Domestic violence professionals have debated whether all physical assaults by partners should be labeled abuse. This study examined the use of labels such as "abuse," "victim," and "battered woman" in a sample of women (n = 78) who had sustained at least one physical assault in their current or most recent relationship. Self-labeling followed a differentiating strategy, that is, women experiencing more frequent and more severe assaults were more likely to apply labels. Lower partner income, being Black, lower relationship commitment, and having ended the relationship also were associated with increased self-labeling. Labeling of hypothetical acts followed an inclusive strategy, that is, all assaults were considered abusive. These results suggest that contextual factors influence labeling. Prevention and intervention programs may be able to increase their effectiveness by including more situational context in their messages.
When is a woman a battered woman? If her partner has physically assaulted her, is she battered as of the first time? The second? Must the assault be of a minimum level of severity? Relatively little empirical attention has been paid to these questions, although it is clear that the answers are a source of controversy. On the one hand, some domestic violence professionals advocate using labels such as "battering" for all physical violence (e.g., Hamberger & Arnold, 1989), what might be called an "inclusive" perspective. Advocates of this perspective assert that if the definition of battering needs to be changed at all, then it should be made broader to incorporate all acts of coercive control (Stark, 1995). They argue that applying a "calculus of harms" (Stark, 1995, p. 980) that discriminates among victims of violence can lead to minimizing some forms of violence. Such a strategy may limit the scope of public policy and intervention, and in the long run, reduce protection for assault victims. Most prevention and intervention programs in western countries adopt an inclusive strategy; that is, the definitions of "abuse" and "violence" tend to be broad and the cessation of all violence, including controlling and abusive behavior, is the treatment goal (e.g., Gondolf, 1995; McMahon & Pence, 1996; Pence & Paymar, 1993; also see Feldman & Ridley, 1995, for a review). Recent policy initiatives tend to adopt an inclusive perspective as well (e.g., Domestic Violence: Probation Act of California, 1995; Department of Public Health, Commonwealth of Massachusetts, 1995).
Other professionals argue that there are equal dangers in labeling all violence the same (e.g., Rouse, 1989). From this "differentiating" perspective, using the same terms for all physical force trivializes the experiences of women who have suffered the most horrific extremes of violence. Those advocating distinctions among levels and forms of violence suggest that their approach will be most helpful in terms of shaping theory, policy, and intervention because of the potential increases in specificity in each of these domains (Jacobson, Gottman, & Shortt, 1995; Johnson, 1995; O'Leary, 1996; Rouse, Breen, & Howell, 1988; Sedlak, 1988). Recent theoretical and empirical work suggests a marked trend in the direction of differentiating among levels of violence. Studies that differentiate among levels of violence are becoming increasingly common (e.g., Gottman et al., 1995). For example, many authors separate violence into minor forms, such as grabbing, versus more severe forms, such as beating up (Hamby, Poindexter, & Gray-Little, 1996; Sugarman, Aldarondo, & Boney-McCoy, 1996). Johnson (1995) has suggested that partner violence be divided into "patriarchal terrorism" and "common couple violence." These authors believe such a shift would have significant implications for the field.
Although the distinction between the inclusive and differentiating perspectives has been fairly well articulated among domestic violence professionals, little is known about the labeling strategies of women who have sustained physical assault by their partners. Much of what is known focuses on perceived deficits in the strategies that women use to label their experience of relationship violence. Apparently, battered women do not adopt the same inclusive labeling strategy as most domestic violence activists. Usually, victims' unwillingness to apply abuse-related labels is seen as pathological and evidence of denial about the seriousness of their situation (e.g., Ferraro, 1983; Graham, Rawlings, & Rigsby, 1994; Henton, Gate, Koval, Lloyd, & Christopher, 1983; Herbert, Silver, & Ellard, 1991; Walker, 1984, 1993). For example, the symptom of denial is a primary focus of research on the Battered Woman Syndrome (Walker, 1984) and the Stockholm Syndrome (Graham et al., 1994). That research, however, is driven by an inclusive perspective and has not specifically examined the link between victimization experiences and labels. A more accurate, less deficit-focused understanding of victims' perspectives on violence is needed. Although denial may lead some victims to reject abuse labels, it is also possible that rejection of abuse labels by some assaulted women reflects a differentiating perspective on violence more than it reflects denial. Some domestic violence researchers reserve abuse-related labels for a subset of physical assaults, and it is possible that this is also the labeling strategy adopted by the broader community.
