The Nature of the Violence—Antisociality Relationship
A recurring theme in this volume will be whether or not violence is just one manifestation of a larger deviant, criminal omnibus trait or whether it is distinct. One possibility, that which is implied by the body of literature in our field, is that the causes of both violent and nonviolent delinquency are the same. There are a variety of patterns in the data that would suggest otherwise. Violence could be on the far end of a continuum in deviant or delinquent activity. If so, we might find a dose-response relationship where a causative factor, such as low socioeconomic position, might cause minor problems at low levels and violence at high levels. In this case, the same factors cause all forms of delinquency, but they are likely to cause violence at high doses or, perhaps, if they are experienced over long time periods. This is approximately the conclusion that Farrington (1991) reached in his attempt to address the question of “whether aggressive or violent behavior is merely one element of a more general antisocial tendency, or whether it reflects a more specific underlying violent tendency” (pp. 7-8). He concluded that the causes of violence are essentially the same as the causes of persistent, extreme delinquency, and criminal behavior. Unfortunately, there are very few studies that look at dose-response directly.
The cleanest type of evidence would be that violence is significantly associated with a risk factor when levels of nonviolent offending have been accounted for; this type of model directly addresses our research question. In some studies, a control variable for frequency of nonviolent offending might be employed; in others, direct comparisons between violent and nonviolent offenders would effectively make the correction.
The most obvious evidence for the differential etiology of violence would be if a risk factor is significantly associated with violent crime but not property crime in studies that disaggregate by dependent variable. Felson and Lane (2009) use this method of “discriminant prediction” to test whether general theories are adequate. It won’t be that simple here because we will be reviewing a complete set of studies. Thus, if we find that the consistency of significant findings predicting violent offending is greater than the consistency for nonviolent offending we will take that as evidence of a differential association with violence. This type of comparison is the most available, though we acknowledge that it takes some large differences and judgment calls to draw conclusions. This does not exhaust all possibilities but it does anticipate the type of studies we will see in the coming chapters.