You are on page 1of 15

Child Abuse & Neglect 26 (2002) 893–907

A social learning theory comparison of the sexual


victimization of adolescent sexual offenders and
nonsexual offending male delinquents夽
David L. Burton∗ , Diane Lynn Miller, Chien Tai Shill
School of Social Work, The University of Michigan, 3734 Social Work Building,
Ann Arbor, MI 48109-1106, USA
Received 18 December 2000; received in revised form 8 October 2001; accepted 11 December 2001

Abstract
Objective: This paper tests hypotheses based on Garland and Dougher’s (1990) formulation of the
“victim to victimizer” relationship which is a social learning theory etiological approach to adolescent
sexual offending.
Method: Two hundred sixteen adolescent sexually victimized sexual offenders and 93 adolescent sexu-
ally victimized nonsexual offending delinquents from three Michigan sexual offender treatment facilities
participated in an anonymous cross-sectional survey regarding their sexual victimization and offending
as well as a social desirability measure.
Results: When comparing the two groups, nonsexually offending delinquents and sexually offending
delinquents, the latter were found to have: closer relationships with their perpetrators; a higher chance of
having a male perpetrator(s); a longer duration of sexual victimization; more forceful sexual victimiza-
tion; and the experience of penetration as part of their sexual victimization. Logistic regression showed
that the gender of the perpetrators being both male and female and the forcefulness of the perpetrators
were the best predictors of being in the sexual offender group.
Conclusions: The social learning theory hypotheses were generally supported. Further multivariate
research on this theory is warranted.
© 2002 Published by Elsevier Science Ltd.

Keywords: Sexual abuse; Intergenerational transmission; Female offenders


This project was funded by the Office of the Vice President of Research of the University of Michigan and
supported by W. J. Maxey Boy’s Training School (Whitmore Lake Michigan), Boysville (Clinton and Saginaw,
Michigan), and Lutheran Home (Bay City, Michigan).

Corresponding author.

0145-2134/02/$ – see front matter © 2002 Published by Elsevier Science Ltd.


PII: S 0 1 4 5 - 2 1 3 4 ( 0 2 ) 0 0 3 6 0 - 5
894 D.L. Burton et al. / Child Abuse & Neglect 26 (2002) 893–907

Introduction

Adolescent sexual offenders are responsible for at least 20% of reported forcible rapes
and child molestations in the United States (Snyder & Sickmund, 1999). This paper explores
whether childhood sexual victimization can be a useful etiological variable in predicting ado-
lescent sexual offending behavior by testing several hypotheses that predict that the nature, or
the characteristics, of sexual victimization will be different and conceptually more severe for
sexually victimized male adolescent sexual offenders than for a comparison group of sexually
victimized nonsexually offending delinquents.
Analyses indicate that the number of adult male sexual offenders who have been sexually
victimized is smaller than generally believed. Hanson and Slater (1988) reported that just over
1/4 (28.2%) of 1700 adult male sexual offenders in 25 studies admitted sexual victimization as
children. However, Hanson and Slater recognize several methodological problems which make
it difficult to determine the number of adult male sexual offenders who have been sexually
victimized, including: (a) different definitions of sexual victimization were used in the various
studies they reviewed; (b) the possibility of under-reporting in these studies, since males are
typically reluctant to reveal sexual victimization; (c) if sexually victimized by males, the
possibility of even further underreporting since the victim may fear being homosexual and/or
being considered homosexual by others and; (d) various methods of data collection used across
the studies reviewed (e.g., interview or survey) which may have led to different reported
results. Similar rates of childhood sexual victimization and a similar review of methodological
problems were reported by the United States General Accounting Office (1996).
While these methodological problems often apply to research on male adolescent sex-
ual offenders, the rate of reported sexual victimization for this group is typically reported
to be 40% (Cooper, Murphy, & Haynes, 1996; Hunter & Figueredo, 2000; Ryan, Miyoshi,
Metzner, Krugman, & Fryer, 1996; Worling, 1995), about 12% higher than the rate reported
for adult male sexual offenders (Hanson & Slater, 1988), and four times higher than the 10%
sexual victimization rates reported for the general male population (Watkins & Bentovim,
1992). For example, in the most nationally representative study in the United States, using
data from 90 agencies, Ryan et al. (1996) report that during the intake process, 39% of 1000
male adolescent sexual offenders reported sexual victimization. Similarly, in an analysis us-
ing eight pre-post design studies, Worling (1995) found that at intake 31% of 1268 youths
reported sexual victimization, and 52% of the same youths disclosed sexual victimization by
the end of treatment. Thus, reported rates of sexual victimization may differ between intake
and completion of treatment. This may be due to increased trust in staff or to learning more
accurate definitions of sexually abusive behavior during treatment. Increased rates may also
serve instrumental purposes such as attempting to get sympathy from judges and/or staff.
Using a social learning theory, Ryan (1989), suggests that sexually abusive behavior by
male adolescents stems from their own sexual victimization and describes the mechanisms by
which the learning of sexual aggression is thought to occur and to be reinforced. She suggests
that a “traumatized child may become fixated on the trauma, recreating the experience in
ritualistic patterns that become more elaborate, more rigid and more secret over time” (p. 10).
An alternative explanation for why victims of sexual abuse may then sexually victimize others
is physiological arousal (Ryan, 1989; Worling, 1995). Such arousal can lead to subsequent
D.L. Burton et al. / Child Abuse & Neglect 26 (2002) 893–907 895