This article focuses on the labeling strategies-the link between act and name-used by women who have sustained physical assault. Specifically, we examined their application of labels such as "abuse," "victim," and "battered" to physical force experienced in intimate relationships. We investigated women's use of labels for their own experiences and for violence experienced by others. Because not all assaulted women see themselves as abused (e.g., Graham et al., 1994; Walker, 1984), the first hypothesis was that labeling would not be universal; that is, that some women would decline to endorse abuse-related labels for physical violence. Second, we hypothesized that attributes of the violence would influence the use of labels; we expected that more frequent and more severe force would be associated with greater use of labels, consistent with a differentiating strategy. Third, we hypothesized that attributes of the women and of their relationships would influence their use of labels. Past research conducted with women seeking assistance for abuse (Babcock, Waltz, Jacobson, & Gottman, 1993; Strube, 1988) has documented that level of commitment, resources, and relationship power influence whether women leave relationships, but these factors have not previously been explored in nonhelpseeking samples. We hypothesized that the same relationship factors will also influence the use of abuse labels in our community sample. Other differences associated with demographic and relationship characteristics were explored.
The respondents were 78 women from a random sample of all nonfaculty female employees at a large southeastern state university. The 78 respondents for this study are the women who indicated that "physical force" had been used against them at least once in their current or most recent relationship (39% of all 200 respondents). The study included questions on responses to violence in addition to those presented here (cf. Hamby & Gray-Little, 1997). The women averaged 38.7 years of age (SD 9.62). They were generally White (79%), had at least some college, and had average to above-average incomes (median income $20,000 to $24,000). Most were married (80%) and 75% had been with the current partner for several years. Most (69%) had at least one child. The respondents' partners had virtually identical demographic characteristics, except that their partners had a higher median income ($25,000 to $29,999 per year). Thus household income for the majority of respondents was more than $45,000. All partners were males.
Respondents for the study were randomly selected from the campus directory and contacted at work. An initial letter of inquiry (with response card) was followed by telephone contact for those who did not respond. The overall response rate was 51%. Respondents were contacted at work to give them the opportunity to complete the questionnaires without their partner's knowledge. Questionnaires, presented in a booklet, were sent to those who agreed to participate. It took approximately 30-40 minutes to complete the questionnaires. Followup letters were mailed 1 week after the questionnaires. In order to reduce the imposition on those who did not wish to participate, the reason for nonparticipation was not solicited, but most who volunteered a reason said that they were not in a relationship. For the analyses below, respondents who had not experienced any partner violence were excluded, as the focus of the study is on the labeling patterns among victims of physical force.
All measures were pretested with a group of 96 females prior to use in the present study.
Self-Labeling (Labeling of Personally Experienced Violence). Three items assessed whether women applied to themselves some of the labels commonly adopted by violence researchers. The items were chosen to tap into the emotionally laden language that often surrounds discourse in this area. Each item was presented as a Likert-type scale that ranged from 1 ("Definitely Not") to 7 ("Definitely"). The abuse label referred to their description of the "worst" forceful incident in their relationship: "Do you believe this event is an instance of physical abuse?" The victim label asked whether they applied the term "victim" to themselves ("Do you think of yourself as a victim of violence?"), and the battered label was similarly phrased ("Do you think of yourself as a battered woman?").
The distributions of all three labels were markedly nonnormal and generally bimodal with peak clusters at Score 1 and Score 7 on the Likert-type scale (see Table 1). Because of nonnormal distributions, these variables were dichotomized for further analyses, which is the only transformation possible for bimodal distributions (Tabachnik & Fidell, 1989). Median splits would have entailed placing individuals with scores of only 2 or 3 into the "high" category, even though substantively they were largely rejecting the label. Thus, variables were dichotomized so that individuals using scores of 3 and below were placed in the "low" group and those with scores of 4 and above into the "high" group.
Finally, these labels were combined into a single self-labeling variable in order to explore factors associated with self-labeling patterns. The self-labeling variable had 3 levels or groups. Women who endorsed no labels (i.e., had "low" scores for all 3 labels) were placed into the No Labels Group. Women who endorsed 1 or 2 but not all 3 labels became the Some Labels Group, while women who endorsed all 3 labels became the All Labels Group.
Violent Acts Scale (VAS, Walker, 1984)-Labeling Violent Acts. A modified version of this scale was used to obtain ratings of respondents' attitudes toward various forms of physical violence for which no information about context was provided. Two modifications were made to the original scale, which asks respondents to make severity ratings of the force. First, we added a separate rating of whether they considered each act to be spouse abuse. Second, 6 of the original 16 items were dropped because a pretest sample showed no variability in rating these 6 acts. The remaining 10 acts were the following: push, claw, verbal abuse, hit with an object, knife, throw an object, slap, kick, wrestle, and punch. For this study, ratings for each act from 0 ("Definitely is not spouse abuse") to 10 ("Definitely is spouse abuse") were examined.
