Differences between Sexually Victimized and Nonsexually Victimized Male Adolescent Sexual Abusers: Developmental Antecedents and Behavioral Comparisons

Abstract

This study compares sexually victimized and nonsexually victimized male adolescent sexual abusers on a number of variables. Self-report measures were administered to 325 male sexually abusive youth (average age 16) in six residential facilities in the Midwest, 55% of whom reported sexual victimization. The results indicate that the sexually victimized sexual abusers have more severe developmental antecedents (trauma, family characteristics, early exposure to pornography and personality) and recent behavioral difficulties (characteristics of sexual aggression, sexual arousal, use of pornography, and nonsexual criminal behavior) than the nonsexually victimized group. Results are contrasted with recent typological research, which found no relationship between sexual victimization and subtype membership. Treatment, research, and theoretical implications are discussed.

Keywords:

• adolescent sexual offending
• sexual victimization
• personality
• sexual recidivism

Examples from the literature of the developmental pathways to sexual abuse as well as current efforts to develop typologies of adolescent sexual abusers highlight the need for additional research on the relationship between sexual victimization and offense characteristics. Moreover, research on the comparison between sexually victimized and nonsexually victimized male adolescent sexual abusers on a number of developmental and behavioral variables is needed. There is an emerging body of recent research investigating the possible correlates and trajectories of adolescent sexual offending. The etiological factors and the ecological contexts postulated to contribute to the emergence of juvenile sexually abusive behavior include examinations of (a) the developmental pathways to sexual aggression and the typological research (Butler & Seto, 2002; Hunter, 2006; Hunter, Figueredo, Malamuth, & Becker, 2003; Knight & Sims-Knight, 2004; Worling, 2001), (b) personality traits (e.g., Hunter, 2006; Hunter et al., 2003; Oxnam & Vess, 2008; Worling, 2001), (c) victimization experiences (e.g., Burton, 2000, 2003; Hunter, 2006; Veneziano, Veneziano, & Legrand, 2000), and (d) comparisons of the characteristics of juvenile sexual abusers with nonsexually offending delinquent youth. Many sexually abusive youth have experienced multiple forms of trauma. However, given that other types of trauma beside sexual victimization have not been found to be risk factors for recidivism and that the focus of this article is on sexual victimization, this article will review the relevant research on sexual trauma.

DEVELOPMENTAL PATHWAYS TO SEXUAL AGGRESSION AND TYPOLOGICAL RESEARCH

Recent examples of typologies of adolescent sexual abusers include Hunter's (2006) preliminary typology, which is based on Moffit's (1993) research on aggressive and delinquent youth. Within two primary offending types, adolescents who sexually abuse children and adolescents who offend against adults and peers, Hunter identified three subcategories similar to Moffit's: Life Course Persistent, Early Adolescent Onset (Paraphilic), and Adolescent Onset (Non-Paraphilic) offenders. In brief, life course persistent juveniles tend to be more antisocial and aggressive. Adolescents developing early onset paraphilic interests are those who are viewed as more likely to sexually recidivate posttreatment, especially on male children, and may pose a long-term recidivism risk. Adolescent Onset (Non-Paraphilic) offenders are considered the best treatment responders, provided they do not develop a negative trajectory of negative peer associations or substance abuse issues. Regarding the role of sexual victimization, Hunter et al. (2003) found that noncoercive sexual victimization by an older male non–family member was associated with later sexual interest in and offending of male children.

PERSONALITY OF ADOLESCENT SEXUAL ABUSERS

Worling (2001) (n = 112) categorized four subtypes of adolescent sexual abusers based on personality characteristics measured by the California Personality Inventory: Antisocial/Impulsive, Unusual/Isolated, Overcontrolled/Reserved, and Confident/Aggressive. Subtype membership in this study was not related to age or a history of sexual victimization. Oxnam and Vess (2008) also recently developed a typology of adolescent sexual abusers. They identified four subtypes using the Millon Adolescent Clinical Inventory: inadequate (internalizers with a dysthymic presentation), antisocial (unemotional and aggressive), conforming (anxious to please others), and passive-aggressive (are more immature and are more prosocial). Interestingly, comparisons of the abused and nonabused adolescents in the Oxnam and Vess study showed that those youth who reported sexual victimization were more likely to have at least one male victim. However, consistent with Worling's (2001) research, subtype membership was not correlated with a history of sexual abuse.

