Correctional Service Canada
Symbol of the Government of Canada

Common menu bar links

The Impact of Experiencing and Witnessing Family Violence during Childhood: Child and Adult Behavioural Outcomes

Warning This Web page has been archived on the Web.

Aggressive Behaviour by Witnesses and/or Victims in Adulthood

Violence in general

The results of studies conducted on generalized violent tendencies by abuse victims/ witnesses are sometimes contradictory. For example, the results of a study by Forsstrom-Cohen & Rosenbaum (1985), utilizing a paper and pencil measure of aggression, are quite different from those obtained when aggression was measured via arrests for violent crime (Widom, 1989a). In the first study, undergraduate students were classified as belonging to one of three types of families based on their reports of the quality of marital interactions viewed in childhood: 44 had witnessed parental violence, 43 viewed discord between parents that stopped short of physical violence, and 77 described their parents' marital relationship as satisfying. Within the witness group, the mean amount of time that had passed since the last violent parental episode was between 4 to 7 years - thus the study was presumed to assess the long term effects of viewing violence. In an effort to isolate the effects of witnessing violence, students who reported that they had been physically abused by parents were excluded from the sample. Aggressive behaviour tendencies were measured via the Buss-Durkee Inventory (Buss & Durkee, 1957). A significant gender by relationship interaction was found wherein women who viewed violence between their parents were more aggressive than either of the other two groups. Surprisingly, there was no discernible relationship between witness status and aggression among male subjects.

In the second study, Widom (1989a) found that victimized males were in fact more likely to be violent. She followed the previously described sample of abused and neglected individuals into adulthood in order to assess the relationship between childhood victimization and adult criminal behaviour. As mentioned earlier, the sample in her longitudinal study was identified through their involvement in court cases involving neglect and/or abuse. Overall, the abused and neglected sample was more likely than the control group to have an adult criminal record (28.6% vs. 21.1%). With respect to violent crime in particular, abuse victims were marginally more likely to have been arrested for a violent offence than were non-abused controls (8.5% vs. 6.1%). When the relationship was reanalysed after excluding those individuals without a "match" in the control group, the differences between the two groups did reach statistical significance (Widom, 1989c). In a reversal of the pattern found during adolescence, the relationship between abuse status and later violent criminal behaviour was largely attributable to differences between abused and control males (15.6% vs. 10.2% arrest rate respectively); the differences between abused and control group females were not significant. The discrepancy in the results of these two investigations is possibly due to the different ways in which "aggression" was measured. The 66-item Buss-Durkee Inventory is used to obtain self-reports on a variety of hostility- related behaviours. However, only 10 of the items directly address the respondents' inclinations toward physical violence. Widom's criterion for aggression was much more narrow, i.e., the prevalence of violent behaviour was determined via documented police contact.

Yet despite its potentially limited usefulness in assessing the degree to which "violence begets [physical] violence," the Buss-Durkee Inventory has nevertheless proven valuable in identifying different types of violent individuals. To illustrate, there has been some debate in the literature as to the degree of overlap between violence directed against non-family members and family-directed violence. The Buss-Durkee has been used in researchers' attempts to determine whether family violence perpetrators are "violent men or violent husbands" (e.g., Cadsky & Crawford, 1988).

This same question has been addressed using a self-report approach; Kandel-Englander (1992) assessed the extent to which men who are violent outside the family are also violent with partners and/or children using Straus' 1985 national probability sample. Although there was no information provided regarding what proportion of this sample had witnessed and/or experienced violence in their family of origin, those men who admitted to having been violent (15% of total sample) were inclined to restrict their violence to one type of victim. Within the violent group, 23% were violent against non-family members only and 67% were batterers only. Only 10% had aggressed against both their wives and non-family members; these men were labelled "pan-violent".

This estimate of pan-violence is considerably lower than that obtained in previous studies using reports of family violence obtained from clinical samples of battered women. Nevertheless, the author admitted that the study's emphasis on non-incarcerated men may have led to underestimation of the size of both the pan-violent and the non-family violent group. Men who were violent against non-family members would be more likely to have been incarcerated and thus excluded from consideration in her study. In a study that specifically focused on an offender sample, Dutton and Hart (1992b) found that the co- occurrence of family-directed and non-family violence was considerably higher; 79% of the men whose files contained references to perpetration of family violence had also committed some form of stranger/non-family violence.