Vignettes-Labeling Violence Experienced by Others. The items from the VAS described above are used to obtain ratings of violent acts without reference to context or to the consequences of the acts. In order to assess reactions to violence occurring in the context of an ongoing relationship, we also asked respondents to rate three brief vignettes. The vignettes depicted a grab, a slap, and a punch during the course of an argument They were intended to provide minimal context and to represent increasing levels of severity. Respondents rated each vignette on the same 0 to 10 scale of abusiveness that was used for the VAS. The text of each of the vignettes follows:
"Kim and Steve were arguing heatedly. Steve grabbed Kim's arm to keep her from leaving. Kim could not get out of his grasp."
"Linda and Michael were arguing heatedly when Michael reached out and slapped Linda in the face. Linda's face stung sharply afterwards."
"David and Laura were arguing heatedly when David pulled back and punched Laura in the face. Laura's nose was broken and required medical attention."
Frequency. Respondents were asked to indicate how many times physical force was used during the course of their relationship, using the following seven categories: 0, 1, 2, 3-5, 6-10, 11-20, and more than 20 (however, individuals with scores of 0 were dropped in this sample as outlined in the Respondents section). This measure of aggregate frequency was shown in another sample to be a very reliable measure of having sustained violence (Hamby, Poindexter, & Gray- Little, 1996). It has the added benefit that, unlike many measures of partner violence, the respondent must herself classify the act as physical force to be judged to have sustained violence.
Severity. Judges rated the severity of the incidents related in participants' narrative descriptions. Judges' severity ratings have been shown to adequately assess the construct of partner violence (Hamby et al., 1996). The directions to the participants were as follows: "We are trying to learn about the ways people act in different relationships, especially if and when partners use force. Please think about the most forceful or physically threatening episode in your relationship, even if you would not normally consider it violent. All respondents should answer these questions. Some examples of forceful acts that sometimes happen are pushing, hitting, shouting, slapping, slamming the door, biting, cursing, threatening with a weapon, and so forth." Examples were included because pretesting indicated that some respondents omitted descriptions of verbal aggression or only described the emotional consequences of the actions and not the acts. The first question following these instructions read, "Describe what you and your partner did (or said), including what kind of force was used."
Two female undergraduate psychology majors were trained to rate the severity of the incident described on a 7-point scale; the judges were blind to other information about the respondents and to the hypotheses of the study. Both judges rated all descriptions and they achieved an effective reliability of .95. This coefficient indicates the aggregate consistency of the two judges' scores and is appropriate for noncategorical ratings (Rosenthal & Rosnow, 1991).
Judges rated about half of the incidences in the lowest two violence categories: grabbing (10%), and pushing or shaking that did not knock the victim off balance (40%). The rest of the acts were judged as moderate-to-severe levels of violence such as slapping (18%), hitting or knocking down (19%), or punching, hitting repeatedly, and more injurious acts (13%).
Respondents were asked to provide their age, educational level, race, and individual yearly income. The same questions were asked about their partners in order to determine if their partner's social resources affected labeling. Relationship length, marital status, relationship status (current or terminated), and number of children were obtained in the demographics section of the questionnaire.
Relationship Commitment Scale (Rusbult, 1980). A 5-item questionnaire developed by Rusbult was included to assess respondents' degree of commitment to their relationship. Questions asked how likely it was that the respondent would end the relationship, for what length of time she thought it would last, for what length of time she wanted it to last, how attached she was to the relationship, and how committed she was to the relationship. Each question was scaled on a 9-point Likert-type scale, with higher scores indicating more commitment. The mean was used in analyses. For this group of women, internal consistency was .93 as measured by coefficient alpha.
Decision Making in Intimate Relationships (DMIR) Scale. The DMIR is an 11-item scale of relationship power (Hamby, 1996). The DMIR was patterned after the scale used by Blood and Wolfe (1960) to study decision-making power, but the original scale was modified so that the items are appropriate for use with both cohabiting and noncohabiting couples. For each of 11 specific content areas (e.g., what to do on weekends, when to have sex), participants indicated on a 5-point scale who usually had the final say (partner always, partner usually, equal, myself usually, myself always). Higher scores indicate that more decisionmaking power resides with the respondent. Internal consistency (alpha) was .79.