SEXUAL VICTIMIZATION OF ADOLESCENT SEXUAL ABUSERS

Adolescents who have been sexually victimized may have different developmental pathways to offending than nonabused sexual offenders (Cooper, Murphy, & Haynes, 1996). Etiological explanations of male adolescent sexual aggression frequently begin with a discussion of a victim-to-victimizer model (Freeman-Longo, 1986; Ryan, 1989) wherein exploration of potential intergenerational transmission of violence based on the youth's childhood sexual victimization is central. ¹

Recent research on the victim-to-victimizer model highlights that the severity of sexually victimized adolescent sexual abusers' own childhood victimization predicts the severity of their later sexual offenses (Burton, Miller, & Shill, 2002) and that sexually victimized adolescent abusers tend to repeat the same sexually abusive acts that they experienced as victims (Veneziano et al., 2000). Additionally, among the various commonly used risk assessment tools (e.g., the Juvenile Sex Offender Assessment Protocol-II, or J-SOAP-II; Prentky & Righthand, 2003), to date there are no empirically validated actuarial methods for assessing the likelihood that a sexually abusive youth will recidivate (Prescott & Levenson, 2007; Viljoen et al., 2008). There is, however, one actuarial measure in development that indicates that a history of sexual victimization may increase the risk of reoffense for male adolescent sexual abusers: the Juvenile Sexual Offense Recidivism Risk Assessment Tool (JSORRAT-II; Epperson, Ralston, Fowers, DeWitt, & Gore 2005).

The rates of childhood sexual victimization of adolescent abusers are quite varied across studies, ranging from less than 10% (Fagan & Wexler, 1988) to over 90% (Veneziano et al., 2000). Notably, these rates vary by study method. For example, studies of youth in residential facilities (Hunter et al., 2003) reveal higher rates of sexual victimization (75%) than studies of youth in community settings (39%; e.g., Murphy, DiLillo, Haynes, & Steere, 2001). Authors of a meta-analysis of more than 50 studies (N > 9,000 youth) reported that the average rate of sexual abuse may be as low as 40% (Burton & Schatz, 2003). Despite this body of research, few investigators have critically explored how nonsexually victimized sexual abusers are understood.

Hummel, Thömke, and Oldenbürger (2000) reported that sexually abusive youth with more severe sexual victimization experiences abused the most children and more often than not abused children of both sexes or only boys (either of which indicates a potentially higher risk of reoffense; see Worling & Curwen, 2001). In relationship to behavioral differences, Cooper et al. (1996) reported that sexually victimized adolescent sexual abusers were more likely to commit hands-on offenses, had a higher number of victims, began offending at a younger age, were less likely to have only female victims, were more likely to have both male and female victims, were less likely to abuse only non–family members, and had more conduct disordered behavior than nonsexually abused respondents. In this study, there were no reported difference between the sexually victimized and nonsexually victimized youth on family dysfunction or in self-reported delinquent behaviors, which differed from Hummel et al.'s (2000) results.

While sexual victimization histories for some adult survivors of childhood sexual assault have been found to be related to elevated personality scores (Elhai, Gold, Mateus, & Astaphan, 2001; Whiffen, Thompson, & Aube, 2000), the extant literature comparing sexually victimized and nonsexually victimized abusers has not yet assessed personality measure differences between these groups. But researchers have reported concerning scores (many psychological measures have a threshold above which a score should be considered “concerning” and may be indicative of problems in the area the scale assesses or of problems in related areas) for adolescent sexual abusers (Carpenter & Peed, 1995; Worling, 2001) and higher scores for sexually abusive youth than for nonsexually abusive delinquent youth (Burton, 2008). These two findings may be clarified via the comparison of personality for sexually victimized and nonsexually victimized abusers.