sexual fantasies and masturbation, resulting in conditioning to sexual acting out in particular
ways and/or on particular types of victims. This proposition has not been empirically tested.
One of the best examples of research on a more general victim to victimizer hypothesis is
Widom’s (1989) longitudinal study of 908 cases, in which she assessed sexual victimization
as a risk factor for further delinquency. She reports that “. . . abused and neglected children
have significantly greater risk of becoming delinquents, criminals and violent criminals” than
nonabused children (p. 246). Another test of the victim to victimizer theory is an application
of social learning theory to 287 sexually aggressive children by Burton, Nesmith, and Badten
(1997). They report significant correlations between the presence of sexual victimization in
the children’s lives and the number of people they sexually abuse.
On other side of the issue, Skuse et al. (1998) report that in a comparison of two groups of
young (11–15 years of age) sexually victimized adolescents (11 who sexually abused others
and 14 who did not), characteristics of their own sexual victimization (e.g., force, duration,
evidence of penetration, number of perpetrators) did not differentiate the two groups in bivariate
or multivariate logistic analyses. They did find that exposure to family violence was a predictor
of being in the sexual abusive group. This study included many other interesting ideas including
the use of Tanner developmental stage, intelligence quotient and network relationships, but is
limited due to it’s small sample size.
Sexual victimization of adolescent sexual offenders also is frequently assumed and tar-
geted in treatment settings across the nation. The 2000 Safer Society Survey (Burton &
Smith-Darden, 2001), polled 209 programs working with male adolescents sexual offenders in
both community and residential settings. The survey results indicate that 89% of community
based treatment programs work with a youth around their own sexual victimization and trauma
of other sorts. Among residentially based programs, 82.4% of programs endorsed this tech-
nique, second only to preassault and assault cycle techniques which help a youth recognize the
cognitive, emotional and behavioral precursors and consequences of their sexually aggressive
behavior. Although empirical support for addressing sexual victimization issues is limited, it
is the basis of clinical practice with many male adolescent sexual offenders.
Although sexual offending for adolescents has been linked theoretically, empirically, and
clinically to sexual victimization, it should be noted that only a small number of sexually
victimized youths become adolescent sexual offenders, and that not all adolescent male sexual
offenders are sexually victimized. Sexual victimization should not be seen as a necessary nor
sufficient causal variable for the development of adolescent sexual offending.
Nonetheless, for sexually offending youths with sexual victimization histories, studies have
investigated the relationship between sexual victimization and the characteristics of their per-
petration. These studies (Cooper et al., 1996; Kaufman, Hilliker, & Daleiden, 1996; Worling,
1995) support the proposition that sexual victimization is related to more varied or more se-
vere sexual offending behavior when compared to the sexual offending behavior of youths
who have not been sexually victimized.
For instance, in a study of 87 male adolescent sexual offenders, Worling (1995) investi-
gated the relationship between the presence of sexual victimization and the gender of their
child victims. He reports that 75% of the male youths that sexually assaulted a male child
reported sexual victimization compared to 25% of those youths that assaulted a female child.
Unfortunately, the study does not report characteristics of the sexual victimization such as, the
896 D.L. Burton et al. / Child Abuse & Neglect 26 (2002) 893–907

relationship of the perpetrator to the victim, the gender of the victimizer, or the duration of the
victimization.
Kaufman et al. (1996) focused on developing a greater understanding of the modus operandi
of adolescent sexual offenders (n = 179). Within this larger study, they analyzed differences
in modus operandi between sexually victimized and nonsexually victimized adolescent sexual
offenders. The χ 2 analyses indicated that sexually victimized adolescent sexual offenders were
likely to pick a younger victim, and a male victim (or male and female victims). However,
there was no difference between the groups regarding whether the offender was related to his
victim. The two significant differences with respect to age and gender of the victim were no
longer clear in a MANOVA, using the 11 scales in their Modus Operandi Questionnaire. While
an important study on modus operandi, this study did not investigate sexual offending in the
context of prior sexual victimization.
Cooper et al. (1996) did address prior sexual victimization in their comparison of sexually
victimized and nonsexually victimized adolescent sexual offenders. Their results indicated a
relationship between prior sexual victimization and characteristics of the sexual offense. In
this study data on other forms of trauma (e.g., physical victimization) were also collected.
They reported that:
Subjects who had experienced abuse [sexual and/or physical] began their offending at a younger
age, were more likely to abuse both males and females and less likely to abuse females only,
and had more victims. (p. 115)