Self-Labeling: Descriptive Data
The patterns of endorsement across the three dichotomized labels ("abuse," "victim of violence," and "battered woman") were examined. Figure 1 shows that a substantial portion (38%) of the respondents did not endorse any labels for their own experiences of force. The next largest group, about a fifth (22%) of respondents, endorsed the abuse label, but not the other two labels. Interestingly, the next largest group consisted of women who endorsed all three labels (18%), while an almost equal number endorsed the abuse and victim labels but not the battered label (17%). If one examines only those women who reported multiple (2 or more) physical assaults (47 of 78 women in the sample, or 60% of the sample), 26% of multiply victimized women still declined all labels and only 26% endorsed all three. Thus not all women who have sustained violence consider themselves to be victims, abused, or battered.
Other-Labeling: Distributions and Association With Self-Labeling
Violence Acts Scale (VAS). Every item on the VAS was rated as abuse (M 8.04 to 9.89). In order to examine the relationship of self-labeling (No, Some, or All labels) to labeling of VAS items, we conducted a multivariate analysis of variance (MANOVA), with selflabeling as the independent variable and the 10 VAS items as dependent variables. This analysis indicated that ratings of the VAS items were not significantly associated with selflabeling (all/?> .15).
Vignettes. A repeated measures analysis in which ratings of the vignettes was the dependent measure revealed that respondents differentiated among levels of abuse depicted in the vignettes; F(2, 74) = 29.21, p < .0001. Mean abusiveness ratings increased from the grab (7.19, SD = 2.86) to the slap (9.40, SD = 1.35), to the punch (9.96, SD = 0.25) vignette. This pattern of increasing abusiveness ratings across the three vignettes was not associated with self-labeling, however (p > .50); nor was there any effect for self-labeling on labeling of any of the individual vignettes, ( all ps > .55).
A similar pattern of differentiation among the vignettes is observed if the distribution of vignette ratings is dichotomized in a manner similar to the self-labeling variables. If the responses are dichotomized so that scores at or above the midpoint are considered to be endorsing the label, then 82% labeled a grab as spouse abuse, 99% labeled a slap as abuse, and 100% labeled a punch as abuse (see Figure 2).
Inferential Analyses: Self-Labeling
Self-labeling (No labels, Some labels, and All labels) was used as the independent variable in a series of multivariate analyses of variance (MANOVA). In the sections below we present a series of analyses examining the relationship of self-labeling to (a) frequency and severity of force experienced by the respondent; (b) demographic characteristics; and (c) relationship characteristics. Significant univariate results were further examined with two orthogonal planned comparisons, one comparing the No Labels Group to both of the other two groups, and one comparing the Some Labels Group to the All Labels. Further, frequency and severity offeree were tested as covariates to see if the association between labeling and demographic and relationship attributes was mediated by differences in degree of force.
Frequency and Severity of Force. See Table 2 for the means and standard deviations of the force variables. The overall MANOVA for the force variables was significant; F(4,146) = 9.34, p < .0001. As expected, both increasing frequency [F (2, 74) = 7.79, p < .001] and severity [F (2, 74) = 12.09, p < .0001] of physical force were associated with a greater endorsement of labels. Planned contrasts indicated that women using no labels reported less frequent [F(l, 74) = 15.38,;? < .001] and less severe [F(l, 74) = 24.10, p< .0001] violence than women using any labels. Contrasts comparing the women in the Some Labels Group to those in the All Labels Group revealed no significant differences in their experience of force; for frequency, F (1, 74) = 2.26, p > . 10, and for severity, F (1, 74) = 1.11, p > .25.
Demographic Characteristics. We found that severity offeree (but not frequency) was associated with demographic characteristics [F (8,59) = 2.58, p < .05]. In order to examine the unique association of demographic characteristics to self-labeling, severity was used as a covariate in the analyses. Overall, there was a significant association between self-labeling and demographic characteristics, F (16, 118) = 3.33,p < .0001. Inspection of univariate effects revealed that labeling differed as a function of respondents' race, F (2,66) = 11.59, p < .0001, and partner's race, F (2,66) = 9.60, p < .001. Planned comparisons indicated that being Black was associated with a greater endorsement of labels. The force covariate was significant, too, because White respondents reported more forceful acts than did Black respondents. A virtually identical pattern was found for partners' race, because the two race variables are highly intercorrelated with each other (r = .96).