CURRENT STUDY AND HYPOTHESES

Based on this literature, in this exploratory project using a large sample of youth we examined developmental antecedents (e.g., trauma, family characteristics, early exposure to pornography, and personality) and behavioral differences (e.g., characteristics of sexual aggression, sexual arousal, use of pornography, and nonsexual criminal behavior) between victimized and nonvictimized incarcerated sexual abusers. In brief, based on the literature review we hypothesized that the sexually victimized youth would report more severe developmental antecedents and more severe behaviors than the nonsexually victimized youth.

METHODS

After State Department of Youth Services Institutional Review Board (IRB) approval, assents and consents were obtained and confidential data were collected from youth in six residential facilities in a Midwestern state, ² all of whom were adjudicated sexual offenders (they also had nonsexual offenses). The battery of self-report measures was administered in a small group format (groups of 8–12). Residential treatment programs prefer this approach for safety; in such facilities, safety is paramount and staff find it easier to monitor youth in groups. All youth were invited to participate, and they were sufficiently separated to ensure they could not view one another's responses. The youth were not provided with an incentive to complete the survey. For those few participants who struggled with reading, trained graduate student research assistants read the surveys aloud to these youth individually. Those youth who refused to participate (approximately 30%) were proportionately distributed across the six facilities. There was no way to compare data on those who declined versus those who consented to participate.

Multipaged pencil and paper surveys were collected from 325 adjudicated juvenile sexual abusers. Of these youth, 179 (55%) reported sexual victimization as children while 146 (45%) denied sexual victimization. While multiple measures of victimization experience were available, this group division was made through a simple question asking if the youth were sexually abused as children. Behaviorally descriptive measures may be best for such a determination; however, based on the assumption that not all abused boys will respond to this question affirmatively, this method was chosen as a conservative method of discrimination between the two groups.

The average age of the sample was 16.60 years (SD = 1.65 years) with no difference between groups, t (314) = .71, p = .472. The age range was 12 to 19. Similarly, there was no difference between the groups on current grade level, t (311) = .01, p = .995; youth averaged a 9th grade (SD = 1.64 grades) level. Racial composition also did not vary between the groups, with 47% of respondents selecting Caucasian, 26% selecting African American, 21% selecting Other, and 6% not selecting any option for race, χ² (3, N = 235) = 3.00, p = .308.

Measures

Measure and results are divided into two categories: developmental antecedents and criminal behaviors. Each set of measures is described in the following sections.

Developmental antecedents

The Childhood Trauma Questionnaire (CTQ) (Bernstein & Fink, 1998) is a 34-item scale that provides a brief and relatively noninvasive screening of traumatic experiences in childhood. All of the subscales have acceptable to good internal consistency in this study. Cronbach's alphas on the five CTQ subscales range from .76 (Physical Neglect) to .92 (Emotional Neglect; see Table 1).

TABLE 1 Cronbach's Alpha and Ranges on Subscales

Scale	Cronbach's alpha	Range victimized sexually	Range nonsexually victimized
CTQ Scales
Sexual Abuse	.83	5–30	1–25
Physical Abuse	.91	1–25	1–25
Emotional Abuse	.90	4–25	4–25
Physical Neglect	.76	7–40	3–38
Emotional Neglect	.92	6–45	3–45
MACI Scales ^†
Introversive	.77	1–113	1–104
Inhibited	.77	1–102	1–97
Doleful	.83	1–100	1–91
Submissive	.73	1–96	1–93
Dramatizing	.77	1–108	1–111
Egotistic	.75	1–81	1–94
Unruly	.68	1–106	1–107
Forceful*	.35	1–115	1–100
Conforming	.78	1–86	1–107
Oppositional	.86	1–115	1–93
Self-Demeaning	.72	1–100	1–95
Self-Reported Delinquency Scales (SRD) ^††
Alcohol Use	.81	0–12	0–12
Drug Use*	.46	0–18	0–18
Felony Assault	.65	0–12	0–12
Felony Theft	.88	.24	.24
General Delinquency	.68	0–36	0–24
Property Damage	.74	0–18	0–18
Public Disorderly*	.52	0–18	0–18
Robbery (1 item)	Not calculated
Selling Drugs	.84	0–12	0–12
^†Scales are ordered by the order of presentation in the instrument
*Not used in further analyses
^††Scales presented alphabetically

