of the characteristics that influence the behavior of offenders and consideration of whether some women have a heightened vulnerability to victimization. Research has sought causal factors at various levels of analysis, including individual, dyadic, institutional, and social. Studies of offending and victimization remain conceptually distinct except in sociocultural analysis in which joint consideration is often given to two complementary processes: those that influence men to be aggressive and channel their expressions of violence toward women and those that position women for receipt of violence and operate to silence them afterwards. Many theorists and researchers have sought to answer the question, "Why does this particular man batter or sexually assault?" by looking at single classes of influences. Among them have been biologic factors such as androgenic hormonal influences; evolutionary theo-
ries; intrapsychic explanations focused on mental disorder or personality traits and profiles; social learning models that highlight the socialization experiences that shape individual men to be violent; social information processing theory concerning the cognitive processes that offenders engage in before, during, and after violence; sociocultural analyses aimed at understanding the structural features of society at the level of the dyad, family, peer group, school, religion, media, and state that encourage male violence and maintain women as a vulnerable class of potential victims; and feminist explanations stressing the gendered nature of violence against women and its roots in patriarchal social systems. Recently, researchers armed with multivariate statistical analysis have tested complex models of violence with multiple factors to explain battering (McKenry et al., 1995) and to model the common roots of verbal, physical, and sexual coercion toward women (Malamuth et al., 1995). Also new are integrative metatheories of intimate violence that consider the impact of historical, sociocultural, and social factors on people, including the processes whereby social influences are transmitted to and represented within individual psychological functioning, including cognition and motivation (White, in press).
Many of the theories about the causes of perpetrating violence against women are drawn from the literature on aggression and general violence. Both the research on general violence and that on violence against women suggest that violence arises from interactions among individual biological and psychosocial factors and social processes (e.g., Reiss and Roth, 1993), but it is not known how much overlap there is in the development of violent behavior against women and other violent behavior. Studies of male batterers have found that some batterers confine their violent behavior to their intimates but others are violent in general (Fagan et al., 1983; Cadsky and Crawford, 1988; Shields et al., 1988; Saunders, 1992; Holtzworth-Munroe and Stuart, 1994). The research suggests that, at least in some cases, there may be differences in the factors that cause violence against women and those
that cause other violent behavior. Much more work is needed in order to understand in what ways violence against women differs from other violent behavior. Such understanding will be particularly important for developing preventive interventions.
Although current understanding suggests that violent behavior is not caused by any single factor, much of the research has focused on single causes. Therefore, in the following sections several salient findings emerging from each single-factor domain are highlighted to illustrate how each contributes something to the causal nexus of perpetration of violence. They are followed by a brief review of efforts to build multifactor models.
Theories of Violent Offending
Evolution From an evolutionary perspective, the goal of sexual behavior is to maximize the likelihood of passing on one's genes. This goal involves maximizing the chances that one will have offspring who themselves will survive to reproduce. In ancestral environments, optimum male and female strategies for successfully passing on one's genes often did not coincide because the amount of parental investment required by males is smaller than that required by females. Males were best served by mating with as many fertile females as possible to increase their chance of impregnating one of them; females, who have the tasks of pregnancy and nurturing the young, are often better served by pair bonding. Sex differences in current human mating strategies may be explained as having been shaped by the strategies that created reproductive success among human ancestors. A number of studies have shown that young adult males are more interested in partner variety, less interested in committed long-term relationships, and more willing to engage in impersonal sex than are young adult females (Clark and Hatfield, 1989; Symons
and Ellis, 1989; Clark, 1990; Landolt et al., 1995). This finding is consistent with the optimum evolutionary strategy for males of mating with as many fertile females as possible.
It is theorized that males who have difficulty obtaining partners are more likely to resort to sexual coercion or rape. Extensive evidence of forced mating among animals has been documented (Ellis, 1989). Evolutionary theory also has been used to explain aspects of intimate partner violence. It is theorized that male sexual jealousy developed as a means of assuring the paternity of their offspring (Quinsey and Lalumière, 1995). Case histories from battered women often mention the extreme sexual jealousy displayed by their batterers (Walker, 1979; Browne, 1987), and extreme sexual jealousy is a common motive of men who kill their wives (Daly and Wilson, 1988).
There is much debate over how much influence evolutionary factors have on modern human beings. Even those who favor evolutionary explanations acknowledge that additional factors are necessary to explain sexual assault and intimate partner violence. For example, Quinsey and Lalumière (1995) suggest that rape and other sexual coercion may be explained by the evolutionary approach that is modified by specific attitudes toward women or by psychopathy, coupled with an erotic interest in coercive sexual behavior. Evolutionary explanations of rape are also criticized as not explaining the proportion of rapes lacking reproductive consequences because they involve oral or anal penetration or victims who are prepubescent or male.
Physiology and Neurophysiology The physiological or neurophysiological correlates of violence and aggression that have received particular attention are the functioning of steroid hormones such as testosterone; the functioning of neurotransmitters such as serotonin, dopamine, norepinephrine, acetylcholine, and gamma-aminobutyric acid (GABA); neuroanatomical abnormalities; neurophysiological abnormalities; and brain dysfunctions that interfere with cognition or
language processing. This literature has been well reviewed in other sources (e.g., Fishbein, 1990; Reiss and Roth, 1993; Brain, 1994; Miczek et al., 1994a,b; Mirsky and Siegel, 1994); this section highlights the overall findings and notes studies that have specifically looked at violence against women. In considering this literature, it should be remembered that much of the evidence comes from animal studies and that generalizing from animals to humans is not straightforward. The evidence that comes from studies of human subjects only shows correlations, so any causal interpretations are tenuous. Furthermore, changes in hormonal, neurotransmitter, and neurophysiological processes may be consequences of violent behavior or victimization, as well as being causes of those behaviors (Reiss and Roth, 1993; van der Kolk, 1994).
A recent comprehensive literature review (Archer, 1991) concluded that the majority of studies showed that high testosterone levels tend to covary with high probabilities of aggressive behaviors, dominance status, and pathological forms of aggression in nonhuman mammals, but that the picture for humans is not as clear. In humans, there appears to be a correlation between testosterone levels and aggression, but it is not clear whether testosterone levels influence aggressive behavior or vary as a result of aggressive behavior. Similarly, the results of human studies of neurotransmitters are not conclusive. For example, low levels of serotonin, the most heavily studied of the neurotransmitters, have been found to be correlated with aggressive behavior, impulsivity, and suicidal behavior (Asberg et al., 1976; Brown et al., 1979; Linnoila et al., 1983; Lidberg et al., 1985; Mann, 1987; Coccaro et al., 1989). More recent studies have found a complex interaction among serotonin, alcoholism, and monoamine metabolism and these behaviors (Linnoila et al., 1989; Virkkunen et al., 1989a,b). Further evidence of the role of neurotransmitters comes from the fact that drugs that act on serotonin receptors or on monoamine oxidase may reduce aggressiveness. Animal and human studies have found trauma and violence to
have effects on hormones, neurotransmitters, and brain function (e.g., van der Kolk, 1994).
Studies have also looked at brain abnormalities and violent behavior. Neuropsychological deficits in memory, attention, and language, which sometimes follow limbic system damage, have been found to be common in children who exhibit violent or aggressive behavior (e.g., Miller, 1987; Lewis et al., 1988; Mungas, 1988). Differences in peripheral measures of nervous system activity, such as heart rate or skin conductance, have been found between control subjects and samples of criminals, psychopaths, delinquents, and conduct-disordered children (Siddle et al., 1973; Wadsworth, 1976; Raine and Venables, 1988; Kagan, 1989; Raine et al., 1990). Langevin (1990:112) found a "link between temporal lobe impairment and sexually anomalous behaviors" that was independent of nonsexual criminality and not explained by learning disabilities or alcohol abuse. Reduced impulse control and personality changes following head injury may lead to an increased risk of battering (Detre et al., 1975; Lewis et al., 1986, 1988). Likewise, studies have found that batterers are more likely to have had head injuries than nonbatterers (Rosenbaum and Hoge, 1989; Rosenbaum et al., 1996).
There is increasing interest in the role played by biological factors in violent behavior; however, most researchers believe it is the interaction of biological, developmental, and environmental factors that is important (Fishbein, 1990). For example, Marshall and Barbaree (1990) speculate that biological factors may set the stage for learning, providing limits and possibilities rather than determining outcomes, and that developmental and environmental factors play the larger role. However, as suggested by a previous study (Reiss and Roth, 1993), preventing head injuries and environmental exposure to toxins, such as lead, that may damage brain functioning could be considered potential avenues for preventing violence.