Partners' income also differed across self-labeling groups, F(2, 66) = 9.43, p < .001. Contrasts indicated that women in the No Labels Group reported significantly more income than those in the other two groups, F(l, 66) = 18.64, p < .0001, but women in the Some Labels and All Labels groups did not differ in income from each other, F(l, 66) = 0.94, p > .30. Increased partners' income was also associated with greater severity of force. Other demographic characteristics did not distinguish the groups. The means and standard deviations for these analyses can also be found in Table 2.
Relationship Characteristics. There were significant overall differences among the labeling groups on relationship variables, F(12,126) = 2.24, p < .05. Neither force covariate was significant, p > .40, and hence neither was included in the final analyses. Relationship status was related to self-labeling. Having terminated the relationship was associated with increased self-labeling, F(2, 66) = 7.41, p < .01; both planned contrasts were significant, F(l, 66) = 9.02, p < .01; and F(l, 66) = 10.45, p < .01. Relationship commitment was lowest among those who used labels, F(2, 66) = 4.03, p < .05, although only the contrast comparing the No Labels Group to the other two was significant [F(l, 66) = 7.27, p < .01]. The contrast between the Some Labels and All Labels group approached but did not achieve significance [F(l, 66) = 2.85, p < .10]. The overall effect for decision-making power only approached significance [F(2, 66) = 2.64, p < .08], but the planned contrast between women in the No Labels Group versus other women was significant, F(l, 66) = 5.10, p < .05. The No Labels group described their relationships as less male dominated than those of the other groups. The contrast between the Some and All Labels Groups was not significant for decision-making power, p > .25. Relationship length, marital status, and number of children were not associated with self-labeling, all p > .15. In summary, increasing endorsement of abuse labels was associated with having terminated the relationship, lower relationship commitment, greater likelihood of view the relationship as male dominated.
Inferential Analyses: Labeling Others
A parallel set of MANOVAs was conducted to examine the association of otherlabeling to the frequency and severity of force as well as to demographic and relationship characteristics. However, neither the labeling of acts on the VAS, nor the vignettes, was associated with frequency nor severity of one's own sustained violence (all p > .05). Furthermore, there were no associations among any of the demographic and relationship characteristics and labeling of the VAS items or for the vignettes, all p > .05.
Use of Both Inclusive and Differentiating Strategies
This study indicates that women who reported that they have experienced assault by their partners generally adopt a differentiating labeling strategy with respect to their own abuse and an inclusive one with respect to abuse that does not personally involve them. Moreover, in this sample, women's labeling of hypothetical acts was not associated with their selflabels, suggesting that two rather different processes may be at work. Women's use of a differentiating labeling strategy for their own experiences is similar to the strategy of researchers who argue that terms like battering should be reserved for more frequent and more severe violence (e.g., Johnson, 1995; Rouse, 1989). Their use of an inclusive strategy for other's experience parallels the practice of other domestic violence professionals who believe that all relationship force should be labeled as violent (e.g., Family Violence Prevention Fund, 1995; Hamberger & Arnold, 1989; Stark, 1995).
Several factors appear to be associated with women's use of labels for their own experiences. Besides the effects of violence frequency and severity, self-labeling is also influenced by the context created by qualities of the relationship. Most of our findings are consistent with the proposition that violence occurring in the context of more valued relationships is less likely to lead to labeling. Thus, women who were more committed to their partners were less likely to adopt labels. Rusbult's commitment measure, used in this study, emphasizes wanting to continue in the relationship and probably reflects a personal valuing of the relationship more than commitment due to moral obligations or structural constraints (Johnson, Caughlin, & Huston, 1999). Similarly, women who were still in a relationship with their partners were less likely to adopt labels. Less male dominant, more egalitarian relationships, which may, in today's society, be associated with the perceived value of the relationship (Risman & Johnson-Sumerford, 1998), were also characterized by less labeling. Similarly, partners' resources influenced labeling; women who declined labels had partners who made considerably more money, on average, than did women who used labels. Finally, the observed racial differences suggest that the broader sociocultural context may also affect labeling, but interpretations should take into account the limited number of Black respondents. These racial differences would need to be replicated before firm conclusions can be drawn.
Do These Findings Indicate Denial?