An unexplored line of inquiry is the relationship between the characteristics of sexual


victimization and characteristics of sexual offending. Garland and Dougher (1990) provide a
conceptual framework leading to research hypotheses in this area. Their work, based on an
extensive review of the related research at that time, speculates on several characteristics of
sexual victimization that might contribute to a youth becoming an adolescent sexual offender.
The variables they suggest, and upon which we expand, include the following:
• The relationship of the perpetrator(s) to the youth: if the perpetrator(s) are relationally
closer to the youth, he may be more likely to attend to, learn from, and later model that
person’s victimizing behaviors. Faller (1989) empirically supports the presence of this
relationship.
• The gender of the perpetrator(s): being victimized by a male may lead to a greater
likelihood of becoming a sexual offender due to the tendency of males to use greater
force and thus reinforce learning. In addition, due to the social responses to male–male
sex in general, and sexual aggression in particular, it may make the victimization more
shameful, making it more difficult for the youth to resolve the trauma in ways other than
repeating it.
• The modus operandi of the perpetrator(s): with greater force, learning is more likely
to take place; youths violated in this way are therefore more likely to pay attention
to the situation and to learn sexually offending behavior from it. Force may also be
related to increased fear and powerlessness during their sexual victimization (Burton &
Fleming, 2000), leading to a need for the victim to act out in order to gain a sense of
control.
D.L. Burton et al. / Child Abuse & Neglect 26 (2002) 893–907 897

• The type of acts the perpetrator(s) commit(s): penetrative acts typically result in more
emotional and physical trauma to the youth, and thus reinforce learning of this behavior.
Youths who have been penetrated also feel more helpless (Burton & Fleming, 2000) and
may feel a greater need to resolve their trauma through the method modeled for them by
their perpetrators.
• The duration of the sexual victimization: the longer the victimization (Bandura, 1986),
the greater the likelihood of sexual offending, perhaps due to the presence of greater
opportunities to learn the behavior and to view its rewards (e.g., reduction of anxiety of
the perpetrator, a perpetrator’s orgasm) (Worling, 1995). Those who have been victimized
longer may gain more knowledge about how to victimize, and perceive greater rewards
for future sexual offending.
• The age at which the sexual victimization occurred: younger age of the child victim
might be related to a greater likelihood of later sexual offending, since sexual aggression
becomes normative at a younger age. Thus, early learning of a norm might increase the
likelihood of sexual offending (Burton et al., 1997).
This study, using an expanded version of Garland and Dougher’s (1990) framework, inves-
tigates the nature of the sexual victimization of a group of adolescent sexual offenders and
compares them to a group of nonsexually offending delinquents in a number of institutional
settings. A number of hypotheses are tested. Compared to sexually victimized nonsexually
offending delinquents, we predict that sexually victimized sexually offending delinquents will:
(a) be more closely related to their perpetrators; (b) be more likely to have been sexually vic-
timized by a male perpetrator; (c) have been subjected to more forceful sexual victimization;
(d) be more likely to have experienced penetrative acts during their sexual victimization; (e)
have a longer duration of sexual victimization; and (f) be younger at the time of their sexual
victimization.

Method

Participants and method

Four hundred seventy-one male delinquent youths in three Michigan treatment facilities
participated in an anonymous cross-sectional survey. Among these were 272 youths either in
sexual offender sites within the agency or were self-reported sexual offenders; 67% (n = 182)
of these youths admitted to penetrative acts of sexual offending as their most severe act, 9%
(n = 24) admitted to fondling as their most severe act, 21% (n = 57) admitted to noncontact
acts as their most severe act (e.g., exhibitionism), and 3% (n = 9) denied committing sexual
offenses. As there were no significant differences between the sexual offenders in the three
agencies on any of the variables described in this paper, the four groups were combined,
for a total of 272 youths in the sexual offender group. Similarly, there were no significant
differences across the three agencies for the remaining 199 youths who denied any sexual
crimes, all of whom were not in sexual offender treatment sites within the facilities They
were convicted of a wide range of nonsexual acts including property damage, drug sales,
898 D.L. Burton et al. / Child Abuse & Neglect 26 (2002) 893–907

aggravated assault and murder. Therefore, by location of service provision (determined by


each agency by legal charge) and by youths’ self-admission or denial of sexually abusive
crimes, two groups were selected for further analysis, 272 sexual offenders and 199 nonsexual
offenders.
There was no significant difference in age between the sexual offenders and nonsexual
offenders. The average age for the all the youths was 16.9 years (SD = 1.47) with a range
from 12 to 21. On average, they reported being in the late 9th grade. Of the youths, 51.1%
were African-American, 29% were Euro-American, and 19.9% belonged to other groups.

Recruitment and selection

In each institution, the administration, clinical leadership, clinical team and on-line staff
were consulted and asked for approval for each boy’s participation. Consent was also obtained
from each participating youth in order to participate. In two of the institutions, familial per-
mission was also required from the small number of children who were not state wards. Across
all three institutions, only 14 youths were not given approval to participate due to extremely
difficult behavior management or clinical concerns (i.e., asking the youth questions about
trauma might upset current clinical work). While all youths in each institution were invited,
102 declined to participate. There was no way to determine differences between those who
were in the study, and those that were not allowed, or did not wish, to participate. An on-site
staff liaison facilitated the process of group administration of the surveys. A staff member was
available for any youth feeling the need to process the material in the survey (n = 3). The
youths were given a pizza party as compensation for their participation.