Alcohol Every category of aggressive act (except throwing objects) has a higher prevalence among people who have been
drinking (Pernanen, 1976). Alcohol use has been reported in between 25 percent and 85 percent of incidents of battering and up to 75 percent of acquaintance rapes (Kantor and Straus, 1987; Muehlenhard and Linton, 1987; Koss et al., 1988). It is far more prevalent for men than their female victims. Considerable research links drinking and alcohol abuse to physical aggression, although adult consumption patterns are likewise associated with other variables related to violence (such as witnessing physical violence in one's home of origin; Kantor, 1993). The relationship of alcohol to intimate partner violence could be spurious, but the relationship of men's drinking to intimate partner violence remains even after statistically controlling for sociodemographic variables, hostility, and marital satisfaction (Leonard and Blane, 1992; Leonard, 1993). Men's drinking patterns, especially binge drinking, are associated with marital violence across all ethnic groups and social classes (Kantor, 1993).
The relationship of alcohol to violence is a complex one, involving physiological, psychosocial, and sociocultural factors. The exact effects of alcohol on the central nervous system remain in question, but nonexperimental evidence indicates that alcohol may interact with neurotransmitters, such as serotonin, that have been associated with effects on aggression (Linnoila et al., 1983; Virkkunen et al., 1989a,b). Studies have found a genetic basis for alcohol abuse and alcoholism (Cloninger et al., 1978; Plomin, 1989) and for antisocial personality traits (Christiansen, 1977; Bohman et al., 1982; Mednick et al., 1984; Cloninger and Gottesman, 1987) that are often found among violent offenders. The fact that alcohol abuse and antisocial personality frequently occur together has led to the speculation of common genetic bases, but the evidence remains inconclusive (Reiss and Roth, 1993).
Alcohol may interfere with cognitive processes, in particular, social cognitions. Recent studies suggest that men under the influence of alcohol are more likely to misperceive ambiguous or neutral cues as suggestive of sexual interest and to ignore or misinterpret cues that a woman is unwilling
(Abbey et al., 1995). The impact of alcohol on behavior has also been linked to a person's expectations about alcohol's effects. For example, Lang et al. (1975) found individuals became more aggressive in laboratory experiments after drinking what they were told was alcohol, even though it was not. Similarly, laboratory studies of penile responses to pornographic stimuli decrease with actual ingestion of alcohol, but increase when participants believe they have drunk alcohol when they have actually received a placebo drink (Richardson and Hammock, 1991). It has also been suggested that alcohol may be used to excuse violent behavior (Coleman and Straus, 1983; Collins, 1986). These deviance disavowal theories ("I wouldn't have done it if I hadn't been drunk") have not been empirically tested, however (Kantor, 1993).
There are methodological weaknesses in the studies of the links between alcohol and violence, including lack of clear definitions of excessive alcohol use and a reliance on clinical samples with an absence of control samples. (For a more complete review of the research and methodological weaknesses see Leonard and Jacob, 1988; Leonard, 1993.) Nonetheless, research has consistently found that heavy drinking patterns are related to aggressive behavior, in general, and to intimate partner and sexual violence. However, exactly how alcohol is related to violence remains unclear. Obviously, many battering incidents and sexual assaults occur in the absence of alcohol, and many people drink without engaging in violent behavior (Kantor and Straus, 1990).
Psychopathology and Personality Traits A number of studies have found a high incidence of psychopathology and personality disorders, most frequently antisocial personality disorder, borderline personality organization, or posttraumatic stress syndrome, among men who assault their wives (Hamberger and Hastings, 1986, 1988, 1991; Hart et al., 1993; Dutton and Starzomski, 1993; Dutton, 1994, 1995; Dutton et al., 1994). A wide variety of psychiatric and personality disorders have also been diagnosed among sexual offenders, most
frequently some type of antisocial personality disorder (Prentky, 1990).
Distinctive personality profiles have been reported for rapists and sexually aggressive men (Groth and Birnbaum, 1979; Abel et al., 1986), and batterers (Geffner and Rosenbaum, 1990). However, personality testing of rapists has found no significant differences between sexual offenders and those incarcerated for nonsexual offenses (Quinsey et al., 1980; Langevin, 1983). Studies of the personalities of incarcerated rapists and court-referred batterers are problematic, these men are typically poorly educated and from low-status occupations. Thus the differences may say more about who gets reported, arrested, tried, convicted, and sentenced than it does about the personalities of violent men. Rape, for example, is one of the most underreported crimes (Bowker, 1979), and only a small proportion of reported rapes result in incarceration (Darke, 1990). Even within the restricted population found in studies of incarcerated sex offenders, most investigators have concluded that there is a great deal of heterogeneity among rapists and that sexual aggression is multiply determined (Prentky and Knight, 1991).
Batterers also seem to be a heterogeneous group (Gondolf, 1988; Saunders, 1992). Because of this heterogeneity, much of the research on incarcerated rapists and known batterers has included attempts to develop typologies to represent subgroups of them. Typologies of batterers have generally used one, or a combination, of three dimensions to distinguish between subgroups: frequency and severity of physical violence and related sexual or psychological abuse; generality of the violence (i.e., violence only in the family or violence in general); and psychopathology or personality disorder (Holtzworth-Munroe and Stuart, 1994). Rapists have been categorized by motivational factors (sexual or aggressive), impulse control factors, and social competence. (For a detailed description of sexual offender taxonomies, see Knight and Prentky, 1990.)
Because incarcerated sexual offenders and batterers in treatment are probably not representative of all sex offenders or batterers, another avenue of research has focused on normal population samples, comparing those who self-report physically or sexually aggressive behavior and those who do not. Sexually aggressive men are said to differ from other men in antisocial tendencies (Malamuth, 1986), nonconformity (Rapaport and Burkhart, 1984), impulsivity (Calhoun, 1990), and hypermasculinity (Mosher and Anderson, 1986). Batterers have been found to show lower socialization and responsibility (Barnett and Hamberger, 1992). It is important to remember, however, that there are potential biases in self-report data, and it is difficult to verify their accuracy other than through consistency of responses. Men may be reluctant to acknowledge that they have engaged in sexually or physically violent behavior or the men who report this behavior may be different from those who have engaged in the behavior but do not report it. Yet, because both intimate partner violence and sexual assault usually take place in private, self-reports play a central role in their study. Self-report measures on sensitive topics, including violent behaviors, have been found to be quite reliable (Straus, 1979; Hindelang et al., 1981; Bridges and Weis, 1989).
Attitudes and Gender Schemas Cultural myths about violence, gender scripts and roles, sexual scripts and roles, and male entitlements are represented at the individual level as attitudes and gender schemas. These hypothetical entities are expectancies that give meaning to and may even bias interpretation of ongoing experience, as well as provide a structure for the range of possible responses. Acceptance of beliefs that have been shown to foster rape has been demonstrated among a variety of Americans, including typical citizens, police officers, and judges (Field, 1978; Burt, 1980; Mahoney et al., 1986). Once a violence-supportive schema about women has developed, men are more likely to misinterpret ambiguous evidence as confirming their beliefs (Abbey, 1991). Sexu-
ally aggressive men more strongly endorse a set of attitudes that are supportive of rape than do nonaggressive men, including myths about rape and the use of interpersonal violence as a strategy for resolving conflict (e.g., Malamuth, 1986; Malamuth et al., 1991, 1995). Beliefs and myths about rape may serve as rationalizations for those who commit violent acts. For example, incarcerated rapists often rationalize that their victim either desired or deserved to experience forced sexual acts. Similarly, culturally sanctioned beliefs about the rights and privileges of husbands have historically legitimized a man's domination over his wife and warranted his use of violence to control her. Men, in general, are more accepting of men abusing women, and the most culturally traditional men are the most accepting (Greenblatt, 1985). Batterers' often excuse their violence by pointing to their wives' ''unwifely" behavior as their justification (Dobash and Dobash, 1979; Adams, 1988; Ptacek, 1988).