The Millon Adolescent Clinical Inventory (MACI) (Millon, 1993) was designed for youth in treatment or correctional facilities. It was normed on 579 adolescents in such facilities with two smaller cross-validation samples. The scales derived from the 160 true–false items are based on Millon's theory of personality (Millon & Davis, 1996). There are 12 personality pattern scales on the MACI, including those measuring Introversive, Inhibited, Doleful, Submissive, Dramatizing, Egotistic, Unruly, Forceful, Conforming, Oppositional, Borderline Tendency, and Self-Demeaning Tendencies. Data from 8 youth were eliminated from the study using Millon's validity scoring procedures. With the exception of the forceful scale (α = .35), which was dropped from further analyses, the remaining scales had acceptable internal consistency with Cronbach's alphas ranging from .68 for the Unruly scale to .86 for the Self- Demeaning scale (see Table 1).

To measure pornography exposure/use, a pencil and paper version of Leguizamo's (2000) interview was developed. These questions evaluate the frequency of childhood (before the age of 10) and adolescent exposure/use of several types of pornography including child pornography and violent pornography via various media including the Internet, television, printed material, etc. on a 48 item scale, with item responses as follows: 0 (never), 1 (1–5 times), 2 (6–25 times), 3 (26–50 times), 4 (51–100 times), 5 (101–500 times), and 6 (over 500 times). The overall pornography scale had respectable internal consistency in this study (α = .93). A subscale was created for pornography exposure before the age of 10 (α = .91).

The youth were surveyed about criminality in the family (3 dichotomous questions—e.g., whether parents sold drugs; α = .60), use of pornography before and after offenses (2 questions on a 5-point scale ranging from never to always), planning of offenses (5-point scale ranging from never to always), exposure to violence both in (3 dichotomous items—e.g., hitting, slapping, punching, etc.; α = .698) and out of the home (11 items on a 4-point scale ranging from never to many times; e.g., seen a stranger get shot, seen a relative or friend get beat up, etc.; α = .91), and a simple yes/no question regarding sexual victimization as a child. Additionally, a set of 17 sexual arousal questions on a 5-point scale ranging from not at all to a great deal (e.g., arousal on a 5-point scale to peers, children under the age of 12, rape, sadism, etc.; α = .71), age group (arousal to children, adolescents, or adults), and number of victims were also used in the study.

Criminal behavior

Elliot, Huizinga, and Ageton's (1985) Self-Reported Delinquency (SRD) measure was used to assess delinquency. The scale has 32 questions using a 7-point frequency scale from 0 (never) to 7 (2–3 times per day) on questions ranging from drug use to aggression. The instrument has several subscales including Alcohol Use, Drug Use, Felony Assault, Felony Theft, General Delinquency, Property Damage, Public Disorderly, Robbery and Selling Drugs. These subscales had acceptable interitem reliability (see Table 1) with the exception of Drug Use (α = .46) and Public Disorderly (α = .52), both of which were dropped from further analyses.

The Self-Report Sexual Aggression Scale (SERSAS) is a multi-item inventory used in prior studies (Burton, 2003; Burton et al., 2002). The scale measures sexually aggressive behaviors over the lifespan. Questions about rape, penetration, etc., are all prefaced with “Have you ever conned or forced someone to … ?” This instrument is essentially a checklist of relationships and acts, with an 8-week test-retest reliability of r = .96 for a small sample (Burton, 2000).

