Witnessing violence toward siblings: an understudied but potent form of early adversity

Abstract

Research on the consequences of witnessing domestic violence has focused on inter-adult violence and most specifically on violence toward mothers. The potential consequences of witnessing violence to siblings have been almost entirely overlooked. Based on clinical experience we sought to test the hypothesis that witnessing violence toward siblings would be as consequential as witnessing violence toward mothers. The community sample consisted of unmedicated, right-handed, young adults who had siblings (n=1,412; 62.7% female; 21.8±2.1 years of age). History of witnessing threats or assaults to mothers, fathers and siblings, exposure to parental and sibling verbal abuse and physical abuse, sexual abuse and sociodemographic factors were assessed by self-report. Symptoms of depression, anxiety, somatization, anger-hostility, dissociation and 'limbic irritability' were assessed by rating scales. Data were analyzed by multiple regression, with techniques to gauge relative importance; logistic regression to assess adjusted odds ratios for clinically-significant ratings; and random forest regression using conditional trees. Subjects reported witnessing violence to siblings slightly more often than witnessing violence to mothers (22% vs 21%), which overlapped by 51-54%. Witnessing violence toward siblings was associated with significant effects on all ratings. Witnessing violence toward mother was not associated with significant effects on any scale in these models. Measures of the relative importance of witnessing violence to siblings were many fold greater than measures of importance for witnessing violence towards mothers or fathers. Mediation and structural equation models showed that effects of witnessing violence toward mothers or fathers were predominantly indirect and mediated by changes in maternal behavior. The effects of witnessing violence toward siblings were more direct. These findings suggest that greater attention be given to the effects of witnessing aggression toward siblings in studies of domestic violence, abuse and early adversity.

Figure 1. Diagram of classic mediation model.

Panel A illustrates the total effect of the independent variable (IV) on the dependent variable (DV) as measured by regression coefficient c. Panel B shows the indirect effect of IV on DV via mediator M. The indirect effect is quantified as path a (IV→M) times path b (M→DV) or ab. The director effect is indicated by c′.

Figure 2. Venn diagram.

Overlapping circles indicating the number of subjects who witnessed threats or assaults to mother, father or sibling and the degree of co-occurrence. There were 937 subjects in the sample who witnessed no threats or assaults to family members.

Figure 3. Relative importance – variance decomposition.

Comparison of the relative importance of witnessing threats or assaults to mother, father or siblings based on regression analyses and variance decomposition with covariates (not shown) for sexual abuse, parental and sibling verbal abuse, parental and sibling physical abuse, sex and demographic factors.

Figure 4. Relative importance – random forest regression.

Comparison of the relative importance of witnessing threats or assaults to mother, father or siblings based on random forest regression using conditional trees with additional regressors (not shown) for sexual abuse, parental and sibling verbal abuse, parental and sibling physical abuse, sex and demographic factors.

Figure 5. Path analysis.

Best-fitting structural equation model showing the interrelationships between independent variables (witnessing violence toward mothers, fathers or siblings, childhood sexual abuse), dependent variables (ratings of depression, anxiety, somatization, anger-hostility, dissociation, limbic irritability) and mediators (maternal or sibling verbal aggression). Only significant paths are shown. Significant covariate relationships between variables of the same type were omitted for clarity. Paths are color-coded to indicate whether the origin of the arrow is from a variable associated with mother (red), father (blue), sibling (blue), or childhood sexual abuse (green). Standardized beta weights are provided with asterisks indicating significance levels. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001, †p<10⁻¹⁰.