Sex and Power Motives Violence against women is widely believed to be motivated by needs to dominate women. This view conjures the image of a powerful man who uses violence against women as a tool to maintain his superiority, but research suggests that the relationship is more complex. Power and control frequently underlie intimate partner violence, but the purpose of the violence may also be in response to a man's feelings of powerlessness and inability to accept rejection (Browne and Dutton, 1990). It also has been argued that rape, in particular, represents fulfillment of sexual needs through violence (Ellis, 1989), but research has found that motives of power and anger are more prominent in the rationalizations for sexual aggression than sexual desires (Lisak and Roth, 1990; Lisak, 1994). Attempts to resolve the debate about sex versus power have involved laboratory studies of men's sexual arousal to stimuli of depictions of pure violence, pure consensual sex, and nonconsensual sex plus violence. These studies have consistently shown that some "normal" males with no known history of rape may be aroused by rape stimuli involv-
ing adult women, especially if the women are portrayed as enjoying the experience (Hall, 1990). However, sexually aggressive men appear to be more sexually arousable in general, either to consenting or rape stimuli (Rapaport and Posey, 1991), and rapists respond more than nonsexual offenders to rape cues than to consenting sex cues (Lalumière and Quinsey, 1994). Sexually aggressive men openly admit that their sexual fantasies are dominated by aggressive and sadistic material (Greendlinger and Byrne, 1987; Quinsey, 1984).
Social Learning Social learning theory posits that humans learn social behavior by observing others' behavior and the consequences of that behavior, forming ideas about what behaviors are appropriate, trying those behaviors, and continuing them if the results are positive (O'Leary, 1988). This theory does not view aggression as inevitable, but rather sees it as a social behavior that is learned and shaped by its consequences, continuing if it is reinforced (Lore and Schultz, 1993). From this perspective, male violence against women endures in human societies because it is modeled both in individual families and in the society more generally and has positive results: it releases tension, leaves the perpetrator feeling better, often achieves its ends by cutting off arguments, and is rarely associated with serious punishment for the perpetrator.
One of the mechanisms through which social learning occurs is social information processing—the decoding or interpreting of social interactions, making decisions about appropriate responses on the basis of the decoding, and carrying out a response to see if it has the intended effect. It has been hypothesized that violent men may be deficient in the skills necessary to accurately decode communications from women. For example, men's judgments of videotapes of male-female interactions are more highly sexualized than women's judgments (Abbey, 1991; Kowalski, 1992, 1993). Batterers appear to be more likely than nonviolent men to attribute negative intentions to their partners' actions and to behave negatively,
for example, with anger or contempt (Dutton and Browning, 1988; Margolin et al., 1988; Holtzworth-Munroe, 1992).
An individual man carries out violence against a woman in a dyadic context that includes features of the relationship, characteristics of the woman, and their communication. The stage of relationship between a man and woman may determine, in part, the probability of violence. Anecdotal evidence from battered women suggest that a man often refrains from physical violence until a women has made an emotional commitment to him, such as moving in together, getting engaged or married, or becoming pregnant (e.g., Walker, 1979; Giles-Sims, 1983; Browne, 1987). It is suggested that the emotional bond between the couple once formed, may contribute to the man's sense of entitlement to control his partner's behavior as well as diminish the facility with which the woman can leave the relationship without ambivalence. Some evidence suggests that women are willing to see the first violent incident as an anomaly, and so are willing to forgive it, although this response may actually reinforce the violent behavior (Giles-Sims, 1983).
Acquaintance or date rape may also be related to relationship stage, with different risk factors for rapes during first dates and rapes in on-going relationships (Shotland, 1992). For example, men who rape on first or second dates may be similar to stranger rapists, while men who rape early in a developing relationship may misperceive their partners' intent (Shotland, 1992). Prior sexual intimacy between partners may increase a man's belief that he has a right to such intimacy any time he desires it, and it may also support his false assumption that a forced sexual encounter in an experienced woman is harmless (Johnson and Jackson, 1988). Completed rapes have been found to be more likely in couples who know each other well than among persons who are acquaintances (Belnap, 1989). As noted in the section on social learning
(above), physically and sexually aggressive men may misinterpret cues from females. It has been found, for example, that male batterers have poor communication skills (Ganley and Harris, 1978; Holtzworth-Monroe and Anglin, 1991).
Family, Schools, and Religion Families are where all socialization begins, including socialization for all types of violent behavior. Studies of violent criminals and violent sex offenders have found these men are more likely than other adults to have experienced poor parental childrearing, poor supervision, physical abuse, neglect, and separations from their parents (Langevin et al., 1985; Farrington, 1991). Increased risk of adult intimate partner violence is associated with exposure to violence between a person's parents while growing up. One-third of children who have been abused or exposed to parental violence become violent adults (Widom, 1989). Sons of violent parents are more likely to abuse their intimate partners than boys from nonviolent homes (Straus et al., 1980). Men raised in patriarchal family structures in which traditional gender roles are encouraged are more likely to become violent adults, to rape women acquaintances, and to batter their intimate partners than men raised in more egalitarian homes (Straus et al., 1980; Gwartney-Gibbs et al., 1983; Fagot et al., 1988; Friedrich et al., 1988; Koss and Dinero, 1989; Riggs and O'Leary, 1989; Malamuth et al., 1991, 1995). Sexual abuse in childhood has been identified as a risk factor in males for sexual offending as an adult (Groth and Birnbaum, 1979; Briere, 1992). Experiences of sexual abuse in one's family may lead to inaccurate notions about healthy sexuality, inappropriate justifications for violent behavior, failure to develop personal boundaries, and contribute to communication and coping styles that rely on denial, reinterpretation of experiences, and avoidance (Briere, 1992; Herman, 1992).
To the extent that schools reinforce sex role stereotypes and attitudes that condone the use of violence, they may
contribute to socialization supportive of violent behavior. Other institutions that have been implicated in contributing to socialization that supports violence against women are organized religion (Fortune, 1983; Whipple, 1987), the workplace (Fitzgerald, 1993), the U.S. military (Russell, 1989), and the media (Linz et al., 1992).
Athletic teams also may socialize children to behavior that is supportive of violence. For example, male athletes may be spurred to greater aggressive efforts by coaches who deride them as "girls." Participation in revenue-producing sports at the collegiate level was found to be a significant predictor of sexual aggression among college students (Koss and Gaines, 1993). It is possible that team sports, particularly revenue-producing sports, attract young men who are already aggressive. Whether team sports encourage aggressive behavior or simply reinforce already existing aggressive tendencies remains to be determined. In either case, it appears that participation in team sports is a risk factor for sexual aggression.
Media Many feminist writers (e.g., Brownmiller, 1975; Dworkin, 1991; Russell, 1993) have suggested that pornography encourages the objectification of women and endorses and condones sexual aggression toward women. Both laboratory research and studies of television lend support to this view. Exposure to pornography under laboratory conditions has been found to increase men's aggression toward women, particularly when a male participant has been affronted, insulted, or provoked by a woman (Linz et al., 1992). Sexual arousal to depictions of rape is characteristic of sexual offenders (Hall, 1990). Even exposure to nonexplicit sexual scenes with graphic violence has been shown to decrease empathy for rape victims (Linz et al., 1988). It appears that it is the depiction of violence against women more than sexual explicitness that results in callousness toward female victims of violence and attitudes that are accepting of such violence (Donnerstein and Linz, 1994).
It is not only pornography that depicts violence against women. Television and movies are filled with scenes of women being threatened, raped, beaten, tortured, and murdered. A number of studies of television point to the deleterious effects of viewing media portrayals of violence (e.g., Eron, 1982; National Institute of Mental Health, 1982; Huston et al., 1992). Eron (1982) found that children who watched many hours of violence on television during elementary school tended to exhibit more aggressive behavior as teenagers and were more likely to be arrested for criminal acts as adults. A meta-analysis of 188 studies found a strong positive association between exposure to television violence and antisocial and aggressive behavior (Comstock and Paik, 1990; Paik and Comstock, 1994). Those who are exposed to television and cinema violence may also become desensitized to real world violence, less sensitive to the pain and suffering of others, and begin to see the world as a mean and dangerous place (Murray, 1995). A recently released national study of violence on television found that context of the violence shown was important: television shows virtually no consequences of violent behavior; victims are not harmed and offenders are not punished (Mediascope, 1996). It seems that many television depictions of violence send the message that violence works.
None of the studies of television violence has focused specifically on violence against women. The National Television Violence Study (Mediascope, 1996) found that 75 percent of the targets of violence in television portrayals are males, while only 9 percent are females (the remainder are nonhuman characters). Research has not yet examined the type of violence directed at female victims on television, how it compares with that directed at male victims, and whether there are differential effects on viewers of violence against women and against men.