In addition, questions regarding sexual interest were asked using a Likert scale that followed the prompt “How much can you be excited by the following?” The scale ranged from 0 (not at all sexually excited) to 4 (a great deal). Questions included sexual interest in underage and peer-age male and females, rapes, frottage, sadism, masochism, etc. The youth were also asked how much time they spent planning their offenses, with the stem question “How much do you agree with the following the statement?” on a 5-point scale, ranging from never (0) to always (4).

RESULTS

PASW Statistics 17 (formally SPSS) was used for all analysis. In order to reduce the likelihood of Type I error inflation, multivariate tests (MANOVA) were conducted between the sexually and nonsexually abused groups on the five trauma variables (see Table 1) on the CTQ, the 11 scales on the MACI, and the 7 SRD scales. Additionally, we reported the results of the individual variable F tests from the MANOVA that were conducted in order to determine which of the individual CTQ, MACI, and SRD subscales were significantly different between the sexually abused and nonabused groups (see Tables 2, 3, and 4). Note that sample sizes in the tables were slightly reduced due to missing data and casewise deletion. We found no significant differences in age, race, or grade level between those who had complete data for each scale and those who did not.

TABLE 2 Results of MANOVA for the CTQ Subscales

Variables	Sexually victimized n = 143** Mean (SD)	Nonsexually victimized n = 177** Mean (SD)	df	F *	p (effect size)
CTQ			5,314	46.620	< 0.001(.426)
Sexual Abuse	15.91(6.49)	7.42(2.12)	1,318	225.35	< 0.001(.415)
Physical Abuse	13.25(6.30)	10.15(5.72)	1,318	21.32	< 0.001(.063)
Emotional Abuse	12.90(6.34)	9.93(5.58)	1,318	19.29	< 0.001(.057)
Emotional Neglect	20.10(9.21)	17.38(9.03)	1,318	7.04	.008(.022)
Physical Neglect	16.54(6.16)	14.28(5.46)	1,318	11.74	.001(.036)
* F-test from MANOVA
**Casewise deletions were used resulting in lower sample sizes for both groups

TABLE 3 Results of MANOVA for the MACI Subscales

Variables	Sexually victimized n = 165*** Mean (SD)	Nonsexually victimized n = 132*** Mean (SD)	df	F *	p (effect size)
MACI**			11,285	2.53	< 0.001(.089)
Introversive	58.97(27.33)	48.38(27.55)	1,295	10.94	.001(.036)
Inhibited	63.61(29.75)	52.10(30.28)	1,295	10.80	.001(.035)
Doleful	66.75(29.91)	53.73(31.97)	1,295	13.07	< 0.001(.042)
Submissive	55.16(17.38)	51.03(17.69)	1,295	4.07	.045(.014)
Dramatizing	48.79(18.74)	52.63(21.16)	1,295	2.74	.099(.009)
Egotistical	45.58(16.95)	47.77(17.89)	1,295	1.17	.281(.004)
Unruly	64.92(21.89)	63.36(24.89)	1,295	.33	.565(.001)
Conforming	46.01(13.86)	46.49(16.44)	1,295	.07	.782(<.001)
Oppositional	64.88(25.58)	59.24(26.55)	1,295	3.45	.064(.012)
BorderlineTendency	54.42(31.58)	47.33(32.18)	1,295	3.64	.058(.012)
Self-DemeaningTendency	63.84(29.68)	51.00(31.19)	1,295	13.12	< 0.001(.043)
* F-test from MANOVA
**The forceful subscale was dropped from the analysis due to low internal consistency
***Casewise deletions were used resulting in lower sample sizes for both groups