Context might partially explain why women adopt inclusive strategies for violence not directed at them and differentiating strategies for their own experiences. Researchers have long noted similar actor-observer differences in perceptions of events (e.g., Jones & Nisbett, 1971). The kind offeree experienced by the women in this study was similar to that depicted in the vignettes and VAS. All of them had experienced something at least as forceful as a grab, yet more than twice as many of the respondents rated the grab vignette as abusive as rated their own experience of "being grabbed" as abusive. It seems likely that disparity in the amount of available information is the important difference between actors and observers. This type of actor-observer difference is often assigned to the fundamental attribution error, the tendency to explain events involving the self in situation-specific terms, but to attribute the behavior of others to transsituational factors (Ross & Sicoly, 1979). This study indicates that relationship-specific circumstances might include strong commitment and egalitarian decision making. Other circumstances, such as the details of precipitating events, might also affect labeling. Without such knowledge of mitigating circumstances, women describe violence experienced by others in a manner that is consistent with abstract societal norms about relationship violence.
Context does affect perceptions in other studies as well. Norris, Nurius, and Dimeff (1996) found that women rated their own risk of rape as lower than the risk of hypothetical vignette characters. Similarly, Crosby and colleagues (1989) found some self-other differences in a study of perceived job discrimination. The authors of both studies interpreted their results as signs of denial and bias, but do not consider that context might affect not only perceived risk but actual risk. For example, Norris and colleagues document that potentially important information about context (for example, feeling comfortable in the social environment) affects rape perceptions, but assume that judgments affected by such information are more inaccurate than judgments made without this information. The Crosby and colleagues study could also be interpreted as reflecting real contextual differences. In a sample composed predominantly of middle- and upper-class lesbian women, participants reported they had personally experienced less discrimination than they perceived to be true for lesbians as a group, consistent with the authors' denial hypothesis. Conversely, a study comprised primarily of working class participants produced findings contrary to their hypothesis: those respondents perceived they had experienced more discrimination than other individuals. Unfortunately, the Crosby and colleagues studies did not measure participants' actual experience of discrimination (e.g., by measuring income differentials). Thus, while those authors conclude that their data are signs of denial of personal disadvantage, it also seems possible that the middle- and upper-class women may be accurately reporting less discrimination and the working class participants accurately reporting more. In other words, their perceptions may be accurately influenced by contextual information that is not available to them in their ratings about other people.
Examples such as these suggest that context does not always affect perceptions in a manner that constitutes bias. Whether the effect of context on attributions ever constitutes an "error" has been disputed by some (e.g., Harvey, Town, & Yarkin, 1981). It is perhaps safest to conclude that the availability of information influences perceptions in a wide array of interpersonal situations (e.g., Jones & Nisbett, 1971; Ross & Sicoly, 1979). The recent focus on how context affects perceptions about victimization experiences can be seen as an extension of previous social psychological understanding of this common phenomenon and not as evidence of some pathological process that uniquely characterizes battered women (e.g., Graham et al., 1994; Walker, 1984). As Harvey et al. have pointed out, determining accuracy in interpersonal perceptions is not always a straightforward matter. Given that many domestic violence professionals themselves adopt a differentiating perspective that takes context into account, it seems wise to consider less deficit-focused interpretations of these processes than those that dominate the victimization literature.
An appreciation of the merits of both the inclusive and differentiating perspectives toward partner violence may help increase the efficiency and effectiveness of prevention and intervention programs, as well as bring additional coherence to research on this topic. Prevention programs might realistically focus on an inclusive goal of preventing all violence. To do otherwise may dangerously promote a perception of tolerance for limited and circumscribed violence. Nonetheless, public service messages about domestic violence should also take into consideration that most victims of violence appear to view their own experiences in the context of their relationships. Unfortunately, many prevention and education programs often present messages without a context. For example, slogans like "There's NO excuse for domestic violence" and "Some men break more than their girlfriends' hearts" (FVPF, 1995) may do an excellent job of communicating general societal attitudes about domestic assault. The lack of context in these ads, however, may be one reason that they are not very effective in changing women's behavior. By contrast, the contextualized formats seen in movies such as "The Burning Bed" or "What's Love Got To Do With It," and wellpublicized incidents such as the murder of Nicole Brown Simpson (Jones, 1995) have led to large increases in help-seeking by victims of domestic violence. These personalized formats are uniquely well suited for conveying the relational and sociocultural contexts that often surround violence.
Intervention programs for victims may benefit from a greater incorporation of a differentiating perspective. Clients of shelters and other services may find it difficult to integrate their own beliefs with the ideology of shelters, which tend to be based on ideas that developed from Western forms of feminism (Timmins, 1995), and which may be seen as antimale or antimarriage. It is possible that a differentiating approach, with a more explicit recognition of context including positive relationship features, could help alleviate some of these problems. For example, providers could encourage women to explore ways of identifying all of the factors that affect their responses to the violence and discuss how adopting abuse, victim, or battered labels affects their ideas about themselves, their relationships, and the violence they have experienced.