For much of recorded Western European and American
history, wives had no independent legal status; they were basically their husbands' property. The right of a husband to physically chastise his wife was upheld by the Supreme Court of Mississippi in 1824 ( Bradley v. State 1 Miss. 157) and again by a court in North Carolina in 1868 (State v. Rhodes, 61 N.C. 453, 353; cited in Pleck, 1989). In 1871 a court ruling in Alabama (Fulgham v. State, 46 Ala. 146-147) made that state the first to rescind a husband's right to beat his wife (Fagan and Browne, 1994). During the 1870s, coinciding with the rise of the child protective movement, there was increased concern that wife beating should be treated as a crime, although few men were ever punished (Pleck, 1989). In the 1890s social casework replaced criminal justice as the preferred system for dealing with family violence and general interest in wife beating waned until the 1960s (Fagan and Browne, 1994).
The status of women as property also can be seen in the development of laws concerning rape. Brownmiller (1975:8) contends that "rape entered the law … as a property crime of man against man. Woman, of course, was viewed as the property." She notes that until the end of the thirteenth century, only unmarried virgins were considered blameless in their victimization; married women who were raped were punished along with their rapist. At that time, the Statutes of West-minister put forward by Edward I of England extended the same penalties to men who raped married women as to those who raped virgins. Rape within marriage, however, was, by definition, impossible. Marriage laws traditionally assumed implied consent to sexual relations by wives and allowed husbands to use force to gain compliance (Fagan and Browne, 1994). It has only been in recent years that laws have begun to recognize marital rape: today every state in the United States has modified or eliminated the marriage exclusion in its rape laws (personal communication, National Clearing-house on Marital and Date Rape, Berkeley, California).
Sexual Scripts Expectations about dating and intimate rela-
tionships are conveyed by culturally transmitted scripts. Scripts support violence when they encourage men to feel superior, entitled, and licensed as sexual aggressors with women as their prey, while holding women responsible for controlling the extent of sexual involvement (White and Koss, 1993). Parents socialize daughters to resist sexual advances and sons to initiate sexual activity (Ross, 1977). By adolescence, both boys and girls have been found to endorse scripts about sexual interaction that delineate a justifiable rape. For example, approximately 25 percent of middle school, high school, and college students state that it is acceptable for a man to force sex on a woman if he spent money on her (Goodchilds and Zellman, 1984; Muehlenhard et al., 1985; Goodchilds et al., 1988).
Since Burt (1980) first defined "rape myths" and developed a scale to measure them, a large body of research has examined the role of attitudes and false beliefs about rape on perpetration of sexual assault and on society's response to sexual assault. Typical rape myths include denial of rape's existence (e.g., most rape claims are false, or women generally lie about rape), excusing the rape (e.g., she led him on, he couldn't help himself, rape only happens to "bad" women), and minimizing the seriousness of rape (e.g., Hall et al., 1986; Briere et al., 1985). Despite psychometrically weak measurement instruments, the study of rape myths has provided important understandings about sexual aggression (Lonsway and Fitzgerald, 1994). Not surprisingly, men are more accepting of rape myths than women (e.g., Muehlenhard and Linton, 1987; Margolin et al., 1989; Dye and Roth, 1990). A number of studies have found a significant association between acceptance of rape myths and self-reported sexually aggressive behavior (Field 1978; Koss et al., 1985; Murphy et al., 1986; Muehlenhard and Linton, 1987; Reilly et al., 1992).
The early studies of rape myths were performed on college campuses and found that 25 percent to 35 percent of the students accepted a variety of them (Giacopassi and Dull, 1986; Gilmartin-Zena, 1987). Since the mid 1980s, many college
campuses have instituted rape awareness and rape education programs. Recent research found fewer than 2 percent of students accepting of sexual aggression or coercion, but up to 36 percent expected that sexual aggression would occur under certain circumstances (Cook, 1995). Cook (1995) surmises that rape education has made it unacceptable to admit to believing rape myths, but that behavioral expectations are still consistent with acceptance of rape myths. It will be valuable for prevention efforts for research to continue to track any changes in rape myth acceptance and sexual script expectations among students, as well as the general public.
Cultural Mores Ethnographic and anthropologic studies determine the critical role that sociocultural mores play in defining and promoting violence against women. Anthropologists have found cultural differences in the amount of and acceptability of intimate partner violence in different societies. A review of 14 different societies (Counts et al., 1992) found that physical chastisement of wives was tolerated in all the societies and considered necessary in many societies, but the rates and severity of wife beating were found to range from almost nonexistent to very frequent. These differences seem to be related to negative sanctions for men who overstepped "acceptable" limits, sanctuaries for women to escape violence, and a sense of honor based on nonviolence or decent treatment of women (Campbell, 1992).
Two general types of rape have been identified. Transgressive or non-normative rape is uncondoned genital contact against the will of the woman and in violation of social norms; tolerated or normative rape is unwanted genital contact that is supported by social norms (Heise, 1993; Rozee, 1993). Normative rape is reported in nearly all societies (97 percent; Rozee, 1993), and all have mechanisms that "legitimate, obfuscate, deny, and thereby perpetuate violence" (Heise et al., 1994:1). Ethnographic studies have found rape in 42 percent to 90 percent of nonindustrial societies, depending on how it is defined and on the cultural and geographic representative-
ness of the sample (Minturn et al., 1969; Bart et al., 1975; Broude and Green, 1976; Sanday, 1981; Levinson, 1989; Rozee, 1993; for a review see Koss et al., 1994). In preliterate societies, there were significantly greater frequencies of rape in those characterized by patrilocality, high degree of interpersonal violence, and an ideology of male toughness. Rape is also prevalent under conditions of marked social inequity and social disorganization, such as slavery and war (Quinsey, 1984).
It is generally accepted that multiple classes of influences—from the individual to the macrolevel—determine the expression of assaultive and sexually aggressive behavior in men (for recent reviews see Ellis, 1989; Sugarman and Hotaling, 1989; Craig, 1990; Hall, 1990; Malamuth and Dean, 1991; Berkowitz, 1992; Shotland, 1992; White and Koss, 1993; White, in press). Although it is possible to model at a general level the causal factors that explain the variance among the forms of violence against women, the heterogeneity of violent men precludes the delineation of a single set of causes that accurately classifies types of offenders. Therefore, researchers have turned to multivariate modeling of violence. Recent efforts include a biopsychosocial model of battering that examines the relative contribution of three domains of predictors including the physical (e.g., testosterone, prolactin, and alcohol), the social (e.g., negative life events, quality of relationships, family income, and social support), and psychiatric symptoms (McKenry et al., 1995). The results showed significant zero-order correlations within each class of predictors, but in multivariate analysis the social variables predicted violence better than the other variables.
Work by Malamuth and colleagues (1991, 1993, 1995) has generated and tested a model to explain both sexual and nonsexual aggression toward women. Their results suggest that
there are common pathways to all forms of aggression, but different specific factors may influence the development of nonsexual versus sexual aggression toward women. Furthermore, some of the same factors that contribute to sexual aggression in early adulthood appear to lead to other conflictual behaviors with women in later life. Male sexual aggression was best predicted by a history of promiscuous-impersonal sex and distrust of women coupled with gratification from dominating them. Physical aggression was best predicted by relationship distress and verbal aggression. General hostility and defensiveness contributed to both types of aggression. This work supports the findings of other researchers (O'Leary and Arias, 1988; O'Leary et al., 1994) that psychological abuse may be a precursor to physical aggression. These findings point to the need for more work that looks at commonalities and differences among all forms of violence against women and general violence.
All this work is a marked improvement over earlier research that focused on single causes or theories. The field appears to be developing toward an integrative, metatheoretical model of violence that considers multiple variables operating at different times in a probabilistic fashion (Leonard, 1993; White, in press). Future work guided by these models can examine the relationship of one form of violence to another; make better connections between macrolevel societal variables and individual variables to establish how culture is expressed; address both structural and contextual causes of violence; use a life-span perspective capable of capturing the processes by which earlier experiences affect later ones; and focus on the gendered nature of violence against women that involves personality and cognitive factors embedded in a social structure that directs and defines the meaning of violence in gendered social relationships. An understanding of the multiple factors that lead to violent behavior in general and to specific forms of violent behavior directed at women is critical to developing effective prevention strategies.
Risk Factors for Victimization
Although most research on the causes of violence focuses on why men use violence and the conditions that support and maintain that violence, some researchers have tried to ask why a particular woman is the target of violence. This line of research has a dismal record of success. A primary problem confronted in trying to identify women's risk factors for violence is the confounding that occurs when traits and behaviors are assessed at some point postvictimization and assumed to represent the previctimization state. An interpretation of current findings is that they represent aftereffects of the violence itself or overly negative self-descriptions triggered by the trauma.
