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Warning This Web page has been archived on the Web.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 Child Witnesses and/or Victims

Most studies testing the cycle of violence hypothesis use one of two basic research designs. The first design is used with adult or adolescent respondents and is usually retrospective in nature. The respondent's current level of violence is considered in conjunction with information about marital violence and/or child abuse perpetrated by the respondent's parents. The second design involves assessing adult respondents' current level of violence and relating it to their school-aged children's concurrent adjustment. Although research using the latter design does not test the intergenerational hypothesis directly, studies of children are valuable given the evidence that aggressive children can remain aggressive into adulthood (Eron, Huesmann & Zelli, 1991; Farrington, 1991). A number of mechanisms by which viewing/experiencing violence may lead to later violence have been proposed (Jaffe, Hurley & Wolfe, 1990). Social learning theorists would suggest that violence in the family may directly affect behaviour given that parents are potent models which children are likely to imitate. An indirect effect on behaviour may also occur via the impact of family violence on attitudes regarding the appropriateness of violence. From an attachment perspective, violence disrupts the internal working model that is developed through interaction with parent; this distorted model of relationships is in turn carried forward as the prototype for future relationships (Sroufe & Fleeson, 1986).

Much of the research addressing the impact of family violence on children has been conducted with agency-identified samples, i.e., children and their mothers are recruited from battered women's shelters or from child protection agencies (Jaffe, Hurley & Wolfe, 1990). For example, a recent study of the long-range effects of witnessing marital violence compared current and former child residents of a battered women's shelter with a non-violent control group (Wolfe, Zak, Wilson, & Jaffe, 1986). The battered mothers were asked to report on their children's aggression using the Child Behaviour Checklist (CBCL; Achenbach & Edelbrock, 1981), a measure designed to identify children who exhibit behaviour problems serious enough to warrant clinical intervention. Former residents had not been exposed to marital violence for at least 6 months while the current resident group had been exposed within the last 6 weeks. According to the mothers' reports, there were no significant behavioural differences between children who had witnessed violence (either in the recent or more distant past) and children who had not.

However, it would be inappropriate to interpret these findings as proof that there are no long-term effects associated with witnessing violence. The sample used in the study was quite unusual in that the comparison group of children scored an average of one standard deviation above the norm on the Child Behaviour Checklist2. Scores as high as these are surprising for a control group, especially given the clinical nature of this instrument. This aberration may have obscured group differences, leading to the determination that there are few differences between witnesses and non-witnesses when such a conclusion may actually be erroneous. Although children in this group were not significantly different from the other two groups with regard to behavioural problems exhibited, it is quite possible that they would differ in other important ways. The Child Witness to Violence Interview (Jaffe, Wilson, & Wolfe, 1988) was designed to assess reactions that are not considered in standard measures of adjustment like the CBCL. Studies conducted using this interview indicate that children exposed to wife battering are significantly more likely to condone violence as a means of resolving conflict in interpersonal relationships in addition to feeling responsible for the violence and possessing inadequate safety skills to deal with a violent event (Jaffe et al., 1988).

2A similar finding was reported in Christopoulos, Cohn, Shaw, Joyce, Sullivan-Hanson, Kraft & Emery (1987), i.e., elevated externalizing scores were found in a community sample of boys.

Davis & Carlson (1987) also looked at the effects of witnessing violence among 66 children residing in a battered women's shelter. The sample ranged in age from 4 to 11 years. It was assumed that by virtue of their coming to the shelter, they had all witnessed marital violence. Half of these children were also known to be victims of abuse as they had been involved with local child protective services. A comparison of witnesses to victim/witnesses using the Child Behaviour Checklist showed higher levels of aggression by the second group, as reflected in the proportion of children whose scores were in the clinical range (24% of witnesses and 47% of victim/witnesses). Gender and age were also significantly related to the aggression factor, with school-aged girls and pre-school boys having the highest mean scores and the highest proportion of children scoring in the clinical range (greater than 50% in both groups). However, there were no analyses involving both gender and abuse status so it is unclear what proportion of aggressive boys and girls had solely witnessed violence and what proportion had also been victimized. This failure to make clear distinctions among children with respect to witness and victim status is a frequent occurrence - in the Wolfe et al. (1986) study described earlier, the authors failed to specify whether any of the children had been abused themselves. Fantuzzo & Lindquist (1989) examined the literature on child witnesses of marital violence and found that the presence of child abuse was not even assessed in 75% of the articles reviewed. This finding is somewhat surprising given the fact that those children who have witnessed their mother being beaten have often themselves experienced direct physical assaults by parents (Walker, 1984; Bowker et al., 1988).

