Tough Guys and Tough Guise: Subtypes of Aggression Among Youth Sexual Abusers

Purpose: There is a great deal of overlap between youth sexual abusers and general delinquents, as youth sexual abusers have high levels of non-sexual criminality. However, empirically supported distinctions exist between these groups, as well as among youth sexual abuser subtypes (Seto & Lalumiere, 2010). The role of nonsexual aggression, which plays a significant role in subtypes of adult sexual offending (Roberts, Doren & Thornton, 2002) is not yet well understood among youthful sexual abusers. In this study, we explore levels of non-sexual delinquency among these youth using subtypes of aggression.

 Methods: Data for these analyses were from a sample of youths (N= 196) incarcerated for a sexual offense in a Midwestern state.  On average, participants were 17 (SD= 2.22) and in the 10^th grade (SD= 1.65 grades).  To estimate overall aggression and its subtypes, The Aggression Questionnaire (AQ) was administered (Warren & Buss, 2000).  An aggression score was computed, along with scores for Physical, Verbal, and Indirect Aggression, as well as Anger and Hostility. Delinquent behaviors were assessed using Elliot, Huizinga and Ageton’s (1985) Self-Reported Delinquency (SRD) scale, which measured a range of delinquent behaviors. Sexual abuse severity scales were included as covariates in the model, with one measure for perpetration and a one for victimization.   

Due to a small amount of arbitrary missing data, multiple imputation (mi) was performed.  Using Stata 11.0, missing values were estimated using Markov Chain Monte Carlo, in which multivariate normal distributions of all variables were used to create linear regression models (Little and Rubin, 2002). Stationary distribution was confirmed by graphing worst linear functions across all iterations and autocorrelation plots were graphed to ensure that enough iterations were left between successive dataset draws. Post-hoc sensitivity analysis showed no meaningful differences with listwise deletion. Regression coefficients were tested with F-tests and violations of regression assumptions were checked using leverage-residual and Cook’s distance plots.

 Results: Controlling for race, age, grade in school and sexual victim and perpetration histories, linear regressions showed that each aggression subscale and the total score are separately associated with delinquency with high significance (p < .001).  However, when all subscales were included in the model together, only physical aggression showed a significant relationship with delinquency (p < .001).  Being African-American was also associated with higher levels of delinquency in many of the models, although F-tests could not confirm this relationship.     

 Implications: This is an important contribution to the field of youth sexual abusers, as aggression permeates multiple domains and is more easily observed, whereas sexual abusing is not. Past studies of aggression among youth sexual abusers have yielded mix results. This is the first study to consider the impact of aggression subtypes on delinquency in this population. Due to the high levels of general delinquency among youth sexual abusers, this finding has the potential for strong clinical implications as well as prevention efforts.