Factors that have been at one time or another linked to women's likelihood of being raped or battered are passivity, hostility, low self-esteem, alcohol and drug use, violence in the family of origin, having more education or income than their intimate partners, and the use of violence toward children. However, based on a critical review of all 52 studies conducted in the prior 15 years that included comparison groups, Hotaling and Sugarman (1986) found that the only risk marker consistently associated with being the victim of physical abuse was having witnessed parental violence as a child. And this factor characterized not only the victimized women, but also their male assailants. Recent studies also found no specific personality and attitudinal characteristics that make certain women more vulnerable to battering (e.g., Pittman and Taylor, 1992). Although alcoholic women are more likely to report moderate to severe violence in their relationships than more moderate drinkers, the association disappears after controlling for alcohol problems in their partners (Miller, 1992, as cited in Leonard, 1993). On the basis of findings such as these, several writers have concluded that the major risk factor for battering is being a woman.
Personality traits and attitudes that could increase vulnerability to rape have also been explored. The earliest studies,
and the only ones to implicate victim personality traits, used different recruitment techniques to obtain subjects: the rape victims were often found among those who had sought help at crisis centers; the nonvictims were college student volunteers (Selkin, 1978; Myers et al., 1984). These methodological differences bias the samples, especially on personality traits like dominance, femininity, and social presence—exactly the variables on which the groups were found to differ. When identical selection procedures were used to select victims and nonvictims, no differences were found in personality characteristics, assertiveness, or identification with feminine stereotyped behavior (Koss, 1985; Koss and Dinero, 1989).
One risk profile did emerge that characterized a small subset (10 percent) of women for whom the risk of rape was twice the rate for women without the profile. Those women were characterized by a background of childhood sexual abuse, liberal sexual attitudes, and higher than average alcohol use and larger number of sexual partners. Researchers presume that having a large number of sexual partners implies short-term relationships and therefore more dating partners, but neither frequency of dates nor number of dating partners has been directly tested as a risk factor. Koss and Dinero (1989) concluded that sexual assault was generally not predictable, but to the extent it could be, was accounted for by variables that represented the aftereffects of childhood sexual abuse, including influences on drinking, sexual values, and level of sexual activity. Recent prospective data support this assertion (Gidycz et al., 1995). Adolescent sexual victimization significantly predicted alcohol consumption at the onset of college, while alcohol consumption during college did not predict subsequent victimization. The link between childhood sexual abuse and adult victimization has been replicated many times across ethnic groups (Wyatt et al., 1992; Gidycz et al., 1993; Urquiza and Goodlin-Jones, 1994; Wyatt and Riederle, 1994). The other certain risk factor for rape (in addition to being female and having been abused previously) is being young: epidemiological data indicate that women
between 16 and 24 years old have the highest rates of sexual assault and rape (Bastian, 1995).
Another line of research has compared the resistance strategies used by women who were raped to those of women whose attack was aborted without penetration. Studies of this type have consistently reported that active strategies such as screaming, fleeing, or physically struggling are associated with higher rates of rape avoidance (Javorek, 1979; Bart, 1981; Quinsey and Upfold, 1985; Levine-MacCombie and Koss, 1986; Siegel et al., 1989; Ullman and Knight, 1991, 1992). Although some of the studies found increased risk of injury among women who resisted, the studies that looked at the actual sequence of events (Quinsey and Upfold, 1985; Ullman and Knight, 1992) found the correlation between resistance and injury disappeared when the violence of the attacker was taken into account. Researchers have uniformly found that offender characteristics are more important than the victim behavior in predicting the outcome of an assault.
The role of alcohol use by victims has also been investigated. Trouble with alcohol and peer pressure to drink have been associated with adolescents' risks of personal victimization, in general, and sexual victimization, in particular (Esbensen and Huizinga, 1991; Windle, 1994; Gidycz et al., 1995). About one-half of college student rape victims report that they were drinking at the time of their assault (Koss and Dinero, 1989), and estimated peak blood alcohol level during the prior 30 days was correlated with lifetime sexual victimization (Norris et al., 1996). Alcohol use is one of the variables that differentiated dates in which sexual aggression occurred from dates involving the same respondents without aggression (Muehlenhard and Linton, 1987).
These studies provide some evidence that the habitual use of alcohol is associated with sexual victimization, but they do not explain the causal pathways. The evidence suggests that alcohol abuse is an aftereffect of earlier victimization, but the effect that alcohol might have on future victimization is unclear. Alcohol may directly increase the risk of victimization
through cognitive and motor impairment that prevents women from recognizing, escaping, or resisting sexual aggression (Nurius and Norris, 1996). Studies of the cognitive effects of alcohol on victims parallel efforts to examine the social information processing of offenders. Rape victims who were drinking report that their judgment was impaired at the time of assault (Frintner and Rubinson, 1993).
It is possible, however, that the effect of alcohol is less direct. Drinking may increase the likelihood of victimization by placing women in settings in which their chances of encountering a potential offender are higher than the average. Several studies have suggested that bar settings increased women's vulnerability to violence independent of the increased vulnerability due to alcohol consumption. For example, exposure to obnoxious behavior, as well as sexual and physical violence, were predicted by the frequency of going to bars (Fillmore, 1985; Lasley, 1989). Alternatively, alcohol consumption by women may be misperceived and misinterpreted by the men they meet as a sexual availability cue. Although scientific evidence suggests that women become less physiologically aroused after drinking, men perceived them as more sexual, more likely to initiate sexual intercourse, and more aroused by erotica (Crowe and George, 1989; George et al., 1990, 1995; Corcoran and Thomas, 1991). In one study, 75 percent of college men admitted to getting a date drunk or high on drugs to try to have sex with her (Mosher and Anderson, 1986).
The consequences of violence against women are far broader than the impact on the women victims. Their families and friends may be affected. In the case of intimate partner violence, there is increasing evidence of the negative impact on children of exposure to violence in the family. Society suffers economically, both in the use of resources and in the loss of productivity due to fear and injury. Understanding the
consequences of violence is necessary for planning and implementing interventions to deal with those consequences. This section examines research findings about the consequences violence against women has on the individual victim, those closest to her, and on society as a whole.
Consequences to Victims
Research in recent years has brought an increased understanding of the impact of trauma, in general, and of violence against women, in particular. Both rape and intimate partner violence are associated with a host of short- and long-term problems, including physical injury and illness, psychological symptoms, economic costs, and death. It should be noted that part of what is known about the consequences of violence against women comes from studies of women who were seeking help, so it may not be representative of all victims. It is possible that these women suffered more severe trauma than women who do not seek help, and so represent the worst cases. The opposite is also possible: that women who come forward have suffered less fear and damage to their self-esteem, and therefore the worst cases remain hidden. Women who agree to participate in research may come from different social, ethnic, and economic backgrounds than those who do not participate. Finally, researchers do not always have the understanding or the resources to reach subgroups of victims who may either be at high risk for violence or face special challenges in recovery.
Virtually absent from the research are studies addressed specifically to the experiences of older women, disabled women, immigrant and refugee women, migrant farm worker women, rural women, Asian American women, American Indian women, homeless women, lesbian and bisexual women, drug-addicted women, and institutionalized women (Eaton, 1995; Gilfus, 1995). Whether or not these groups differ in the overall level of violence they experience, the evidence suggests that the descriptive characteristics of the as-
saults are very similar (Torres, 1991; Wyatt, 1992). However, the same act can have very different meanings depending on many features that shape perceptions and behavior, including the age of the victim, her relationship with the perpetrator, culture, social class, sexual orientation, previous history of violence, perceived intent of the violence, and perceived causes and effects of the violence (Murphy and O'Leary, 1994). Victims from oppressed racial, ethnic, or cultural groups or who are lesbian or bisexual face additional challenges that may influence their strategies and resources for recovery (Brown and Root, 1990; Sue and Sue, 1990; Wyatt, 1992; Garnets and Kimmel, 1993; Schriver, 1995). Most studies of the consequences of violence look at impairments; only a few studies examine resilience and strengths as protectors against untoward outcomes or as alternative results to impairment (Gilfus, 1995).
Also missing in the literature is a developmentally oriented approach that follows the outcomes of exposure to violence into later stages of adult development. Little is known of the impact of trauma on social roles, life patterns, and timing of life transitions. A life-span perspective would look at differential effects on women's lives when violence involves multiple types and perpetrators, is ongoing, cumulative, and becomes a chronic feature of the environment. Many social and public health consequences of violence are unstudied, including labor force participation, economic well-being, fertility decisions, divorce rates, and health status (Gilfus, 1995).