To understand fully how victims make sense of violence, researchers should include individuals who have sustained all levels of violence. A limitation of the present study, for example, is that the most extreme forms of violence were not well represented. The levels of violence included in this sample, however, are typical of most forms of relationship violence. National data suggest that minor violence is approximately 4 times more common than severe violence (Straus & Gelles, 1990), and other avenues of research (e.g., Aldarondo & Sugarman, 1996; Feld & Straus, 1989) suggest that a significant number of relationships do not remain violent over time. Sample selection strategies focusing on only the most extreme forms of violence will tend to promote portrayals of dysfunction that may be erroneously generalized to represent all couples who experience violence. Those failing to include the severe abuse may miss the opportunity to examine qualitative differences occurring in women's response to severe violence. Thus a framework that accounts for all levels of violence is needed, and it seems likely that such a framework will incorporate both the inclusive and differentiating perspectives.
Aldarondo, E., & Sugarman, D. B. (1996). Risk marker analysis of the cessation and persistence of wife assault. Journal of Consulting and Clinical Psychology, 64, 1010-1019.
Babcock, J. C, Waltz, I, Jacobson, N. S., & Gottman, J. M. (1993). Power and violence: The relation between communication patterns, power discrepancies, and domestic violence. Journal of Consulting and Clinical Psychology, 61, 40-50.
Blood, R. O., & Wolfe, D. M. (1960). Husbands and wives: The dynamics of married living. Glencoe, IL: Free Press.
Crosby, F. J., Pufall, A., Snyder, R. C., O'Connell, M., & Whalen, P. (1989). The denial of personal disadvantage among you, me and all the other ostriches. In M. Crawford & M. Gentry (Eds.), Gender and thought (pp. 79-99). New York: Springer-Verlag.
Department of Public Health, Commonwealth of Massachusetts. (1995). Massachusetts Guidelines and Standards for the Certification of Batterers Intervention Programs. Boston, MA: Author.
Domestic Violence: Probation Act of California, Statutes of California and Digests of Measures, Chapter 641. (1995).
Feld, S. L., & Straus, M. A. (1989). Escalation and desistance of wife assault in marriage. Criminology, 27, 141-161.
Feldman, C. M., & Ridley, C. A. (1995). The etiology and treatment of domestic violence between partners. Clinical Psychology: Science and Practice, 2, 317-348.
Ferraro, K. J. (1983). Rationalizing violence: How battered women stay. Victimology, 3-4, 203-212.
Gondolf, E. W. (1995, July). Discharge criteria for batterer programs. Presented at the 4th International Family Violence Research Conference.
Gottman, J. M., Jacobson, N. S., Rushe, R. H., Shortt, J. W., Babcock, J., LaTaillade, J. J., & Waltz, J. (1995). The relationship between heart rate reactivity, emotionally aggressive behavior, and general violence in batterers. Journal of Family Psychology, 9, 227-248.
Graham, D. L. R., Rawlings, E. I., & Rigsby, R. K. (1994). Loving to survive: Sexual terror, men's violence, and women's lives. New York: New York University Press.
Hamberger, L. K., & Arnold, J. (1989). Dangerous distinctions among "abuse," "courtship violence," and "battering": A response to Rouse, Breen, and Howell. Journal of Interpersonal Violence, 4, 520-522.
Hamby, S. L. (1996). The Dominance Scale: Preliminary psychometric properties. Violence and Victims, 11, 199-212.
Hamby, S. L., & Gray-Little, B. (1997). Responses to partner violence: Moving away from deficit models. Journal of Family Psychology, 339-350.
Hamby, S. L., Poindexter, V. C., & Gray-Little, B. (1996). Four measures of partner violence: Construct similarity and classification differences. Journal of Marriage and the Family, 58, 127-139.
Harvey, J. H., Town, J. P., & Yarkin, K. L. (1981). How fundamental is the "fundamental attribution error"? Journal of Personality & Social Psychology, 40, 346-349.
Henton, J., Cate, R., Koval, J., Lloyd, S., & Christopher, S. (1983). Romance and violence in dating relationships. Journal of Family Issues, 4, 467-482.
Herbert, T. B., Silver, R. C., & Ellard, J. H. (1991). Coping with an abusive relationship: I. How and why do women stay? Journal of Marriage and the Family, 53, 311-325.
Jacobson, N. S., Gottman, J. M., & Shortt, J. W. (1995). The distinction between Type 1 and Type 2 batterers-further considerations: Reply to Ornduff et al. (1995), Margolin et al. (1995), and Walker (1995). Journal of Family Psychology, 9, 272-279.