Materials

After the University of Michigan’s Institutional Review Board and the Attorney General of
Michigan’s approval and the completion of several pilot tests, demographic data were collected
along with an instrument based upon the Sexual Abuse Exposure Questionnaire (SAEQ) (Ryan,
Rodriguez, Anderson, & Foy, 1992). This instrument was used to assess the type(s) of sexual
victimization experienced and other dimensions of victimization, including relationship and
gender of the youth’s victimizer, duration of the youth’s sexual victimization, modus operandi
of the youth’s perpetrator, age of the youth at onset of sexual victimization, and severity of the
sexual victimization. Dichotomous questions were asked about each of 18 possible sexually
victimizing acts (e.g., having some one show you their sexual parts); if a youth answered yes to
one of the acts, he was then asked which of 26 possible relationships (and the genders of each)
of people he was victimized by in that way. He was also asked his age when the victimization
behavior began and stopped as well as the degree of force used when he was victimized or
the modus operandi (method of operation) of his offender(s). To assess the modus operandi
the youths were asked about the degree of force used during the abuse. Degree of force was
measured by three levels, the first level was coercion by the youth’s perpetrator(s) by means
of playing games with the youth, doing favors for the youth or baby-sitting the youth. The
second level was threats to the youth and/or to the youth’s family while the third and most
serious level was the use of physical force during sexual victimization.
D.L. Burton et al. / Child Abuse & Neglect 26 (2002) 893–907 899

Severity of the act was determined by the nature of the acts themselves, which were collapsed
into three levels of severity—exhibitionism, fondling or penetration. Reliability testing for
this entire instrument yielded an alpha of .86, with an 8 week test–retest agreement of 79%
(Burton & Fleming, 1998) for a subsample.
Finally, social desirability was assessed using a measure designed for adult sexual offenders,
the Balanced Inventory of Desirable Responding (BIDR) (Kroner & Weekes, 1996). This
instrument uses 42 questions with a 7-point Likert scale with responses from “not true” (1)
to “very true” (7). Two sample questions, for which most people would not respond on either
endpoint of the scale are “I rarely appreciate criticism” and “I am very confident of my
judgments.” A Cronbach’s alpha of .92 was calculated for this sample.

Analysis

Data entry was verified by a random check across all instruments in the study of 10% of
the data. Less than .01% error was found. SPSS was used for data entry and analysis. Each
of the research questions is examined using a simple χ 2 or t test analysis to compare the
differences between the two groups of youths. Then a logistic regression is used to examine
how the combined variables help predict to which of the two groups a youth belongs based on
the characteristics of his sexual victimization (Selvin, 1991). Following the logistic regression,
odds ratios are calculated to help clarify the meaning of the regression in terms of specific
numerical combinations of variables.

Results

Of the 272 sexual offenders, 79.4% (n = 216) reported having been sexually victimized,
while a much smaller percentage, 46.7% (n = 93) of the 199 nonsexual offending delinquents
reported being sexually victimized (χ 2 = 54.39, df = 1, p < .000). Based on these rates of
prior victimization, the two groups of sexually victimized youths were compared on a number
of variables resulting in a total of 309 youths. The results of the analysis of the data from
these youths are reported in both Figure 1 and Table 1 (with the exception of duration) and are
discussed below. Not all youths answered all questions.
Using the BIDR, there were no significant differences between the groups on either of the
scales, nor were the scores correlated with any of the other variables in this study; therefore,
the social desirability measure is not used in further analysis.
The sexual offenders’ sexual perpetrators (potentially across many perpetrators) were more
likely to be related to the youth than the nonsexual offending delinquents’ perpetrators (χ 2 =
24.04, df = 3, n = 269, p = .000). As shown in Figure 1, the sexual offending group is more
likely to have been sexually victimized by a relative or a parent.
When comparing the 176 youths for whom the gender(s) of their perpetrator(s) was(were)
either male or female, the sexual offenders were much more likely to have been offended
against by men (χ 2 = 9.7, df = 1, p = .002).
While Garland and Dougher (1990) discussed male versus female perpetrators, a large
number of youths in this sample were sexually victimized by both males and females. As
900 D.L. Burton et al. / Child Abuse & Neglect 26 (2002) 893–907

Figure 1. Percentages of youths: relationships with perpetrators.

indicated in Table 1, the sexual offenders were significantly more often victimized by both
men and women compared to the nonsexual offending delinquents (χ 2 = 39.19, df = 2,
n = 269, p = .000).
Reported modus operandi of the youth’s perpetrator(s) during their sexual victimization was
collapsed into three categories for the 253 youths for whom data were available: (1) coercion

Table 1
Significant χ 2 analyses
Variable Nonsexual offending Sexual offending
delinquents (n = 93; %) delinquents (n = 216; %)

Gender of perpetrator (two groups)a


Male(s) only 19.7 42.7
Female(s) only 80.3 57.3
Gender of perpetrator(s) (three groups)b
Male(s) and female(s) 9.6 43.9
Male(s) only 17.8 24.1
Female(s) only 72.6 32.0
Modus operandic
Physical force 19.7 48.4
Threats 9.8 7.3
Games, favors, baby-sitting 70.5 44.3
Worst victimization reportedd
Penetration 46.2 75.0
Fondling 10.8 9.3
Exhibitionism 43.0 15.7
a
χ 2 = 9.7, df = 1, n = 176, p = .002.
b
χ 2 = 39.19, df = 2, n = 269, p < .000.
χ = 15.9, df = 2, n = 253, p < .000.
c 2