Rape and Sexual Assault Surveys of adult females have found that women characterize the ''typical" rape as entailing a high risk of physical injury and of death (Warr, 1985; Gordon and Riger, 1989). However, the data show that between one-half and two-thirds of rape victims sustain no physical injuries (Beebe, 1991; Koss et al., 1991; Kilpatrick et al., 1992); and
only about 4 percent sustain serious physical injuries (Kilpatrick et al., 1992). Genital injuries are more likely in elderly victims (Muram et al., 1992). It appears that very few homicides are associated with rape: in 1993 only 106 of the 5,278 female homicide victims were also raped (Federal Bureau of Investigation, 1993). Even though serious physical injury is relatively rare, the fear of injury or death during rape is very real. Almost one-half of rape victims in a recent national study (Kilpatrick et al., 1992) feared serious injury or death during the attack. Rape can also result in transmission of a sexually transmitted disease (STD) to the victim, or in pregnancy. STD infection has been found in up to 43 percent of rape victims (Jenny et al., 1990), with most studies reporting STD infection rates between approximately 5 and 15 percent depending on diseases screened for and type of test used (Lacey, 1990; Murphy, 1990; Beebe, 1991). The rate of the human immunodeficiency virus (HIV) transmission due to rape is unknown (Koss et al., 1994), but it is of concern to a sizable proportion of rape victims (Baker et al., 1990). Pregnancy is estimated to result from approximately 5 percent of rapes (Beebe, 1991; Koss et al., 1991).
Rape has health effects that extend beyond the emergency period. Self-report and interview-administered symptom checklists routinely reveal that victims of rape or sexual assault experienced more symptoms of physical and psychological ill health than nonvictimized women (Waigant et al., 1990; Koss et al., 1991; Golding, 1994; Kimerling and Calhoun, 1994). Sexual assault victims, compared with nonvictimized women, were more likely to report both medically explained (30 percent versus 16 percent) and medically unexplained symptoms (11 percent versus 5 percent). Consequently, rape and sexual assault victims also seek more medical care than nonvictims. In longitudinal data, rape victims seeking care at a rape crisis center were initially similar to matched nonvictims in their self-reported physician visits, but at 4 months and 1 year after the rapes they were seeking care more frequently (Kimerling and Calhoun, 1994). These
findings are consistent with studies using population data on medical use: women in primary care populations with a history of severe sexual and physical assault had nearly twice as many documented physician visits a year as nonvictimized women (6.9 versus 3.5; Koss et al., 1991). Utilization data across 5 years preceding and following victimization ruled out the possibility that the victims had been high users of services prior to their attacks.
A number of long-lasting symptoms and illnesses have been associated with sexual victimization including chronic pelvic pain; premenstrual syndrome; gastrointestinal disorders; and a variety of chronic pain disorders, including headache, back pain, and facial pain (for reviews see Koss and Heslet, 1992; Dunn and Gilchrist, 1993; Hendricks-Mathews, 1993). Persons with serious drug-related problems and high-risk sexual behaviors were also characterized by elevated prevalence of sexual victimization (Paone et al., 1992). These findings suggest that victimized women may become inappropriate users of medical services by somaticizing their distress; however, the number of sexual assault victims who qualify for the psychiatric diagnosis of somatization disorder is small. In a comparison of sexual assault victims with matched nonvictimized women on nine psychiatric diagnoses and a sample size of more than 3,000, too few cases of somatization disorder were identified to analyze statistically (Burnam et al., 1988).
Intimate Partner Violence A woman is more likely to be injured if she is victimized by an intimate than by a stranger (Bachman and Saltzman, 1995). Victims of battering suffer from a host of physical injuries, from bruises, scratches, and cuts to burns, broken bones, concussions, miscarriages, stab wounds, and gunshot wounds to permanent damage to vision or hearing, joints, or internal organs to death. Bruises and lacerations to the head, face, neck, breasts, and abdomen are typical. Review of emergency room medical records in one urban hospital revealed that 50 percent of all injuries to
women seen in the emergency room and 21 percent of the injuries that required emergency surgery could be attributed to battering. The review also found that 50 percent of the rapes of women over age 30 had been committed by the woman's intimate partner (Stark et al., 1981). Victims of partner violence were 13 times more likely to have injuries to the breast, chest, or abdomen than were accident victims (Stark et al., 1979), and three times as likely as nonbattered women to sustain injuries while pregnant (Stark and Flitcraft, 1988). Assaults directed at the abdomen can be associated with injuries both to the victim and the fetus (Helton et al., 1987a,b). In a representative national sample, 15 percent of pregnant women were assaulted by their partners at least once during the first half of pregnancy and 17 percent during the latter half (Gelles, 1988). A study of women attending prenatal clinics also found 17 percent of them suffered physical or sexual abuse during pregnancy (McFarlane et al., 1992). Several studies have found that white women experience more abuse during pregnancy than African American or Hispanic women (Berenson et al., 1991; McFarlane et al., 1992).
Women involved with a violent partner may be frequent users of medical services even if they do not identify the reason for their visit as the violence. They are likely to show evidence of injuries in various stages of healing, indicating the ongoing nature of the abusive behavior (Burge, 1989). Among women patients in a community-based family practice clinic who were living with a partner, recently separated, or divorced, 25 percent were assaulted by their partners during the previous year, and 15 percent sustained injuries from a partner (Hamberger et al., 1992). Some of this violence is lethal. Between 1976 and 1987, 38,468 people were killed by their intimate partners; 61 percent involved men who killed women. Among white couples, 75 percent of the victims were women (Browne and Williams, 1989, 1993).
Victims of intimate partner violence and rape exhibit a variety of psychological symptoms that are similar to those of victims of other types of trauma, such as war and natural disaster. Following a trauma, many victims experience shock, denial, disbelief, fear, confusion, and withdrawal (Burgess and Holmstrom, 1974; Walker, 1979; Browne, 1987; Herman, 1992; Janoff-Bulman, 1992; van der Kolk, 1994). Assaulted women may become dependent and suggestible and have difficulty undertaking long-range planning or decision making (Bard and Sangrey, 1986). Although a single victimization may lead to permanent emotional scars, ongoing and repetitive violence is clearly highly deleterious to psychological adjustment (Follingstad et al., 1991). In one national study, the more a woman had been assaulted, the more psychological distress she experienced (Gelles and Harrop, 1989).1
A large empirical literature documents the psychological symptoms experienced in the aftermath of rape (for reviews see Frieze et al., 1987; Resick, 1987, 1990; McCann et al., 1988; Roth and Lebowitz, 1988; Hanson, 1990; Lurigio and Resick, 1990). Rape (with the exception of marital rape) is more likely than partner violence to be an isolated incident, which creates a somewhat different course of recovery. For many victims, postrape distress peaks approximately 3 weeks after the assault, continues at a high level for the next month, and by 2 or 3 months later recovery has begun (Davidson and Foa, 1991; Rothbaum et al., 1992). Many differences between rape victims and nonvictimized women disappear after 3 months with the exception of continued reports of fear, self-esteem problems, and sexual problems, which may persist for up to 18 months or longer (Resick, 1987). Approximately one-fourth of women continue to have problems for several years (Hanson, 1990).
Women who have sustained sexual or physical assault have been found to disproportionately suffer from depression, thoughts of suicide, and suicide attempts (Hilberman and
Munson, 1978; Hilberman, 1980; Kilpatrick et al., 1985; Stark and Flitcraft, 1988; McGrath et al., 1990; Dutton, 1992a,b; Herman, 1992). In one community sample, 19 percent of rape victims had attempted suicide in comparison with 2 percent of nonvictims (Kilpatrick et al., 1985). In other studies, 13 percent of rape victims suffered from a major depressive disorder sometime in their life, compared with only 5 percent of nonvictims (Burnam et al., 1988; Sorenson and Golding, 1990). Depression scores for victims of intimate partner violence on a widely used epidemiological measure (Radloff, 1977) were twice as high as the standard norms and well above the high-risk cutoff scores (Walker, 1984).
Other psychological symptoms reported by both victims of rape and partner violence include lowered self-esteem, guilt, shame, anxiety, alcohol and drug abuse, and posttraumatic stress disorder (PTSD) (Walker, 1979; Burnam et al., 1988; Winfield et al., 1990; Herman, 1992). Even when evaluated many years after they were sexually assaulted, survivors were more likely to receive several psychiatric diagnoses, including major depression, alcohol abuse and dependence, drug abuse and dependence, generalized anxiety, obsessive-compulsive disorder, and PTSD (Kilpatrick et al., 1985; Burnam et al., 1988; Winfield et al., 1990). Women who were both beaten and sexually attacked by their partners were at particular risk of the most severe psychological consequences (Shields and Hanneke, 1983; Pagelow, 1984; Browne, 1987).
