Although we began this chapter with the argument that traumatic abuse is most likely to lead to serious behavior problems, the number of studies on trauma and consequent violent behavior is not large enough to make a serious case regarding an association between psychological trauma and violent behavior in its victims. In our review, 3 out of 3 studies reported a statistically significant PoC linking trauma and violent behavior (Maschi et al., 2008; McNulty & Bellair, 2003; Peacock et al., 2003). The number of studies of nonviolent offending was even lower (k = 2), and the findings were split between 1 study that was in the predicted direction and another that was in the predicted direction and also statistically significant (Maschi et al., 2008 and Peacock et al., 2003, respectively). In offender samples, 2 out of 2 studies comparing violent to nonviolent offenders report that violent offenders had experienced significantly more trauma (Neller et al., 2006; Ramoutar & Farrington, 2006).
Studies that provided estimates of coefficients for the association between trauma and violent and nonviolent offending separately do not offer much in favor of our hypothesis. Maschi et al. (2008) report that the association is significant for violence, but not for nonviolent offending, but Fagan et al. (1983) report the opposite pattern, Peacock et al. (2003) report a significant coefficient for both, and findings reported by Ramoutar and Farrington (2006) do not support an association for either.
For this topic, there are many studies that look at violent and nonviolent offending among offenders. If violent offending is associated with abuse in an offender sample, this can be seen as one approach to controlling for nonviolent offending. Several studies suggest that traumatic experiences predict violence among the criminally-prone (e.g., Fagan et al., 1983; Lewis et al., 1979;). Neller et al. (2006) reported that most items on the trauma questionnaire were associated with violence in the predicted direction, but a preponderance was not statistically significant. All in all, however, that study was supportive of an association. Byrd and Davis (2009) report partial evidence; their disaggregation suggests that female offenders who experienced both interpersonal and noninterpersonal trauma (such as car accidents), and those who experienced both sexual and physical abuse, reported greater violent behavior (note that there were no differences in violence for some of their other, narrow categories such as “sexual assault only”). Ramoutar and Farrington (2006) found that neighborhood victimization was positively associated with frequency of violence (though the PoC were not statistically significant)-and it had a negative association with property offending.
A variety of definitions of trauma were used in these studies. For example, some studies used a broad definition of trauma. Byrd and Davis (2009) used the Trauma Assessment for Adults questionnaire which includes questions about 18 types of traumatic events, including physical and sexual abuse, natural disasters, and car accidents. Neller et al. (2006) used a traumatic events questionnaire including items about witnessing an accident, natural disaster, being a victim of a violent crime, experiencing physical or sexual child abuse or unwanted adult sexual experiences, and a few other items. Others focused more on human violence. Maschi et al. (2008) used a questionnaire consisting of items about physically abusive punishment, physical assault, sexual assault, and witnessing violence. A few used fairly narrow definitions of trauma, which could be seen as limited to victimization. Fagan et al. (1983) operationalized trauma based on whether the participant had ever been victim of serious property or violent offense. Others used witnessing extreme violence between parents (Lewis, Shanok, Pincux, & Glaser, 1979), level of exposure to violence in neighborhoods (Peacock et al., 2003), and neighborhood victimization (Ramoutar and Farrington, 2006). These narrower definitions are likely to miss participants who may have been traumatized in other ways, which is likely to reduce statistical power. However, even the use of trauma questionnaires may have problems capturing the extent of trauma due to a failure, in most cases, to determine whether or not the individual’s response to the events indicate that traumatization had occurred. Noting this, we believe the literature thus far leans slightly in favor of the differential etiology hypothesis, but tighter definitions should be used as this line of research evolves so that a clearer understanding can emerge.