χ = 28.44, df = 3, n = 309, p < .000.


d 2
D.L. Burton et al. / Child Abuse & Neglect 26 (2002) 893–907 901

by the youth’s perpetrator(s) by means of playing games with the youth, doing favors for
the youth or baby-sitting the youth; (2) threats to the youth and/or to the youth’s families;
or (3) the use of physical force during sexual victimization. Youths were placed in only one
category and were placed in the category of the most severe modus operandi they reported
across all victimization incidents. The sexual offenders were much more likely to report that
their perpetrators used physical force than the nonsexual offending delinquents (χ 2 = 15.9,
df = 2, p < .000).
When comparing the most severe type of sexual victimization reported by the 309 youths, the
sexual offenders reported more severe victimization than the nonsexual offending delinquents
(χ 2 = 28.44, df = 2, p = .000).
The average duration of sexual victimization for the sexual offenders (n = 203) was
5.6 years. For the nonsexual offending delinquents (n = 80) the average duration was
3.9 years, a significant difference between the two groups (t = −2.13, n = 283, p =
.034).
Reported age at the onset of sexual victimization did not differ between the two groups.
Average age at onset of sexual victimization (n = 309) was 8.3 years (SD = 3.85) for both
groups, 7.6 years (SD = 3.40) for the nonsexual offenders and 8.8 years (SD = 3.97) for the
sexual offenders. The modal response for both groups was 5 years.
In order to examine the relative strength of each of the variables in predicting whether a
youth would be a sexual offender, a logistic regression was conducted. All of the variables
above, except for age at onset of victimization, were included in the analysis. Age at onset was
not included because in the bivariate analysis it was not related to the dependent variable. The
three ordinal variables are treated as interval in this analysis.
As illustrated in Table 2, two of the variables were significant in predicting whether an
individual was a sexual offender: modus operandi of the victimizer and the gender(s) of the
perpetrators being both male and female. Being victimized by force and being victimized
by both males and females were critical in discriminating between these groups. The model
successfully predicted sexual offender status for 78.3% of the youths. The model is quite
effective at correctly predicting which group the sexual offenders belong to (97.2% correctly
classified), but poor at predicting to which group the nonsexual offenders belong to (6.7%
correctly classified).
To illustrate these results, three case scenarios were examined and relative odds ratios were
calculated. The following reference case was used to compare other victimization scenarios.
In this reference case, the duration of the sexual victimization was 1 month long, the youth
was offended against by a stranger, the youth’s perpetrator used favors, games or baby-sat and
did not use any physical force, the youth suffered exhibitionism as the worst offense, and the
youth was offended against by a woman.
A second case is one in which the duration of sexual victimization was 10 years, the youth
was offended against by a parent, the youth’s perpetrator used physical force, the youth suffered
penetration, and the youth was offended against by men. According to the logistic regression
a youth with this profile would be 14.4 times as likely as the reference case to be a sexual
offender. However, a youth whose profile was the same as the second case except that he was
offended against by both men and women would be 23.1 times more likely to be a sexual
offender than the reference case.
902 D.L. Burton et al. / Child Abuse & Neglect 26 (2002) 893–907

Table 2
Logistic regressiona
Independent variablesb Regression Standard Lower Upper Wald
coefficient error confidence confidence statistics
interval interval
Constant −1.2897 .7838 .06 1.28 2.70
Modus operandic .4723 .2308 1.02 2.52 4.19∗
Gender of perpetrator (two groups)d .9339 .4752 1.00 6.46 3.86∗
Closeness of perpetratore .3817 .2662 .87 2.47 2.05
Durationf −.0505 .0528 .86 1.05 .9155
Gender of perpetrator (male only)d .2747 .5774 .42 4.08 .22
Severity (worst victimization reported)g .3517 .3974 .65 3.10 .78
a
χ 2 = 24.86, df = 6, n = 221, p = .0004.
b
The dependent variable is coded as 1 for a sexual offender and 0 for a nonsexual offender.
c
For modus operandi, a higher score is more severe force; favors and so forth was coded as 1, coercion as 2 and
physical force as 3.
d
For gender, two dichotomous variables were each dummy coded as 1. The omitted category for these two
variables is “victimized by a woman only.”
e
For relationships, stranger was coded as 1, friend/neighbor as 2, relative as 3, and parent as 4.
f
Duration was entered as a continuos variable in years and fractions of years (.083 = 1 month).
g
For severity of victimization, a higher score is worse with exhibitionism coded as 1, fondling as 2 and penetration
as 3.

p < .05.