Violence directed at partners

Detailed analyses of risk markers for spousal abuse indicate that the association between experiencing child abuse and later perpetration of spousal abuse is somewhat inconsistent but the observation of marital violence in childhood has reliably emerged as a risk marker among men for violence against their partners (Hotaling & Sugarman, 1986; Tolman & Bennett, 1990). This latter relationship is supported by data from the recent Statistics Canada survey (1993). Women whose partners grew up with a violent father were three times as likely as women with nonviolent fathers-in-law to be assaulted by their partners. The association between men's viewing violence and battering is also apparent in Rouse's (1984) survey of 79 males regarding violence received and observed during childhood and the relationship to battering. Curiously, the measure used to assess violence in the marital relationship, the Conflict Tactics Scale (CTS), was sent only to a subsample (n=55). No explanation was given for this decision. Nevertheless, the observation of violence in childhood was strongly related to CTS scores in adulthood.7 The effect of victimization on CTS scores was mediated by the effect of observing violence; the correlation between victimization and CTS scores became negative after observation was statistically controlled.8 This result suggests that being the target of violent acts in childhood reduces the tendency to use abusive tactics with a female partner among witnesses to violence. Still, the author warns that there are limitations in the study that serve to qualify the results obtained: the wording of the items dealing with violence made reference to very mild types of exposure, there was no information available regarding the frequency of exposure, and the context in which violence occurred was not specified (i.e. it was not necessarily parent to child violence).



The link between witnessing marital violence and subsequent violence in the adult family is further supported by the results of Stith and Farley's (1993) study, designed to generate a predictive model of male spousal violence. The sample was comprised of a group of men participating in a battering treatment program and another group of men involved in a program for alcoholism. Using path analysis, the authors found that approval of marital violence, measured using a modification of the Inventory of Beliefs about Wife Beating (Saunders, Lynch, Grayson & Linz, 1987), was one of two variables having a direct effect on the use of severe marital violence (traditional sex-role attitudes also contributed to wife assault). Observation of marital violence in childhood had a further indirect effect on marital violence via its positive effect on approval of violence and its negative effects on self-esteem and egalitarian attitudes. However, given the sample used in the development of this model, Stith and Farley cautioned that it might only be applicable to men who were in treatment programs.

The importance of witnessing violence in the family of origin can also be seen in research conducted with offender samples. The results of studies using offenders also highlight the important role that experiencing violence in childhood plays in the perpetration of family violence in adulthood (perhaps more so than do studies conducted with the general population). Dutton & Hart (1992b) conducted a file- review of the records belonging to 597 federally incarcerated male offenders in CSC's Pacific Region. Regardless of the type of crime for which they were currently incarcerated, these men were categorized according to the violent acts they had committed and the target of those violent acts (stranger or family member). Files were searched for references to charges or convictions for offences against persons or references to allegations of violence reported or investigated but not resulting in charges. Using this system, 74 men were classified as non-violent, 346 as violent with strangers or acquaintances only and 177 as violent with family members (included spouse- and child-directed violence). The extent to which the offenders had been victimized in childhood was also assessed via reports in the files of neglect, physical or sexual abuse or inter-parental violence. Forty-one percent of offenders had been victimized in some way during childhood; 31.4% had been physically abused. The investigators found that men who were violent in their adult families were significantly more likely to have been physically abused in their family of origin (41.4%) than were inmates who had records of being violent with non-family members only (29.9%) and non-violent offenders (14.9%). The family violent group was also more likely to have witnessed abuse in their family of origin (20.3%) than were Stranger Violence offenders (11%) or non-violent offenders (5.4%).