TABLE 4 Results of MANOVA for the Self-Reported Delinquency (SRD) Subscales

Variables	Sexually victimizedn = 143***Mean (SD)	Nonsexually victimizedn = 120***Mean (SD)	df	F *	p (effect size)
SRD			7, 255	2.21	.034(.057)
Alcohol Use	3.52(4.04)	2.96(3.55)	1, 261	1.39	.239(.005)
Felony Assault	2.39(3.51)	1.49(2.50)	1, 261	5.54	.019(.021)
Felony Theft	5.88(7.17)	3.96(5.74)	1, 261	5.53	.019(.021)
GeneralDelinquency	7.94(7.76)	5.51(5.87)	1, 261	7.94	.005(.030)
PropertyDamage	3.69(4.51)	2.22(3.33)	1, 261	8.74	.003(.032)
Robbery	.85(1.70)	.73(1.54)	1, 261	.31	.576(.001)
Selling Drugs	2.61(4.18)	2.54(4.07)	1, 261	.02	.896(<.001)
* F-test from MANOVA
**The Drug Use and Public Disorderly subscales were dropped due to low internal consistency
***Casewise deletions were used resulting in lower sample sizes for both groups

The rationale for using MANOVA is that the measures making up the CTQ, MACI, and SRD should be correlated with one another, so the analysis should compare the measures as a whole between the abused and nonabused groups. Since the measures as a whole were significant, it is usually the case that one would investigate whether all individual items were significant or just some of them. The other measures that were analyzed by t-tests were “stand alone” measures, not part of a sequence of related measurements, or were constructed scales (summative) from other items.

Developmental Antecedents

Measured by a cumulative variable of family criminality, sexually victimized juvenile sexual abusers were more likely to have witnessed criminality (i.e., drug sales, nonsexual violence in the home, and illegal acts by family members), t (319) = 3.28, p = .001; experienced (in a cumulative variable) more exposure to violence at home (i.e., punching, slapping, hitting at home and physical abuse of children at home), t (320) = 5.24, p < .001; and experienced more exposure to violence in the community (i.e., someone threatened to stab, shoot, or kill the youth; the youth has been beaten up; youth has seen a stranger or relative, shot, beaten, stabbed, or killed), t (322) = 2.86, p = .004. Sexually victimized sexually abusive youth also had significantly more exposure to pornography before age 10 (i.e., adults having sex with other adults or with children and adults forcing sex with other adults or with children, in pictures, movies, or on the Internet), t (313) = 4.22, p < .001.

The results of the overall MANOVA tests indicated that all five CTQ variables of the sexually abused group were significantly different than the nonabused group, F (5,314) = 46.62, p < 0.001. Moreover, using the CTQ scales sexually victimized juvenile sexual abusers had significantly more experiences of childhood victimization, with higher scores on all five measures of the CTQ (see Table 2 effect size = .426). While some youth answered “no” to the question about being sexually abused, they responded “rarely” or “sometimes” on the CTQ to questions related to sexual abuse (e.g., “I believe I was sexually abused”). If some of the youth in the nonvictimized group were actually sexually abused as children and the group differences are statistically significant, the actual differences will be larger than reported.

Sexually victimized juvenile sexual abusers also had higher scores on several scales of the MACI, including the Introversive, Inhibited, Doleful, Submissive, Unruly, Oppositional, and Borderline Tendency scales, F (11,285) = 2.53, p < 0.001 (effect size = .089; see Table 3).

Behavior

Sexually victimized juvenile sexual abusers were younger when they started sexually abusing others (M = 11.75 years, SD = 2.87 years versus M = 13.35 years, SD = 3.18 years; t [286] = 4.47, p < .001) and thereby had a greater number of years of abusing others (M = 2.17, SD = 2.68 vs. M = .95, SD = 1.78; t [282] = 19.89, p < .001). Sexually victimized juvenile sexual abusers were also more likely to commit incest, defined as any male or female relative (49% versus 35% in the nonsexually victimized group), χ² (1, n = 325) = 6.15, p = .013.