Discussion

While not surprising, the fact that a significantly larger percentage of sexually offend-
ing youths reported sexual victimization than the nonsexually offending delinquent
youths confirms both theory and previous empirical data. The percentage of sexual offend-
ers reporting sexual victimization is higher than that found in previous studies, even when
youths were asked post-treatment (the youths in this study were, on average, in the middle
of treatment). These strong results may be due to a number of factors unique to this
study.
First, this study was anonymous, which may have resulted in more willingness to disclose
victimization. Second, the sample reflects relatively severe offenses. The primary institution in
this study houses the state’s oldest, most noncompliant, and most severely offending youths. It
is the largest high-security facility in the state and is also a residential facility, unlike many of
the settings in which other research was conducted. Finally, of note, all of the youths had been
in an intensive treatment program for some time and may have been more willing to disclose
their victimization.
All of the research hypotheses in this study were supported with the exception of age at
onset of sexual victimization. The juvenile sexual offenders in the sample were more likely to
have had perpetrator(s) who are related to them, to have had men offending against them, to
have suffered greater duration of sexual victimization, to have a perpetrator use force against
them, and to be penetrated orally or anally during their victimization. It should be noted that
D.L. Burton et al. / Child Abuse & Neglect 26 (2002) 893–907 903

the classification resulted in many false positives for the nonsexual offender group, limiting
its application.
The fact that there was no difference between the groups on age at the onset of the sexual
victimization may be due to the nature of retrospective reporting and the response format
of the question. Many of the youths had difficulty reporting the year and month when the
victimization started; many youths did not offer months or answered incorrectly.
The logistic analysis suggests that two variables may be significant in predicting whether a
sexually victimized youth would belong in the sexual offender group: the force of the youth’s
perpetrator(s) and the genders(s) of the youth’s perpetrators. The sexual offenders were more
likely to have male and female perpetrators, although why this should contribute so strongly
to the prediction of sexual offending is not explained or predicted in the current literature.
Perhaps, such youths were more often sexually victimized than were others in the study. They
may have been victimized by more than one adult simultaneously, thus leading to an increase
in trauma. Or, victimization in such situations may be more ritualistic, reinforcing patterns
that might become more normalized in the child’s life. Such youths may have learned that
neither men nor women could be trusted, and thus they felt less protected in their environment,
leading to a greater need to control their worlds through sexual aggression. In cases where
both perpetrators were family members, support for this proposition may be even greater,
since trusted adults were the perpetrators. Finally, if the males and females involved in the
perpetration were indeed relatives, it may be that sexual aggression was more likely to be seen
by the victim as a “normal” behavior to be emulated. The proposition is supported through a
clinical study by Faller (1991) who reports that among a group of 48 polyincestuous families,
“In close to half the cases, a belief system supporting sex between adults and children was
found” (p. 320). Clearly, much more research on perpetrator gender(s) as related to sexual
offending behaviors is needed.

Theoretical implications

Social learning theory is very complex (Bandura, 1986), but some of the principles of
learning, such as repetition of a stimulus and greater duration increasing the likelihood of
one’s ability to repeat what was observed, are very clear as explained by Bandura: “The
more often and the longer the children attended to their model’s behavior, the higher their
level of observational learning” (Bandura, 1986, p. 53). Sexual victimization and trauma are
experiences that result in learning. If the sexual victimization of the youth is characterized by
greater repetition and relative severity, as indicated in the results of this study, we posit that
the youth then has knowledge that is different from the nonvictimized youth and has had a
different learning experience that is more likely to result in sexual offending behavior.
Additionally, perhaps, the youth saw his perpetrator rewarded for his, her or their behav-
ior(s), thereby increasing the “. . . children’s attentiveness to what the models were doing”
(Bandura, 1986, p. 53). The youth may have learned that an offender’s anxiety is reduced or
resolved when an offender offends. Or the youth may have learned that sexual victimization
results in orgasmic pleasure for the offender. Sexual orgasm, as stated earlier, may lead to
further conditioning (Ryan, 1989; Worling, 1995) and to further sexual offending. Finally, a
youth may have learned, from observation of his own perpetrator(s), that sexual offending is a
904 D.L. Burton et al. / Child Abuse & Neglect 26 (2002) 893–907

way to feel powerful or to feel in control of others (Burton & Fleming, 2000). For all of these
reasons, if the youth’s sexual victimization was more severe and there also was an opportunity
to offend, a youth may choose sexual offending himself as a way to deal with his own feelings
and repeat what he learned from his perpetrator(s).

Research implications

Several research questions are implied by these results. The effects of being victimized by
men and women, versus by men or by women alone, should be explored further. Additionally,
given the high rates of physical and other forms of abuse that these youths experienced in child-
hood, as well as the potential witnessing of domestic, sexual and other forms of violence and
neglect, these factors should also be evaluated for potential contributions to etiological theory
of sexual offending by youths. Multivariate models of analysis, as well as interaction models
using these types of data, are needed due to the relationships demonstrated between the vari-
ables explored above (e.g., gender of perpetrator(s) and modus operandi of the perpetrator(s)
of the youth).
Future research should also include a nondelinquent, nonsexually abusing, sexually victim-
ized sample for comparison (Breiling, personal communication, June 2001). Perhaps those
youths have better early attachment, more resiliency, more positive life experiences or more
supportive families than the youths in this study. Understanding such differences may guide
treatment and prevention. Additionally, samples matched on life experience (e.g., poverty,
social experiences) and other learning variables (e.g., age, IQ, etc.) would provide a better
comparison of these variables. Future studies should also include further analysis of related
questions implied by Bandura.
Finally, a question that must be raised and researched when considering social learn-
ing theory relates to the youths who do not report sexual victimization, but who are sex-
ually aggressive nonetheless. How can their learning be explained? Have they experienced
other forms of victimization and then somehow sexualized them so that they attempt to
resolve their trauma sexually? Are they exposed to more pornography (Friedrich, personal
communication, October 1997) or to harder core pornographic materials (Leguizamo,
2000)?

