In studies where measures of intellectual function are taken before measures of violent behavior, associations have been very consistent (e.g., Barker et al., 2007; Barker et al., 2011; Beaver et al., 2010; Bellair & McNulty, 2005, 2010; Campbell et al., 2010; Farrington, 1989; H0gh & Wolf, 1983; Loeber et al., 2012; Nagin & Tremblay, 2001; NICHD, 2004). There have been some exceptions, but we believe that the models used were not ideal for the purposes of this review. Brownlie et al. (2004) report a prospective association between age 12 language impairment and young adult parent-rated aggression; verbal IQ was in the same model and was not statistically significant (Brownlie et al., 2004). It could be the case that including both measures in the model at the same time led to a type II error (see our note about model “overspecification4”). Piquero (2000) reported that frequent nonviolent offenders were not different than violent offenders on earlier measured
IQ tests in the Philadelphia Collaborative Perinatal Project. Piquero reports very few significant relationships in this paper, and we see his model specification as quite conservative. Salekin et al. followed up a sample of adolescent offenders over a 3-year period and found that intelligence measures were not associated with violent recidivism (Salekin, Lee, Dillard, & Kubak, 2010). Predicting recidivism in violence among violent offenders is different than predicting any violence.