Sexually victimized juvenile sexual abusers had a higher number of reported victims (M = 3.97, SD = 5.54 vs. M = 1.41, SD = 1.39; t [307] = 28.94, p < .001), were more likely to abuse males (70.1% versus 49.3%; χ² [1, n = 325] = 11.88, p = .001), and had more severe offenses on a 7-point scale derived from the SERSAS (1 = exposure or voyeurism, 2 = fondling, 3 = exposure or voyeurism and fondling, 4 = penetration, 5 = penetration and exposure or voyeurism, 6 = penetration and fondling, 7 = all three acts), t (273) = 4.48, p < .001.

Sexually victimized juvenile sexual abusers reported significantly higher sexual arousal to a number of stimuli, including males under 12, males between the ages of 13–18, masturbation in public, voyeurism, rape, frottage, bestiality, and exhibitionism, t (323) = 3.53, p < .001. They were also more likely to look at pornography before (t [319] = 2.30, p < .001) and after their sexual offenses (t [321] = 10.93, p = .037) and to spend more time planning their sexual offenses, t (319) = 1.67, p = .001.

The results of the MANOVA indicated that sexually victimized juvenile sexual abusers had significantly higher scores on almost all of the scales of the SRD measure, including Alcohol Use, General Delinquency, Property Damage, Felony Theft, and Felony Assault, F (7,255) = 2.21, p = 0.034; effect size = .057 (see Table 4). This is true for violent and nonviolent crimes, with the summative scores of t (302) = 1.95, p < .05 for nonviolent crimes and t (297) = 2.45, p = .015 for violent crimes.

DISCUSSION

Based on a comparison of youth who were divided into victimized sexual abusers and nonvictimized sexual abusers (55% were sexually abused as children), as hypothesized the victimized group had more traumatic childhoods, had elevated scores on personality measures, lived in more criminogenic environments, and exhibited more antisocial behaviors, which started younger and lasted longer. In fact, on all the self-reported instruments the victimized group indicated greater developmental and behavioral challenges compared with the nonvictimized group.

While this study is not focused on personality measures per se, it is clear that the victimized youth had higher scores on several of the personality scales of the MACI. This finding is especially interesting in light of the fact that previous research on adolescent sexual abusers based on personality characteristics (Oxnam & Vess, 2008; Worling, 2001) could not classify subtype memberships as a result of sexual victimization. While the findings regarding personality score differences are a new addition to the literature, the remaining results concur with previous research conducted on this population (Cooper et al., 1996; Craissati, Falla, McClurg, & Beech, 2002) with one exception: Cooper et al. (1996) did not find differences in penetrative offenses between the groups. However, this may be explained by the use of a continuous measure of severity in the current study as opposed to use of a dichotomous variable in the prior research.

Results such as these pose additional questions for researchers and clinicians to examine in order to better understand and treat these youth. How can these findings and observations of the literature be interpreted from a meaningful theoretical perspective? Given this list of differences and recent research on adolescent sexual abusers using Moffit's framework with adolescent sexual offenders (Hunter, 2006), it seems conceivable that these two groups of delinquent youth tentatively may be categorized by Moffitt's (1993) well-researched typology of offenders across the lifespan (D. Finkelhor, personal communication, July 7, 2004).

While analyzing data from youth in longitudinal studies (Moffitt, Caspi, & Harrington, 2002), Moffitt hypothesized and validated that another group of the youth (Moffitt, 1993; Moffitt et al., 2002) were “life-course-persistent.” This group begins their criminality at a younger age and have different developmental antecedents, which in culmination create a more challenging picture: their families have poor mental health and are of low socioeconomic status, the youth tend to have challenging neurological developments (Moffitt & Caspi, 2001), and they are typically raised in a criminogenic environment. Although unconfirmed (see Hunter, 2006), Moffit's work has interesting implications for the present study and fits the research on the typologies of adolescent sexual abusers discussed earlier in this article. Life-course-persistent and early-onset adolescent sexual abusers are likely youth whose early core relational experiences were marked by abuse and neglect and whose overall neuropsychological health was impacted by problematic and abusive early caregiving experiences. Perhaps nonsexually victimized abusers are better viewed as adolescent limited, delinquent youth who commit sexual offenses.