In subsequent analyses with a similar sample, Dutton & Hart (1992a) made further subdivisions within the family violence and stranger violence groups by differentiating between physically and sexually violent crimes. Finer distinctions were also made with respect to type of abuse suffered in childhood, i.e. physical abuse, sexual abuse and witnessing marital violence. The results indicated that various types of abuse were associated with specific patterns of adult offending. The men tended to commit the same type of crime that had been perpetrated against them in childhood, lending support to the cycle of violence hypothesis within an offender population. Of the men who had been physically abused as children, 65% committed crimes of physical aggression in adulthood and of the men who had been sexually abused, 58% later committed sexual crimes. The experience of physical abuse in childhood increased the odds for physical violence in the adult family by 5 times and the odds of stranger-directed violence twofold.

In their national study of offenders, Robinson & Taylor (1994) found results that echoed Dutton & Hart's regional findings with respect to the proportion of men having family violence histories. Almost half (46.2%) had been victimized in childhood, experiencing or witnessing some form of family violence; 24.6% had been physically abused whereas 23.8% had witnessed the abuse of parents or siblings. As Widom (1989b) would predict, not all offenders who had official histories of abuse showed evidence of being perpetrators. Nevertheless, those who were victims or witnesses to abuse were almost twice as likely to be violent with family members than those who were not. Of the files containing references to violence in the family of origin, 42.4% also contained evidence of violence within the adult family (23.9% of non-victimized offenders' files had references to family violence perpetrated). Looking at partner abuse in particular, the base rate among non- victimized offenders was 22.6% whereas 33.7% of victimized offenders were abusive with their wives. Witnessing abuse appeared to be the most important family of origin correlate of partner abuse.

Other studies using non- incarcerated samples have underscored the association between childhood abuse and later marital violence. Marshall & Rose (1988) surveyed a sample of 330 undergraduate witnesses and victims of violence in childhood using a modified version of the Conflict Tactics Scale. Forty percent reported that they saw their fathers hit their mothers, 40.6% reported seeing mothers hit fathers and 76.7% received violence at the hands of their parents. Violence by and against the respondent in adult relationships was also assessed. Seventy-five percent of the sample reported that they had expressed violence in an intimate relationship while 64% received violence from a partner. The results of the study indicated that greater violence in any particular relationship was associated with more violence in all of the other relationships. Significant correlations emerged between expressing violence in adulthood and being abused as a child, expressing violence and viewing father to mother violence, and expressing violence and witnessing mother to father violence. Experiencing child abuse also predicted expressed violence for men and received violence for women.

Focusing specifically on batterers, Caesar (1988) studied a group of 44 men in therapy for discordant marital relationships. Within this sample, 26 had Conflict Tactics Scale (Straus, 1979) scores that identified them as batterers. Batterers were more likely to report having been abused by one or both parents (38% vs. 11%) and to have been disciplined harshly with a switch, belt, etc. (58% vs. 31%). Overall, batterers were more likely to have been exposed to some type of family violence, either witnessing parental violence or having been themselves abused (62% vs. 28%), and to get a "double dose" of family violence (27% vs. 0%). Furthermore, among men with a childhood history of violence, the batterers were more likely to be involved in marital strife and also more likely than the non-violent group to idealize the violent parent or rationalize his behaviour. This latter result is consistent with findings obtained by attachment researchers in the context of child abuse, i.e., abusive mothers who were themselves victimized tended to idealize their past (Egeland, 1993; Oliver, 1993).

The studies reviewed thus far deal with violence inflicted by those victimized in childhood, however, a great deal of research effort has also been expended on exploring the so-called "cycle of victimization." In their 3 -generation study of marital violence and child abuse (described earlier), Doumas and her associates (1994) found that not only was viewing marital aggression in childhood related to husbands' aggression against their wives, witnessing marital aggression in the family of origin was significantly related to wives' victimization by husbands. These findings appear to suggest that women who have experienced family violence in the past may be prone to choose abusive partners. The "cycle of victimization" represents an interpretation of events that is controversial in the family violence literature because it connotes passivity and helplessness, even an acceptance of violence by the battered women. Despite the controversy, Cappell & Heiner (1990) explicitly framed their results relating childhood witnessing and victimization to later spousal and child abuse in terms of an "intergenerational transmission of vulnerability." Straus' national probability sample (1980) was used; respondents reported on the frequency of having been hit by either parent during their teens and of witnessing violence by one parent against the other. The Conflict Tactics Scale was used to assess current family relations between spouses. Cappell and Heiner found that marital violence in family of origin increased the likelihood that a respondent (male or female) was the target of aggression from his/her spouse.