Johnson, M. P. (1995). Patriarchal terrorism and common couple violence: Two forms of violence against women. Journal of Marriage and the Family, 57, 283-294.
Johnson, M. P., Caughlin, J. P., & Huston, T. L. (1999). The tripartite nature of marital commitment: Personal, moral, and structural reasons to stay married. Journal of Marriage and the Family, 61, 160-177.
Jones, C. (1995, October 15). Nicole Simpson, in death, lifting domestic violence to the forefront as national issue. The New York Times, p. A28.
Jones, E. E., & Nisbett, R. E. (1971). The actor and the observer: Divergent perceptions of the causes of behavior. Morristown, NJ: General Learning Press.
McMahon, M., & Pence, E. (1996). Replying to Dan O'Leary. Journal of Interpersonal Violence, 11, 452-455.
Norris, J., Nurius, P. S., & Dimeff, L. A. (1996). Through her eyes: Factors affecting women's perception of and resistance to acquaintance sexual aggression threat. Psychology of Women Quarterly, 20, 123-145.
O'Leary, K. D. (1996). Physical aggression in intimate relationships can be treated within a marital context under certain circumstance. Journal of Interpersonal Violence, 11, 450-452.
Pence, E., & Paymar, M. (1993). Education groups for men who batter: The Duluth model. New York: Springer Publishing.
Risman, B. J., & Johnson-Sumerford, D. (1998). Doing it fairly: A study of postgender marriages. Journal of Marriage and the Family, 60, 23-40.
Rosenthal, R., & Rosnow, R. L. (1991). Essentials of behavioral research: Methods and data analysis (2nd ed.). New York: McGraw-Hill.
Ross, M., & Sicoly, F. (1979). Egocentric biases in availability and attribution. Journal of Personality & Social Psychology, 37, 322-336.
Rouse, L. P. (1989). Reply to Hamberger and Arnold. Journal of Interpersonal Violence, 4, 523-527.
Rouse, L. P., Breen, R., & Howell, M. (1988). Abuse in intimate relationships: A comparison of married and dating college students. Journal of Interpersonal Violence, 3, 414-429.
Rusbult, C. E. (1980). Commitment and satisfaction in romantic associations: A test of the investment model. Journal of Experimental Social Psychology, 16, 172-186.
Sedlak, A. J. (1988). Prevention of wife abuse. In V. B. Van Hasselt, R. Morrison, A. Bellack, & M. Hersen (Eds.), Handbook of family violence (pp. 319-358). New York: Plenum Press.
Stark, E. (1995). Re-presenting woman battering: From battered woman syndrome to coercive control. Albany Law Review, 58, 973-1026.
Straus, M. A., & Gelles, R. J. (1990). How violent are American families? Estimates from the National Family Violence Resurvey and other studies. In M. A. Straus & R. J. Gelles (Eds.), Physical violence in American families: Risk factors and adaptations to violence in 8,145 families (pp. 95-112). New Brunswick, NJ: Transaction Publishers.
Strube, M. J. (1988). The decision to leave an abusive relationship: Empirical evidence and theoretical issues. Psychological Bulletin, 104, 236-250.
Sugarman, D. B., Aldarondo, E., & Boney-McCoy, S. (1996). Risk marker analysis of husband- to-wife violence: A continuum of aggression. Journal of Applied Social Psychology, 26, 313-337.
Tabachnik, B. G., & Fidell, L. S. (1989). Using multivariate statistics (2nd ed.). New York: Harper & Row.
Timmins, L. (Ed.). (1995). Listening to the thunder: Advocates talk about the battered women's movement. Vancouver, Canada: Women's Research Centre.
Walker, L. E. (1984). The battered woman syndrome. New York: Springer Publishing.
Walker, L. E. (1993). The battered woman syndrome is a psychological consequence of abuse. In R. Gelles & D. Loseke (Eds.), Current controversies on family violence (pp. 133-153). Newbury Park, CA: Sage Publications.
Acknowledgments. Support for this study was provided by the Smith Fund of the University of North Carolina, the Family Research Laboratory of the University of New Hampshire, and NIMH grant 5-T32-MH15161. The authors would like to thank Lauren Duncan, Murray Straus, and the members of the Family Research Seminar for their comments on an earlier draft of this article.
Sherry L. Hamby
University of North Carolina at Chapel Hill
Offprints. Requests for offprints should be directed to Sherry L. Hamby, Family Research Laboratory, University of New Hampshire, 126 Horton, Durham, NH 03824.…