Practice implications

The large majority of both residential and community based programs for adolescent sexual
offenders provide treatment for the victimization of sexually aggressive youths (Burton &
Smith-Darden, 2001). Most programs provide this service based upon the assumption that
victimization is one of the etiological factors of sexual offending. But few, if any, empirically
based practice models clearly explicate the relationship between victimizing and offending in
a way that indicates exactly how clinical resolution of the effects of sexual victimization can
be achieved, or for whom. Methodologically sophisticated research about the effectiveness of
such treatment compared to other treatments is very much needed. In addition, if victimization
is related to nonsexual criminal behavior as well, clinical work on sexual victimization for
nonsexual offending delinquents may need revisiting.
D.L. Burton et al. / Child Abuse & Neglect 26 (2002) 893–907 905

Limitations

Limitations which should be considered when reading this study include the use of self-
reporting without confirming data from a different source, although a social desirability mea-
sure was used to counter this to whatever degree possible. In addition, the sample is not a
random selection of all possible youths. While it is impossible to obtain a random selection of
all adolescent sexual abusers (as many are not caught), a random selection from incarcerated
youths would be possible. Another consideration is that retrospective reporting has to depend
on the integrity of memory of childhood, which is always in question. Finally, a lack of multiple
comparison groups, including nondelinquent youths limits the applicability of the study.

Conclusions

The development of etiological theory for male adolescent sexual offenders is relatively new.
This project found strong support for some of the tenants offered by Garland and Dougher’s
(1990) work. Thus, social learning theory seems a sound approach to pursue for understanding
the behaviors of adolescent sexual offenders.

Acknowledgments

The authors would like to express appreciation to all of the youths in the study for their time
and effort. Appreciation is extended to Michael Fleming, Stacey Frisinger, Diane Miller, Anne
Burton, Tom Restrick, Linda Schwartz, Judith Block, and Karl Hagstrom and all the graduate
students who interviewed the youths. The authors would also like to thank Drs. William
Birdsall, William Meezan, and Tamara Lyn for suggestions, support, and reading of earlier
drafts of this paper. Finally, thanks to Joanne Smith-Darden for her unwavering assistance.

References

Bandura, A. (1986). Social foundation of thought and action: A social cognitive theory. Engelwood Cliffs, NJ:
Prentice-Hall.
Burton, D., & Fleming, W. (1998). The trauma project initial results report. Unpublished manuscript.
Burton, D., & Fleming, W. (2000, October). “[I felt] ashamed, dirty, used, hurt, confused, excited, and cared for”:
Feelings of adolescent sexual offenders and delinquents about their own victimization and perpetration. Paper
presented at the Association for the Treatment of Sexual Abusers Annual Conference, Vancouver, BC.
Burton, D., Nesmith, A., & Badten, L. (1997). Clinician’s views of sexually aggressive children: A theoretical
exploration. Child Abuse & Neglect, 21, 157–170.
Burton, D., & Smith-Darden, J. (2001). The 2000 safer society survey report. Rutland, VA: Safer Society Press.
Cooper, C. L., Murphy, W. D., & Haynes, M. R. (1996)). Characteristics of abused and nonabused adolescent sexual
offenders. Sexual Abuse: A Journal of Research and Treatment, 8, 105–119.
Faller, K. C. (1989). The role relationship between victim and perpetrator as a predictor of characteristics of
intrafamilial sexual abuse. Child and Adolescent Social Work, 6, 217–229.
Faller, K. C. (1991). Polyincestuous families: An exploratory study. Journal of Interpersonal Violence, 6, 310–322.
906 D.L. Burton et al. / Child Abuse & Neglect 26 (2002) 893–907

