Controlling for Child Effects
Another potential source of spuriousness in the association between abuse victimization and violent behavior, is the “child effect.” A strong child effect could completely confound the association between abuse and violence. Biological factors have been predictive of violence in many studies, and the manifestation of biological predispositions may include personality traits or behavior that might elicit abuse. In this example, the abused child commits violence due to a biological predisposition to commit violence, and being abused by parents is only coincidental (i.e., the variables are correlated with no causal influence). Thus, controlling for a child effect would be expected to nullify the association between abuse and violence. In the second instance, the child’s predisposition has an indirect effect on violence through its evocation of abuse. The child elicits abuse, and abuse causes his violent behavior. In this case, abuse is still causally associated with later violent behavior, but a structural model is required to test the indirect effect. To be sure, both processes may occur at the same time.
Confounding child effects are most expected in studies of physical violence. If child effects are relevant, we would expect to see them most obviously in studies of corporal punishment because children are generally spanked because of their naughty behavior. In addition, children with conduct disorder (CD) may be underresponsive to reinforcement (Lytton, 1990) eliciting punishment instead. Other studies have demonstrated that CD children are more likely to receive negative parenting, even when the parents are randomly assigned to mind them (Anderson, Lytton, & Romney, 1986). Parents are predictably critical and punitive when children are aggressive, even when the parents have been instructed to behave in a highly nurturant fashion (Yarrow, Waxler, & Scott, 1971).
The operationalization of child effects is fraught with difficulty. Genetically- informed designs are probably the most conservative, assigning the greatest proportion of variability to “genetic” factors. Some genetically-informed studies have reported that genetics play a small role in the receipt of physical discipline (Jaffee, Caspi, Moffitt, Polo-Tomas et al., 2004; Schulz-Heik et al., 2009), but some have also reported that the contribution of genetics to maltreatment was not statistically significant (Jaffee, Caspi, Moffitt, & Taylor, 2004). This is actually surprising because of the ubiquity of significant “heritability” in genetically- informed designs. A slightly less conservative approach is to use child behavior as a proxy for a child effect. Some studies have demonstrated that children’s aggressive, externalizing, antisocial behavior is followed by parents’ harsh discipline (e.g., Lansford et al., 2011; Linares, Montalto, Rosbruch, & Li, 2006; Riggins- Caspers, Cadoret, Knutson, & Langbein, 2003).
Some authors have used longitudinal designs in conjunction with controls for child effects in an attempt to disentangle the problem of temporal order and mutual influences, and these studies have consistently reported that harsh overreactive discipline influences later child antisocial behavior (Jaffee, Caspi, Moffitt, & Taylor, 2004; Lynch et al., 2006; Nix et al., 1999), and the effects of the child’s behavior on subsequent discipline have not been different from zero (Lansford et al., 2011; O’Leary, Slep, & Reid, 1999). More work remains to confirm whether the findings would hold (1) for other forms of abuse, (2) for violent behavior rather than the more general “externalizing,” and (3) among samples of adolescents.