There are few reliable predictors of positive readjustment among rape survivors (Hanson, 1990; Lurigio and Resick, 1990). In general, those assaulted at a younger age are more distressed than those who were raped in adulthood (Burnam et al., 1988). Some research has suggested that Asian and Mexican American women have more difficult recoveries than do other women (Williams and Holmes, 1981; Ruch and Leon, 1983; Ruch et al., 1991). Victims of these ethnic backgrounds, as well as Moslem victims, face cultures in which intense, irremediable shame is linked to rape. However, recent direct comparisons have revealed no ethnic differences
in the psychological impact of rape as measured by self-report and interview-assessed prevalence of mental disorders among Hispanic, African American, and white women (Burnam et al., 1988; Wyatt, 1992).
The actual violence of an attack may be less important in predicting a woman's response than the perceived threat (Kilpatrick et al., 1987). The fear that one will be injured or killed is equally as common among women who are raped by husbands and dates as among women who are raped by total strangers (Kilpatrick et al., 1992). Likewise, acquaintance rapes are equally as devastating to the victim as stranger rapes, as measured by standard measures of psychopathology (Koss et al., 1988; Katz, 1991). However, women who know their offender are much less likely to report the rapes to police or to seek victim assistance services (Stewart et al., 1987; Golding et al., 1989). The impact of rape may be moderated by social support (Ruch and Chandler, 1983; Sales et al., 1984). Unsupportive behavior, by significant others in particular, predicts poorer social adjustment (Davis et al., 1991), and proceeding with prosecution appears to prolong recovery (Sales et al., 1984).
One way of systematizing some of the psychological responses evidenced by women victims of partner assault and rape is the diagnostic construct of posttraumatic stress disorder (PTSD) (Burge, 1989; Kemp et al., 1991; Dutton, 1992a). This construct has been used to understand a range of psychological responses to traumatic experiences, from natural disaster or military combat to rape and other forms of criminal attack (Figley, 1985; van der Kolk, 1987; Herman, 1992; Davidson and Foa, 1993). On the basis of clinical and empirical inquiries, a growing number of clinicians now suggest that PTSD may also be the most accurate diagnosis for many survivors of interpersonal and family violence (Herman, 1986, 1992; Bryer et al., 1987; van der Kolk, 1987; Burge, 1989; Gondolf, 1990; Koss, 1990; Davidson and Foa, 1991; Kemp et al., 1991; Koss and Harvey, 1991; Walker, 1991, 1992; Browne, 1992; Dutton, 1992a).
As early as 1974, Burgess and Holmstrom described what they termed "rape trauma syndrome" to describe the psychological aftermath of rape. Today, many assaulted women, like other victims of trauma receive diagnoses of PTSD. Among victims of intimate partner violence recruited from shelters and therapist referrals, 81 percent of those who had experienced physical attacks and 63 percent of those who had experienced verbal abuse were diagnosed with PTSD. Most rape victims (94 percent) who are evaluated at crisis centers and emergency rooms meet the criteria for PTSD within the first few weeks of the assault, and 46 percent still do so 3 months later (Rothbaum et al., 1992). Rape and physical assault are both more likely to lead to PTSD than other traumatic events affecting civilians, including robbery, the tragic death of close friends or family, and natural disaster (Norris, 1992).
Although the concept was initially constructed to explain reaction patterns in survivors of natural disasters and combatants in war, it is not surprising to find a high prevalence of PTSD among survivors of intimate violence. The most common trauma suggested for PTSD in the Diagnostic and Statistical Manual of Mental Disorders (American Psychiatric Association, 1994:427) is "a serious threat to one's life or physical integrity; [or] a serious threat or harm to one's children …," experiences known to characterize the lives of women in relationships with violent mates. Factors most often associated with the development of PTSD include perception of life threat, threat of physical violence, physical injury, extreme fear or terror, and a sense of helplessness at the time of the incident (March, 1990; Herman, 1992; Davidson and Foa, 1993). Moreover, some researchers suggest that PTSD is most likely to develop when traumatic events occur in an environment previously deemed safe (Foa et al., 1989), another dimension clearly applicable to violence occurring in one's home.
Many of the psychological aftereffects of violence against women can be understood as elements of a PTSD diagnosis
(but see below). The PTSD construct has the advantage of providing a framework for recognizing the severe impact of events external to the individual (van der Kolk, 1987; Herman, 1992). However, for reactions to be seen as expectable responses to severe stressors, the trauma must be known. Unfortunately, in most mental health settings, routine screening for a history of family violence is almost never done; thus, serious or chronic psychological and physical conditions are treated without knowledge of the core trauma that may underlie current symptoms.
Finally, PTSD sufferers can become aware of the potential links between the symptoms that plague them and the exposure to an extreme external stressor. Clinical researchers consistently note how abused women internalize the derogatory attributions and justifications of the violence against them (Walker, 1979, 1984; Pagelow, 1984; Browne, 1987). An enhanced understanding of the range of responses manifested by all types of people who are faced with physical or sexual danger or attack expands the interpretation of symptoms beyond internal or gender explanations and empowers both survivors and providers to proceed with focused goals of safety, symptom mastery, reintegration, and healing (Herman, 1992).
Yet there are problems with the PTSD conceptualization. First, it doesn't account for many of the symptoms manifested by victims of violence. For example, thoughts of suicide and suicide attempts, substance abuse, and sexual problems are not among the PTSD criteria. Second, the diagnosis better captures the psychiatric consequences of a single victimization than the consequences of chronic abusive conditions (Herman, 1992). Third, the description of traumatic events as outside usual human experience is not accurate in describing women's experiences with intimate violence. Fourth, the diagnosis fails to acknowledge the cognitive effects of this kind of violence. People who have been untouched often maintain beliefs (or schema) about personal invulnerability, safety, trust, and intimacy, that are incom-
patible with the experience of violence (McCann and Perlman, 1990; Norris and Kaniasty, 1991).
In recent years, the notion of a battered woman syndrome has been used in a variety of legal proceedings, including criminal prosecutions of batterers, criminal prosecutions of women who have attacked their batterers, and divorce and child custody proceedings. The idea of the battered woman's syndrome developed as an attempt to explain the psychological effects of being in a battering relationship and has similarities with the PTSD conceptualization, but it is not a recognized psychiatric syndrome. Rather, it refers to the consequences of being battered as those consequences are represented in expert testimony in legal settings. The use of "battered woman syndrome" has been criticized for making those consequences of intimate partner violence for women a pathology and ignoring differences among battered women's responses to violence (e.g., Dutton, 1993, Schopp et al., 1994). Furthermore, because expert testimony about the experiences of battered women often encompasses more than just a discussion of psychological consequences, the term battered woman syndrome is misleading (Dutton, 1993).
Consequences to Family and Friends
Children in families in which the woman is battered are at risk of both physical (Walker, 1984; Straus and Gelles, 1990) and sexual abuse (Herman and Hirschman, 1981; Paveza, 1988). Even if children are not themselves abused, living in a family in which there is violence between their parents puts children at risk. These children have been found to exhibit high levels of aggressive and antisocial, as well as fearful and inhibited, behaviors (Jaffe et al., 1986a; Christopherpoulos et al., 1987). Other studies have shown that children who have experienced parental violence have more deficits in social competence (Jaffe et al., 1986b; Wolfe et al., 1986) and higher levels of depression, anxiety, and temperament problems than children in nonviolent homes (Jaffe et al., 1986b; Christopher-
poulos et al., 1987; Holden and Ritchie, 1991). Jaffe et al. (1990) also found that children exposed to family violence see violence as an acceptable and useful means of resolving conflict.
Interpreting these findings should be done with caution. Not only is there debate about what constitutes exposure to violence (e.g., actually seeing the violent acts or seeing the results of the violence), but some of the studies have methodological weaknesses. For example, samples are often drawn from among children residing in shelters for battered women. These children are under a lot of stress—beyond that of witnessing violence—related to dislocation and family crisis that may influence their behaviors and feelings. The source of the information may influence the findings; mothers report more behavior problems in children than children self-report (Sternberg et al., 1993). However, these studies suggest that children exposed to parental violence are at potential risk of emotional and behavioral difficulties that may be long lasting.