The relationship between witnessing violence and being battered found in these two investigations coincide with the conclusions drawn in an extensive review of the risk markers for husband to wife violence (Hotaling & Sugarman, 1986). More than forty female partner characteristics were examined but only one, witnessing violence between parents in the family of origin, was consistently associated with being battered. Although the female partner's characteristics were of limited value for explaining marital violence, nine of 38 abuser characteristics and five couple-level variables were identified as consistent risk markers for husband to wife violence. Given this situation, it appears that researchers' efforts would be better spent concentrating on perpetrators of family violence than on the "cycle of victimization."

Violence directed at children in the adult family

In her longitudinal study of abused individuals, Widom (1989c) found no significant differences between abused group and controls with respect to arrests for child abuse. But as alluded to earlier, child abuse is an underreported crime. Furthermore, arrests for child abuse are rare, therefore these results are not very compelling. In another study pertaining to family violence by offenders, the Virginia Department of Corrections (1983) surveyed a random sample of inmates (n=202) regarding their childhood history of abuse and regarding child abuse perpetrated by them. One-quarter (25.2%) reported having been abused in childhood. However, it was not possible to calculate the rate of intergenerational transmission since only one offender in this sample admitted to having abused his own child. Further checking through the records of all 9,131 inmates incarcerated in Virginia at the time showed that 240 (2.7%) were identified as having been charged with or suspected of child abuse. Because a substantial proportion of the offenders did not have children (23.2%), the actual percentage of child abusers in this sample is even higher. In fact, Robinson & Taylor's (1994) research with offenders indicates that the proportion of victimized individuals who go on to abuse their own children is quite a bit higher than the Virginia data would suggest. Twenty percent of offenders with histories of witnessing or experiencing abuse had records of perpetrating child abuse in their files compared to 6.4% of the non-abused men.

Turning to the general population, it seems that in recent years there has been a move toward assessing child abuse potential rather than actual occurrences of abuse. This trend seems to be related to a change in research focus from clinical samples to samples drawn from the larger community. Because of mandatory child abuse reporting laws, investigators increasingly seem to be turning to measures of potential in the hopes that respondents will respond more truthfully to the hypothetical scenarios presented in such instruments than they would to direct questions about abuse perpetrated. In one such study of abuse potential, Zaidi, Knutson & Mehm (1989) utilized the Assessing Environments III Scale (AEIII; Berger, Knutson, Mehm & Perkins, 1988) to uncover the punishment history of 86 undergraduates. Forty-nine had severe physical punishment histories while 37 had histories involving only mild forms of punishment. The Analog Parenting Task (APT; Larrance & Twentyman, 1983) was used to assess the disciplinary responses that students endorsed in hypothetical child discipline scenarios. Punishment history was significantly related to choice of discipline but only up to a point. The severely punished group was more likely to respond to an initial incident of child misbehaviour with potentially injurious discipline, however no group differences emerged if misbehaviour was prolonged.

Another inventory that appears to be receiving a great deal of use is the Child Abuse Potential Inventory. Doumas and her associates (1994) found that fathers' scores on this measure were related to a history of marital aggression between their own parents (i.e., second generation males' potential for abusing their children was linked to viewing marital aggression between their own parents (child's grandparents)). Eleven percent of the variance in child abuse potential was explained. This result is in line with the notion that the impact of family violence is greater on males. However, there was no significant relationship between violence in mothers' family of origin and their child abuse potential. This finding is in contrast to the results obtained when Cappell & Heiner (1990) assessed the occurrence of child abuse rather than abuse potential; they found that marital violence in the mother's family of origin predicted child-directed aggression by the mother.

