Studies Accounting lor Abuse
Although we have argued that associations among abuse, parental rejection and child outcomes are likely to be confounded, deciding whether or not to control for abuse is not straightforward. While it is possible that abuse and warmth are inversely correlated, and a control should be applied, it is also possible that the impact of abuse is not wholly due to the physical reality of abuse, but due, at least in part, to the feeling of rejection experienced by the abused child. As such, a control for abuse may be overly conservative because the common variance shared with the parental rejection measure would be removed from partial parameter estimates. Progressive models comparing R2 have been recommended to look into this possibility.
Two studies included in this review used models that included spanking or abuse as covariates. As mentioned above, Stacks et al. (2009) employed a conservative model and did not find an association between maternal warmth
(at 24 months) and age 36 month aggression for any of their racial subgroups in models that included spanking. In addition, the interaction between spanking and warmth in the model was not significant, either. Using Add Health data, however, Tyler et al. (2011) found a direct effect of parental warmth on dating violence perpetration in a model controlling for basic demographics and three forms of abuse. They tested whether the effect of warmth on dating violence was mediated by its association with delinquency, and that was not the case.
It might be instructive to add some information from analyses that were not part of this review. There are numerous studies where authors test the possibility that indicators of attachment (not warmth) mediate or moderate the association between abuse and violent behavior. Rebellon and van Gundy (2005) found that the quality of relationships between parents and adolescents was significantly associated with violence measured a year later controlling for many factors including abuse. Several other studies have found significant associations between measures of attachment/caregiver sensitivity and violence in models where measures of abuse or harsh parenting are included (e.g., Alink et al., 2009; Haskett, Allaire, Kreig, & Hart, 2008; Salzinger et al., 2007; Sousa et al., 2011). A few studies of general delinquency (not violence specifically) also report these associations. Hipwell et al. (2008) report reciprocal relationships among girls where parent- child warmth and use of harsh punishment independently predict girls’ conduct problems; furthermore, girls’ conduct problems predicted changes in harsh punishment over time.
Thus, the question is not settled, and future research disentangling associations between abuse, rejection and aggression is needed.