Implications

Research

Further research on multiple etiologic or multiple pathway analyses (Knight & Sims-Knight, 2004) using large samples are needed to analyze heterogeneity among sexual abusers on various characteristics, including sexual victimization history and severity. Research projects investigating differences in risk of sexual reoffense and differences in treatment needs and family therapy models are all needed. Longitudinal studies following sexually abusive youth from the inception of their sexually aggressive behavior into adulthood might work toward understanding the course of these behaviors but are very difficult to conduct. Research on desistance, resilience, and protective factors are needed as well. This includes studies that examine the protective factors related to desistance of not only sexual aggression but also the more frequently seen nonsexual criminal recidivism of adolescent sexual abusers. Nonvictimized abusers should also be compared to nonsexually aggressive delinquent youth to see if these two groups are similar in antecedents and in criminal paths; they may be the same group of youth with similar treatment needs. Additionally, Moffitt's typology now includes neurobiological and genetic implications, much of which should be researched with adolescent abusers. Finally, and as implied previously, typological research that describes criminal behavioral paths combined with antecedents may help guide treatment with this population.

Treatment

These results are from a relatively large group of youth and imply that childhood sexual trauma may distinguish abusive youth's antecedents (e.g., arousal patterns), treatment responsivity (e.g., personality scores), and severity and type of their sexual aggression (e.g., victim type). This research illustrating the relationship between the severity of sexual victimization and the severity of sexual aggression has important implications for treatment. For example, the incorporation of a trauma history may be a useful contribution to placement decisions and may help guide treatment types for these diverse groups of youth. For example, some victimized youth may be at higher risk to reoffend and therefore require trauma-informed treatment (Greenwald, 2009; Williams, Smith, An, & Hall, 2008).

These results indicate that victimized youth may have a greater need for risk assessment for nonsexual criminality, although given that nonsexual criminal recidivism rates for adolescent sexual abusers tend to be 3–5 times greater than sexual recidivism rates (Worling & Curwen, 2000), this should be occurring in all programs already. Treatment dosage, intensity, and psychoeducational topics for this group may also need to be different, with more treatment of diverse types related to trauma.

Limitations

This is a simple cross-sectional study that raises interesting questions. Data in this study are based on self-report, which has both weaknesses and strengths, but they are from only a single source. The question used to divide the groups (inquiring as to whether or not the youth were sexually abused) may not be the most reliable. Nevertheless, since all youth were involved in trauma therapy for their victimization and for their sexually aggressive behavior, the question should have good psychometric performance, given their familiarity with the construct (sexual abuse). Additionally, as stated previously, this division was considered conservative, such that if differences between the groups existed, and even if some of the sexually victimized youth were placed in the nonsexually victimized group, actual differences between these two groups may be larger than those reported.

Finally, while the differences between the two groups in this study are quite clearly unidirectional and meaningful, a history of childhood sexual victimization may be only one of a plethora of events creating these differences. A related point is that even if we assume that childhood trauma is causing these behavioral challenges, the mechanisms of behavioral production cannot in any way be assumed. For example, in families in which violence and criminality occur and are rewarded, youth may be led to be more violent and more criminal. Alternately, certain delinquent acts among youth may be in response to posttraumatic stress and traumatic experiences. Clearly, tests of these mechanisms and further theoretical developments are needed. As discussed, MANOVA was used to reduce Type I error for analyses of the instruments with correlated subscales. Although there were many analyses, the probability values were almost always .01, therefore further correction for type I error was not taken. However, as in all studies, type I error should be carefully considered in further replication of this research.

Notes

1. Nearly all victim-to-victimizer research is retrospective (investigates the childhood sexual victimization of adolescent sexual offenders rather than the prospective study of outcome). This limitation should be carefully considered reading this research.

2. For the sake of potential meta-analysis, this is the first paper using this data.