Milner, Robertson & Rogers (1990) asked undergraduate students about their childhood history of receiving and observing physical abuse and compared these responses to scores obtained on the Child Abuse Potential Inventory. A substantial proportion (21.1%) had been physically abused to the point of injury or had observed acts that resulted in injury (14.7%). The correlation between abuse received and/or observed and inventory scores was moderate9. Respondents were then divided into 4 groups based on the researchers' determinations of the severity of abuse received. The "moderate" and "severe" groups had significantly higher scores on the CAP than did the "no abuse" and "mild" groups, a finding that provides support for the cycle of violence hypothesis. Taken together with the Zaidi et al. (1989) study, these results suggest that potential for child abuse can be measured even before an individual becomes a parent. Earlier studies conducted by the developers of the Child Abuse Potential Inventory seem to support this view.

9 r=0.30

Milner, Gold, Ayoub & Jacewitz (1984) assessed 190 parents from the At Risk Parent-Child program using the Child Abuse Potential Inventory -interestingly, 90% of the clients assessed were female, indicative of a possible bias toward the study of abusive mothers. Assessment were conducted when the children were 6 months old or less and it was established that no abuse had occurred yet in these families. Unfortunately, no information was provided regarding the childhood abuse history of the "at risk" parents. Child abuse allegations were investigated for a period of 6 months following the initial testing sessions. One hundred and three parents had scored above the designated cut-off score for abuse whereas there were 42 confirmed reports of child maltreatment. Because the program included a therapeutic component, the authors felt that individuals who obtained elevated scores but no abuse allegation might be considered successfully treated rather than "inventory prediction errors." The correlation between abuse scores and subsequent confirmed reports was r=.34, thought by the authors to be a conservative estimate of the association. Retaining the focus on female perpetrators, Caliso & Milner (1994) studied 26 abusive mothers with a self-reported history of childhood abuse; 26 non abusive mothers with a history of being abused and 26 non abusive mothers with no history of violence. Using discriminate analysis techniques, 84.6% of abusers and 90.4% of non abusers were correctly identified.

The developers' research on the Child Abuse Potential Inventory seem to indicate that this is a very promising measure. However, prior research has shown that the connection between attitude and behaviour is not always direct. Dibble & Straus (1980) carried out research on attitude- behaviour consistency with respect to how "necessary," "normal" and "good" it was to slap a 12-year- old. A large majority (81.5%) expressed at least some endorsement for one or more of the parental violence attitudes and of these respondents, 72% had actually used violence against their children. The surprising finding was that even among those who did not endorse attitudes condoning violence, 37% had engaged in violence against their children. Both mothers and fathers were more likely to be violent with their children when their partners had used physical punishment on their children. In short, parents' behaviour tended to be in agreement with their spouse's behaviour even if it was in disagreement with their own attitudes. It appears that more investigations need to be conducted with measures of abuse potential before firm conclusions can be drawn about how this research fits in to the "cycle of violence" literature.

Senior abuse

The problem of senior abuse is an area within family violence that is receiving increasing attention. Research in this area is still in the early stages therefore a detailed overview will not be presented here. However the topic is relevant to this review because in some cases, former witnesses and/or victims of abuse may become violent toward parents when the dynamics of the parent-child relationship change. For example, Carlson's (1990) study of adolescent witnesses to family violence revealed what might be a precursor of senior abuse. Boys from violent homes had a greater tendency to hit their mothers than did other boys although the difference was of marginal significance. None of the remaining relationships between witnessing marital aggression and use of violence against parents reached statistical significance; interpretation of the results may have been complicated by the fact that respondents' abuse and witness status were confounded -one-third of the witnesses to marital aggression were also victims of child abuse.

Other family violence researchers have speculated about the relationship between childhood abuse received and later senior abuse perpetrated. Johnson (1979; cited in Kashani et al., 1992) put forth a theory suggesting that the abused child, when an adult, abuses his or her parent in return. Schlesinger (1984; cited in Appleford, 1989) says that there is a 50% chance that abused children will themselves abuse their dependent parents10.

10Still, based on the limited information available on this form of family violence, it appears that in the majority of cases, senior abuse is "wife abuse grown old" (National Clearinghouse on Family Violence fact sheet, 1992).