Garland, R., & Dougher, M. (1990). The abused/abuser hypothesis of child sexual abuse: A critical review of theory
and research. In J. Fierman (Ed.), Pedophilia: Biosocial dimensions (pp. 488–509). New York: Springer.
Hanson, R., & Slater, S. (1988). Sexual victimization in the history of child sexual abusers: A review. Annals of
Sex Research, 1, 485–499.
Hunter, J., & Figueredo, A. (2000). The influence of personality and history of sexual victimization in the prediction
of juvenile perpetrated child molestation. Behavior Modification, 24(2), 241–263.
Kaufman, K., Hilliker, D., & Daleiden, E. (1996). Subgroup differences in the modus operandi of adolescent sexual
offenders. Child Maltreatment, 1, 17–24.
Kroner, D. G., & Weekes, J. R. (1996). Balanced inventory of desirable responding: Factor structure, reliability,
and validity with an offender sample. Personality & Individual Differences, 21, 323–333.
Leguizamo, A. (2000). Differences in object relations between sexual offenders and sexual offending delinquents.
Unpublished dissertation, University of Michigan, Ann Arbor, Michigan.
Ryan, G. (1989). Victim to victimizer: Rethinking victim treatment. Journal of Interpersonal Violence, 4, 325–341.
Ryan, G., Miyoshi, T., Metzner, J., Krugman, R., & Fryer, G. (1996). Trends in a national sample of sexually
abusive youths. Journal of the American Academy of Child & Adolescent Psychiatry, 35(1), 17–25.
Ryan, S., Rodriguez, J., Anderson, R., & Foy, D. (1992). Psychometric analysis of the sexual abuse exposure
questionnaire (SAEQ). Paper presented at the American Psychological Association, Washington, DC.
Selvin, S. (1991). Statistical analysis of epidemiologic data. NY: Oxford University Press.
Skuse, D., Bentovim, A., Hodges, J., Stevenson, J., Andreou, C., Lanyado, M., New, M., Williams, B., & Macmillan,
D. (1998). Risk factors for development of sexually abusive behavior in sexually victimised adolescent boys:
Cross sectional study. British Medical Journal (International ed.), 317(7152), 175–180.
Snyder, H., & Sickmund, M. (1999). Juvenile offenders and victims: 1999 national report. Washington, DC:
National Center for Juvenile Justice.
United States General Accounting Office. (1996). Cycle of sexual abuse (GAO/GGD–96–178). Washington, DC:
United State Government Printing Office.
Watkins, B., & Bentovim, A. (1992). The sexual abuse of male children and adolescents: A review of current
research. Journal of Child Psychology an Psychiatry, 33, 197–248.
Widom, C. (1989). Does violence beget violence? A critical examination of the literature. Psychological Bulletin,
106, 3–28.
Worling, J. R. (1995). Sexual abuse histories of adolescent male sex offenders: Differences on the basis of the age
and gender of their victims. Journal of Abnormal Psychology, 104, 610–613.

Résumé

Objectif: Cet article vérifie des hypothèses basées sur le concept de Garland et Dougher (1990) à savoir
la relation entre la victime et l’agresseur. Plus précisément, c’est une approche qui explique l’agression
de l’adolescent à partir de la théorie de l’apprentissage social.
Méthode: Ont participé à une étude, 216 adolescents agresseurs qui avaient eux-mêmes été agressés
sexuellement et 93 adolescents agressés qui n’étaient pas des agresseurs. Ces participants, qui faisaient
partie d’un programme thérapeutique dans l’état du Michigan aux E-U., ont été interviewés par le biais
d’un sondage anonyme à propos de leurs agressions subies et de leurs actes en tant qu’agresseurs. On
les a aussi testés au moyen d’une échelle mesurant leur comportement social.
Résultats: Ayant comparé les deux groupes, on a noté que (1) les adolescents agresseurs avaient eu
des liens plus étroits avec leur agresseur; (2) une plus grande probabilité que l’agresseur fut du sexe
masculin; (3) les agressions étaient d’une plus longue durée; (4) il y avait eu agression par la force;
et (5) les adolescents avaient vécu un acte de pénétration. Une régression logique indique que deux
facteurs prédisent de façon importante l’appartenance au groupe des agresseurs: si leurs agresseurs ont
appartenu aux deux sexes, et si l’agression était accompagnée par de la force.
D.L. Burton et al. / Child Abuse & Neglect 26 (2002) 893–907 907

Conclusions: L’étude a, de façon générale, appuyé l’hypothèse basée sur la théorie de l’apprentissage
social. Des recherches à variables multiples devront approfondir nos connaissances sur cette théorie.

Resumen

Objetivo: Este artı́culo presenta una investigación que pone a prueba hipótesis basadas en la formulación
de Garland y Dougher (1990) sobre la conversión “de vı́ctima en agresor.” Se trata de una aproximación
etiológica sobre los agresores sexuales basada en la teorı́a del aprendizaje social.
Método: Participaron en el estudio 216 adolescentes ofensores sexuales y vı́ctimas de abuso sexual
y 93 adolescentes vı́ctimas de abuso sexual y delincuentes pero no ofensores sexuales, todos ellos
seleccionados en tres recursos para ofensores sexuales de Michigan. Se les aplicó una encuesta anónima
relacionada con su victimización sexual y sus agresiones sexuales, ası́ como una medida de deseabilidad
social.
Resultados: Los delincuentes ofensores sexuales se diferencian de los delincuentes no ofensores sexu-
ales en los siguientes aspectos: tienen una relación más cercana con sus agresores, una mayor probabil-
idad de que el agresor fuera un varón, una duración mayor de las agresiones sexuales y mayor presencia
de la fuerza y mayores tasas de penetración en dichas agresiones. Los análisis de regresión logı́stica
señalan que las variables “género de los agresores” y “utilización de la fuerza fı́sica” son los mejores
predictores de que las vı́ctimas de abuso sexual se conviertan en agresores sexuales.
Conclusiones: Se apoyan en general las hipótesis de la teorı́a del aprendizaje social. Se considera
necesario avanzar en la utilización de diseños multivariados en relación con esta teorı́a.

You might also like