Another drawback associated with Davis & Carlson's study is the reliance on mothers to provide information about their children's aggressive behaviour. Studies have shown that the presence of marital discord and abuse in the family can bias parents' descriptions of their children's behaviour (Hughes & Barad, 1982). Nevertheless, the mother is often the sole source of information in many studies of children's reactions to family violence. Recognizing this limitation, Dodge, Bates & Pettit (1990) utilized a variety of methods to assess aggressive behaviour in their sample of 309 kindergarten children. Teachers were asked to rate children on the school version of the Child Behaviour Checklist while peers made nominations regarding which children tended to start fights, get angry and be mean toward others. Direct observations of the children's behaviour were also made. Abuse status in this study was determined from an interview with mothers regarding disciplinary practices and deliberate physical harm done to the child. Based on the mothers' responses, the researchers subjectively determined the probability that any particular child had been physically abused. This procedure resulted in 15% of the sample being classified as "harmed." On average, these children were rated by teachers and peers as significantly more aggressive than comparison children. Group differences with respect to direct observations of aggressive behaviour failed to reach statistical significance although the rate at which aggressive acts were committed was 30% higher for the harmed children. The main effect of child's abuse status on aggressive behaviour remained significant even when other variables, including severity of marital conflict as rated by interviewer, were statistically controlled.

The results of these three studies are not directly comparable since the first contrasted witnesses with controls, the second compared victim/witnesses with victims and the third compared victims to non-abused children. Such lack of consistency in group composition is not uncommon in studies of family violence sequelae. A study by Hughes (1988) was slightly better in that it included children from three of the four possible groups: children who had witnessed marital violence (n=40), children who had been abused and had also witnessed violence (n=55) and children from non-violent families (n=83). Hughes found a significant interaction between age and abuse status with respect to the number of externalizing problems reported by mothers on the Eyberg Child Behaviour Inventory (Eyberg & Ross, 1978). Abused/witness children in the preschooler and primary school age groups displayed significantly more problems than witness and comparison children. It appears that children who receive a "double dose" of family violence, i.e., they witness violence between their parents and they also experience child abuse, show the greatest tendency to display aggressive behaviour. This relationship appears to make intuitive sense; it has been uncovered in subsequent child research by the same investigator (Hughes, Parkinson & Vargo, 1989) and it emerges again in research addressing adult witness/victims' tendencies toward marital violence (Kalmuss, 1984).

Given the considerable amount of overlap between the witness and victim groups, it seems inadequate to look at the impact of either form of family violence alone. The results would likely reflect the confounding of the variables and not the unique effect of either type of violence. In an attempt to address this issue, both marital aggression and child abuse were traced across three generations in 181 families (Doumas, Margolin & John, 1994). Current family functioning was assessed via a number of standardized questionnaires completed by parents, including the Domestic Conflict Index (Margolin, Burman, John & O'Brien, 1990; cited in Doumas et al., 1994), the Child Hostility Inventory (Kazdin, Rodgers, Colbus & Siegel, 1987) and the Child Behaviour Checklist. Unfortunately, child abuse currently occurring was not recorded. Only the potential for child abuse was assessed using a paper and pencil survey presumed to measure abusive disciplining style3 (Child Abuse Potential Inventory; Milner, 1986). Doumas and her associates found that elevated scores by parents on the Child Abuse Potential Inventory predicted aggressive behaviour in boys but not in girls. Unexpectedly, marital violence between parents was not significantly related to boys' or girls' aggression. The findings of this study are consistent with other research showing a stronger link between problem externalizing behaviours and exposure to family violence for boys (Jaffe, Wolfe, Wilson & Zak, 1985, 1986; Hughes & Barad, 1983; Hurley & Jaffe, 1990).

3The problems encountered when attempting to infer behavioural tendencies from attitudinal measures will be described in a later section.

Hughes (1986) found that among children residing in shelters, the effects of viewing violence depended on such variables as the child's gender and abuse/victim status, as well as the child's age, mother's mental health functioning, and amount of violence observed. Comparisons between children who have observed marital violence and those who have not show that serious behaviour and emotional problems are 17 times more frequent among boys who witness violence and 10 times more frequent among female witnesses (Myers Avis, 1992). Male victims/observers are thought by some to be at greater risk for becoming violent. In fact, one researcher declared that "male children over the age of 12 are frequently, as a matter of policy, not allowed to stay in shelters for battered women because of the aggressive and violent behaviours, they have learned at home" (Hofford, 1991; p. 13). The issue of gender-linked differences in violent behaviour win be raised again with adolescent and adult samples given that Miller & Challas (1981; cited in Saunders, 1994) found that among those who had been abused in childhood, men were almost twice as likely as women to be rated at high risk for becoming abusive parents.

In summary, the studies described in this section lead us to the same conclusion that Widom (1989b) came to when she surveyed the research linking child victimization to later aggression: abused children seem to display more aggressive behaviour than comparison children. For some children, witnessing marital violence is as detrimental to healthy adjustment as experiencing physical abuse (Jaffe, Wolfe & Wilson, 1990; Widom, 1989b). Generally though, being abused seems to lead to more severe behavioural outcomes than does witnessing violence, while the experience of both forms of violence represents the most potent predictor of aggression (Cooper, 1992). However, many of the research design flaws pointed out by Widom five years ago are present in investigations currently being conducted. Studies often involve convenience or opportunity samples of mothers and children recruited from women's shelters or social service agencies. Many investigators rely solely on descriptions of children's behaviour given by mothers who are themselves experiencing abuse, a practice which may detract from the validity of the findings. Moreover, the time between admission to the shelter/service agency and laboratory testing is often quite short. It is possible that some of the behavioural differences detected after admission are actually reflective of difficulties in adjusting to the shelter. The upcoming sections of the report address the issue of whether violent behaviour by family violence victims/witnesses persists beyond an initial adjustment phase and into adolescence and adulthood.