Depression, developmental problems, acute and chronic physical and mental health problems, and aggressive or delinquent behavior are characteristic of children exposed to battering. An unknown number of the 3 million children exposed to battering each year (Jaffe et al., 1990) end up in foster care. Increased costs for schools, counseling, and juvenile justice programs have not been calculated. There are also unknown long-term costs associated with young boys who are learning how to be future batterers by modeling their fathers' behavior.
Longitudinal investigations that are both labor intensive and expensive are an important way to investigate how witnessing violence between one's parents during childhood is related to violence in one's own intimate relationships during adulthood. Widely cited assertions of intergenerational relationships in intimate partner violence are based on cross-sectional studies, and the findings are open to multiple explanations, including biases inherent in self-report data. There
is evidence that longitudinal research following child victims may be needed to overcome possible problems with forgetting of childhood experiences (L. M. Williams, 1994).
Physical and sexual assaults may also affect other family members and friends, making them into secondary victims. Davis and colleagues (1995) found that rape, attempted rape, and aggravated assault of women all had negative psychological consequences on their friends, family members, and romantic partners, regardless of the victim's level of distress. Female friends and family members were more affected than male friends and family members, particularly in regard to increased fear of violent crime. Some rape victims also experience sexual dysfunction and difficulties with interpersonal relationships, both of which can have negative effects on their family relationships. Sexual dysfunction may be long lasting: Burgess and Holmstrom (1979) found that 30 percent of rape victims reported that their sexual functioning had not returned to normal as long as 6 years after the assaults.
Consequences to Society
Fear of Crime
Criminologists recognize that one social consequence of crime that affects many people beyond those who have been directly victimized is fear of crime (Hindelang et al., 1978; Skogan and Maxfield, 1981). The consequences of fear of crime are real, measurable, and potentially severe (Conklin, 1975; Skogan and Maxfield, 1981). Because women fear crime more than men (Warr, 1985; Gordon and Riger, 1989; Federal Bureau of Investigation, 1991), these consequences are disproportionately borne by women.
Women's fear of crime seems to be driven primarily by their fear of rape (Warr, 1985; Gordon and Riger, 1989; Klodawsky and Lundy, 1994; Softas-Nall et al., 1995). Women perceive rape as a very serious crime—at least as serious, if not more so, than murder (Warr, 1985; Softas-Nall et al., 1995).
The perceived risk of being raped is also high. Warr (1985) found young, urban women believed they were three times as likely to be raped as murdered and equally as likely to be raped as to suffer a less serious offense, such as theft of an auto. Similar ratings of seriousness and a high perceived risk of rape have been found in studies of women in Canada (Gomme, 1986), Great Britain (Smith, 1989), Germany (Kirchhoff and Kirchhoff, 1984), Holland (Van Dijk, 1978), and Greece (Softas-Nall et al., 1995). All these studies also found that women curtail their activities because of this fear: 42 percent of women in Warr's (1985) sample avoided going out alone (compared with only 8 percent of men), and 27 percent of women even refused to answer their door in response to fear.
Existing data give some indication of the social consequences and attendant costs of violence. Straus (1986) estimated that intrafamilial homicide cost $1.7 billion annually; Meyer (1992) calculated the medical costs and lost work productivity of domestic violence at $5 to $10 billion per year; and the Bureau of National Affairs (1990) estimated the annual cost of domestic violence to employers for health care and lost productivity at $3 to $5 billion. Though alarming, the limited data available on women victims of violence and exclusion of sexual violence from these studies suggest that these figures may significantly underestimate the economic toll of violence.
It is estimated that between 12 percent and 35 percent of women visiting emergency rooms with injuries are there because of battering (Randall, 1990; Abbott et al., 1995). Outside of emergency departments, there is practically no information on a myriad of other health costs related to battering and sexual assault, such as treatment for depression and PTSD, drug and alcohol abuse, prenatal complications, sui-
cide attempts, and other chronic physical and psychological conditions.
Estimates of the number of women who are homeless because of battering range from 27 percent (Knickman and Weitzman, 1989) to 41 percent (Bassuk and Rosenberg, 1988) to 63 percent of all homeless women (D'Ercole and Struening, 1990). In New York City, homeless shelters cost $125-130 per day per family; battered women's shelters with a variety of services cost more than $200 a day (Lucy Friedman, personal communication). But there is little information about other social service costs resulting from battering, such as the number of women and children on welfare because of abuse or the total costs of providing battered women with job training and placement, victim assistance services, and child care.
Battering and sexual assault puts an enormous burden on the criminal justice system; a study in the District of Columbia found that 22 percent of 911 calls were from victims of battering (Baker et al., 1989). Yet the full extent of costs to the courts—civil and family, as well as criminal—and law enforcement generally have not been calculated. These include costs associated with getting and enforcing orders of protection; divorce, child custody, and support proceedings; and prosecutions for assault, sexual assault, stalking, trespassing, harassment, and murder, all of which involve personnel costs for prosecutors, judges, defense lawyers, court staff, and police, among others. In addition, anecdotal evidence suggests that some battered women may be forced into performing criminal acts by their batterers (Browne, 1987).
Researchers are just beginning to look at the indirect costs of battering and sexual assault—costs that result not from using services but from reduced productivity and changes in quality of life. For example, a study by Victim Services in New York City found that 56 percent of working battered women had lost a job as a direct result of the violence, and 75
percent had been harassed while they were at work by their partners (Friedman and Couper, 1987). Resick et al. (1981) found women's work performance to suffer up to 8 months after rape. The costs of such reduced productivity or of constricted opportunity are unknown. How many women are prohibited from working by jealous partners or cannot concentrate at work because of battering or sexual assault? How many days are missed by women embarrassed to come to work with a black eye, afraid that the batterer will harass them at the office, or fearful of leaving their homes after being raped? Do partners or family members of rape victims lose time from work because of caring for injured victims or accompanying them to court?
Diminished quality of life is another unexplored indirect cost. What are the costs associated with the isolation, fear, and lack of freedom that plague the lives of battered women and their children? How many activities and opportunities do women forsake out of fear of sexual assault? What are the long-term costs to society of batterers'—and victims'—inability to parent their children? Information on the direct and indirect costs of violence against women would provide a useful guideline for evaluating the cost-effectiveness of intervention programs.
Conclusions And Recommendations
Better understanding of the causes of violence against women will be useful in designing both prevention programs and interventions with offenders. Research has begun to identify childhood precursors to later violent aggressive behavior, and criminological research has studied the progression of criminal careers. Yet little research has considered the development of violence against women and whether pathways to violence against women are similar to the development of other violent behaviors. Nor is it known if physical and sexual violence against women develop in a similar manner and what the nature and extent of the relations among them
are. Identifying precursors to violence against women may be important for early intervention and prevention efforts.
Most of the information on violence against women comes from either clinical samples or general population surveys. Clinical samples are most likely not representative of either victims or perpetrators; in general population surveys, the numbers of ethnic, racial, cultural, and other subgroups are too small for analysis. Differences among subgroups in the causes of violence against women could have important implications for prevention and intervention strategies. Subgroups about which information is lacking include racial and ethnic minorities, lesbians, migrant workers, immigrants, the homeless, the disabled, and the elderly.
Recommendation: Longitudinal research, with particular attention to developmental and life-span perspectives, should be undertaken to study the developmental trajectory of violence against women and whether and how it differs from the development of other violent behaviors. Particular attention should be paid to factors associated with the initial development of violent behavior, its maintenance, escalation, or diminution over time, and the influence of socioeconomic, cultural, and ethnic factors. Funding is encouraged for identification and analysis of existing data sets that include relevant information. In addition, research on the causes and consequences of violent behavior should include questions about violence against women.
Although some of the direct effects of physical and sexual violence (and psychological abuse) on individual women have been fairly well documented, understanding indirect effects to victims, the consequences to women in general, and consequences to the society as a whole is only beginning. Research suggests that women who have been victims of violence seek physicians' care not directly related to the violence nearly twice as often as other women. Some preliminary data indicate that intimate partner violence may play a role in
women's need to receive and remain on welfare. As mandatory arrest laws continue to be passed, and as more jurisdictions encourage filing charges in cases of sexual assault, the criminal justice system faces increased costs. Some research on rape has found reduced job performance for up to 8 months after an assault. There is very little information on lost productivity and reduced performance, on the job and at home, of victims of violence.
Recommendation: Research is needed on the consequences of violence against women that includes intergenerational consequences and costs to society, including lost productivity and the use of the criminal justice, medical, and social service systems. Such research should address the effects of race and socioeconomic status on consequences of violence.